Home |
Search |
Today's Posts |
#1
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
I am toying with a mew amplifier design incorporating several unusual design features, and the output transformer I am planning on using will not tolerate unbalanced DC in the primary. Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif A couple of points to note, the two tubes are connected in series for DC thereby insuring equal DC currents through the two halves of the OPT. For AC (audio) the tubes are connected in the normal push-pull configuration. The current through the both tubes is determined by the bias applied to the lower tube, this bias is provided by a cathode bias resistor in the example schematic, but could fixed bias could also be used. The grid of the upper tube is biased at approximately half of the B+ supply voltage by two resistors so that the B+ voltage is split equally between the two tubes and each tube operates with normal B+ voltage across it. This scheme does have several unusual requirements which my make it unsuitable for some applications. 1. It requires an output transformer where each half of the primary has separate leads, with no common "center tap" lead. 2. It requires twice the normal B+ voltage. 3. It requires a separate heater transformer for the upper tube to avoid undue heater to cathode voltage stress. Comments or thoughts? -- Regards, John Byrns Surf my web pages at, http://fmamradios.com/ |
#2
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
John Byrns wrote: I am toying with a mew amplifier design incorporating several unusual design features, and the output transformer I am planning on using will not tolerate unbalanced DC in the primary. Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif A couple of points to note, the two tubes are connected in series for DC thereby insuring equal DC currents through the two halves of the OPT. For AC (audio) the tubes are connected in the normal push-pull configuration. The current through the both tubes is determined by the bias applied to the lower tube, this bias is provided by a cathode bias resistor in the example schematic, but could fixed bias could also be used. The grid of the upper tube is biased at approximately half of the B+ supply voltage by two resistors so that the B+ voltage is split equally between the two tubes and each tube operates with normal B+ voltage across it. This scheme does have several unusual requirements which my make it unsuitable for some applications. 1. It requires an output transformer where each half of the primary has separate leads, with no common "center tap" lead. 2. It requires twice the normal B+ voltage. 3. It requires a separate heater transformer for the upper tube to avoid undue heater to cathode voltage stress. Comments or thoughts? The circuit would best suit pure class A performance. But your'e right, fixed bias on the btm tube sets the bias for both. But in class AB with fixed bias the top tube hasn't got fixed bias so the Ea will vary, increasing THD etc. The use of an 800V supply isn't needed for a typical class A circuit with 2 x EL34 or 6550. You could have a +400V and -400V supply which lessens the large Pd between 1/2 the Pri and 0V of the sec. But you have to reference the top tube cathode to -400V at ac with a large electro. Thus the btm 1/2 of the Pri is at near 0V potential but we cannot connect it directly to 0V because then the Ia dc for each tube could vary. Its very easy to see why few manufacturers ever sold samples of this circuit that has been around for at least since 1950. Too much effort and parts wheras a normal Leak type output stage works well enough. If you really want to prevent effects of dc imbalance without using SS, then just use much larger values of cathode resistors in a normal class A output stage by means of having a -100V supply for the low end of the Rk, while biasing the grids at 0V and allowing the Ek to be about 45V, with Ea at say 400V. Then if Ia was 50mA, then the Rk would be 2,900 ohms and dissipating 7.3W. But once class AB commences then the Ek will vary easily. If you want to make the OPT less prone to dc offsets then consider using an air gap. Now this may sound like heresy in a PP design but consider that if the RLa-a was say 10k for class A and that we should have the reactance of Lp = 10k at 20Hz at least, to avoid most effects of adverse reactive loading at bass F. Therefore Lp should be 80H. In the oroginal Williamson, with GOSS, the maximum Lp at high Bm was maybe 600H, but was lower at low Bm where say Va-a was only 5Vac. DTN W said you should have 100H at the low Bm level to ensure LF stability with the 20dB of NFB. It was barely ever stable though. If you use a gain reduction network to reduce gain and phase shift below 20Hz then you can get away with a lot less Lp. If you have GOSS core material you will perhaps have maximum µ of the iron when fully interleaved = 17,000. It is with the stuff I buy here. Such a high µ is entirely unecessary, allowing for the fact it will be perhaps only 1,700 at say 5Va-a levels of output, and still give an Lp with reactance = RLa-a at 20Hz. If you bunched the Es and the Is in sub-piles of say 15 lams thick each and stack in the bundles in alternate directions, then the core becomes partially air gapped, and µ max is drastically reduced. Even with just close butting all E and I like an SE OPT will give a µ of around 1,000 with GOSS, and with staggered stacking in 15 lam thick bunches maybe you get µ = 2,000, still enough to have high enough Lp at high Bm, ie, and high output levels. At low output levels the lower Lp and low shunt reactance would not matter too much because the NFB corrects it all. C-cores also allow easy gapping, and in my 50W class A mosfet based amp with an OPT, I have C-cores with highest µ = 4,500 with neat butted iron, but with a film of plastic taken from a shopping bag the µ fell to about 2,000. This made the OPT less prone to any dc offset, although I found that with fixed bias the mosfets don't vary their bias much, and if the they get hot they bias offwards a bit to ease matters. So I have fixed bias and no adjust pots or Vbe stage. The other potential problem with all PP OPT coupled OP stages is the tendency for saturation as F goes low no matter what you do. If for example you make a very nice design which has BW of say 3Hz to 60kHz at low levels, and then you test the amp with true pink noise with a BW of say 2Hz to 20kHz, then you will find that the amp makes a terrible knocking noise as you raise the tests input signal. Its a knocking noise that is irregular, and caused by the OPT saturating as signal with plenty of random F below 14Hz are present. Such huge current peaks in the tubes or mosfets cause massive IMD to all other F. Basically, while the saturation lasts, the load is just the winding resistance, and there is no inductance. Having low P inductance would be preferable to just allowing the saturation. Hence the need to reduce the iron µ by fiddling with lamination assembly or putting in a slight gap. But you also need to have an input filter to limit the BW to say 14Hz at LF. This will immediately stop most of the knocking noise you'll here with a pink noise signal while testing, and you'll be able to raise the input level and you'll find the clipping will first begin to occur occur on F above 14Hz, and your amp will be about OK. The Williamson had 4,400 turns around Afe = 1,408 sq.mm. 16W into 10k a-a means Vaa = 400Vrms. But suppose Rl was 4ka-a, and Vaa was 400V, the PO = 40W, and in my designs I might use a typical Afe = 2,728, ie, Afe is twice the size of the DTN amp. I will then use typical Np = 2,200t, and inductance is 1/2 the W OPT. In one samle I used plain NOSS with max µ = 3,500 only, and yet the amp has fabulous bass. because the turns are less the winding R is lower and losses are lower, but there isn't any problem with saturation from sub-sonics. In summary, the effects of dc offset AND related LF saturation efects can be largely overcome during high level use by careful core arrangement, and a sensible input filter. (In Pa amps the LF cut off was often 75Hz.) At low level use, core saturation rarely if ever occurs, even with very low cut off and no attempt to reduce the core µ. I know this thus raises the question of why use you pet series arrangement, but you did ask for comments. Patrick Turner. -- Regards, John Byrns Surf my web pages at, http://fmamradios.com/ |
#3
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
John Byrns wrote:
I am toying with a mew amplifier design incorporating several unusual design features, and the output transformer I am planning on using will not tolerate unbalanced DC in the primary. Patrick says it's not new, I notice, but he doesn't cite a particular precedant. Anyway, what are the unusual design features? Why are you planning to use an intolerant OPT? It seems perverse to deliberately create a problem to fit a solution. If there is a good alternative reason for your intolerant transformer plan, perhaps you could say what it is? Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. Why? Is this pollution of your idea of purity of design, or of the signal? Also, there may be other alternatives. Perhaps routine manual bias resetting? Just how intolerant are you planning this transformer to be? After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif What's that odd resistor for? Don't it need adjusting for Vak balance? What's the penalty for Vak being out of balance as the valves drift? Are there any other problems that might arise from asymmetrical grid resistances? The only good reason for planning transformer intolerance is the hope of achieving better low frequency performance AFAIK. That is, you sacrifice flexibility by maximising primary inductance. I wonder therefore how good your design is at maintaining perfect full power AC balance at LF? Particularly if it's running in AB. If LF AC balance isn't perfect, then surely your intolerant transformer will complain? A couple of points to note, the two tubes are connected in series for DC thereby insuring equal DC currents through the two halves of the OPT. For AC (audio) the tubes are connected in the normal push-pull configuration. The current through the both tubes is determined by the bias applied to the lower tube, this bias is provided by a cathode bias resistor in the example schematic, but could fixed bias could also be used. The grid of the upper tube is biased at approximately half of the B+ supply voltage by two resistors so that the B+ voltage is split equally between the two tubes and each tube operates with normal B+ voltage across it. This scheme does have several unusual requirements which my make it unsuitable for some applications. 1. It requires an output transformer where each half of the primary has separate leads, with no common "center tap" lead. 2. It requires twice the normal B+ voltage. 3. It requires a separate heater transformer for the upper tube to avoid undue heater to cathode voltage stress. Comments or thoughts? -- Regards, John Byrns Surf my web pages at, http://fmamradios.com/ |
#4
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
In article ,
"Ian Iveson" wrote: John Byrns wrote: I am toying with a mew amplifier design incorporating several unusual design features, and the output transformer I am planning on using will not tolerate unbalanced DC in the primary. Patrick says it's not new, I notice, but he doesn't cite a particular precedant. I'm sure Patrick is correct; I would be surprised if it hadn't been done before, there is nothing new under the sun. Anyway, what are the unusual design features? The actual amplifier design features are largely orthogonal to this idea, anyway I discussed them in this forum a few years back but never built the amplifier because of potential problems with the output transformer from my junk box. Why are you planning to use an intolerant OPT? Because I have one in my junk box and the original plan was to build the amplifier largely with parts I already had on hand. It seems perverse to deliberately create a problem to fit a solution. I'm not sure I follow you here, I had the problem and tried to create a solution to fit the problem, not vice versa. If there is a good alternative reason for your intolerant transformer plan, perhaps you could say what it is? See above, I already had the transformer in my junk box. Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. Why? Is this pollution of your idea of purity of design, or of the signal? Purity of design. Also, there may be other alternatives. Probably, however I would like a set it and forget it solution that will work for years without the need for readjustment when tubes are replaced or between tube replacements. Perhaps routine manual bias resetting? See above, I don't like the requirement to have to reset things. Just how intolerant are you planning this transformer to be? I'm not planning how intolerant the transformer will be, it is what it is, it was hot designed as an output transformer and I would like to minimize any potential problems with core saturation. After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif What's that odd resistor for? Could you identify the "odd" resistor, it isn't clear which resistor you are speaking of? Don't it need adjusting for Vak balance? No, that is why the grid bias voltage on the upper tube is set to approximately one half of the B+ voltage by the resistive voltage divider so that the Vak is balanced. What's the penalty for Vak being out of balance as the valves drift? The Vak should not drift by more than a couple of volts as the tubes drift due to the way the upper tube is biased. Therefore any penalty for Vak being out of balance should not be invoked. Are there any other problems that might arise from asymmetrical grid resistances? The grid resistances are equal so I haven't considered what problems might arise if they weren't equal. The only good reason for planning transformer intolerance is the hope of achieving better low frequency performance AFAIK. Another good reason is the desire to make use of a part you already have on hand. That is, you sacrifice flexibility by maximising primary inductance. I wouldn't know about that one way or the other. I wonder therefore how good your design is at maintaining perfect full power AC balance at LF? Maintaining perfect full power AC balance at LF is not an issue that I am concerned with, my design is not atypical of push-pull valve amplifiers in this regard. Particularly if it's running in AB. If LF AC balance isn't perfect, then surely your intolerant transformer will complain? It is a class A design. A couple of points to note, the two tubes are connected in series for DC thereby insuring equal DC currents through the two halves of the OPT. For AC (audio) the tubes are connected in the normal push-pull configuration. The current through the both tubes is determined by the bias applied to the lower tube, this bias is provided by a cathode bias resistor in the example schematic, but could fixed bias could also be used. The grid of the upper tube is biased at approximately half of the B+ supply voltage by two resistors so that the B+ voltage is split equally between the two tubes and each tube operates with normal B+ voltage across it. This scheme does have several unusual requirements which my make it unsuitable for some applications. 1. It requires an output transformer where each half of the primary has separate leads, with no common "center tap" lead. 2. It requires twice the normal B+ voltage. 3. It requires a separate heater transformer for the upper tube to avoid undue heater to cathode voltage stress. Comments or thoughts? -- Regards, John Byrns Surf my web pages at, http://fmamradios.com/ |
