A conversation with Charles Hansen, Gordon Rankin, and Steve Silberman

[Sam Lord]

Wow, thanks for the conceptual explanation of that Sam. Fascinating. Now I get to got and ask my engineering partner to explain that in more detail--which will mostly sail over my head!

 

Can you cite some examples of DACs that DO use "gain-ranging?"

No, I haven't heard of any, and I don't know whether the technique is feasible in any real-world audio DAC. I only mentioned it because 140dB is rather impressive for any kind of circuit, and gain-ranging is primarily a noise-reduction technique. In a general sense, every DAC and ADC performs gain-ranging: it switches to or from a different analog part as it modulates. But "gain-ranging" in the circuits I mentioned indicates multiple (two or more) switched modulators. That is different from though not necessarily better than interleaved or paralleled modulators. I think the Stagetec ADC uses four exclusive modulators, each for a separate range of output. Some amplifiers switch between two major circuits, though they are usually supporting circuits like the power rails.

[Sam Lord]

Because of the light-weight 6th order polynomial interpolator with low 3.125 MHz output rate. Since it has the 0.001 Hz word clock pull, it's practically a polynomial ASRC to reduce jitter at the same time.

Thank you Miska.

 

Not enough processing muscle, resulting in fewer coefficients, greater quantization error and greater distortion? I have only begun to study numerical methods (Dormand-Prince, etc.) so this stuff is way out of my league. Bruno did mention in his blog the need to "shoehorn" his interpolation calculations into one SHARC processor per channel. I wonder what the tradeoffs were: latency, electrical noise from processing, routing of the clock signal...? Anyway, I highly recommend that folks read his design blog at the mola-mola.nl.com website, it is really illuminating.

[Miles]

It's not that simple Eloise.

 

This is a good introductory article on USB Audio, with a particular focus on getting the right bits at the right time:

 

Fundamentals of USB Audio | EDN

 

Isochronous USB uses checksums so you know if the bits are the right ones, but as it's real time, you can't resend the data if it's wrong. The error rate is really low though, as that slide shows. The error manifests itself as a 'tick' or a 'pop', which Charles Hansen mentions in the original article. I've experienced this with an Audioquest Dragonfly when running on an overloaded USB bus (it wasn't getting enough power).

 

I think the main takeaway from the article is sending a digital encoding on a wire is not trivial, but good engineers have overcome the problems and can do it reliably.