Re: [buildcheapeeg] Amplitude modulation

From: Jim Peters (jim_at_uazu.net)
Date: 2001-12-27 22:04:39


Andreas Robinson wrote:
> I did some calculating on AM transmission about a month ago...

I'm not going to push AM as a solution, because I know almost nothing
about electronics, and the modular EEG looks the way to go, with lots
of advantages.

But I'm not sure about all the calculations below:

> 2. The signal is filtered so that you get 96 dB
> attenuation at 9kHz. A third order Chebyshev I filter
> is enough.

Why filter at 9kHz ? We're talking about sampling at 256Hz in the
modular EEG design, so I reckon a 1kHz carrier would be fine. If
making fancy filters is difficult, use a 4kHz carrier. This leaves
plenty of space between 128Hz and 4kHz for the filter to operate.

To my simple-minded way of looking at it, you get a data point for
every carrier cycle. All you have to do is look for a maximum (with
maybe some fiddling due to the inversions for -ve signal levels). If
the software tracks the signal to stay in sync with the carrier, then
accurately spotting maxima/minima the right way around should be no
trouble. Or is this too `cheap' an approach ?

I was imagining a sine-wave carrier, by the way.

> 4. The signal is amplitude modulated to 10kHz. This is
> the tricky part. If it is done badly, a lot of
> resolution is lost.

> * Modulate using a switch, flipping it back and forth
> between the signal and a constant voltage at a rate of
> 10kHz. Then filter the signal using a 6th order
> elliptic filter. If you don't do that, you'll get
> aliasing in the sound card.

I'm cautious about this approach. For a slow 1Hz wave, for example,
this means that we would be switching between a +ve voltage and 0V for
0.5 sec, then -ve voltage and 0V for the next 0.5sec. A soundcard
20Hz high-pass filter would make the 0V level appear to `sag' or
`rise' over the half-wave periods. You need a balanced +ve/-ve signal
to avoid this.

For example, this:

* * +V
* *
*** * * * *** * * * *** 0V
* *
* * -V
0 0.5 1.0 sec

Gets distorted to something like this, which is impossible to
interpret on the software side:

* * * * * * *
*** *** *** 0V
* * * * * * *

0 0.5 1.0 sec

So, the carrier would have to be a sine-wave, or maybe a balanced +/-
square wave. If a fairly accurate balanced +/- wave can't be
generated easily, or if a modulator is hard to do well, then I guess
the whole thing is not going to work at all anyway.

> * Modulate with a sine wave, the old fashioned way. Requires an
> experienced analog designer to get it right.

Okay, not me then.

Anyway, I agree -- stick with the serial RS232 solution that is
currently being developed. There's no point in wasting development
effort on a soundcard solution if it will just slow things down,
especially when the ideas are untested and there are other known
limitations to the approach.

Serial is also much cleaner for the software -- you don't have to
worry about duplex audio (as someone pointed out), or soundcard mixer
settings or any of that stuff.

I'll not mention AM again !

Jim

-- 
Jim Peters (_)/=\~/_(_) Uazú
(_) /=\ ~/_ (_)
jim@ (_) /=\ ~/_ (_) www.
uazu.net (_) ____ /=\ ____ ~/_ ____ (_) uazu.net


This archive was generated by hypermail 2.1.4 : 2002-07-27 12:28:36 BST