DAC performance at very low frequency (1-20 Hz)

[maxwell3e10]

Hello,
I just stumbled onto this message board and was really amazed by the depth of the analysis. My interest in DACs is actually for scientific experiments to generate very clean signals and it seems that audio DACs are well ahead of other test equipment in this frequency range. But I am curious about lower frequency range, less than 20 Hz. I know that some DACs are DC-coupled, but I haven't been able to find any FFTs that extend down to 1 Hz. Would anyone be willing to run a test on APx555 for some high-performance DC-coupled DACs with something like a 10 Hz test signal.

 

[pkane]

Hello,
I just stumbled onto this message board and was really amazed by the depth of the analysis. My interest in DACs is actually for scientific experiments to generate very clean signals and it seems that audio DACs are well ahead of other test equipment in this frequency range. But I am curious about lower frequency range, less than 20 Hz. I know that some DACs are DC-coupled, but I haven't been able to find any FFTs that extend down to 1 Hz. Would anyone be willing to run a test on APx555 for some high-performance DC-coupled DACs with something like a 10 Hz test signal.

Hi maxwell3e10,

What's important for you in these measurements? Frequency response? Linearity? Distortion? Noise? I'm not sure what a 10Hz test signal will tell you if you're looking down to 1Hz. I don't have an AP. My DAC and ADC are both in a simple pro-recording interface (Apogee Element 24). Here's the result down to 2Hz. If you really want to see down to 1Hz, I could do that too (even below 1Hz), but will need to use different software to capture and plot it.

 

[maxwell3e10]

Thanks, Paul. This is already encouraging in that there is not a lot of 1/f noise, at least down to 2 Hz. But the harmonics seem to be pretty high. Are they actually higher with test signal at 10 Hz than at usual 1 kHz?

Both the noise and harmonics are important, basically the goal is to reproduce a given waveform as accurately as possible. The frequency response, if its really DC coupled, should be sufficiently flat.

 

[pkane]

Thanks, Paul. This is already encouraging in that there is not a lot of 1/f noise, at least down to 2 Hz. But the harmonics seem to be pretty high. Are they actually higher with test signal at 10 Hz than at usual 1 kHz?

Both the noise and harmonics are important, basically the goal is to reproduce a given waveform as accurately as possible. The frequency response, if its really DC coupled, should be sufficiently flat.

It’s about 12dB better at 1kHz, which is not surprising for an audio device — our hearing is much, much more sensitive at 1kHz than it is below 20Hz.

But, this is also not a state of the art device, it’s fairly inexpensive, and is outperformed by many other devices, many measured here on ASR

 

[maxwell3e10]

Just wondering what is the physical origin of the higher non-linearity at lower frequencies. Perhaps it is due to heating of the chip components that can happen at 10 Hz? But most measurements of THD+N vs. frequency don't show increase at low frequency.

 

[Blumlein 88]

I wonder if you could run a 1000 hz square wave in an AC coupled DAC, and modulate the square wave to create precise sub-10hz frequencies by filtering the result.

 

[trl]

I usually test this with my PicoScope 2204A. It has a built-in generator and I compare 10 Hz square-waves from the generator vs. the DAC-output. With all of the DC-coupled DACs tested the 10 Hz was perfect or almost perfect. The DACs tested by me were: PLAY, PLAYMATE, Conductor V2+ and Conductor 3X, but I guess all DC-coupled DACs should measure about perfectly at 10 Hz.

Below is the PLAY at 10 Hz with IN and OUT signals superimposing:

For 1 kHz I get the same squares, without any volume difference and without phase shifting, without touching the volume on both input generator and output DAC, so the levels for 10 Hz were not manually adjusted just to match the squares (the volume matching was done at 1 kHz).

 

[JohnYang1997]

What exactly are you trying to accomplish and what's the exact specifications you need? And your budget.

 

[maxwell3e10]

Basically, what everyone else here is trying to do, find the absolute best performance possible, not really limited by the budget (a few $k is fine). So white noise level at or below -160dB and as little harmonic distortion as possible.

 

[trl]

white noise level at or below -160dB

I guess you are referring to the background noise of the DAC relative to full-scale? Or relative to 2V RMS?

 

[Joachim Herbert]

Nobody will design an audio dac that is optimized for frequencies below 20 Hz, because there is close to no market for this.

So if this important to you, just roll your own. You might start a kickstarter project.

