SE Headphones Amp
On Oct 1, 7:58*am, Ian Bell wrote:
Patrick Turner wrote:
On Sep 30, 8:00 am, Ian *wrote:
I have been thinking a bit about Patrick's suggestion of using something like an EL84 in an SE
headphones amp. I was intrigued that he had built one using a 'cheap' output transformer so I have
looked at specs from various suppliers of such transformers. These transformers typically have a
primarily inductance of around 10 Henries (at 20 or 40mA) for a nominal 5K match. Now my transformer
rule of thumb is you want the primary inductance to be twice the plate load at the lowest frequency
of interest (say 20Hz). So for 5K ohms and 20Hz the inductance needs to be getting on for 40 Henries
so is not the bass response of these transformers going to suffer somewhat??
Cheers
Ian
Most old radio OPS designed for a 6V6 operating as a beam tetrode with
no NFB have low Lp because they are designed to give from only 100Hz
upwards as the little speaker used just can't cope with any below
100Hz.
You are right about the 40H. If there is that much L and Ra was 2k2
with EL84 in triode and the RL was a lot higher or not even present,
then response will be -3dB when XLp = Ra or at 8.75Hz, so Lp could be
16H to give -3dB at 22Hz. FB can flatten the resonse further. The real
other problem is that if Lp is low, it is often because the core Afe
size is too small and there are not enough P turns to prevent core
saturation, ie, Bdc + Bac exceeding about 1.6 Tesla at an F that is
too high.
In other words, we want Fsat to be low as possible at the 1kHz
clipping signal voltage.
If one designs for Fsat at below 20Hz then the size of a 5 watt rated
SE OPT can be 4 times heavier than some silly bean counter designed
crap from an old radio. Some such old radios or old hi-fi sets do have
just enough Lp and freedom from saturation at too high an F but they
should be carefully measured before use.
Most old radi OPT just have one sec section *wound over the top of a
single primary which gives a poor HF response.
But even if Fsat s at say 50Hz for full Vo, for phone use the Vo may
only be 1/4 full Vo max and Fsat will then be 12.5Hz which s
acceptable.
Which is an interesting point. Say we have a SET OP transformer with 10H and a 16 ohm secondary.
One watt into 16 ohms need 4V rms whcih is plenty to drive the phones I am intersted in. So suppose
I connect a 32 ohm headphone to this 16 ohm winding and it requires no more than 2V rms (for 125mW)
I would expect lower distortion because of the lower level and higher reflected ac load. What would
happen to the bass response with a 10H primary?
Its all very well to consider a "SET OP transformer with 10H and a 16
ohm secondary" but you have not mentioned the primary load nor type of
device driving the primary load.
If you have a single EL84 in SE class A1 triode then consider a
typical op point of Ea = +300Vdc and Ia = 30mAdc, and Ra = 2k0 approx..
Total RL + Rw for maximum PO = ( Ea / Ia ) - ( 2 x Ra ) =
( 300/0.03 ) - 4k0 = 6k0 and let us say Rw total is 10% of the totalo
RL so actual RL = 6k0 = 600R = 5k4.
Max PO = RL x ( Idc squared ) / 2 = 2.43W, and anode load Va = 0.707 x
5,400 x 0.03 = 114.5Vrms.
Your OPT has a 5k4 : 16 ohm ratio which gives ZR = 337.5:1 so the Ra
appears at the output as 2k0 / 337.5, plus the total Rw / 337.5 =
5.96R + 1.77 = 7.73R.
This is a not very good outcome unless we add at least about 12dB GNFB
from speaker connection to the driver tube cathode to reduce Rout to
about 2 ohms, giving a DF of 8 with 16 ohm load.
The response at low levels is given as the pole formed by Lp in
parallel to RA, which is Ra + RL in parallel. RDH4 spells it out.
Anyway, assuming the load is resistive at low F where LF cut off
occurs, then RA = 6,000 // 2k0 = 1k5, and 10H has XL = 1k5 at 23.9Hz,
not too bad. If the load around cut off was very high the low triode
Ra manages to keep the source R low so the Fco remains low.
