Excuse me... Do you have the correct time? Do you care?

[solderdude]

A 440Hz tone would become 440.002Hz (assuming the ADC had proper timing)
Try to find a vinyl or tape system that good !

 

[voodooless]

A 440Hz tone would become 440.002Hz (assuming the ADC had proper timing)
Try to find a vinyl or tape system that good !

Actually, it would be about 439.9795 Hz. The clock is presumed to be too slow

 

[tengiz]

In this application it means over engineering. Not nicer engineering.

And so is 32 bit DAC vs 24 bit DAC and even vs 16 bit DAC. For consumers use case there is no practical difference.

 

[NTK]

And so is 32 bit DAC vs 24 bit DAC and even vs 16 bit DAC. For consumers use case there is no practical difference.

How much cheaper is a 16 bit audio DAC versus a 24 bit one?

 

[Sokel]

In this application it means over engineering. Not nicer engineering.

Over engineering would be an ever better clock if the previous one didn't miss a beat or missed some over days.
One that misses but it's inaudible is the bare adequate, cost effective probably.

"Bare adequate" and "hobby" rarely mix, specially at the engineering side of things.

 

[ssashton]

Over engineering would be an ever better clock if the previous one didn't miss a beat or missed some over days.
One that misses but it's inaudible is the bare adequate, cost effective probably.

"Bare adequate" and "hobby" rarely mix, specially at the engineering side of things.

Well, let's be specific, how good is good enough not to be worth improving it? There is always the possibility of better.

 

[danadam]

Out of curiosity, I used 10 minutes files consisting of 10-bit dither and ending with 20 bandlimited pulses every 1 ms.

Here are a few DACs playing at 48 kHz rate, captured with E1DA ADCiso at 96 kHz:


DACs playing at 44.1 kHz, captured at 96 kHz:


And for some estimation of repeatability and ADC clock differences, Apple dongle playing at 48 kHz and captured at 96 kHz, twice with E1DA ADCiso and twice with Adi-2 Pro:

 

[tengiz]

How much cheaper is a 16 bit audio DAC versus a 24 bit one?

I wouldn’t know - I’m just a consumer. But spec sheets are a riot: 32-bit DACs promise one-in-a-billion precision, while the clock feeding them drifts at 50 ppm - one-in-twenty-thousand. The mismatch is huge, but who cares? 16 bits and 50 ppm are already more than enough. Shiny bit-depth numbers sell, boring clock specs don’t.

 

[NTK]

I wouldn’t know - I’m just a consumer. But spec sheets are a riot: 32-bit DACs promise one-in-a-billion precision, while the clock feeding them drifts at 50 ppm - one-in-twenty-thousand. The mismatch is huge, but who cares? 16 bits and 50 ppm are already more than enough. Shiny bit-depth numbers sell, boring clock specs don’t.

With today's semiconductor manufacturing technology, it doesn't cost any more money to make a 24 bit DAC that performs better than the max performance achievable by a 16 bit DAC than to make a 16 bit DAC.

Note that a 32 bit DAC doesn't mean it has 32 bit signal resolution. It only means when you communicate with the DAC or the DAC chip, you (can) use 32 bits for each sample. (Most 32 bit DACs will also support 16, 24 bit modes.) AFAIK, no current audio DAC has an effective number of bits (ENOB) approaching 24 bits. However, today's entry level DAC chips easily have ENOB greater than 16, so there is no practical disadvantage in going to 24 bit.

Also note that not all errors are the same. For clock errors for audio, the absolute frequency accuracy is far less critical than the consistency in the timing between clock pulses, which is jitter.

To give an analogy, I'll compare the clock frequency accuracy of the DAC to its gain accuracy. How much does it matter if there is a 1% difference in the digital full scale output voltage between 2 different DACs (or even between channels of the same DAC), e.g. 2.00 V vs 1.98 V ? However, if the half scale output of a DAC (same channel) is off by 1% relative to its full scale, e.g. half scale = 0.495x full scale instead of 0.5x, this DAC is very broken.

 

[tengiz]

...

I don't disagree at all with the technical side of it. Still, the "better" DAC did have better clock accuracy.

 

[NTK]

I don't disagree at all with the technical side of it. Still, the "better" DAC did have better clock accuracy.

Sure. But you'll have to pay for the "better" oscillator, unlike 16 bit vs 24 bit DAC.

 

[voodooless]

Still, the "better" DAC did have better clock accuracy.

We don’t know that, because the clock accuracy of the ADC is unknown.

 

[tengiz]

We don’t know that, because the clock accuracy of the ADC is unknown.

Ok, use the proper term instead of "accuracy" in this case. I did not mean jitter. I meant when it’s used to measure time by counting at least tens of thousands of cycles, not cycle-to-cycle variations.

 

[voodooless]

Ok, use the proper term instead of "accuracy" in this case. I did not mean jitter. I meant when it’s used to measure time by counting at least tens of thousands of cycles, not cycle-to-cycle variations.

That doesn’t matter…

 

[tengiz]

That doesn’t matter…

So, what is the correct tribal term?

 

[voodooless]

So, what is the correct tribal term?

"Accuracy" seems fine to me. "Jitter", we call the other thing, which you can also call "precision":

 

[ssashton]

Ok, use the proper term instead of "accuracy" in this case. I did not mean jitter. I meant when it’s used to measure time by counting at least tens of thousands of cycles, not cycle-to-cycle variations.

He just means we don't know if the ruler being used to measure things is at all accurate. If the ruler has not been checked against a trusted reference the results are meaningless.