RME ADI-2 FS Version 2 DAC and Headphone Amp Review

[KSTR]

Hi @MC_RME or @amirm

Can you do the same measurement as per the slide below, with your ADI-2 DAC ?

Rob Watts says his current DACs are the only ones in the world with no measurable noise floor modulation (below is his Hugo TT2 measured with his own APx555b)

Would be really nice to show him he is not correct

Hugo TT 2 by Chord Electronics - The Official Thread

I used to think DAC/amp that measures incredibly well would equal sterile and analytical sound, obviously not the case with Chord. Yes using traditional measurements there is truth in that observation. If you reduce 2nd harmonic distortion, and reduce higher order harmonics, things sound...

www.head-fi.org

While this plot really is remarkably clean, IME such tests must be done at various levels. Near full level alone might be too optimistic.
For example, I've found that the AKM AK4490 and AK4493 develop very(!) low level non-harmonic tones when a sine is played in a certain very limited level range. Many other D/S DAC chips are suspect to show this as well, including ESS (the hump, you know).

But first, let's see what an AK4493 in the RME ADI-2 Pro FS R does in loopback at -1dBFS, in an attempt of a comparison. I'm using the lowest sample rate, 44.1kHz and the largest FFT size, 4Meg points plus tons of averaging to reduce the noise floor in the FFT as much as any possible.
With 32 averages, the "1kHz@-1dBFS" plot looks like this:

We can see maybe a handful tiny little peaks just above -165dB on frequencies off the 1kHz grid, therefore non-harmonic. -160dB is already only a fraction of an 24 bit LSB and analog noise is another 20dB higher (at -120dBFS thereabouts) so we are looking at stuff at 1/100th of the RMS noise level, only visible by the extreme noise reduction from the FFT and samplerate setup.
The idle noise (not shown) is exactly the same BTW, is just the same noise without any frequency peaks.

The AK4493 get a bit weaker around -25dBFS with regard to modulation products... let's max out the averaging at 200x (took almost an hour to capture) and finally see something:

The effective FFT noise floor has now dropped to -175dB, exposing the train of artifacts nicely (but at a high cost).
But again it's all still below -150dB and therefore extremely deep down in the actual noise floor, I hardly can imagine any sonic degradation from that, even when considering a 20dB penalty for (alleged) dynamic worst-case errors with complex signals.

 

[Music1969]

While this plot really is remarkably clean, IME such tests must be done at various levels. Near full level alone might be too optimistic.
For example, I've found that the AKM AK4490 and AK4493 develop very(!) low level non-harmonic tones when a sine is played in a certain very limited level range. Many other D/S DAC chips are suspect to show this as well, including ESS (the hump, you know).

But first, let's see what an AK4493 in the RME ADI-2 Pro FS R does in loopback at -1dBFS, in an attempt of a comparison. I'm using the lowest sample rate, 44.1kHz and the largest FFT size, 4Meg points plus tons of averaging to reduce the noise floor in the FFT as much as any possible.
With 32 averages, the "1kHz@-1dBFS" plot looks like this:
View attachment 157162
We can see maybe a handful tiny little peaks just above -165dB on frequencies off the 1kHz grid, therefore non-harmonic. -160dB is already only a fraction of an 24 bit LSB and analog noise is another 20dB higher (at -120dBFS thereabouts) so we are looking at stuff at 1/100th of the RMS noise level, only visible by the extreme noise reduction from the FFT and samplerate setup.
The idle noise (not shown) is exactly the same BTW, is just the same noise without any frequency peaks.

The AK4493 get a bit weaker around -25dBFS with regard to modulation products... let's max out the averaging at 200x (took almost an hour to capture) and finally see something:
View attachment 157164
The effective FFT noise floor has now dropped to -175dB, exposing the train of artifacts nicely (but at a high cost).
But again it's all still below -150dB and therefore extremely deep down in the actual noise floor, I hardly can imagine any sonic degradation from that, even when considering a 20dB penalty for (alleged) dynamic worst-case errors with complex signals.

Thanks for doing this.

Rob makes a lot of dubious claims about audiblility but we won't go there in this thread.

I simply wanted to know how ADI-2 goes against his technical performance claim that his Chord DACs are the only in the world that don't have measurable noise floor modulation.

I wonder if loopback is the cause of the noise floor modulation you are getting here?

Would be great for @amirm or @MC_RME to confirm with APx555 measurement of ADI-2 DAC section standalone.

