Previous Entry Share Next Entry

Shaving a Yak

I have gotten distracted shaving a yak.

I've been looking to put into a more permanent form a number of the little circuits I use for testing audio projects. They all tend to have the same class of problem in common-- getting a signal, be it collected from a circuit or calibrated microphone, and getting it into an A/D box with as little hassle as possible. The two recurring problems with just whipping up a circuit on a protoboard or grabbing the output of a mixer are calibration and noise. Exposed circuits are noisy and the unbalanced inputs on the A/D box don't help. Using a mixer where a deflection of a few degrees on a knob could bean a difference of 10dB means checking the calibration of every step very carefully using the oscilloscope.

The biggest reason for wanting to get signals into the computer is that digital analysis seriously outstrips anything I can do outside of software. Complex spectral analysis and all the standard audio tests are damn near trivial in the digital domain. So I've started working on a little box that puts the calibrated, noiseproof basics of collection in one place. Suddenly I'm glad I bought all those Russian rotary switches.

First step is a nice, bulletproof input buffer. I can't just plug an RCA cable into the sampler, chop the other end off and start touching wires to things.

  1. For one, the sampler input only has a 2kOhm input impedence. That's far too low to collect data without affecting the circuit I'm watching. The input impedence needs to be 1M minimum, 10-100M is better.
  2. It would be noisy as heck due to all the ground loops.
  3. Everything would have to be ground referenced; no arbitrary differential testing.
  4. No protection: Touch the wrong thing and pop goes the weasel^H^H^H^H^H^H$300 sampler box.
  5. Limited to sampling signals on the scale of 10mV-1V. No attenuation, no gain.

So... a floating input buffer with overvoltage clamping and gain selection. Here's a first attempt that floats completely, although it will only work within the absolute range of the power supply (post-attenuation). There needs to be a ground referenced somewhere--- but that would happen at the source or the sink. The ground would not be carried in the signal path. Source and sink are completely isolated. In addition, several can be used and remain isolated from each other.

Barring finding any flaws, this is a good circuit but overkill. My A/D box has unbalanced inputs, so a fully differential drive probably gives me nothing while sinking an extra opamp. An unbalanced drive must, however, reference ground. And since the grounds on each input are all tied together inside the A/D box, referencing the power supply ground anywhere inside the test circuit introduces ground loops when more than one buffer is used at the same time. The buffer still needs to float-- One possible solution is to only reference signal ground from the individual cable fed to the A/D box:

Both circuits test nicely in Spice, which is a very different thing than good performance real-world. Part of the point of floating the ouput is improved noise performance, and only testing will tell. I'm also hoping some of my readers have suggestions for obvious things I've missed.

Comments Disabled:

Comments have been disabled for this post.


Log in

No account? Create an account