Digital, but not by the numbers

[abraxalito]

Qusp and I found that we've got similar objectives about a fully-fledged digital system (digital XO and multichannel DACs, potentially integrated with amps), even though we might not yet be able to agree on particular DAC chips so that inspired me to start a thread on digital from a systems perspective, with the following constraints to help maintain focus:

1) Redbook (44k1/16) rather than hires. That's because most software is still redbook though this does not preclude expansion to include higher rates and bit depths in future.

2) Analog rather than digital sound. Seeing as the word 'analog' is something of a catch all, by this I mean no false detail, no grit or harshness, no typical "audiophile show" digital sound. Rather mid-range purity, huge soundstage depths and great dynamics. The aim being to approach (or even exceed) the level of satisfaction enjoyed by vinylphiles.

3) Cost effective, avoiding use of boutique bits and ignoring audiophile approved parts where possible. This means fancy components can be retrofitted by individuals if they like, the designs in the main won't call for them.

4) A focus on whole system integration and synergy rather than endless details of technical measurement, in particular (but by no means limited to) jitter. This doesn't mean the traditional measurements no longer matter, but that they're not so appropriate to focus on when the aim is subjective enjoyment.

I intend to throw in details about the DAC design I've been working on which so far has only been outlined on my blog. As such this thread could be seen as like a hardware complement to the mainly software focus of the 'Open Source DSP XOs' thread.

[TNT]

I think You are on to something.... system optimization rather than silo thinking is the way forward. Subscribed!

[qusp]

sweet, OK i'll finish off my PM to you when I get up and think of something to add here.

[fas42]

A big thumbs up from me of course, I'll follow this with keen interest ...

Frank

[twest820]

There's some discussion over here about standardizing I2S impedances (and, presumably by extension, those of clock lines) to enable better integration between boards. But I've also been wondering about defining some kind of power interface since the layout of a linear supply board is enough different from DSP and DAC boards that it's kind of interesting to break it out separately. Would such power distribution be on topic for this thread?

[abraxalito]

It could be if you have done (or plan to do) some experiments with different power topologies that involve listening to how they sound.

I recently changed a power supply to an opamp in an integrated amp over from linear regs to the same regs operating as (relatively distant) current sources into local TL431 shunts. This made a huge improvement to the SQ - ever since then I've been building power supplies in this manner. However the improvement could have been down to uncontrolled factors I admit but conceptually this makes sense in that it keeps the larger loops high impedance so hopefully more immune to noise pick-up.

[twest820]

My usual approach is if you can hear the supply there's an objective bug in the design. I'll find a more appropriate thread.

[fas42]

Originally Posted by twest820

My usual approach is if you can hear the supply there's an objective bug in the design. I'll find a more appropriate thread.

I would say on most systems you can hear the supply, but everyone is so used to it they perceive nothing unusual: "that's how all hifi's sound when you turn up the volume" - that sort of thing ...

Frank

[marce]

Curious as to how a well designed DC supply can alter the sound, it is DC after all.

[abraxalito]

Well in the case I cited, because I managed to improve the sound considerably, I'd say it wasn't so well-designed.

In particular it used an LM337 running from the main rail of the power stage, with no passive filtering on the input. The opamp in question was the AD712. When you go to the datasheet you notice (if you look carefully enough) that the LM337's line rejection sucks big-time at HF and - lo and behold - the AD712's -ve PSRR also sucks. On the +ve supply things are much, much better as the LM317 has perhaps 20dB better rejection and the PSRR of the AD712 is superior to a similar degree. So which PS rail was more DC than the other?

So no, DC is a myth, all supplies have AC noise on them