#5
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
Ian Iveson wrote: John Byrns wrote: I am toying with a mew amplifier design incorporating several unusual design features, and the output transformer I am planning on using will not tolerate unbalanced DC in the primary. Patrick says it's not new, I notice, but he doesn't cite a particular precedant. Anyway, what are the unusual design features? Why are you planning to use an intolerant OPT? It seems perverse to deliberately create a problem to fit a solution. The circuit John has presented has been invented probably before 1960 and I saw it cited in some ancient magazine, probably an old copy of Wireless World. However, maybe it is different in detail to old ideas of preventing dc imbalance with series tube PP amps. If one really wants to keep out the the Idc and use series tubes, then you can have a series pair with an OPT which is cap coupled to one end of a single primary, with the other end taken to 0V. No dc in the winding at all. If there is a good alternative reason for your intolerant transformer plan, perhaps you could say what it is? Just about all PP OPT set up conventionally are intolerant of dc imbalance because they have ungapped cores and with a high max µ, so the slightest dc imbalance causes serious increases in THD/IMD. Again if nobody objects to weight and size and cost then there you can have choke feed to a pair of output tubes using a choke with CT and then cap couple the OPT which has its CT grounded. This allows normal drive, but then the choke suffers the imbalance if it occurs. But then if the coke is gapped, and had a massive sive and lots of turns then it may withstand far more dc imbalance than any normal OPT. Or one might use TWO individual air gapped chokes, but you can't get the large wanted inductance to prevent inductance shunting the load. Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. Why? Is this pollution of your idea of purity of design, or of the signal? Also, there may be other alternatives. Perhaps routine manual bias resetting? Just how intolerant are you planning this transformer to be? After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif What's that odd resistor for? Don't it need adjusting for Vak balance? What's the penalty for Vak being out of balance as the valves drift? Are there any other problems that might arise from asymmetrical grid resistances? The only good reason for planning transformer intolerance is the hope of achieving better low frequency performance AFAIK. That is, you sacrifice flexibility by maximising primary inductance. I wonder therefore how good your design is at maintaining perfect full power AC balance at LF? Particularly if it's running in AB. If LF AC balance isn't perfect, then surely your intolerant transformer will complain? Indeed. The OPT performance between dc and say 20Hz is a grey area which needs careful consideration. Where one does have two series tubes with say an 800V supply, one can cap couple a single winding OPT as I stated above. Its an easy OPT to wind compared to a conventional PP OPT with CT because it has half the P turns. The load is a lot lower than a normal PP RLa-a load. In the late 1950 Philips made a range of amps using 2 x EL86 in series with a supply = +400Vdc, and the anode cathode junction was at +200V and there was no OPT. The speaker had an 800 ohm voice coil impedance and was driven by an electro cap off the a-k join. But there still was a kind of small transformer to get the screen voltage of the top tube to follow its cathode voltage while being held at +400V. A choke would have worked fine though with cap bypass coupling to the cathode. The bottom tube just has normal fixed screen supply and at 1/2 the supply voltage for both tubes, ie, +200V. While working in class A the load of 800 is shared between the two output tubes and each tube sees 1,600 ohms. If one runs such an amp only in class A then with 12W Pda in each EL85, each tube puts 5W into 1,600 ohms or 10W into the 800 ohms. If you wanted to use the EL86 in a normal PP amp under the same Ea/Ia conditions then the OPT would have a primary load of 3k2 a-a, ie, with twice the turns of the single winding OPT. To get drive meant a special bootstrapped circuit was used with a 12AX7. But where the top tube has to have a large grid signal delivered, one may have a 1:1 IST with two windings. One winding is driven with a CF tube which is driven by a gain triode to make say 30Vrms. The other winding has one end connected to the a-k join of the two power tubes and thus the drive to the top tube is always equal to the bottom tube and there is symetrical drive conditions. NFB is applied the normal way, but with some care about gain/phase shift compensation networks because you have an IST in the signal path. To avoid having huge drive voltage and be able have an OPT with much lower RL, and also have no screen voltage to worry about then the 6AS7 or 6C33C come to mind as excellent canditates. But one could run EL34 or KT88 with Ea at 250V OK if one doesn't mind having a screen drive choke. 4 x KT88 would give 40W PP class A and load would be 560 ohms. The cap to couple it can be 100uF, and Lp would need to be at least 20H. The LF pole for resonance is 3.6Hz, and the load to damp the resonance is about 600ohms or less, so when loaded the OPT won't have a peaked LF response. Such a peak can make NFB difficult to apply. The higher the LP, the lower is the Fo, and one could also have the coupling cap value a lot higher, say 470uF, but then onmce you move Fo down to say 0.36Hz, then one is i the grey area with the cap charging and discharging an the core of the OPT is subject to slow ac changes and the resulting temporary saturation effects. As I mentioned in a previous post, If one doesn't like series PP tubes, then one may as well bite the bullet and have a large PP OPT set up conventionally, and then just air gap the PP core. Oh, and all series output tube connections require a biased heater supply for one of the pair of tubes. Philips didn't bother though because the EL86 has a high heater-cathode voltage rating. Patrick Turner. A couple of points to note, the two tubes are connected in series for DC thereby insuring equal DC currents through the two halves of the OPT. For AC (audio) the tubes are connected in the normal push-pull configuration. The current through the both tubes is determined by the bias applied to the lower tube, this bias is provided by a cathode bias resistor in the example schematic, but could fixed bias could also be used. The grid of the upper tube is biased at approximately half of the B+ supply voltage by two resistors so that the B+ voltage is split equally between the two tubes and each tube operates with normal B+ voltage across it. This scheme does have several unusual requirements which my make it unsuitable for some applications. 1. It requires an output transformer where each half of the primary has separate leads, with no common "center tap" lead. 2. It requires twice the normal B+ voltage. 3. It requires a separate heater transformer for the upper tube to avoid undue heater to cathode voltage stress. Comments or thoughts? -- Regards, John Byrns Surf my web pages at, http://fmamradios.com/ |
#6
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
On Jun 18, 6:58*pm, John Byrns wrote:
In article , *"Ian Iveson" wrote: John Byrns wrote: (snip) Why are you planning to use an intolerant OPT? Because I have one in my junk box and the original plan was to build the amplifier largely with parts I already had on hand. Hi, John, you've stolen my shtick... that's what I do! However, I'd rather not compromise (too much!) on OPT's. They may not be Partridge but they're not that bad... I've a pair of Hammond 1650F's and a salvaged MacIntosh unit standing by for when I get a "round tuit". Very best, Roger |
#7
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
Patrick Turner wrote in message
The circuit John has presented has been invented probably before 1960 and I saw it cited in some ancient magazine, probably an old copy of Wireless World. However, maybe it is different in detail to old ideas of preventing dc imbalance with series tube PP amps. If one really wants to keep out the the Idc and use series tubes, then you can have a series pair with an OPT which is cap coupled to one end of a single primary, with the other end taken to 0V. No dc in the winding at all. If there is a good alternative reason for your intolerant transformer plan, perhaps you could say what it is? Just about all PP OPT set up conventionally are intolerant of dc imbalance because they have ungapped cores and with a high max µ, so the slightest dc imbalance causes serious increases in THD/IMD. Tolerance is quite a complicated thing when it comes to transformers. The story goes that the more evenly a transformer magnetises the core, the greater the inductance that can be achieved for a given amount of iron, the more linear the relationship between inductance and current, and the less leakage, particularly if the core has no sharp corners. Because of the more effective use of the iron, less can be used, making for a comparitively dainty transformer with lower capacitance than an EI of the same inductance. So if you want plenty inductance for bottom end grunt combined with low capacitance and leakage for a top end extended enough to use stacks of feedback, then you should use a toroid, or an "R-core", or C-core, in that order. The story continues with a downside: the more even the field distribution, the more sharply the core saturates, especially if less iron has been used. This is where the toroid's reputation for DC intolerance comes from, I think. Perhaps you could make a better toroid? If you were to use the same core area as you would for an EI, would that make it less intolerant? If you aimed for the same HF performance as you would for an EI, you could make a humungous toroid, I would guess. Vanderveen, of toroid fame, shows a sketch of the output resulting from poor AC balance. From memory it looks like the signal disappears around the voltage zero crossing points. He doesn't explain how this happens though...it's just shown to illustrate how to adjust it. So I have a question. Taking John's transformer as an example, each half of the primary winding will on its own have a certain inductance, which will be doubled if they are both operating in unison, because twice the windings gives four times the inductance. Consequently, any unbalanced part of the signal will see half the load, and so result in twice the current for a given voltage amplitude (and so a zero at twice the frequency?). Question is, will an intolerant transformer find this more of a problem at low frequencies than a tolerant one? This is part of a more generic question because a conventional PP output stage is never quite balanced anyway, obviously particularly when operating in AB. I was set off on this train of thought by John's assymetric cathodes. I guess he can just make those caps really big, so their effects are both well below a bandwidth determined elsewhere. Again if nobody objects to weight and size and cost then there you can have choke feed to a pair of output tubes using a choke with CT and then cap couple the OPT which has its CT grounded. This allows normal drive, but then the choke suffers the imbalance if it occurs. But then if the coke is gapped, and had a massive sive and lots of turns then it may withstand far more dc imbalance than any normal OPT. Or one might use TWO individual air gapped chokes, but you can't get the large wanted inductance to prevent inductance shunting the load. And by this time your amp weighs half a ton and costs a fortune. Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. Why? Is this pollution of your idea of purity of design, or of the signal? Also, there may be other alternatives. Perhaps routine manual bias resetting? Just how intolerant are you planning this transformer to be? After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif What's that odd resistor for? Don't it need adjusting for Vak balance? What's the penalty for Vak being out of balance as the valves drift? Are there any other problems that might arise from asymmetrical grid resistances? The only good reason for planning transformer intolerance is the hope of achieving better low frequency performance AFAIK. That is, you sacrifice flexibility by maximising primary inductance. I wonder therefore how good your design is at maintaining perfect full power AC balance at LF? Particularly if it's running in AB. If LF AC balance isn't perfect, then surely your intolerant transformer will complain? Indeed. The OPT performance between dc and say 20Hz is a grey area which needs careful consideration. Where one does have two series tubes with say an 800V supply, one can cap couple a single winding OPT as I stated above. Its an easy OPT to wind compared to a conventional PP OPT with CT because it has half the P turns. The load is a lot lower than a normal PP RLa-a load. In the late 1950 Philips made a range of amps using 2 x EL86 in series with a supply = +400Vdc, and the anode cathode junction was at +200V and there was no OPT. The speaker had an 800 ohm voice coil impedance and was driven by an electro cap off the a-k join. Interesting, thanks. A better proposition now than then, because of the improvement in price and performance of electro caps. But there still was a kind of small transformer to get the screen voltage of the top tube to follow its cathode voltage while being held at +400V. A choke would have worked fine though with cap bypass coupling to the cathode. The bottom tube just has normal fixed screen supply and at 1/2 the supply voltage for both tubes, ie, +200V. While working in class A the load of 800 is shared between the two output tubes and each tube sees 1,600 ohms. If one runs such an amp only in class A then with 12W Pda in each EL85, each tube puts 5W into 1,600 ohms or 10W into the 800 ohms. If you wanted to use the EL86 in a normal PP amp under the same Ea/Ia conditions then the OPT would have a primary load of 3k2 a-a, ie, with twice the turns of the single winding OPT. To get drive meant a special bootstrapped circuit was used with a 12AX7. But where the top tube has to have a large grid signal delivered, one may have a 1:1 IST with two windings. One winding is driven with a CF tube which is driven by a gain triode to make say 30Vrms. The other winding has one end connected to the a-k join of the two power tubes and thus the drive to the top tube is always equal to the bottom tube and there is symetrical drive conditions. NFB is applied the normal way, but with some care about gain/phase shift compensation networks because you have an IST in the signal path. To avoid having huge drive voltage and be able have an OPT with much lower RL, and also have no screen voltage to worry about then the 6AS7 or 6C33C come to mind as excellent canditates. But one could run EL34 or KT88 with Ea at 250V OK if one doesn't mind having a screen drive choke. 4 x KT88 would give 40W PP class A and load would be 560 ohms. The cap to couple it can be 100uF, and Lp would need to be at least 20H. The LF pole for resonance is 3.6Hz, and the load to damp the resonance is about 600ohms or less, so when loaded the OPT won't have a peaked LF response. Such a peak can make NFB difficult to apply. The higher the LP, the lower is the Fo, and one could also have the coupling cap value a lot higher, say 470uF, but then onmce you move Fo down to say 0.36Hz, then one is i the grey area with the cap charging and discharging an the core of the OPT is subject to slow ac changes and the resulting temporary saturation effects. As I mentioned in a previous post, If one doesn't like series PP tubes, then one may as well bite the bullet and have a large PP OPT set up conventionally, and then just air gap the PP core. Conventional PP amps throughout history haven't bothered much about DC balance, beyond offering a means of measuring and adjustment. It shouldn't be hard, these days, to automate that process of periodic maintenance. One interesting thing about this idea of cap coupling, when applied to valves operating in DC parallel, is that the idle currents of the two valves can be independently adjusted for optimum AC performance. Oh, and all series output tube connections require a biased heater supply for one of the pair of tubes. Philips didn't bother though because the EL86 has a high heater-cathode voltage rating. Anyone remember the chap here who made *massively* parallel amps with hundreds of EL86? On the grounds, IIRC, that OPTs become relatively smaller as amps get bigger, so in terms of kilos per kilowatt, the bigger the better? Now EL86 aren't as cheap as they were, and electricity is much more expensive . cheers, Ian |
#8
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
http://www.tubecad.com/2009/04/blog0163.htm
Just a more modest proposal ... not Broskie's Wunderkind, just the simple garters Happy Ears! Al |
#9
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
Ian Iveson wrote: Patrick Turner wrote in message The circuit John has presented has been invented probably before 1960 and I saw it cited in some ancient magazine, probably an old copy of Wireless World. However, maybe it is different in detail to old ideas of preventing dc imbalance with series tube PP amps. If one really wants to keep out the the Idc and use series tubes, then you can have a series pair with an OPT which is cap coupled to one end of a single primary, with the other end taken to 0V. No dc in the winding at all. If there is a good alternative reason for your intolerant transformer plan, perhaps you could say what it is? Just about all PP OPT set up conventionally are intolerant of dc imbalance because they have ungapped cores and with a high max µ, so the slightest dc imbalance causes serious increases in THD/IMD. Tolerance is quite a complicated thing when it comes to transformers. The story goes that the more evenly a transformer magnetises the core, the greater the inductance that can be achieved for a given amount of iron, the more linear the relationship between inductance and current, and the less leakage, particularly if the core has no sharp corners. Because of the more effective use of the iron, less can be used, making for a comparitively dainty transformer with lower capacitance than an EI of the same inductance. I don't know in what context "evenly" rests with regard to core magnetization. Maybe you don't either. But anyway, considering what you are saying, having toroidal cores, R-core or Ccore means you can have a much higher maximum permability and less hysteresis so the distortion at all F is lower. However, to get good bass you need a low Bm and that means a larger core and more turns than most commercial bean counter propelled "dainty" designs you must be referring to. To reduce LL and Csh, one has to do what these commercial people won't do, ie, use more sections and use thicker insualtion material. Were you to design a few OPT according to my logic steps all listed out at my website, you'd understand the need to have a certain number of Kg per watt of audio power. Many makers put only 1/2 the required weight into their OPT, and Quad-II is a perfect example of shoddy OPT design that might be called "dainty" for want of a better word. I have used a 63mm stack of 44mm tongue NOSS E&I lams for a 100W design, the 8585 at my website. Its been made along ther lines of the rest of the theory at my site and it performs excellently and better than most commercial maker OPTs. The NOSS has a max µ of only 3,500, and thus it is more tolerant of dc imbalance than a GOSS core of the same size with max µ of 17,000. The higher µ material gives less THD though. But if the higher µ material was air gapped to give a µe of 3,500, or perhaps even lower, then the tolerance of dc imbalance would be very much improved. A toroidal core is THE WORST type of core one might use if there is any Dc ever likely to be present because its µ max is around 40,000, and the slightest dc offset current will cause the dc magnetization to easily exceed 1.3T and then there is any headroom for ac magnetization because its "hard over" one way, so for 1/2 the wave form the OPT is a virtual short circuit. So if you want plenty inductance for bottom end grunt combined with low capacitance and leakage for a top end extended enough to use stacks of feedback, then you should use a toroid, or an "R-core", or C-core, in that order. I think you know very little if you say that. Perhaps you didn't realise, but if you remove the core of an OPT and run the OPT as an air cored transformer it should work perfectly well above say 5kHz, and the HF stability with FB is determined by the geometry of the copper, and the iron is functionless above 10kHz. The frequencies of HF oscillation if they occur are usually above 30kHz and due to phase shift due to high LL or Csh, or both. At LF, for stability, you could use enormous amounts of inductance, as is the case of the Williamson. But you still get an ultimate 90 degree phase shift with Ra and Lp, and with the other CR couplings there is always a tendency for instability. So to me its utterly pointless using high µ iron for an OPT, and pointless to use huge numbers of P turns because all that makes the dc saturation all the more likely, and the design prone to LF transients, ie, a range of signals between dc and 20Hz which magnetize the iron too much. Huge numbers of P turns increases LL because the LL like Lp is proportional to the square of turns. Twice the turns means 4 times Lp and 4 times the LL. Its far better to have **adequate Lp** and no more. See my website and design an axample to understand more. The story continues with a downside: the more even the field distribution, the more sharply the core saturates, especially if less iron has been used. This is where the toroid's reputation for DC intolerance comes from, I think. Perhaps you could make a better toroid? If you were to use the same core area as you would for an EI, would that make it less intolerant? If you aimed for the same HF performance as you would for an EI, you could make a humungous toroid, I would guess. You could make a toroid with a cut and glue in some gapping material. Its difficult and fragile. Clamped C-cores are better and achieve the same thing. Once the air gap is there it dominates the magnetic path length. Vanderveen, of toroid fame, shows a sketch of the output resulting from poor AC balance. From memory it looks like the signal disappears around the voltage zero crossing points. He doesn't explain how this happens though...it's just shown to illustrate how to adjust it. Just set the bias very low on one tube and high on the other then observe the dreadful even order distortion. Any signal above 1/2 a watt sounds awful. So I have a question. Taking John's transformer as an example, each half of the primary winding will on its own have a certain inductance, which will be doubled if they are both operating in unison, because twice the windings gives four times the inductance. John's circuit works with the same dc current flowing in both tubes and through two P windings. But at ac the operation is the same as any normal PP OPT. It can be made to work in class AB where all Ia in either tube is cut off and each tube does only the crests of each wave. During the class A work the inductance is mutual and high, and once the Ia cuts off the inductance reduces but so does the load. In a class B amp the core still saturates if F goes low enough. And in a class B amp there is no DC present. Devices are biases off. Consequently, any unbalanced part of the signal will see half the load, and so result in twice the current for a given voltage amplitude (and so a zero at twice the frequency?). Question is, will an intolerant transformer find this more of a problem at low frequencies than a tolerant one? Depends on saturation and the amount of dc offset. Once saturation occurs, the winding acts as though it has no L and the dc winding R shunts the load. If you air gap the OPT, the Lp will be much lower, but the OPT will behave without its Lp dissapearing to leave just the Rw. Its better to have some Lp rather than to have plenty of Lp and have it vanish suddenly at say 14Hz. You can make your dainty little OPT if *you* want to. I won't. To make the OPT tolerant to dc imbalance and dc offset, you have to use lots of iron and more turns and maybe an air gap. Thus to get 60 watts of SE power from tubes requires an E&I core 72mm stack 51mm tongue and an air gap and very low µ of perhaps only 500. An SE amp is one which thrives on having a HUGE dc offset. In my SE55 amp with a pair of 845, load was 6k, and Lp abot 40H, Ia = 150mA. If I had a PP amp, the same core could be used to make a 200W AB amp, based on having core saturation occuring at 20Hz at full PO, ie, at the same F as in the case of the SE amp. If you want less iron, and you cut the stack height in half, then the Fsat for PP goes up from 20Hz to 40Hz. Thus is done routinely by most makers to save construction costs and freight costs etc. Dc imbalance worsens... But at 5 watts, a 200W tube amp might have a huge DC imbalance and nobody will hear anything. This is part of a more generic question because a conventional PP output stage is never quite balanced anyway, obviously particularly when operating in AB. I was set off on this train of thought by John's assymetric cathodes. I guess he can just make those caps really big, so their effects are both well below a bandwidth determined elsewhere. Again if nobody objects to weight and size and cost then there you can have choke feed to a pair of output tubes using a choke with CT and then cap couple the OPT which has its CT grounded. This allows normal drive, but then the choke suffers the imbalance if it occurs. But then if the coke is gapped, and had a massive sive and lots of turns then it may withstand far more dc imbalance than any normal OPT. Or one might use TWO individual air gapped chokes, but you can't get the large wanted inductance to prevent inductance shunting the load. And by this time your amp weighs half a ton and costs a fortune. In RDH4 the figures for pounds per watt for hi-fi are given. Most makers laughed at what the text books suggest they should have done. Nobody's amp need conform to the woeful lowest common denominator standards of the mass produced amps. Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. Why? Is this pollution of your idea of purity of design, or of the signal? Also, there may be other alternatives. Perhaps routine manual bias resetting? Just how intolerant are you planning this transformer to be? After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif What's that odd resistor for? Don't it need adjusting for Vak balance? What's the penalty for Vak being out of balance as the valves drift? Are there any other problems that might arise from asymmetrical grid resistances? The only good reason for planning transformer intolerance is the hope of achieving better low frequency performance AFAIK. That is, you sacrifice flexibility by maximising primary inductance. I wonder therefore how good your design is at maintaining perfect full power AC balance at LF? Particularly if it's running in AB. If LF AC balance isn't perfect, then surely your intolerant transformer will complain? Indeed. The OPT performance between dc and say 20Hz is a grey area which needs careful consideration. Where one does have two series tubes with say an 800V supply, one can cap couple a single winding OPT as I stated above. Its an easy OPT to wind compared to a conventional PP OPT with CT because it has half the P turns. The load is a lot lower than a normal PP RLa-a load. In the late 1950 Philips made a range of amps using 2 x EL86 in series with a supply = +400Vdc, and the anode cathode junction was at +200V and there was no OPT. The speaker had an 800 ohm voice coil impedance and was driven by an electro cap off the a-k join. Interesting, thanks. A better proposition now than then, because of the improvement in price and performance of electro caps. Interesting, but I have no plans to build an amp with sereies tubes. Philips did it solely to avoid the cost of an OPT. But the OPT which Philips was using was a walnut sized POS sourced from the cheapest supplier and with highg windinmg losses at 15% in may radio grams. The savings to be had were miniscule, and offset by the cost of winding voice coils with extremely thin wire to get 800 ohms of speaker impedance. Using a pair of EL86 like Philips makes a fairly efficient 10W amp. But gimme a pair of EL34 in triode and a decent sized OPT. It'll **** all over the Philips idea, and over most other 19 watters. Dc imbalance is simply not a huge problem. But there still was a kind of small transformer to get the screen voltage of the top tube to follow its cathode voltage while being held at +400V. A choke would have worked fine though with cap bypass coupling to the cathode. The bottom tube just has normal fixed screen supply and at 1/2 the supply voltage for both tubes, ie, +200V. While working in class A the load of 800 is shared between the two output tubes and each tube sees 1,600 ohms. If one runs such an amp only in class A then with 12W Pda in each EL85, each tube puts 5W into 1,600 ohms or 10W into the 800 ohms. If you wanted to use the EL86 in a normal PP amp under the same Ea/Ia conditions then the OPT would have a primary load of 3k2 a-a, ie, with twice the turns of the single winding OPT. To get drive meant a special bootstrapped circuit was used with a 12AX7. But where the top tube has to have a large grid signal delivered, one may have a 1:1 IST with two windings. One winding is driven with a CF tube which is driven by a gain triode to make say 30Vrms. The other winding has one end connected to the a-k join of the two power tubes and thus the drive to the top tube is always equal to the bottom tube and there is symetrical drive conditions. NFB is applied the normal way, but with some care about gain/phase shift compensation networks because you have an IST in the signal path. To avoid having huge drive voltage and be able have an OPT with much lower RL, and also have no screen voltage to worry about then the 6AS7 or 6C33C come to mind as excellent canditates. But one could run EL34 or KT88 with Ea at 250V OK if one doesn't mind having a screen drive choke. 4 x KT88 would give 40W PP class A and load would be 560 ohms. The cap to couple it can be 100uF, and Lp would need to be at least 20H. The LF pole for resonance is 3.6Hz, and the load to damp the resonance is about 600ohms or less, so when loaded the OPT won't have a peaked LF response. Such a peak can make NFB difficult to apply. The higher the LP, the lower is the Fo, and one could also have the coupling cap value a lot higher, say 470uF, but then onmce you move Fo down to say 0.36Hz, then one is i the grey area with the cap charging and discharging an the core of the OPT is subject to slow ac changes and the resulting temporary saturation effects. As I mentioned in a previous post, If one doesn't like series PP tubes, then one may as well bite the bullet and have a large PP OPT set up conventionally, and then just air gap the PP core. Conventional PP amps throughout history haven't bothered much about DC balance, beyond offering a means of measuring and adjustment. It shouldn't be hard, these days, to automate that process of periodic maintenance. Conventional designs were designed by bean counters foisting their POS efforts onto the unsuspecting public. You can have a 20% dc imbalance and hardly anyone notices some thing is wrong because most ppl use 1/2 a watt average only even though the amp is capable of 25W. I have witnessed Quad-II owners enduring years where the OP tubes have drifted well apart so one tube has 40mA and the other has 90mA with a bit of red anode. Music is absolute crap above the 1/2 W level. The measured THD is many times what it should be at all levels. Go to my website to see what I do about the incompetence used by Quad to build their tube amps. One interesting thing about this idea of cap coupling, when applied to valves operating in DC parallel, is that the idle currents of the two valves can be independently adjusted for optimum AC performance. Oh, and all series output tube connections require a biased heater supply for one of the pair of tubes. Philips didn't bother though because the EL86 has a high heater-cathode voltage rating. Anyone remember the chap here who made *massively* parallel amps with hundreds of EL86? On the grounds, IIRC, that OPTs become relatively smaller as amps get bigger, so in terms of kilos per kilowatt, the bigger the better? Now EL86 aren't as cheap as they were, and electricity is much more expensive . If one has a big enough number of any tubes in series, one can couple the anode-cathode junction to an 8 ohm speaker and have class A action. But its easier to use a few power mosfets. If you had 300 x EL34 on top, and 300 on the bottom, and Ea = 250V and Ia = 100mA each, then the class A load will be conventional speaker type load and a nice load match for all the EL34. Or if you had 200 x EL86 instead of only 2, then you'd get 2,000W into 8 ohms without the OPT instead of 10W into 800 ohms. Nobody needs 2,000W capability, or to have to drain 5kW from the mains to get it. The easiest path to good dc imbalance tolerance is to examine how you think about the core µ during the design process, and just cut any corners. Patrick Turner. cheers, Ian |
#10
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
The short and clean answer is to use only transformers that have a
reasonable tolerance to unbalance and get rid of those that don't. Instead of being a cheapass, buy what's required once and be happy thereafter. If people bought the good stuff and shopped carefully for it, pretty soon they'd be making the good stuff at a reasonable price. Instead, people spend a lot of time-and time is always money to some or other extent-making do with piles of ****. Partridge demonstrated pretty well that the best push pull transformers must have a small air gap, but not as big as that as SE transformers or chokes, to linearize the core's BH curves as well as because in the real world there is always some, small, imbalance. In a guitar amp say, no one really cares and a small imbalance is perhaps preferable. Now of course there are servo circuits, but I like to avoid them. Toroids, unless wound on s slotted core, just do not make very good output transformers and that's that. The smaller EI types are fine for guitar or hobby projects but for the serious high end amp get a good transformer. |
#11
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
Bret L wrote: The short and clean answer is to use only transformers that have a reasonable tolerance to unbalance and get rid of those that don't. Instead of being a cheapass, buy what's required once and be happy thereafter. If people bought the good stuff and shopped carefully for it, pretty soon they'd be making the good stuff at a reasonable price. Instead, people spend a lot of time-and time is always money to some or other extent-making do with piles of ****. Partridge demonstrated pretty well that the best push pull transformers must have a small air gap, but not as big as that as SE transformers or chokes, to linearize the core's BH curves as well as because in the real world there is always some, small, imbalance. I have never seen what Partridge may have had to say about OPT design. But he'd be right about the small air gap. I have used air gaps in PP OPT. Like everything, you have to make the design allow for the lower Lp with the gap. The Williamson OPT spec in RDH4 does not include for any air gap, and old Will said for stability and NFB you had to have Lp = 100H and when using only 5Va=a for the KT66 triode amp with 20dB of NFB. But the Will amp didn't have phase/gain tweaking network included in the open loop circuit; if that is put in then you don't need anywhere near as much Lp as Will suggested. Will had bias balance adjust pot and ac drive balance adjust pot but of course who ever remembered to keep adjusting their damn amp? Now we are able to make a little board with a two transistor LTP which sense the Ia of a pair of tubes and the balance of Ia is displayed with a pair of LED. Imbalance may be adjusted by a very small pot knob and then everything is fine for another 6 months. In a guitar amp say, no one really cares and a small imbalance is perhaps preferable. Now of course there are servo circuits, but I like to avoid them. Toroids, unless wound on s slotted core, just do not make very good output transformers and that's that. Well, if you wound a toroid core with NOSS instaed of GOSS, then the µ would be a lot lower and yet sufficiently high enough for enough Lp. But you can't get a NOSS toroid core; they are allwound now using GOSS, and with a much higher µ than anything Will might have been able to get in 1947. are fine for guitar or hobby projects but for the serious high end amp get a good transformer. Hardly anyone gets serious, because serious costs money. And for every man who actually goes beyound the gunner do stage and finishes his amps, there are twenty gunner doos that don't complete. Patrick Turner. |
#12
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
"Patrick Turner" wrote in message ... Ian Iveson wrote: Patrick Turner wrote in message The circuit John has presented has been invented probably before 1960 and I saw it cited in some ancient magazine, probably an old copy of Wireless World. However, maybe it is different in detail to old ideas of preventing dc imbalance with series tube PP amps. If one really wants to keep out the the Idc and use series tubes, then you can have a series pair with an OPT which is cap coupled to one end of a single primary, with the other end taken to 0V. No dc in the winding at all. If there is a good alternative reason for your intolerant transformer plan, perhaps you could say what it is? Just about all PP OPT set up conventionally are intolerant of dc imbalance because they have ungapped cores and with a high max µ, so the slightest dc imbalance causes serious increases in THD/IMD. Tolerance is quite a complicated thing when it comes to transformers. The story goes that the more evenly a transformer magnetises the core, the greater the inductance that can be achieved for a given amount of iron, the more linear the relationship between inductance and current, and the less leakage, particularly if the core has no sharp corners. Because of the more effective use of the iron, less can be used, making for a comparitively dainty transformer with lower capacitance than an EI of the same inductance. I don't know in what context "evenly" rests with regard to core magnetization. Maybe you don't either. Evenly? Why isn't that obvious? Evenly magnetised is magnetised the same all over. In the case of a toroid, this is true in two important ways: geometrically, because the windings can be evenly distributed around the core; and with respect to grain orientation, because the GOSS the core is wound from ends up with its grain in constant alignment with the field. The first sense of "evenly" is rather less true with R- and C-cores, but all three share the second characteristic. Hence, for a given winding, Bmax occurs throughout the core for the same current. Consequently it saturates all at once. What you miss in the rest of your advertising copy is the fact that this even distribution of magnetisation also leads to much higher inductance than for an EI of the same size, so for a given signal voltage and frequency, less magnetising current flows, and hence B is lower. Consequently, Bmax may occur at a lower frequency. Or a wound core can achieve the same low frequency performance with less iron. That's where your half-truths unravel. Assuming all are equally optimised and well-built, wound cores give wider bandwidth, period. No point in beating around the bush. The "quality factor" of a good toroid beats the pants off *any* EI. I recommend Menno van der Veen's "Modern High End Valve Amplifiers". He puts the case very well for the view that a wound core and an RC power supply are defining features of a modern hi-fi valve amp. OTOH, modern amps are generally made of transistors, so the toroidal OPT may be a sad historical anachronism. Manual bias adjustment is perfectly adequate for a careful owner, as all DIYers are, unlike your half-witted customers. Manual adjustment is also the choice of the hi-fi purist. But anyway, considering what you are saying, having toroidal cores, R-core or Ccore means you can have a much higher maximum permability and less hysteresis so the distortion at all F is lower. However, to get good bass you need a low Bm and that means a larger core and more turns than most commercial bean counter propelled "dainty" designs you must be referring to. Dainty because the higher inductance per kilo allows a smaller core to be used. To reduce LL and Csh, one has to do what these commercial people won't do, ie, use more sections and use thicker insualtion material. Or make a more dainty transformer. Swings and roundabouts, as I am sure you know, but all in all a wound core achieves a wider bandwidth by increasing the ratio of Lp/Cp.Lleakage. Were you to design a few OPT according to my logic steps all listed out at my website, you'd understand the need to have a certain number of Kg per watt of audio power. No. Depends on the core material and geometry. Also, even for a given material and shape, the relationship between weight and power is not linear as you suggest. Yes, for a given type of transformer and low frequency limit, weight is a function of power, but the function is not a straight