Design of a High-Performance DAC

[amirm]

I missed this talk at Rocky Mountain Audio Fest (RMAF) 2016. Fortunately they just put it online. It is a talk by a team of four on the design of a new Boulder 2120 DAC. It is about an hour long so you may want to speed it up to watch it in less time.

It hits on things like atomic clocks, clock fidelity, synchronous transmission versus not, resampling, selection of DAC chips, discrete versus monolithic DAC chips, software stacks, modularity, etc. It is not a super deep talk but still, highly worth a watch

 

[amirm]

Here is the DAC by the way:

It is the top box with streaming and large high-resolution display support.

 

[amirm]

Here is my thread on it from RMAF: http://audiosciencereview.com/forum...boulder-soundings-analysis-plus-rockport.917/

 

[fas42]

Nice design!

 

[Cosmik]

But in order to achieve real quality, we need to multiply it by three to build active speakers. That's going to be quite a big box.

 

[amirm]

I was not recommending the product as much as the video of building it. Personally I am not a fan of putting streaming technology inside DACs. Much better to do that on other solutions than inside the DAC and be at the mercy of a small company to maintain it over time, using their app, etc.

 

[dallasjustice]

The main engineer discussed atomic clock jitter. He's correct about what he's saying. But I think he is incorrect in how he communicates his point. He uses piano tuning as an example of how oscillator jitter ultimately affects frequency response tones on a DAC.

How the jitter from a crystal oscillator inside a DAC affects the output of a DAC is much more complex than that. He also refers to jitter as one number and that's not very helpful either.

 

[Phelonious Ponk]

And in the end, the difference between the high-end DAC and the one built in your iPhone is quite small. Especially to aging, male, audiophile ears. So many angels on the head of this pin.

Tim

 

[Thomas savage]

How do the various diffrent 'clock' implementations drift over time?

 

[amirm]

The main engineer discussed atomic clock jitter. He's correct about what he's saying. But I think he is incorrect in how he communicates his point. He uses piano tuning as an example of how oscillator jitter ultimately affects frequency response tones on a DAC.

How the jitter from a crystal oscillator inside a DAC affects the output of a DAC is much more complex than that. He also refers to jitter as one number and that's not very helpful either.

I don't think he was talking about jitter in that context but rather timebase accuracy which would relate to speed of playback. But I may be mistaken.

 

[amirm]

How do the various diffrent 'clock' implementations drift over time?

Due to aging and temperature they drift. You can pay more of course and get better accuracy.

 

[Cosmik]

Due to aging and temperature they drift. You can pay more of course and get better accuracy.

I must confess I haven't yet watched the vid yet, but crystal clock modules with a temperature-compensated accuracy of 0.5 ppm are available for a few dollars. I think that works out at about 15 seconds per year.

But if you feed the output of your atomically-calibrated DAC into a non-time aligned speaker you get a grossly staggered signal at your ear, exacerbated by the effects of bass reflex. Feed it into a two-way speaker and the mid range is being heavily Doppler modulated by the bass.

Clearly a mega-DAC is pitched at users of non-DSP, non-active, systems. Obsession with timing of the DAC to absurd precision while allowing massive errors in the speakers seems like a failure to think in terms of the overall system.