If the sec load is higher than 16 ohms, say 32 ohms or 320 ohms, the
triode Ra remians the main dominant source resistance and Fco will
remain fairly low. My example presumes anode RL with 16 ohm sec =
total of 6k0, and this is 3 x Ra and maybe you get 4% THD at 2 Watts.
With anode RL = high as possible, ie, CCS, then THD will be about 1%
at the maximum Vo but because there is no Ia change there is no PO, so
THD/IMD will vary between about 4% and maybe 2% at maximum *useful*
Vos for RL above 6k0.
But should you use ther EL84 in pentode with Ea = 250V and Ia = 40mA,
Ra = maybe 50k0 and then the Fco is MUCH higher and more dominated by
total RL which may be 0.9 x 250/0.04 = 5k6, if in fact there is a an R
load present at Fco. With 10H and RA = approx 5k0, Fco = 79.6Hz.
The lower the Rsource the lower is the distortions caused by the damn
iron.
EL84 THD in SE mode at near clipping without any NFB anywhere is
usually well over 10%, and fulla odd order H.
Triodes are the gold standard for driving OPTs but one may completely
overcome the shortcomings of pentodes by using NFB in the form of
global, or local CFB or maybe UL plus global.
When I made a pair od SET amps for a customer a couple of years ago I
used 2 x 845 in parallel to get PO max = 60W.
The noise was only barely discernible with head held against a
speaker, and noise was estimated at 0.25mV.This illustrates that one
may use any power tube one wishes to use for either driving a speaker
*or phones*. With a luxury of a 60W output max to 5 ohms, or 17.3Vrms,
one might only want say 3Vrms max at the phones of 32 ohms which is
280mW. The R divider might have R across phones output = 2R2, and feed
R of 10R, so that noise is then reduced to 0.25 x ( 2.2/12.2 )
0.045mV, and if the average phones level was 0.2Vrms, then unweighted
SNR = 20 log 0.000045/0.2 which I guess = -72dB approx and possibly
acceptable, because the background noise of the recording venue is
likely to be higher if nothing else is.
So a single EL34 for phones would be OK, or a 1/2 of 6AS7, and so on,
providing you think it through. PP use of 6SN7 was common with a 20k:
16 OPT.
If one wants to drive down to 8 ohms without an OPT, I'd suggest a
single power mosfet in source follower mode which may be driven by CR
coupling from EL84 in triode. The triode needs a B+ of say +350V and a
CCS feed to anode so that Ea = 250V and Ia = 25mAdc. The EL84 will be
**extremely linear** while producing only up to say 3V into its very
high R loading which is merely the biasing resistors of the mosfet
gates.
Diode clamps would be needed to prevent excessive drive voltage
appearing across Vgate-source. The mosfet would have Ed at +20Vdc and
Idc = 0.5A, and a CCS sink taken to -20vdc, with load connected via
4,700uF.
But better would be to use NPN and PNP mosfets in a complementary pair
and biased for class A1 between the +/- 20Vdc rails.
Direct coupling to a load is then possible if DC offset is controlled.
The Pdd of each mosfet = 10W, and somewhat high to drive phones - 20W
total per channel.
But you can get about 9W of max PO into 18 ohms, 12.7Vrms pure class A
with THD approx 1%. But the noise problem may still be there and a
resistance divider needed. No phones should ever require more than say
1V, so R divider can be 22R + 2R2 is OK and probably class AB action
is OK so Id may be reduced to 250mAdc.
Rout from mosfet followers is about 2 ohms without any NFB, so the
Rout is about determined by the R divider lower resistance.
At 0.2Vrms from the divider, there may be 2V from the mosfets where
THD may be as low as 0.17%. Reducing the THD further means using
global loop NFB and more tube gain. The complementary pair of mosfets
will produce far more undistorted Vo than will one mosfet with
CCS.
Since an EL84 has a gain = approx 20 with CCS loading, then gain with
R divider will be slightly above unity, and about right for use if the
source is a CD player.
One's imagination may easily be employed to make a headphone amp much
simpler and better than the dreadful range of insipid bean counter
inspired designs currently available at excessive prices in hi-fi
stores.
Patrick Turner.
Cheers
Ian
Patrick Turner.- Hide quoted text -
- Show quoted text -- Hide quoted text -
- Show quoted text -
|