line. Many makers put only 1/2 the required weight into their OPT, and Quad-II is a perfect example of shoddy OPT design that might be called "dainty" for want of a better word. I have used a 63mm stack of 44mm tongue NOSS E&I lams for a 100W design, the 8585 at my website. Its been made along ther lines of the rest of the theory at my site and it performs excellently and better than most commercial maker OPTs. The NOSS has a max µ of only 3,500, and thus it is more tolerant of dc imbalance than a GOSS core of the same size with max µ of 17,000. The higher µ material gives less THD though. But if the higher µ material was air gapped to give a µe of 3,500, or perhaps even lower, then the tolerance of dc imbalance would be very much improved. A toroidal core is THE WORST type of core one might use if there is any Dc ever likely to be present because its µ max is around 40,000, and the slightest dc offset current will cause the dc magnetization to easily exceed 1.3T and then there is any headroom for ac magnetization because its "hard over" one way, so for 1/2 the wave form the OPT is a virtual short circuit. So if you want plenty inductance for bottom end grunt combined with low capacitance and leakage for a top end extended enough to use stacks of feedback, then you should use a toroid, or an "R-core", or C-core, in that order. I think you know very little if you say that. Perhaps you didn't realise, but if you remove the core of an OPT and run the OPT as an air cored transformer it should work perfectly well above say 5kHz, and the HF stability with FB is determined by the geometry of the copper, and the iron is functionless above 10kHz. The frequencies of HF oscillation if they occur are usually above 30kHz and due to phase shift due to high LL or Csh, or both. At LF, for stability, you could use enormous amounts of inductance, as is the case of the Williamson. But you still get an ultimate 90 degree phase shift with Ra and Lp, and with the other CR couplings there is always a tendency for instability. So to me its utterly pointless using high µ iron for an OPT, and pointless to use huge numbers of P turns because all that makes the dc saturation all the more likely, and the design prone to LF transients, ie, a range of signals between dc and 20Hz which magnetize the iron too much. Huge numbers of P turns increases LL because the LL like Lp is proportional to the square of turns. Twice the turns means 4 times Lp and 4 times the LL. Its far better to have **adequate Lp** and no more. See my website and design an axample to understand more. The optimum is different depending on the core material and geometry. If you think it is necessary to make transformers in order to understand them, then why do you claim to understand wound cores, when you don't make them or even use them? The story continues with a downside: the more even the field distribution, the more sharply the core saturates, especially if less iron has been used. This is where the toroid's reputation for DC intolerance comes from, I think. Perhaps you could make a better toroid? If you were to use the same core area as you would for an EI, would that make it less intolerant? If you aimed for the same HF performance as you would for an EI, you could make a humungous toroid, I would guess. You could make a toroid with a cut and glue in some gapping material. Or one with a larger area core than the commercial winders use. That would sacrifice some top end for a bit more bottom. Its difficult and fragile. Clamped C-cores are better and achieve the same thing. Once the air gap is there it dominates the magnetic path length. Gapped toroids intended for SE OPT are commercially available. Vanderveen, of toroid fame, shows a sketch of the output resulting from poor AC balance. From memory it looks like the signal disappears around the voltage zero crossing points. He doesn't explain how this happens though...it's just shown to illustrate how to adjust it. Just set the bias very low on one tube and high on the other then observe the dreadful even order distortion. Not the same thing. Menno and I were talking about AC imbalance between the two halves of a PP output stage. Any signal above 1/2 a watt sounds awful. Of course. So stop messing about and reset the bias. So I have a question. Taking John's transformer as an example, each half of the primary winding will on its own have a certain inductance, which will be doubled if they are both operating in unison, because twice the windings gives four times the inductance. John's circuit works with the same dc current flowing in both tubes and through two P windings. But at ac the operation is the same as any normal PP OPT. No, because the cathode circuits are not symmetrical, as they are in conventional PP OP stages, and because Va may not be the same for each side. LF AC imbalance could result, but it's OK...John says he doesn't care. It can be made to work in class AB where all Ia in either tube is cut off and each tube does only the crests of each wave. During the class A work the inductance is mutual and high, and once the Ia cuts off the inductance reduces but so does the load. In a class B amp the core still saturates if F goes low enough. And in a class B amp there is no DC present. Devices are biases off. Consequently, any unbalanced part of the signal will see half the load, and so result in twice the current for a given voltage amplitude (and so a zero at twice the frequency?). Question is, will an intolerant transformer find this more of a problem at low frequencies than a tolerant one? Depends on saturation and the amount of dc offset. Once saturation occurs, the winding acts as though it has no L and the dc winding R shunts the load. Which doesn't answer my question. Never mind, I think I can do that myself. Just trying to raise the issues. If you air gap the OPT, the Lp will be much lower, but the OPT will behave without its Lp dissapearing to leave just the Rw. Its better to have some Lp rather than to have plenty of Lp and have it vanish suddenly at say 14Hz. Depends. If you can get enough Lp to last down to, say 7Hz, then the suddenness of its demise may not be an issue. This is the nub of the matter. Then the only remaining issue is DC intolerance, which is not a problem for DIYers. It's a matter of convenience versus hi-fi. You can make your dainty little OPT if *you* want to. Not just me. Those who make good wound cores, such as Plitron, tend to spread the benefit between top and bottom end considerations. That's why their transformers are relatively dainty. Read Menno. I won't. OK. There are good reasons to use EI too. Ultimately it depends on whether you want a modern, or a historical, valve amp. That in turn is related to what kind of speakers you have, and to some degree on what kind of music you like. To make the OPT tolerant to dc imbalance and dc offset, you have to use lots of iron and more turns and maybe an air gap. Thus to get 60 watts of SE power from tubes requires an E&I core 72mm stack 51mm tongue and an air gap and very low µ of perhaps only 500. An SE amp is one which thrives on having a HUGE dc offset. In my SE55 amp with a pair of 845, load was 6k, and Lp abot 40H, Ia = 150mA. If I had a PP amp, the same core could be used to make a 200W AB amp, based on having core saturation occuring at 20Hz at full PO, ie, at the same F as in the case of the SE amp. If you want less iron, and you cut the stack height in half, then the Fsat for PP goes up from 20Hz to 40Hz. Thus is done routinely by most makers to save construction costs and freight costs etc. Dc imbalance worsens... But at 5 watts, a 200W tube amp might have a huge DC imbalance and nobody will hear anything. This is part of a more generic question because a conventional PP output stage is never quite balanced anyway, obviously particularly when operating in AB. I was set off on this train of thought by John's assymetric cathodes. I guess he can just make those caps really big, so their effects are both well below a bandwidth determined elsewhere. Again if nobody objects to weight and size and cost then there you can have choke feed to a pair of output tubes using a choke with CT and then cap couple the OPT which has its CT grounded. This allows normal drive, but then the choke suffers the imbalance if it occurs. But then if the coke is gapped, and had a massive sive and lots of turns then it may withstand far more dc imbalance than any normal OPT. Or one might use TWO individual air gapped chokes, but you can't get the large wanted inductance to prevent inductance shunting the load. And by this time your amp weighs half a ton and costs a fortune. In RDH4 the figures for pounds per watt for hi-fi are given. Most makers laughed at what the text books suggest they should have done. Nobody's amp need conform to the woeful lowest common denominator standards of the mass produced amps. EI is assumed, presumably. Figures are different depending on core material and geometry. Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. Why? Is this pollution of your idea of purity of design, or of the signal? Also, there may be other alternatives. Perhaps routine manual bias resetting? Just how intolerant are you planning this transformer to be? After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif What's that odd resistor for? Don't it need adjusting for Vak balance? What's the penalty for Vak being out of balance as the valves drift? Are there any other problems that might arise from asymmetrical grid resistances? The only good reason for planning transformer intolerance is the hope of achieving better low frequency performance AFAIK. That is, you sacrifice flexibility by maximising primary inductance. I wonder therefore how good your design is at maintaining perfect full power AC balance at LF? Particularly if it's running in AB. If LF AC balance isn't perfect, then surely your intolerant transformer will complain? Indeed. The OPT performance between dc and say 20Hz is a grey area which needs careful consideration. Where one does have two series tubes with say an 800V supply, one can cap couple a single winding OPT as I stated above. Its an easy OPT to wind compared to a conventional PP OPT with CT because it has half the P turns. The load is a lot lower than a normal PP RLa-a load. In the late 1950 Philips made a range of amps using 2 x EL86 in series with a supply = +400Vdc, and the anode cathode junction was at +200V and there was no OPT. The speaker had an 800 ohm voice coil impedance and was driven by an electro cap off the a-k join. Interesting, thanks. A better proposition now than then, because of the improvement in price and performance of electro caps. Interesting, but I have no plans to build an amp with sereies tubes. Philips did it solely to avoid the cost of an OPT. But the OPT which Philips was using was a walnut sized POS sourced from the cheapest supplier and with highg windinmg losses at 15% in may radio grams. The savings to be had were miniscule, and offset by the cost of winding voice coils with extremely thin wire to get 800 ohms of speaker impedance. Your logic doesn't stand up. How can you say they did it for cheapness, and also say it's not cheaper? Generally, you say *everyone* does *everything* for cheapness, except you. This makes you appear a bit grumpy and stupid. Using a pair of EL86 like Philips makes a fairly efficient 10W amp. But gimme a pair of EL34 in triode and a decent sized OPT. It'll **** all over the Philips idea, and over most other 19 watters. Dc imbalance is simply not a huge problem. But there still was a kind of small transformer to get the screen voltage of the top tube to follow its cathode voltage while being held at +400V. A choke would have worked fine though with cap bypass coupling to the cathode. The bottom tube just has normal fixed screen supply and at 1/2 the supply voltage for both tubes, ie, +200V. While working in class A the load of 800 is shared between the two output tubes and each tube sees 1,600 ohms. If one runs such an amp only in class A then with 12W Pda in each EL85, each tube puts 5W into 1,600 ohms or 10W into the 800 ohms. If you wanted to use the EL86 in a normal PP amp under the same Ea/Ia conditions then the OPT would have a primary load of 3k2 a-a, ie, with twice the turns of the single winding OPT. To get drive meant a special bootstrapped circuit was used with a 12AX7. But where the top tube has to have a large grid signal delivered, one may have a 1:1 IST with two windings. One winding is driven with a CF tube which is driven by a gain triode to make say 30Vrms. The other winding has one end connected to the a-k join of the two power tubes and thus the drive to the top tube is always equal to the bottom tube and there is symetrical drive conditions. NFB is applied the normal way, but with some care about gain/phase shift compensation networks because you have an IST in the signal path. To avoid having huge drive voltage and be able have an OPT with much lower RL, and also have no screen voltage to worry about then the 6AS7 or 6C33C come to mind as excellent canditates. But one could run EL34 or KT88 with Ea at 250V OK if one doesn't mind having a screen drive choke. 