Cheap Electronics Dissection Project
Kragen Javier Sitaker
kragen at pobox.com
Mon Oct 30 17:49:47 EST 2006
On Fri, 13 Oct 2006 18:48:29 -0500, Shae Matijs Erisson wrote:
> Kragen Javier Sitaker <kragen at pobox.com> writes:
> > But the FCC doesn't rule the world.
> Well, they rule the largest market...
> (Or do they? What's the total market for radio using devices in Japan, etc?)
I think it's bigger than the US, because the device lifetimes are
shorter. But even the biggest market is only a small part of the
> > I think the USRP is still a little pricey, and a paperback book is
> > a heck of a big cell phone these days :)
> Yeah, true... I guess the profile should either be backpack or
> pocket sized. Nothing else is really portable. One good solution
> might be backpack size with the new Marvell combo bluetooth & wifi
> chip (assuming they'll let people write non-stupid drivers, see
> http://lwn.net/Articles/203562/ for details )
I'll read that when I get a chance. I'm still thinking that if you
want to sell the thing in, say, Perú rather than Japan, you can't
afford the amount of circuitry needed for Wi-Fi.
> Nokia is making a stereo bluetooth headset, model BH-501. That would
> let you keep your device in your backpack and use it without direct
> keyboard access.
> > A wonderful list.
> Did any of these ideas seem like good candidates for killer apps?
In the mass-market sense? Voice chat is the main one, I think.
> After fifteen minutes of thinking about it, I don't see how it would be
> possible to use SDR to detect ionizing radiation. Neither scintillating nor
> charged electrode detectors would benefit from SDR.
Maybe if the ionized radiation resulted in lightning --- which
wouldn't necessarily have to be the big visible kind.
> >> Maybe 'social distance' could be a way to discover a good voice mesh
> >> encoding?
> > What kind of encoding are you talking about? Presumably not an audio
> > compression algorithm.
> No, more than that.
> With SDR you could easily tune the design of a signal to its intended use.
What do you mean in this case?
> >> Cheap $50 mesh phones would be popular in places where cellphone networks
> >> will never be setup due to economic or terrain problems.
> > I'm not sure about this $50 thing. Maybe. I suspect it might need to cost
> > less than that.
> I think you're right, any ideas on how to make that happen?
A few: CPUs without MMUs, low RF data rates, mesh networking to lower
needed RF power, maybe single-chip shared-nothing multi-CPU designs
like Chuck Moore's 25x, pulse radio, pure-audio or audio+morse user
interfaces, store-and-forward voice (so the radio data rate doesn't
have to keep up with real-time, and to tolerate transient reception
problems, which become worse when you use fewer watts), no
rechargeable battery, maybe no battery but just a capacitor. Maybe
use ultrasound or light instead of radio. Solar's reasonable too.
> > Good point, although sand dunes move pretty fast, so your node
> > would be completely buried in a few weeks. You could probably
> > just use thermocouples, or small solar panels.
> If your entire device is the size of a pringles can you can just
> keep it in your backpack or attached to your belt, and then give it
> a thermal gradient when you're using it. Even half in direct
> sunlight half in shadow would give some power.
True. How does the available power per dollar compare to amorphous
solar panels? How about power per unit area?
> > What's the appeal of liquid?
> Spheres, disks, etc all force a particular shape on the component. If liquid
> could flow through tubes and generate power, you'd be able to pack the
> generator into most any device without affecting its overall shape.
> That would also allow you to have a flexible generator.
Hmm, suppose you took a piezoelectric mercury delay line and removed
some of the mercury, leaving a vacuum, like a thermometer. (Ideally
it'd be a little thicker than a thermometer because of viscosity, but
still thin enough that the surface tension kept the mercury filling
the whole width of the tube.) Whenever the mercury slid from one end
to the other, it would generate an impulse of electrical energy from
the piezoelectric terminal at that end. It sounds like a workable
device, but it might have a short lifetime in use due to the shock of
the mercury hitting the end.
The Forever Flashlight power plant might be a more practical version:
a tube with a permanent magnet that can slide (or fall) up and down
the tube, with a coil of wire in the middle of the tube. When the
magnet falls through the wire, it generates a pulse of electrical
energy, but without the shock waves of the mercury version.
> > Yup. Also other kinds of development tools: simulation tools,
> > perhaps development environments (esp. for things like the USRP
> > where you need to program FPGAs.)
> Yeah, it's too bad there aren't any FPGAs with publically documented
> specifications. Netlists suck.
There was a recent US court case that might help that change.
I assume the reason for the lack of public specs was to make it harder
for competitors to clone the FPGA, e.g. as an ASIC. But some FPGA
company just won a case asserting that its "mask work" rights in its
FPGA were infringed by a company that produced a netlist-compatible
ASIC, so the FPGA company's customers could turn their designs into
ASICs easily. They won the case.
I wish I remembered the parties to the case.
Anyway, given that, I don't know why FPGA companies would now want to
keep their specs secret. So that they can make more money on the
per-seat licenses for their synthesis software, at the expense of
selling more FPGAs? Doesn't seem reasonable (although I don't have
> > That's a problem with radios in general. In the Wi-Fi case, my
> > best-case received signal strength is something like -55dBm, or
> > three millionths of a milliwatt, three nanowatts, and my card
> > claims it can still at less than a picowatt. But when I'm
> > transmitting, that same antenna emits as much as 100 milliwatts,
> > 11 orders of magnitude greater. It's hard to detect a signal when
> > the noise is 10 orders of magnitude greater.
> Nifty encodings like the one GPS uses would probably be quite
> popular if we had SDR. I hope people would just try new stuff and
> record the results.
How does the GPS encoding work?
> > So my Wi-Fi card doesn't try to send and receive at the same time,
> > and the 802.11 PHY standard doesn't even try to do CSMA/CD,
> > relying on RTS/CTS packets, acknowledgement, and PHY-layer
> > retransmission instead.
> And there's probably some nifty trick that will let you do CSMA/CD...
> If only we had SDR so we could search for such a thing!
Like I said, CSMA/CD isn't really feasible, because the "carrier"
you'd like to sense in order to detect collisions is 11 orders of
magnitude weaker than the signal you're putting out on the same wire.
And that's not even getting into the geometry problems --- it might be
that you're 100 meters to the west of the AP, the guy you're colliding
with is 100 meters to its east, and so while the AP can hear you both,
his signal is 6dB lower at your antenna than at the AP, so you can't
hear him at all.
> Yeah, switching time. Have you read "The Final Nail in WEP's Coffin" ?
I don't think so.
> >> Another important piece of infrastructure would be a library of
> >> antenna designs and ways to build your own, extending on the
> >> cantenna idea.
> > Possibly, especially when we have more automated fabrication. I
> > think there's probably a lot you can do without it, though.
> Yeah, it's not nearly as important as being able to send/receive
> signals in the first place!
> > Does that make it synthetic aperture radar if your GPS subtracts
> > successive readings to tell you which way you're walking? :)
> Right now GPS uses successive readings to get heading, but I wish
> for more...
I know, I just meant that that's analogous to how SAR uses successive
readings to get more detail.
> >> FPGAs are good for reconfigurable hardware, but does that exist for
> >> antennas?
> > Not that I know of. I wonder if it's possible.
> Sure it is. At worst you could use a MEMS system like the color
> projectors that have the zillion micromirrors inside them. Then you
> could build antennas out of 'pixels' just like anything else.
Of course you're correct, once you have big enough MEMS, or small
enough signal wavelength.
More information about the Kragen-discuss