• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DAC Noise Modulation: Chord DAVE vs Topping DX7 Pro+

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
39,430
Likes
179,341
Location
Seattle Area
In one of the Chord TT2 review thread, I was asked to comment on noise modulation claims by the product designer Rob Watts. I attempted to recreate his measurements and compare them to Topping DX7 Pro+ which I just reviewed.
Chord Dave Noise Modulation vs Topping DX7 Pro+ Review Study.jpg


DAC Noise Modulation
Noise modulation refers to noise floor of the DAC changing with signal. To the extent the signal changes, if the noise floor changes with it, it is said to be "modulated." An ideal DAC would keep its noise floor constant as one has (ideally) nothing to do with the other. In reality activities of the DAC can manifest itself as extra noise, raising that component of the signal.

Chord DAC Claimed Lack of Noise Modulation
Rob Watts claims his DAVE DAC to be completely free of noise modulation and being the only DAC to be so. He backs it with the following measurement:

Chord Dave Claimed Noise Modulation Measurement.png


Notations in red are mine. This is the text below the graph:
Chord Dave Claimed Noise Modulation Measurement interpretation.png


Before getting into the details, the graph shows two overlaid measurements. One is when the DAC is producing a 2.5 volt signal, and the other, when it is producing nothing. From the graph it appears that the noise floor is the same in both measurements, backing what he says.

I wanted to replicate his measurements so that I could test other DACs to see how well they do. As I note above in red, this is made difficult by lack of documentation in the above measurements. The noise floor that you see is the result of both DAC noise and how much FFT has reduced it (called "FFT Gain"). I can make that noise floor as low as I want as long as I keep increasing the number of FFT points. As a result, the comments Rob makes about how low the noise floor is, i.e. -180 dB, is useless. That is NOT the actual noise floor of the DAC. So that claim is wrong although in the context of this comparison, it can be ignored. That is, both the no signal and 2.5 signal measurement are subject to the same FFT gain.

For some odd reason, the output voltage is picked to be 2.5 volt for a DAC that can go up to 6 volts. I suspect this may have been picked because it shows least distortion. So on that front, the claim of low distortion is also misleading especially since 0 dB is NOT set to 2.5 volt. Instead, it is set to 6 volt. In reality then, distortion is NOT -150 dB but something close to -142 dB. But again, in the context of comparing noise floors, we can ignore this.

Noise Modulation Comparison
I took the above graph and applied it to measurements of Topping DX7 Pro+. I played with the FFT samples and measurement bandwidth until I got something similar to DAVE DAC. Something still bothered me though. The measurements I performed of the DAVE DAC did not produce such a clean output. So I stepped back and ran the test against DAVE DAC which I happen to still have (owner is on long vacation). Here are the results for DAVE:
Chord Dave Noise Modulation Measurement.png


We see very different results. Distortion products are much higher (in relative terms) and so is the noise floor. What is more, there is noise modulation although in reverse. Noise floor actually goes up instead of down with no signal! Strange. To measure how much it is changing, I first compensated for FFT gain of 48 dB and then smoothed the two graphs:
Chord Dave Noise Modulation Zoomed Measurement.png


Smoothing screws up the 1 kHz tone so ignore that as all we care about is the noise floor differential which is 2.8 dB. It is changing from -114 dB to -111 dB. This is barely above best case threshold of hearing. So in an extreme case of 1 kHz tone NOT being audible, and with suitable amount of amplification, one may be able to hear that modulation.

Now let's run the exact same test but simply moving cables from DAVE DAC to Topping DX7 Pro+:
Chord Dave Noise Modulation vs Topping DX7 Pro+ Measurement.png


We immediately see confirmation of my reviews of both products: Topping DX7 Pro+ despite costing 20 times less, has much lower noise floor. It seems to have more distortion spikes but that is because the noise floor is so low, allowing them to peak through. In absolute levels, it is still superior to DAVE DAC by 6 dB.

It does show noise floor modulation and this time, as expected noise floor goes up with signal. Smoothing and compensating for FFT gain we get:

Topping DX7 Pro+ Noise Modulation Zoomed Measurement.png


The average noise modulation is 4.2 dB which is just slightly more than DAVE DAC. However, in this case, the modulation is occurring at -130 dB to -127 dB. With threshold of hearing at -115 dB, no way this is remotely audible no matter what contrived test we create for it. It is a completely non-issue and the reason I don't measure it.