4 x KT88 would give 40W PP class A and load would be 560 ohms. The cap to couple it can be 100uF, and Lp would need to be at least 20H. The LF pole for resonance is 3.6Hz, and the load to damp the resonance is about 600ohms or less, so when loaded the OPT won't have a peaked LF response. Such a peak can make NFB difficult to apply. The higher the LP, the lower is the Fo, and one could also have the coupling cap value a lot higher, say 470uF, but then onmce you move Fo down to say 0.36Hz, then one is i the grey area with the cap charging and discharging an the core of the OPT is subject to slow ac changes and the resulting temporary saturation effects. As I mentioned in a previous post, If one doesn't like series PP tubes, then one may as well bite the bullet and have a large PP OPT set up conventionally, and then just air gap the PP core. Conventional PP amps throughout history haven't bothered much about DC balance, beyond offering a means of measuring and adjustment. It shouldn't be hard, these days, to automate that process of periodic maintenance. Conventional designs were designed by bean counters foisting their POS efforts onto the unsuspecting public. You can have a 20% dc imbalance and hardly anyone notices some thing is wrong because most ppl use 1/2 a watt average only even though the amp is capable of 25W. I have witnessed Quad-II owners enduring years where the OP tubes have drifted well apart so one tube has 40mA and the other has 90mA with a bit of red anode. Music is absolute crap above the 1/2 W level. The measured THD is many times what it should be at all levels. Go to my website to see what I do about the incompetence used by Quad to build their tube amps. There is no reason to go to your site, so I've never been. I assume it's like you write here, which is quite enough biased half-truths and lies for me, thanks. If I need reliable information, I go to the original source. One interesting thing about this idea of cap coupling, when applied to valves operating in DC parallel, is that the idle currents of the two valves can be independently adjusted for optimum AC performance. Oh, and all series output tube connections require a biased heater supply for one of the pair of tubes. Philips didn't bother though because the EL86 has a high heater-cathode voltage rating. Anyone remember the chap here who made *massively* parallel amps with hundreds of EL86? On the grounds, IIRC, that OPTs become relatively smaller as amps get bigger, so in terms of kilos per kilowatt, the bigger the better? Now EL86 aren't as cheap as they were, and electricity is much more expensive . If one has a big enough number of any tubes in series, one can couple the anode-cathode junction to an 8 ohm speaker and have class A action. But its easier to use a few power mosfets. If you had 300 x EL34 on top, and 300 on the bottom, and Ea = 250V and Ia = 100mA each, then the class A load will be conventional speaker type load and a nice load match for all the EL34. Or if you had 200 x EL86 instead of only 2, then you'd get 2,000W into 8 ohms without the OPT instead of 10W into 800 ohms. Nobody needs 2,000W capability, or to have to drain 5kW from the mains to get it. The easiest path to good dc imbalance tolerance is to examine how you think about the core µ during the design process, and just cut any corners. I don't need dc imbalance tolerance, thanks. I adjust my bias currents routinely. Ian |
#13
Posted to rec.audio.tubes
|
|||
|
|||
Alternative OPT current balance method
Ian Iveson wrote: "Patrick Turner" wrote in message ... Ian Iveson wrote: Patrick Turner wrote in message The circuit John has presented has been invented probably before 1960 and I saw it cited in some ancient magazine, probably an old copy of Wireless World. However, maybe it is different in detail to old ideas of preventing dc imbalance with series tube PP amps. If one really wants to keep out the the Idc and use series tubes, then you can have a series pair with an OPT which is cap coupled to one end of a single primary, with the other end taken to 0V. No dc in the winding at all. If there is a good alternative reason for your intolerant transformer plan, perhaps you could say what it is? Just about all PP OPT set up conventionally are intolerant of dc imbalance because they have ungapped cores and with a high max µ, so the slightest dc imbalance causes serious increases in THD/IMD. Tolerance is quite a complicated thing when it comes to transformers. The story goes that the more evenly a transformer magnetises the core, the greater the inductance that can be achieved for a given amount of iron, the more linear the relationship between inductance and current, and the less leakage, particularly if the core has no sharp corners. Because of the more effective use of the iron, less can be used, making for a comparitively dainty transformer with lower capacitance than an EI of the same inductance. I don't know in what context "evenly" rests with regard to core magnetization. Maybe you don't either. Evenly? Why isn't that obvious? Because you didn't spend the time to specify eaxactly what you meant. We are not ****in mind readers, so kindly post on electronic engineering subjects with prcision and lack of hot air or ambiguity please. Evenly magnetised is magnetised the same all over. In the case of a toroid, this is true in two important ways: geometrically, because the windings can be evenly distributed around the core; and with respect to grain orientation, because the GOSS the core is wound from ends up with its grain in constant alignment with the field. The first sense of "evenly" is rather less true with R- and C-cores, but all three share the second characteristic. Yes, OK, but in fact the inner turns of the core magnetize more than the outer turns because the magnetic path length varies. So a toroidal core which has thin build up with respect to the width of wound strip has a more even distribution of magnetic field strength. But in a PP toroid OPT with no gap, the slightest difference in Idc in either 1/2 primary causes saturation because the permability, µ, is very high, maybe 40,000. Bdc in Tesla = 12.6 x µe x Np x Idc / (10,000 x Iron ML) where 12.6 is a constant, µe is the effective µ with or without an air gap, Np = total P turns, Idc is the net dc imbalance, 10,000 is a constant, and Iron ML is the iron magnetic path length without an air gap. ML is measured in mm. In a normal SE OPT, the core is allowed to be magnetized to about 0.6Telsa by dc, leaving another 0.6T to cover the action of the Bac, or ac caused magnetization. If you calculate the Bac using formulas at my website, then you should only allow 0.6T at full power at 20Hz. It appears that even with a big air gap, the Bac is still proportional to 1/F, 1/Np, 1/Afe and to Vac across the coil; the µ does not appear in the formulas. But is does for the Bdc. There is quite a lot more that *you* have to read and understand before you are qualified to discuss this usefully at this discussion group. I have more to learn on the sunbject of course. But what I do know is that if I apply my formulas to anything I make and sell, I get text book performance or better, and the ghosts of the authors of RDH4 rest easy in their graves because I make stuff which complies to all the known rules good design which have been ignored and laughed at by most major and minor manufacturiers in the interests of profits. Hence, for a given winding, Bmax occurs throughout the core for the same current. Consequently it saturates all at once. What you miss in the rest of your advertising copy is the fact that this even distribution of magnetisation also leads to much higher inductance than for an EI of the same size, so for a given signal voltage and frequency, less magnetising current flows, and hence B is lower. I don't write advertizing copy. But my website has all the answers for you and anyone else interested in making decent quality audio amps. Its a one stop site where everything can be found. Not just **** for sale with bull**** and lies to sell it. I presently have about 100MB downloaded each day from my site. Apart from your own stubborn self, a huge number of ppl find my site a great resource, and so do I because it saves me having to explain things in detail here repeatedly. Consequently, Bmax may occur at a lower frequency. Or a wound core can achieve the same low frequency performance with less iron. That's where your half-truths unravel. You are going to have to spend a lot more time and effort to convince the world I am telling half lies. While ever you don't then everyone will see what a ****ingh ignorant jerk and bag of ****in wind that you are. Assuming all are equally optimised and well-built, wound cores give wider bandwidth, period. No point in beating around the bush. The "quality factor" of a good toroid beats the pants off *any* EI. Only one trouble here. You don't substantiate your claims. I recommend Menno van der Veen's "Modern High End Valve Amplifiers". He puts the case very well for the view that a wound core and an RC power supply are defining features of a modern hi-fi valve amp. OTOH, modern amps are generally made of transistors, so the toroidal OPT may be a sad historical anachronism. The toroidal OPT is not necessarilly better than a normal E&I cored tranny also using GOSS. Its well known that dc imbalance in an ungapped toroidal core is lethal for the music fidelity; see my reasoning about Bdc above. Most amps now provided with toroidal OPT are bleedin awful. A typical example id the CR Audio Developments 5050 Woodham with two channels with a pair of KT88/6550, Ea at 600V, bias Ia is real low, and load value way too low. Idc need only vary slightly, and opps, the music turns to mud. I re-wired one 2 years ago. I much improved the way the bias was set so an owner had no trouble setting the bias for each tube to get balance of idc within 3 mA, and the OPT then coped with music OK at normal levels averaging 1 watt or less. Manual bias adjustment is perfectly adequate for a careful owner, as all DIYers are, unlike your half-witted customers. Manual adjustment is also the choice of the hi-fi purist. Manual adjustment of anything ona tube amp is a nightmare for many owners, as well as having fuses that are owner replaceable. Inevitably, no matter who the owners are or what their levels of undersanding are, a large % are simply unable to cope with the slightest technical repair of adjustment job of anything at all in their homes or at work. One friend I have is a brilliant chess player, and his powers at the chess board have led him to club tournament championship wins each year for decades. But he doesn't understand movie plots, gets upset easily, and if ever I catch him with a screw driver or spanner in his hand I quietly remove them and do the job for him to avoid him damaging some poor peice of gear. The owner of the CR Developments Woodham had just about zero chance of ever adjusting bias correctly because the makers made it so damn difficult to get it right. I coud quote dozens of other examples to make you look completely ill-informed. In the case of my 300W amps with 12 x 6550 within, there was no way I was going use individual fixed bias because it meant 24 adjustments for the two channels just to get the bias right. Individual cathode bias works just fine for most owners and fixed bias is a nightmare for most. There must be active protection against excessive Ik in each case and afaik, I am the ONLY maker world wide to add such protection measures to prevent smoke in the loungerooms of anyone who buys an amp from me, or who has me re-engineer someone else's amp such as made by Quad, ARC, Manley Labs, CR Devepments, VAC, just to name a few I have had to re-design and re-wire to make them sing without so much noise and distortion and without ever smoking. I went as far as inventing an active form of bias stabilisation for cathode biased amps so that even with sine wave signals taken right up to clipping the N&D was the same as was the case with a well adjusted fixed bias amp. How to achieve it is all at my website. And BTW, the bleedin awful toroidal OPT made by CR Developments had way too few P turns, or if you like, had Afe about 1/2 what it sould have been for Fsat to get below 20Hz at full PO. Sure the CR amp was what *you* might think is a nice UK made "dainty design". In terms of how well it complied with good design methods as clearly spelt out in RDH4 and other good books, it was in fact a complete POS!!!! The toroid transformer does offer a maker the opportunity to bring leakage inductance to very low levels because the traverse width of the winding goes right around the toroid, whereas in E&I or C-core bobbins there has to be layered sectioning to reduce LL. However, with the CR developments, they completely forgot about the effects of the shunt C between P and earthy S windings and they used attrociously thin insulation between layers of P and S windings and layers of P to P. Naturally, makers of toroidal OPT are mainly sloppy fuctards who don't do things well. They generally use one thin mylar tape strip wound over a winding before winding another over on top. There is typically just a P-S-P wind up interleaving arrangement. The trouble is that you cannot easily varnish the winding to stop howl due to microscopic winding movements. The way to do it properly is to use woven tape material and with thickness at least 0.5mm or greater, and this allows easy varnish penetration just by soaking the tranny in a vat of varnish and baking afterwards. Toroidal PT also are far better when done this way than the the usual muck I see too often. With PT, the extremely high µ of the toroidal core with GOSS menas that the primary inductance is nice and high and the magnetizing current is quite negligible. But the high µ doesan't mean that you can have a bigger than normal Bac. About 0.9 Tesla is about right for any PT used in quality hi-fi gear. But the idiots keep pumping out crap running 1.25Tesla. So it hums and vibrates, despite the low core losses. But anyway, considering what you are saying, having toroidal cores, R-core or Ccore means you can have a much higher maximum permability and less hysteresis so the distortion at all F is lower. However, to get good bass you need a low Bm and that means a larger core and more turns than most commercial bean counter propelled "dainty" designs you must be referring to. Dainty because the higher inductance per kilo allows a smaller core to be used. This is simply bull****. The core size is determined mainly by the Fsat, and has very little to do with primary inductance. In crappy "dainty" OPT, you will find that the fuctards who design their POS OPT have Fsat = 40Hz at full PO. Plenty of Lp though. To get the Fsay downwards to 13Hz means you have to use 3 times the P turns for the same size of core, and the turns won't fit, or the wire has to be made dangerously thin with very high winding losses. The other solution also hated by fuctard bean counter designers is to increase the Afe of the core by 3 times but leave Np the same. This may double or triple the wieght of the OPT and the result is anything but ****ing dainty, OK! But its better engineering, right? To reduce LL and Csh, one has to do what these commercial people won't do, ie, use more sections and use thicker insualtion material. Or make a more dainty transformer. Swings and roundabouts, as I am sure you know, but all in all a wound core achieves a wider bandwidth by increasing the ratio of Lp/Cp.Lleakage. The wound core does NOT necessarily result in better BW. You have obviously never designed any OPT for which