[time-nuts] EFC tracking

[Oz-in-DFW]

On 6/26/2010 7:12 AM, Steve Rooke wrote:
> <Deletia> 
> I next thought about turning the DC into AC by chopping it, IE.
> inverting 50% of the voltage via an oscillator. This way I could pass
> the square wave directly into an unmodified sound card, take
> measurements and then do an RMS calculation on them (really just need
> to flip the sign on, say, the negative readings).
>   
I've done similar stuff in work projects, but never written code.  I've
thought about this some as well.  I'd consider a few things;

   1. Use the sound card output as the chopper control signal instead of
      the discrete unit.   You'll have more control and phase sync will
      be easier.
          * I'd be temped to take the sound card output and run it
            through a comparator to square it up, but I'm almost certain
            this isn't needed.
   2. Buffer the input so that your waveform is not so dependent on
      source impedance.
   3. Make the input buffer differential so that you can get some small
      amount of ground isolation and CMRR
   4. look at the 4053 mux, it might make your interconnect life easier.
   5. The probelm with chopping is that signal levels around zero don't
      have much amplitude and are a challenge to extract from noise.
   6. If you mix (in the RF receiver sense, not sum in the audio studio
      sense) rather than chop the DC offset becomes a phase shift,
      generally pretty easy to calibrate for and decode from the output
      samples of a sound card. See
      http://en.wikipedia.org/wiki/Frequency_mixer

> I wonder if anyone has done something like this before and could share
> their experiences. I've attached a diagram image (hope it is accepted
> by the list) which is my first go with Eagle so I'm not exactly very
> familiar with it, sorry. The R's and C's in the astable would be set
> to a clock frequency that enables this to work without bias given the
> sampling frequency. I'm not sure if this clock should be slower than
> the sampling frequency or higher, just haven't got my head around that
> yet. 
The clock needs to be much higher than the highest frequency of the
input waveform to keep Nyquist happy and things simple.  You can do this
inband, but you don't want to. 

If you chop very close to half the soundcard sample rate I suspect
you'll get no output because you'll be in the roofing filter cutoff and
your waveform will integrate to zero.  I suspect you want to be 5 - 10X
below that to make waveform recovery easier, and even lower is better.

So, if you use a 44.1 ksps default rate, Nyquist is 22.05.  I'd run the
chopper at less than 1 kHz. The good news is that your input waveform
period is hours (maybe ~100 microhertz) and chopping at 1 Khz will make
100 Hz response easy and 500 Hz possible with great care and some effort. 
> The R's around the op-amp would need to be set in a ratio that
> transforms the EFC voltage into the range that the sound card can
> handle (that is yet to be calculated by measuring the limits). 
Most sound cards I've seen are ~ 1V pk to peak, though some are MUCH
higher. 
> If you
> have any suggestions or ways of doing this in a better way, I'd be
> very grateful for the advice.
>   
It's worth exactly what you've paid for it...
> Thanks,
> Steve
>   
>
Oz

-- 
mailto:oz at ozindfw.net    
Oz
POB 93167 
Southlake, TX 76092 (Near DFW Airport)