Yggdrasil Glitch Fixed

[Speed Racer]

37 minutes ago, Don Hills said:

Yes. So?

 

Oh boy.....

 

opus101 was suggesting that all Mike did was hide the glitch. That is not what Mike said he did.

[Don Hills]

Yes. So? I've been waiting for the other shoe to drop. Or was that it, simply an observation? If so, I apologise, I thought there was more to it.

Anyway, Mike did what he said he did. Got rid of the glitch. If you repeat the measurement, you see the glitch isn't there any more.  (I agree with Opus101, in theory it just got moved somewhere else, but good luck trying to find it by measurement.)

[mindbomb]

So was this major carry transition error? I don't understand how that could be fixed simply with additional calibration.

[opus101]

With an R2R ladder DAC the zero crossing is when all switches transition at once. But timing errors between switches mean their switch instants can never be truly simultaneous, hence the output code isn't correct until the last one has settled, resulting in a glitch.

 

No 'additional calibration' was used, merely the zero crossing of the DAC was moved some distance away from zero volts in audio terms so that low level signals no longer incur MSB transitions and hence fewer switches are changing state simultaneously.

 

Burr Brown's 'CoLinear' idea is rather similar - the PCM63 instead of using a single 20bit DAC uses two 19bit ones. The zero crossings of those 19bit DACs then move to -6dB, well away from low level signals.

[mindbomb]

To me, it sounded like he was describing dac calibration that is typically done to fix gain and zero scale errors, so that's why I was confused. But you're saying he is using a new coding system to avoid msb transitions in these problem cases?