the public might be proud to buy. In my 300W amps, the core is a 110mm stack of 51 tongue GOSS E&I lams. Max µ = 17,000, and I have Np = 1,058 turns. From formulas found at my website I am sure you can work out the Lp and Fsat where the load is 1,000 ohms a-a. I have open loop BW = 270kHz without any NFB and when using 50% UL connection, where the source resistance approx = load resistance. There is an interleaving geometry with 5P sections each with two layers of thin wire and 6 sections of thick wire single layer secondaries. Insulation between P and S layers = 0.75mm, with dielectric constant of 2.5. Winding losses are well below 6%. It might be possible to produce similar reults with a toroidal, but proneness to Fsat is much worse if DC imbalance occurs. Using the same turn count would mean the Afe for the toroid would have to be 110mm x 51mm. Now LL is approximately proportional to Np squared, so the fewer P turns you have, the lower the LL becomes. Reffered to a primary load, you should have an equivalant 10mH of LL in relation to 10k of load resistance. In my case with RL = 1k, I needed to get LL = 1mH. Many makers such as Leak had LL = 50mH for a 10k a-a load. CRAP. Its very hard to make such an amp unconditionally stable at HF. My web pages on Leak amp mods spells out the circuit required. Were you to design a few OPT according to my logic steps all listed out at my website, you'd understand the need to have a certain number of Kg per watt of audio power. No. Depends on the core material and geometry. No? Its all explained at my website. I know how long you resented me telling you to buy a copy of RDH4 and learn it all by heart. Its correct what they say in that book. I repeat it all at my website, but give formulas to avoid having ton read the absurdly difficult to follow graphs in RDH4 with regard to checking LL and Csh. Core material isn't critcal once you have enough of it. The very best core material is never much good if there isn't enough of it. Kondo of Audio Note Japan fame used air gapped E&I lams and 50% nickel and 50% GOSS core material in the Ongaku amps of the 1990s which cost about aud $140,000 a pair, for 21 watts of class A2 PO. Now I bet Kondo found that with the nickel in the core and the air gap, the core caused artifacts were minimised for the turns he would have been using, and he used silver wire, so the less turns the better because silver wire is expensive. The gap dominates the path length and core behaviour in SE designs. A gap in hi µ material is used to reduce a maximum very high µ from say 40,000 to not more than say 500. Inductance has to be adequate so that at least the Lp reactance in ohms at 20Hz = RL, to avoid loading the amp down with an inductive load at LF. Afe still has to be high to give toatl Bdc + Bac = 1.2Tesla at 20Hz. You simply cannot avoid this fact unless you insist on making "dainty" crap quality like so many makers out there. Of course, once an air gap is selected to reduce max µ to 500, it becomes less critical what the max µ might have been without an air gap at all. After you have wound a tonne of chokes with varying core quality and with air gaps, you begin to understand this. for a given material and shape, the relationship between weight and power is not linear as you suggest. Indeed weight and power are not linearly proportional. Doubling the stack height will double the weight. But with the same turns, the Fsat will be halved. Winding losses increase slightly due to slightly longer turn length. But with twice the stack height you could havle the Np if you have Fsat where you want it, ie, below 20Hz. This means the wire is 1.4 times the diametre which means the I allowable doubles, and PO = I squared x RL, so you get 4 times more PO. So you can increase n, if you All this becomes crystal clear if you apply ideas and formulas so clearly set out at my website. Yes, for a given type of transformer and low frequency limit, weight is a function of power, but the function is not a straight line. RDH4 has an approximate calculation of the power to weight ratio required for hi-fi OPT. Its only a very approximate rough guide, and one mostly completely ignored by major makers such as Quad and many others on a big long sorry dismal list. At my website I have made comments about the power to weight ratio, then got down to the nitty ****in gritty aspects of finding out just how bloody heavy you really have to make your cores according to what you really want your OPT to be able to do. Many makers put only 1/2 the required weight into their OPT, and Quad-II is a perfect example of shoddy OPT design that might be called "dainty" for want of a better word. I have used a 63mm stack of 44mm tongue NOSS E&I lams for a 100W design, the 8585 at my website. Its been made along ther lines of the rest of the theory at my site and it performs excellently and better than most commercial maker OPTs. The NOSS has a max µ of only 3,500, and thus it is more tolerant of dc imbalance than a GOSS core of the same size with max µ of 17,000. The higher µ material gives less THD though. But if the higher µ material was air gapped to give a µe of 3,500, or perhaps even lower, then the tolerance of dc imbalance would be very much improved. A toroidal core is THE WORST type of core one might use if there is any Dc ever likely to be present because its µ max is around 40,000, and the slightest dc offset current will cause the dc magnetization to easily exceed 1.3T and then there is any headroom for ac magnetization because its "hard over" one way, so for 1/2 the wave form the OPT is a virtual short circuit. So if you want plenty inductance for bottom end grunt combined with low capacitance and leakage for a top end extended enough to use stacks of feedback, then you should use a toroid, or an "R-core", or C-core, in that order. I think you know very little if you say that. Perhaps you didn't realise, but if you remove the core of an OPT and run the OPT as an air cored transformer it should work perfectly well above say 5kHz, and the HF stability with FB is determined by the geometry of the copper, and the iron is functionless above 10kHz. The frequencies of HF oscillation if they occur are usually above 30kHz and due to phase shift due to high LL or Csh, or both. At LF, for stability, you could use enormous amounts of inductance, as is the case of the Williamson. But you still get an ultimate 90 degree phase shift with Ra and Lp, and with the other CR couplings there is always a tendency for instability. So to me its utterly pointless using high µ iron for an OPT, and pointless to use huge numbers of P turns because all that makes the dc saturation all the more likely, and the design prone to LF transients, ie, a range of signals between dc and 20Hz which magnetize the iron too much. Huge numbers of P turns increases LL because the LL like Lp is proportional to the square of turns. Twice the turns means 4 times Lp and 4 times the LL. Its far better to have **adequate Lp** and no more. See my website and design an axample to understand more. The optimum is different depending on the core material and geometry. If you think it is necessary to make transformers in order to understand them, then why do you claim to understand wound cores, when you don't make them or even use them? Because I understand how badly wond cores behave which have unecessarily high µ. You have to make and test OPT to unsterdtand them. And you have to understand them before you make them. In order to save the Planet's amp workers the agonies of ignorance, I have set up a website which any amp worker or amp designer may forecast what performance he will get once he has wound some given design. My website is as simple as it can be to explain it all as well as I think it needs to e explained. It does appear horrendously complex to those without sufficient ability or any willingness to comprehend my message. You just don't get good at making OPT by reading a book once, or a website once. BTW, two young IT student guys did try to convert all my website steps of design logic of OPT into a program which then produces a diagram on the screen which is the section through the bobbin showing layers and wire sizes and interleaving et all. Both these fairly brainy individuals failed to get past making a program which did any more than calculate the turns for P and for S. One gave up after 6 mths and lots of emails with me and the other after 3 mths. I doubt you'd have less difficulty reading my site on OPT, but feel welcome to try to make a program to accurately design something that can be given to a tradesman or tradeswoman in a factory with the instruction, "Hey Joe, would you mind winding 100 of these this week?" I have seen the attrocious offerings of mass made amps using toroidal OPT. I have no intention to ever emulate the utter crap offered these days. I really don't give a **** if my amps are heavier than most that are sold in the shops. This doesn't mean for a second that toroidals cannot ever be good. When they are good enough they will be heavier than most "dainty" crap products. Plitron make a good range of toroidals afaik. But they are very expensive. I have never seen any reason to invest heavily in toroidal winding machines because I know how easy it is to wind accurately layered conventional bobbin wind ups with neat layers and giving many taps at ends of windings to allow a great variety of wasteless impedance matches. This is very difficult with toroids. Uncut strip wound cores are totally unecessary and damned un-welcome. C-cores are excellent because they allow a gap to easily be placed while its fairly easy to keep them clamped tight. E&I are slightly more difficult to clamp tight, but with partial air gapping where the Es and Is are inserted in sub-stacks of say 15 lams thick in opposite directions to reduce µ to say 1,500, the assembled pile becomes stable once the bolts are tightened. The story continues with a downside: the more even the field distribution, the more sharply the core saturates, especially if less iron has been used. This is where the toroid's reputation for DC intolerance comes from, I think. Perhaps you could make a better toroid? If you were to use the same core area as you would for an EI, would that make it less intolerant? If you aimed for the same HF performance as you would for an EI, you could make a humungous toroid, I would guess. You could make a toroid with a cut and glue in some gapping material. Or one with a larger area core than the commercial winders use. That would sacrifice some top end for a bit more bottom. Just because you increase the core Afe doesn't mean you will increase LL or Cshunt and thus lower the HF cut-off. See my website for all details about why, and see what is possible. Its difficult and fragile. Clamped C-cores are better and achieve the same thing. Once the air gap is there it dominates the magnetic path length. Gapped toroids intended for SE OPT are commercially available. Well indeed. Once gapped, there isn't any superiority of a toroid. As a niche market maker, it better suits me to keep using E&I GOSS lams or to use C-cores. But in this country C-cores are never stocked by anyone. AEM used to make them to order. Max µ varied between 4,500 and 11,000. This was low, and I suspect the 4,500 was done by simply using quite low grade SiFe strip. So I have settled for the E&I cores from Sankey Aust. Max µ is 17,000, and this gives excellent low distortion operation in OPT and better results than anything published in RDH4 where they have a formula for calculating distortion in an OPT, based on available material for GOSS cores of the 1950s and where µ max possible was only about 5,000. But 5,000 was good in comparison to crap grade irn available then which maybe went 2,000. I know because I have measured cores taken from old trannies made in the 1950s and 60s. Vanderveen, of toroid fame, shows a sketch of the output resulting from poor AC balance. From memory it looks like the signal disappears around the voltage zero crossing points. He doesn't explain how this happens though...it's just shown to illustrate how to adjust it. Just set the bias very low on one tube and high on the other then observe the dreadful even order distortion. Not the same thing. Menno and I were talking about AC imbalance between the two halves of a PP output stage. OK, ac imbalance is not good. One can set the tubes to have equal Idc. Then set the ac balance so that Iac in each tube is the same. Due to slghtly different gm and Ra of each of two healthy OP tubes, you will then not have equal Vgrid applied to each side. Fortunately, imbalances of up to 10% of applied Vg at normal levels are virtually undetectable by anyone. Any signal above 1/2 a watt sounds awful. Of course. So stop messing about and reset the bias. Not always possible, especially if its a Quad-II amp. So I have a question. Taking John's transformer as an example, each half of the primary winding will on its own have a certain inductance, which will be doubled if they are both operating in unison, because twice the windings gives four times the inductance. John's circuit works with the same dc current flowing in both tubes and through two P windings. But at ac the operation is the same as any normal PP OPT. No, because the cathode circuits are not symmetrical, as they are in conventional PP OP stages, and because Va may not be the same for each side. LF AC imbalance could result, but it's OK...John says he doesn't care. Because he like me knows it isn't very critical in a PP class A circuit. It can be made to work in class AB where all Ia in either tube is cut off and each tube does only the crests of each wave. During the class A work the inductance is mutual and high, and once the Ia cuts off the inductance reduces but so does the load. In a class B amp the core still saturates if F goes low enough. And in a class B amp there is no DC present. Devices are biases off. Consequently, any unbalanced part of the signal will see half the load, and so result in twice the current for a given voltage amplitude (and so a zero at twice the frequency?). Question is, will an intolerant transformer find this more of a problem at low frequencies than a tolerant one? Depends on saturation and the amount of dc offset. Once saturation occurs, the winding acts as though it has no L and the dc winding R shunts the load. Which doesn't answer my