On Wed, 6 Mar 2013 14:05:56 -0800 (PST), RichD
wrote:
Who's designing these things? In my experience, RF
designers are a rare breed, and with the digital market
vastly larger, they're even rarer.
Most such gadgets have very little RF inside. Todays BlueGoof, Wi-Fi,
all digital AM/FM receivers, SRD radios, and 433/900MHz weather
station chips are almost all digital with maybe a MMIC RF amp or
receiver pre-amp on the PCB. The design is being done by digital
chips designers, not RF engineers. However, one place where the RF
engineer is required is when the product has to pass FCC
certification.
I'll guess, the IC have been perfected to the no-brainer
level. But still, you need need amps, filters, antenna, plus
issues of noise and layout, yes/no? That stuff isn't obsoleted.
Yes and no. Each of the items you mention are somewhat separate from
the basic function of the radio and are most often a purchased part
from a company that specializes in the device. It is conceivable,
that someone with only a minimal knowledge of RF can assemble a
sellable and certifiable product using off the COTS modules and
components, including the actual radio. I've cleaned up the design on
a few such attempts. To be honest, the problems I fixed were
oversights due to lack of experience, which will eventually be
overcome by the designer.
I assume by "noise and layout" you mean digital noise trashing the
receiver sensitivity. Yes, that happens, but with clever design,
careful layout, and decent grounding, such problems can be minimized.
(Notice I didn't include shielding). The trick is that traces (wires)
radiate, while components do not. By simply reducing the size of the
device or PCB to the point where the radiating traces are sufficiently
small that they don't radiate enough to matter, many such "noise"
problems solve themselves. In addition, the power levels found on
todays radios are much lower than what was common even a few years
ago, making noise pickup less of an issue. Clever protocol design
also helps. For example, a GPS receiver with a processing gain of
43dB will not have a noise problem until the noise maybe 30dB above
the receive signal. Another example is the common 60KHz WWVB
receiver, where the 1 baud data rate results in such a narrow
bandwidth that the atmospheric noise inside the approximately 3Hz
receiver bandwidth is sufficiently low that it can almost be ignored.
I don't work in this area, but I'm curious, so can anyone
elaborate on what's going on, from a system viewpoint?
No. I can't. I might be able to provide some insight into current
trends in a specific product area, but not the entire world of RF
design.
What are the chip functions, options, price, trade-offs?
In which situations would you reject them, to roll your own?
Same problem as before. Too broad a question. As I mumbled, it is
possible to assemble a working product out of COTS (commercial off the
shelf) parts and pieces. Some volume production areas have been
heavily integrated, with plenty of mostly working chips available.
Others are specialty products, which are less well integrated. Today,
I would roll my own only when I have the projected volume to justify a
custom design, or when I want to protect the IP with a custom chip.
Is it simple on/off keying, or more sophisticated?
There's quite a bit of OOK (on-off keying) modulated products
available. TV remotes, WWVB receivers[1], wireless weather stations,
car security dongles, etc. OOK has the advantages of being very low
power, cheap, simple, and reliable. Sensitivity and efficiency
(bits/baud) are terrible, but for many applications, you don't need
the speed.
Currently,
in communications theory, sensor networks are a hot topic,
where thousands of sensors are competing for bandwidth,
but for mundane consumer apps, I doubt those issues arise.
Sensor mesh networks add a new set of challenges. The search for the
ultimate routing protocol for store and forward mesh networks is the
holy grail of sensor knotworks. This brings the level of complexity
well beyond the RF level and into the realm of queuing theory and
statistics. There's also the problem of scaling mesh networks. Too
few nodes, and the path could easily dead end. Too many and mutual
interference, loops, airtime consumption, and bottlenecking near the
backhaul point become issues. Incidentally, one of my favorite tests
with mesh networks is to put all the nodes in one room, turn them all
on, and try to pass some data. I've seen some that fall over badly
where nothing moves. They're not competing for bandwidth, but rather
for air time. As long as all the devices are using the same frequency
hopping code and RF channel, only one radio in view can be
transmitting at the same time.
As for "mundane" consumer apps, Wi-Fi mesh networks have all the
problems of sensor networks with the added enjoyment of multiple
incompatible protocols, overkill tx power, monster antennas, and
plenty of possible interference sources.
I'm looking to pick the brains of any gurus here -
Sure, but try to be more specific. What are you trying to accomplish?
[1] Yes, I know that it's not really OOK because the carrier is
reduced by 10db at the beginning of each UTC second.
--
Jeff Liebermann
150 Felker St #D http://www.LearnByDestroying.com
Santa Cruz CA 95060
http://802.11junk.com
Skype: JeffLiebermann AE6KS 831-336-2558