Here is the same claim, this time for the more affordable Chord Qutest which amir has measured too:

Chord Electronics Qutest DAC - Official Thread

Will the Qutest handle DSD native on a mac too? (as oppose to Mojo's DoP)

www.head-fi.org

 

[panther]

Does anyone know what the warranty period on this in Canada would be. All I can see in the manual is that it is 6 months or what your distributor tells you.

 

[Rja4000]

Thanks for doing this.

Rob makes a lot of dubious claims about audiblility but we won't go there in this thread.

I simply wanted to know how ADI-2 goes against his technical performance claim that his Chord DACs are the only in the world that don't have measurable noise floor modulation.

I wonder if loopback is the cause of the noise floor modulation you are getting here?

Would be great for @amirm or @MC_RME to confirm with APx555 measurement of ADI-2 DAC section standalone.

Here is the same claim, this time for the more affordable Chord Qutest which amir has measured too:

Chord Electronics Qutest DAC - Official Thread

Will the Qutest handle DSD native on a mac too? (as oppose to Mojo's DoP)

www.head-fi.org

I suggest you read this comment from Amir:

Important note: you may be tempted to read the DAC noise level from the noisy part of the graph/floor in the FFT. DO NOT DO THIS! The true noise floor of a DAC is much higher. When we perform FFT, the more points we use, the lower the noise floor becomes. This is called “FFT gain” and is a good thing as it allows us to see distortion products that may be well buried inside the noise floor of the DAC (and ADC in the measurement gear). The gain can be as much as 30 dB to 40 dB lower than reality! Math can be used to back out this gain using the number of points.

(From his explanation about FFT measurements)

You can, basically, set the APPARENT noise level on a FFT graph to the level you want.
The "FFT gain" can be computed by 10*Log10(<FFT window size>), as Amir explains here.

In Amir's review of the Chord Qutest, he measured noise at 117dB below max level.
SINAD is 114dB.
Those are respectable figures, but no miracle.
In line with the RME ADI-2 DAC equivalent measures of 119dB and 115dB.

 

[Music1969]

I suggest you read this comment from Amir:

(From his explanation about FFT measurements)

You can, basically, set the APPARENT noise level on a FFT graph to the level you want.
The "FFT gain" can be computed by 10*Log10(<FFT window size>), as Amir explains here.

In Amir's review of the Chord Qutest, he measured noise at 117dB below max level.
SINAD is 114dB.
Those are respectable figures, but no miracle.
In line with the RME ADI-2 DAC equivalent measures of 119dB and 115dB.

Noted but I believe Rob's comment is about the modulation of noise floor (between signal levels), not just a noise floor at one signal level like Amir's measurements?

Like KSTR demonstrated above for ADI-2 Pro (in loopback).

Rob's claim is about noise floor modulation specifically.

 

[KSTR]

I wonder if loopback is the cause of the noise floor modulation you are getting here?

Most likely not. The ADC was checked with an independent analog generator as well as with a notched version of the 1kHz output of the DAC and the tones remained the same in the latter case whereas they were absent with the analog generator. So I'm confident we see the contribution of the AK4493 pretty well (except for the higher/different distortion components when the ADC is seeing the the full 1kHz swing).

But I may repeat the bottom line of this: These artifacts are so low, so deeply buried in the normal noise floor that it requires specialized techniques to make them visible at all. A clean graph here might look nice and is ideal for marketing a possible "technical superiority" but that's it. These artifacts are most certainly inaudible in any kind of controlled listening test.

 

[pkane]

While this plot really is remarkably clean, IME such tests must be done at various levels. Near full level alone might be too optimistic.
For example, I've found that the AKM AK4490 and AK4493 develop very(!) low level non-harmonic tones when a sine is played in a certain very limited level range. Many other D/S DAC chips are suspect to show this as well, including ESS (the hump, you know).

But first, let's see what an AK4493 in the RME ADI-2 Pro FS R does in loopback at -1dBFS, in an attempt of a comparison. I'm using the lowest sample rate, 44.1kHz and the largest FFT size, 4Meg points plus tons of averaging to reduce the noise floor in the FFT as much as any possible.
With 32 averages, the "1kHz@-1dBFS" plot looks like this:
View attachment 157162
We can see maybe a handful tiny little peaks just above -165dB on frequencies off the 1kHz grid, therefore non-harmonic. -160dB is already only a fraction of an 24 bit LSB and analog noise is another 20dB higher (at -120dBFS thereabouts) so we are looking at stuff at 1/100th of the RMS noise level, only visible by the extreme noise reduction from the FFT and samplerate setup.
The idle noise (not shown) is exactly the same BTW, is just the same noise without any frequency peaks.