question. Never mind, I think I can do that myself. Just trying to raise the issues. If you air gap the OPT, the Lp will be much lower, but the OPT will behave without its Lp dissapearing to leave just the Rw. Its better to have some Lp rather than to have plenty of Lp and have it vanish suddenly at say 14Hz. Depends. If you can get enough Lp to last down to, say 7Hz, then the suddenness of its demise may not be an issue. This is the nub of the matter. Then the only remaining issue is DC intolerance, which is not a problem for DIYers. It's a matter of convenience versus hi-fi. The LP can become lower at lower applied voltage across the OPT pri. Williamson suggested minimun Lp = 100H for RLa-a = 10k. You can if you want work out what the Fsat is at 16W. At 1.6W, the Fsat will have moved down by a factor = 1/3.16, or about 1/3 of maximum PO. In the Williamson design you will discover that Fsat is excellently always low enough, and so is the Lp quantity But if you halve the Np and double the Afe, Fsat says the same for the voltages applied at LF. The increase of wire dia means load can be much reduced and PO much increased. Lp does not have to be 100H minimum. See my design example OPT No1 at my website. You can make your dainty little OPT if *you* want to. Not just me. Those who make good wound cores, such as Plitron, tend to spread the benefit between top and bottom end considerations. That's why their transformers are relatively dainty. Read Menno. I won't. OK. There are good reasons to use EI too. Ultimately it depends on whether you want a modern, or a historical, valve amp. That in turn is related to what kind of speakers you have, and to some degree on what kind of music you like. Nope, the use of GOSS E&I lams is still a very valid technique for wide BW OPT. History, type of speakers amd musical program all just don't matter one little bit. Its not hard to make an OPT suitable for use in a subwoofer amp. You do have to use a large Afe, and high turns per volt ratio. But if only for subh woofer, the HF does not matter at all above 200Hz. But then some idiot will try to use it for full range music.... So then you should make all tube amps capable of running a sub and also able to go to 65kHz, -3dB. I have a 25W amp in my shed with 2 x 6CA7, UL, and its fine to use with a sub. I've build nmany subs for ppl and tested them with my UL amp. Stuff on top of my metal shed roof rattles when i pump a big 30Hz note. It has the first OPT in it that I made in about 1994, and with Fsat at about 20Hz at 25W, and subs rarely have to handle signals below 20Hz, and anyway I found you don't need huge power with a sub to just make the missing bits between 20Hz and 50Hz appear with normal music. If you want to have movie sound effects up high or absurd bass levels from be-bop or rap crap and keep all the 17 year olds happy then you need more sub power than the amp power used for the main band. To make the OPT tolerant to dc imbalance and dc offset, you have to use lots of iron and more turns and maybe an air gap. Thus to get 60 watts of SE power from tubes requires an E&I core 72mm stack 51mm tongue and an air gap and very low µ of perhaps only 500. An SE amp is one which thrives on having a HUGE dc offset. In my SE55 amp with a pair of 845, load was 6k, and Lp abot 40H, Ia = 150mA. If I had a PP amp, the same core could be used to make a 200W AB amp, based on having core saturation occuring at 20Hz at full PO, ie, at the same F as in the case of the SE amp. If you want less iron, and you cut the stack height in half, then the Fsat for PP goes up from 20Hz to 40Hz. Thus is done routinely by most makers to save construction costs and freight costs etc. Dc imbalance worsens... But at 5 watts, a 200W tube amp might have a huge DC imbalance and nobody will hear anything. This is part of a more generic question because a conventional PP output stage is never quite balanced anyway, obviously particularly when operating in AB. I was set off on this train of thought by John's assymetric cathodes. I guess he can just make those caps really big, so their effects are both well below a bandwidth determined elsewhere. Again if nobody objects to weight and size and cost then there you can have choke feed to a pair of output tubes using a choke with CT and then cap couple the OPT which has its CT grounded. This allows normal drive, but then the choke suffers the imbalance if it occurs. But then if the coke is gapped, and had a massive sive and lots of turns then it may withstand far more dc imbalance than any normal OPT. Or one might use TWO individual air gapped chokes, but you can't get the large wanted inductance to prevent inductance shunting the load. And by this time your amp weighs half a ton and costs a fortune. In RDH4 the figures for pounds per watt for hi-fi are given. Most makers laughed at what the text books suggest they should have done. Nobody's amp need conform to the woeful lowest common denominator standards of the mass produced amps. EI is assumed, presumably. Figures are different depending on core material and geometry. Several solid state current servos designed to solve this problem have been posted here recently, however I would like to avoid polluting my design with a solid state current servo. Why? Is this pollution of your idea of purity of design, or of the signal? Also, there may be other alternatives. Perhaps routine manual bias resetting? Just how intolerant are you planning this transformer to be? After a bit of head scratching I came up with the following scheme which I hope I will be able to integrate into my overall amplifier design. http://fmamradios.com/stuff/CurrentBalance.gif What's that odd resistor for? Don't it need adjusting for Vak balance? What's the penalty for Vak being out of balance as the valves drift? Are there any other problems that might arise from asymmetrical grid resistances? The only good reason for planning transformer intolerance is the hope of achieving better low frequency performance AFAIK. That is, you sacrifice flexibility by maximising primary inductance. I wonder therefore how good your design is at maintaining perfect full power AC balance at LF? Particularly if it's running in AB. If LF AC balance isn't perfect, then surely your intolerant transformer will complain? Indeed. The OPT performance between dc and say 20Hz is a grey area which needs careful consideration. Where one does have two series tubes with say an 800V supply, one can cap couple a single winding OPT as I stated above. Its an easy OPT to wind compared to a conventional PP OPT with CT because it has half the P turns. The load is a lot lower than a normal PP RLa-a load. In the late 1950 Philips made a range of amps using 2 x EL86 in series with a supply = +400Vdc, and the anode cathode junction was at +200V and there was no OPT. The speaker had an 800 ohm voice coil impedance and was driven by an electro cap off the a-k join. Interesting, thanks. A better proposition now than then, because of the improvement in price and performance of electro caps. Interesting, but I have no plans to build an amp with sereies tubes. Philips did it solely to avoid the cost of an OPT. But the OPT which Philips was using was a walnut sized POS sourced from the cheapest supplier and with highg windinmg losses at 15% in may radio grams. The savings to be had were miniscule, and offset by the cost of winding voice coils with extremely thin wire to get 800 ohms of speaker impedance. Your logic doesn't stand up. How can you say they did it for cheapness, and also say it's not cheaper? Generally, you say *everyone* does *everything* for cheapness, except you. This makes you appear a bit grumpy and stupid. Using a pair of EL86 like Philips makes a fairly efficient 10W amp. But gimme a pair of EL34 in triode and a decent sized OPT. It'll **** all over the Philips idea, and over most other 19 watters. Dc imbalance is simply not a huge problem. But there still was a kind of small transformer to get the screen voltage of the top tube to follow its cathode voltage while being held at +400V. A choke would have worked fine though with cap bypass coupling to the cathode. The bottom tube just has normal fixed screen supply and at 1/2 the supply voltage for both tubes, ie, +200V. While working in class A the load of 800 is shared between the two output tubes and each tube sees 1,600 ohms. If one runs such an amp only in class A then with 12W Pda in each EL85, each tube puts 5W into 1,600 ohms or 10W into the 800 ohms. If you wanted to use the EL86 in a normal PP amp under the same Ea/Ia conditions then the OPT would have a primary load of 3k2 a-a, ie, with twice the turns of the single winding OPT. To get drive meant a special bootstrapped circuit was used with a 12AX7. But where the top tube has to have a large grid signal delivered, one may have a 1:1 IST with two windings. One winding is driven with a CF tube which is driven by a gain triode to make say 30Vrms. The other winding has one end connected to the a-k join of the two power tubes and thus the drive to the top tube is always equal to the bottom tube and there is symetrical drive conditions. NFB is applied the normal way, but with some care about gain/phase shift compensation networks because you have an IST in the signal path. To avoid having huge drive voltage and be able have an OPT with much lower RL, and also have no screen voltage to worry about then the 6AS7 or 6C33C come to mind as excellent canditates. But one could run EL34 or KT88 with Ea at 250V OK if one doesn't mind having a screen drive choke. 4 x KT88 would give 40W PP class A and load would be 560 ohms. The cap to couple it can be 100uF, and Lp would need to be at least 20H. The LF pole for resonance is 3.6Hz, and the load to damp the resonance is about 600ohms or less, so when loaded the OPT won't have a peaked LF response. Such a peak can make NFB difficult to apply. The higher the LP, the lower is the Fo, and one could also have the coupling cap value a lot higher, say 470uF, but then onmce you move Fo down to say 0.36Hz, then one is i the grey area with the cap charging and discharging an the core of the OPT is subject to slow ac changes and the resulting temporary saturation effects. As I mentioned in a previous post, If one doesn't like series PP tubes, then one may as well bite the bullet and have a large PP OPT set up conventionally, and then just air gap the PP core. Conventional PP amps throughout history haven't bothered much about DC balance, beyond offering a means of measuring and adjustment. It shouldn't be hard, these days, to automate that process of periodic maintenance. Conventional designs were designed by bean counters foisting their POS efforts onto the unsuspecting public. You can have a 20% dc imbalance and hardly anyone notices some thing is wrong because most ppl use 1/2 a watt average only even though the amp is capable of 25W. I have witnessed Quad-II owners enduring years where the OP tubes have drifted well apart so one tube has 40mA and the other has 90mA with a bit of red anode. Music is absolute crap above the 1/2 W level. The measured THD is many times what it should be at all levels. Go to my website to see what I do about the incompetence used by Quad to build their tube amps. There is no reason to go to your site, so I've never been. So you have said plenty of times. You remain fairly ignorant. You took lots of telling before you'd even consider reading RDH4. People laughed at you, and finally you got a copy of this very fine book. I assume it's like you write here, which is quite enough biased half-truths and lies for me, thanks. If I need reliable information, I go to the original source. RDH4 and other books with excellent reputations provided me with enough info coupled with years of apprenticeship so that I was able to distill it all into what is my site. After many years of having a large website there has not been one complaint from the reading public about me telling half lies. Not one knowledgeable engineer has informed me of any errors in technical issues. But here you are bull****ting about half truths. You really have to wake up. I won't hold my breath. One interesting thing about this idea of cap coupling, when applied to valves operating in DC parallel, is that the idle currents of the two valves can be independently adjusted for optimum AC performance. Oh, and all series output tube connections require a biased heater supply for one of the pair of tubes. Philips didn't bother though because the EL86 has a high heater-cathode voltage rating. Anyone remember the chap here who made *massively* parallel amps with hundreds of EL86? On the grounds, IIRC, that OPTs become relatively smaller as amps get bigger, so in terms of kilos per kilowatt, the bigger the better? Now EL86 aren't as cheap as they were, and electricity is much more expensive . If one has a big enough number of any tubes in series, one can couple the anode-cathode junction to an 8 ohm speaker and have class A action. But its easier to use a few power mosfets. If you had 300 x EL34 on top, and 300 on the bottom, and Ea = 250V and Ia = 100mA each, then the class A load will be conventional speaker type load and a nice load match for all the EL34. Or if you had 200 x EL86 instead of only 2, then you'd get 2,000W into 8 ohms without the OPT instead of 10W into 800 ohms. Nobody needs 2,000W capability, or to have to drain 5kW from the mains to get it. The easiest path to good dc imbalance tolerance is to examine how you think about the core µ during the design process, and just not cut any corners. I don't need dc imbalance tolerance, thanks. I adjust my bias currents routinely. I do too, where it is easy with just a pair of tubes in the OP stage. Very few makers ever provide air gaps in their PP OPTs. They just bung in the Es and Is with maximal lamination interleaving, or have the C-cores right up tight. Toroids have no cut gap. But some slight gapping to reduce µ down to about 2,000 does wonders for dc imbalance tolerance. It should be done using Afe and Np large enough to still have ZLp = RL at below 20Hz at full PO. Patrick Turner. Ian |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Is AC balance more important or DC balance? | Vacuum Tubes | |||
The Rock House Method | Pro Audio | |||
DELTA console - balance trim method | Pro Audio | |||
Best FLAC to CD method? | General | |||
Is there a name for this drive method? | Vacuum Tubes |