Proposal for grouping DACs by measured performance

[Blumlein 88]

What I have in mind is something similar to Spurious Free Dynamic Range. You have maximum signal and any other signal whether noise, harmonic or intermod distortion or jitter spurs or idle tones or aliasing or anything that should not be there with the signal.

I would propose the clear to -100 db group, to -110 db group, and if we get any, the clear to -120 db group. Anything less than a 100 db of free and clear range these days should be considered a fail.

The only thing I can think of that wouldn't show up here is some linearity errors. Those would only occur with multi-bit DACs and a note could be attached to the results of those.

Now there are features and other things worth considering, but perhaps this would allow us to group DACs by basic performance to narrow the field some and make it simple for non-techies. All such simplifications have problems and leave out details. The question is whether it is helpful short hand, and what it leaves out is less important than what it gets right.

I'll post a proposed way of viewing these in the next few posts using spectrograms.

 

[Blumlein 88]

As an example here is a spectrogram for a DAC I have playing a 1 khz tone at - 1dbFS.

Notice other than the 1 khz tone everything else is clear. This spectrogram goes to gray at -100 db. So there are no distortions or other tones above - 100 db. I am using a 2k FFT to generate the spectrogram.

Now the same only the range of the spectrogram has expanded to -110 db. We see there is a touch of 3rd harmonic distortion showing up at 3 khz. Otherwise still clear.

And now with the spectrogram coving 120 db. We see the 3 khz harmonic distortion more clearly and some very low level tones of some sort showing up just above the 16 khz area. Plus one faintly at 8 khz. Maybe USB bus related?

So if this DAC clears a 100 db for the various normal tests like jitter, thd, noise, and IMD it would be 100 db clear. Yet it does not meet the criteria for being clear to 110 db or 120 db.

 

[Blumlein 88]

Here the spectrograms for a twin tone IMD sweep. Two tones separated by 1 khz are swept from 1 khz to 20 khz. You would expect a 1 khz difference tone at some level and some sum and difference tones around the twin tones.

Here at a 100 db range to the spectrogram we see nothing other than the twin tones with the exception of some very, very faint lines in the upper frequencies. So this isn't quite fully clear, but on the edge of being clear to 100 db.

Now the same conditions only with a range to the spectrogram of 120 db. We see some sum and difference distortion showing up each side of the twin tones. We also see the spurious tones just above 16 khz. So those apparently have nothing to do with the signal. We also see faintly some 3rd harmonic distortion which are the faint blue lines that are steeper, but follow the sweep tones upward. This DAC is not clear to 120 db.

So would such categories be helpful?

Are the spectrograms helpful or confusing?

Any reason to pursue this further or just a bad idea? Please give your comments.

 

[Blumlein 88]

Here is an example of a low quality DAC. This is the DAC built into my HDMI switcher which cost $40 and does video switching as well as audio decoding.

Here we see something that fails badly at being CLEAR at 100 db of range. This is a single tone sweep. We see harmonic distortion, noise, noise floor modulation and aliasing. In fact this DAC doesn't show clear until the range covered by the spectrogram is reduced to 65 db. Not high fidelity performance. On par with LP or cassette tape.

 

[Blumlein 88]

reserved

 

[Dimitri]

It's not a "bad idea", unfortunately ranking DACs on any one parameter doesn't tell the whole story.
Being able to sort them by parameter a first, the n b then c then d (aka parametric search - a'la analog devices) is going to be more interesting, and more work , where you can see that for a smal bump in any one measured performance the price difference is X,xxxx.xx additional dollars.
And no matter the objective parameters, there is no such thing as a listenability index
Now that would be something.

 

[Blumlein 88]

Well what I would hope we can agree upon is a 120 db clear DAC is transparent to the signal it is fed. If it doesn't max the listenability index, well listenability and fidelity are two different scales. I fear listenability scales become preference scales. And while there may be generally agreed upon preferences you enter a swamp chasing those too far. Once you have transparency, you can season to taste with control.

I believe even 100 db clear DACs will very rarely, and only under peculiar conditions be less than transparent.

PS-I dig the avatar Dimitri.