 

[JohnYang1997]

Nobody will design an audio dac that is optimized for frequencies below 20 Hz, because there is close to no market for this.

So if this important to you, just roll your own. You might start a kickstarter project.

If he can buy one that happens to suit the purpose it will be extremely convenient isn't it?

 

[maxwell3e10]

Nobody will design an audio dac that is optimized for frequencies below 20 Hz, because there is close to no market for this.

So if this important to you, just roll your own. You might start a kickstarter project.

I said I can spend a few $k on it, not a few years There are dozens of companies making DACs with no cost spared, so even if they are not designed specifically for low frequency, there bound to be some that work quite well.

 

[Mnyb]

Is the RME ADI2 good enough ? It also have eq and an ADC to and a host of other practical features as an audio interface ? If you can generate test signals and record with the same device ?

In thier spec it goes <1hz -1dB at the higher sampling rates .

 

[JohnYang1997]

Take a look at D70s maybe. https://www.l7audiolab.com/f/measurements-of-topping-d70s-dac/
https://www.audiosciencereview.com/forum/index.php?threads/topping-d70s-mqa-review-dac.18728/
I can make some measurements for you since we have D70s.


Or X16 from Gustard.
https://www.audiosciencereview.com/...ds/gustard-x16-balanced-mqa-dac-review.17419/
I can't measure this for you because we don't have one.

 

[maxwell3e10]

I guess you are referring to the background noise of the DAC relative to full-scale? Or relative to 2V RMS?

It should be relative to full scale, doesn't matter exactly what it the maximum voltage. Actually white noise specification is ambiguous, since FFT bin width is not always kept constant. For a 22kHz spectrum with 32000 points, the bin width is roughly 1 Hz. But if one uses the same number of points for a narrower spectrum, the bin width would get smaller and the white noise level appears lower. Also the total noise RMS level would appear lower if the bandwidth is smaller. But what one cares about is spectral noise density, which should stay constant. So the best way to specify white noise would be something like -160dBFS/Hz.

 

[maxwell3e10]

Take a look at D70s maybe. https://www.l7audiolab.com/f/measurements-of-topping-d70s-dac/
https://www.audiosciencereview.com/forum/index.php?threads/topping-d70s-mqa-review-dac.18728/
I can make some measurements for you since we have D70s.
View attachment 102190

Or X16 from Gustard.
https://www.audiosciencereview.com/...ds/gustard-x16-balanced-mqa-dac-review.17419/
I can't measure this for you because we don't have one.

Thanks. This is getting interesting, there is no evidence of 1/f noise! And the 3rd harmonic (-120dB) is not much worse than for 1 kHz (-125dB). But with white noise level one has to be careful (see previous post). It looks like it is less than -160dB. But in the usual audio spectrum range (20 Hz to 20 kHz), the white noise level is about -150 dB for this DAC in 20-50 Hz range. I think the difference must be from the FFT bin width. So for fair comparison, one either needs to make the number of FFT points roughly equal to the spectrum width or specify the number of points. If the number of points here is 32000, then there is a correction by 10*log(sqrt(22000/50))=13dB

 

[JohnYang1997]

It should be relative to full scale, doesn't matter exactly what it the maximum voltage. Actually white noise specification is ambiguous, since FFT bin width is not always kept constant. For a 22kHz spectrum with 32000 points, the bin width is roughly 1 Hz. But if one uses the same number of points for a narrower spectrum, the bin width would get smaller and the white noise level appears lower. Also the total noise RMS level would appear lower if the bandwidth is smaller. But what one cares about is spectral noise density, which should stay constant. So the best way to specify white noise would be something like -160dBFS/Hz.

Specifying accumulated noise across certain frequency range in VRMS would be better. "Noise floor" is not really that useful.

 

[JohnYang1997]

I made a mistake the ap was set at 10hz bandwidth. Will remeasure later

 

[F1308]

I said I can spend a few $k on it, not a few years There are dozens of companies making DACs with no cost spared, so even if they are not designed specifically for low frequency, there bound to be some that work quite well.

Good idea...
So many loudspeakers with crossovers and no DACs with any.

We want a DAC with ten cores or more (crossovers in loudspeakers, cylinders in cars, thrusters in astronautics) : one for frequencies 1-20, another for 20-40, one more for 40-80...and so on till 50000 Hz.
We could sell it as The Octave DAC.
One core for every doubling the frequency...

Sure Topping gets the idea....and by Christmas 2021 we buy one...an easy waiting while enjoying D90.