DAC output amplitude stability

[solderdude]

For example, opening the top cover of the SMSL box makes amplitude fluctuations much worse due to air convection currents causing temperature fluctuations. Putting a finger on various components indicates that the ES9311Q is the most sensitive element.

I would suggest that opening the box lets in RF. Putting a finger on the chip works as an antenna. The reaction time suggest it is not heat at all.
If that were the case the drift would be slow downwards and slow upwards.

External clocks are for studio usage (word clocks) so various ADCs and DACs sample at the exact same time. The fact that some idiots demand it and claim sonic superiority is another matter.

0.0005dB level difference over 2 minutes time (at any time) is inaudible even when a constant tone is used.
In music the level changes constantly and filter errors alone (due to the way filters are constructed) there will be differences at any time between the described sample value and the actual output voltage that is bigger than 0.0005dB anyway.

 

[KSTR]

I've found the same, natural 1/f noise dominated, microscopic gain fluctuations in my gear, notably my two ADI-2 Pro's. When untreated, they can spoil the results of deep null tests.
In normal, even precision, audio application this is as completetly a non-issue as it ever gets. As are microscopic clock drifts.

 

[chris719]

Yeah, this is of no concern for audio, which is why the ESS regulator isn't designed with reference stability or 1/f in mind and why it's cheaper than a real precision reference, filter, and buffer combination.

 

[maxwell3e10]

I suppose it's the difference between a mid-priced DAC like SMSL and some of the fancier ones. For example, Okto uses LT3045 regulators. People spend a lot of money on all kinds of things, like cables, that don't even make any measurable difference at all. In this case, there is a clear measurable difference, even if not audible. It just hasn't caught marketer's attention. Otherwise, one could add a much better reference for not much money. If it catches people's attention, it's just a matter of time before some company will advertise a DAC with LTZ1000 reference. Will it make audible difference? No. Will it make DACs technically more perfect- yes.

 

[JohnYang1997]

I suppose it's the difference between a mid-priced DAC like SMSL and some of the fancier ones. For example, Okto uses LT3045 regulators. People spend a lot of money on all kinds of things, like cables, that don't even make any measurable difference at all. In this case, there is a clear measurable difference, even if not audible. It just hasn't caught marketer's attention. Otherwise, one could add a much better reference for not much money. If it catches people's attention, it's just a matter of time before some company will advertise a DAC with LTZ1000 reference. Will it make audible difference? No. Will it make DACs technically more perfect- yes.

It really doesn't take LT3045 to make improvement. 9311 isn't a great solution anyway.
But you need a lot of current for 9038pro. So it's not some trivial task if both cost and performance are in mind.

 

[maxwell3e10]

It is still on the order of a few dollars, unless one uses LTZ1000, which would take a couple hundred dollars to implement property.

 

[Katji]

So do any of the super-expensive DAC use it?

 

[maxwell3e10]

Not that I know. Like I said, it hasn't caught people's attention. There are just a few mentions on diyaudio, but I don't think anyone has implemented it.

 

[solderdude]

When/why would 0.0005dB level differences over a long time period catch people's attention when used for audio ?

 

[mansr]

I suppose it's the difference between a mid-priced DAC like SMSL and some of the fancier ones. For example, Okto uses LT3045 regulators.

With LT3045 the long-term stability depends on the external reference used to set the voltage. Typically a resistor, its temperature coefficient is likely the dominant factor. For better stability, some kind of temperature-compensated reference can be used. I have no idea if anyone does that in a DAC. I do know it would be pointless.

 

[preload]

We should also measure the reflectivity of the paint used on the chassis.

 

[Blumlein 88]

A change in barometric pressure of .18 inches of mercury over 3 hours is considered rising or falling. That amount is just about equivalent to .05 db SPL over a 3 hour time period. Obviously this new metric of voltage stability is an over-looked secret to great sound.

 

[Katji]

So this doesn't qualify for the [difference is the implementation] thing, but what does, though? Surely that's something that could be defined [somewhat]? (It could be a useful reference thread, given the ongoing stream of "new-to-DAC" questions.)

 

[KSTR]

When/why would 0.0005dB level differences over a long time period catch people's attention when used for audio ?

They won't. That is, unless you're a measurements guy and the level difference/drift shows up in subtractive analyses. I actually needed to find a solution for the problem and the only way I can have any impact there is to use heavy (100x) time-domain averaging, which also gives a desirable ~20dB reduction of any strongly non-correlated noise/error. And careful user control of environmental circumstances like temperature and airflow.

Therefore, professional DACs/ADCs also aimed at precision measurements should (and do) strive for low noise, low 1/f-noise especially, and low temperature drift references. Lowest wideband impedance, too.

 

[maxwell3e10]

So this doesn't qualify for the [difference is the implementation] thing, but what does, though? Surely that's something that could be defined [somewhat]? (It could be a useful reference thread, given the ongoing stream of "new-to-DAC" questions.)

This is actually easy to measure with a medium-grade AC voltmeter. Simply setup the DAC to output a continuous tone and measure the AC output voltage. It will drift at start-up and with temperature. Another effect I found is that when the signal is alternated between silence and constant high output. Every time the signal turns on, there is a small drift in amplitude. I believe that is due to drift in the DAC itself or the resistors used in I/V conversion.