The AK4493 get a bit weaker around -25dBFS with regard to modulation products... let's max out the averaging at 200x (took almost an hour to capture) and finally see something:
View attachment 157164
The effective FFT noise floor has now dropped to -175dB, exposing the train of artifacts nicely (but at a high cost).
But again it's all still below -150dB and therefore extremely deep down in the actual noise floor, I hardly can imagine any sonic degradation from that, even when considering a 20dB penalty for (alleged) dynamic worst-case errors with complex signals.

ADI-2 Pro FS (AK4490) in loopback mode + some experimental software

 

[KSTR]

@pkane, ha, already with REW's distortion feature you can dial in counter harmonics to get a nice clean loopback plot with the first 10 harmonics at around -150dBFS ;-)

 

[MC_RME]

HpW could do this forever, but nobody ever used it. Here is something I did 2018. Analysis via a self built-notch filter (-60 dB @ 1 kHz) into the (old) Pro.

 

[pkane]

@pkane, ha, already with REW's distortion feature you can dial in counter harmonics to get a nice clean loopback plot with the first 10 harmonics at around -150dBFS ;-)

This is actual compensation, not a trick I can listen to music with it, not just a single tone

 

[KSTR]

@pkane,
Interesting... I seem to recall you mentioned this project a while back. It'll be worth its own thread when time is due, right?
Some starter questions, if I may:
- So you have a scheme that undoes/reduces a DAC's distortion (by "intelligent" predistorting?) even with arbitrary unknown signals? Say you could use a multitone and that is also lower distortion than without comp, same goes for single sines or two-tones at any frequency and level? That would be truly ground-breaking...
- How do you -- assuming you do so -- isolate and factor out the ADC distortion?

 

[bennetng]

@pkane,
Interesting... I seem to recall you mentioned this project a while back. It'll be worth its own thread when time is due, right?
Some starter questions, if I may:
- So you have a scheme that undoes/reduces a DAC's distortion (by "intelligent" predistorting?) even with arbitrary unknown signals? Say you could use a multitone and that is also lower distortion than without comp, same goes for single sines or two-tones at any frequency and level? That would be truly ground-breaking...
- How do you -- assuming you do so -- isolate and factor out the ADC distortion?

+1 for thread split when necessary.
How about something like this? Make a pink noise, poking holes here and there, and see how much noise are filled in the analog recording.

 

[pkane]

@pkane,
- So you have a scheme that undoes/reduces a DAC's distortion (by "intelligent" predistorting?) even with arbitrary unknown signals?

Correct!

- How do you -- assuming you do so -- isolate and factor out the ADC distortion?

I don't currently, instead correcting the DAC->ADC loopback in total. Planning on using a very low distortion ADC to correct DACs and amps in the future.

My last experiment was to use the IR derived from complex music signals to correct for frequency and phase errors. That worked well, the RMS null error with ADI-2 loopback went from -53dBFS to around -80dBFS with such a correction (requires a convolver). This new experiment is complimentary to that, and can be combined with IR correction, if desired.

 

[Matias]

That is why MC used a notch filter on the ADC input, so that the ADC harmonics are not added to the DAC harmonics.

 

[Paul_d]

Sorry for stupid question, but how I can see that device really is Version 2, but not first ?

 

[pkane]

That is why MC used a notch filter on the ADC input, so that the ADC harmonics are not added to the DAC harmonics.

That doesn't help what I'm doing, as I can't apply a notch filter that will remove only ADC harmonics while letting DAC harmonics go through to be captured by the ADC.

 

[Rja4000]

Sorry for stupid question, but how I can see that device really is Version 2, but not first ?

See this post from Matthias

More details about serial number here.

 

[Jimbob54]

Sorry for stupid question, but how I can see that device really is Version 2, but not first ?

The remote is bigger, more controls. V2 pic attached

 

[carlymart]

Looking for a little advise. Would the rme adi2 fs benefit for, something like the topping d10s usb converter or would it be a waist of money ?

 

[Jimbob54]

Looking for a little advise. Would the rme adi2 fs benefit for, something like the topping d10s usb converter or would it be a waist of money ?

Would the rme benefit from having a d10s feeding it? No. Not unless your source usb device was incredibly noisy, in which case your best bet is get a better source.