Note: these tests are pushing limits of physics and instrumentation. Distortion measurements at -150 dB and lower is just crazy! It is unknown how accurate the AP is in this regard. Ditto for its own noise floor/modulation. When I first started to test the DAVE dac it had small train of pulses which disappeared after warming up. So some variation is to be expected in such tests.

Conclusions
This investigation of noise modulation shows that not only does the DAVE DAC perform worse than shown, but it also suffers from some noise modulation. While this noise modulation is slightly (0.5 dB) less than Topping DX7 Pro+, it happens at the threshold of hearing which may make it audible in pathological situations. Topping DX7 Pro+'s noise floor and modulation thereof is so far below audible threshold that it simply is not an audible concern in any contrived situation.

Maybe company's claim that it is the best there is was due to understanding of DAC performance years back (Rob Watts' post is from 2015 although renewed in 2021). This is certainly not true today where a $699 Topping DX7 Pro+ easily outperforms the DAVE DAC on both distortion and noise performance.

Company needs to provide comparative measurements to other current DACs before continuing to make such objective performance claims. And certainly not push points that are not the strength of its DAC, i.e. noise performance.

Bottom line, noise modulation is not a performance metric to worry about in well implemented DACs.

------------
As always, questions, comments, recommendations, etc. are welcome.

It was bound to happen: the panthers have unionized and now demanding raises through collective bargaining. They are threatening a walk-out leaving me with no panthers for product reviews. In other words, they have me by the neck. So, please donate generously to keep them modeling for us using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/
 

Phorize

Major Contributor
Forum Donor
Joined
Apr 26, 2019
Messages
1,349
Likes
1,779
Location
U.K
Thanks @amirm. This goes to show how important it is that manufacturer claims are empirically tested.
 

dartinbout

Active Member
Joined
Jan 15, 2020
Messages
211
Likes
208
Congrats! Once again, you have seemed to "Struck a Chord". /s
(I love multi-layer puns:cool:)
 

ZgSTar

Member
Joined
Oct 26, 2019
Messages
56
Likes
58
To be completely honest here, if I had the money and bought a Dave back in 2015 I would have been extremely happy. Not until the last couple of years we have seen the inexpensive competition getting to those levels of performance. A Dave today is not a good deal at all, but was it state of the art back then? Probably yes.
 

Rja4000

Major Contributor
Forum Donor
Joined
May 31, 2019
Messages
1,804
Likes
2,160
Location
Liège, Belgium
these tests are pushing limits
It's obviously much harder to provide relevant measurements at such low levels, below the standard limits of the instrument itself.
Every detail matters, and each result has to be challenged and checked carefully.

It's easier with the Chord though, since values are significantly higher.
 

voodooless

Master Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
5,666
Likes
9,316
Location
Netherlands
I played with the FFT samples and measurement bandwidth until I got something similar to DAVE DAC. Something still bothered me though. The measurements I performed of the DAVE DAC did not produce such a clean output. So I stepped back and ran the test against DAVE DAC which I happen to still have
Something doesn’t add up here ;)

Otherwise, great article as always. Would be nice to test a few more DAC’s and see if the Topping is special in any way (which I do not suspect it is).
 

srkbear

Addicted to Fun and Learning
Joined
Nov 16, 2021
Messages
775
Likes
1,007
Location
Dallas, TX
The review of my dreams! It’s like my Harold and the Purple Crayon Day—I wanted something just like this and poof! I’m deliriously happy :p
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
16,386
Likes
28,186
I've not seen too bad a noise floor modulation (NFM). Quite a few DACs show more in the octave above 20 khz. So you might want to expand the bandwidth you are looking at. Of course that wasn't how Chord was doing it either.

One device which had plenty was the ADC of the Quant Asylum Q401.

When you tested the Chord with no signal was it truly no signal? I wonder if it turns off with all zeroes and this is somehow causing the odd result. Maybe some shaped dither is turned off with zeroes fed into it.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
39,430
Likes
179,341
Location
Seattle Area
When you tested the Chord with no signal was it truly no signal? I wonder if it turns off with all zeroes and this is somehow causing the odd result. Maybe some shaped dither is turned off with zeroes fed into it.
It was with no signal which is what Rob stated.
 

voodooless

Master Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
5,666
Likes
9,316
Location
Netherlands
It was with no signal which is what Rob stated.
That can be interpreted in many ways of course. No signal doesn't really make sense anyway, because that's never how you would listen to it. So testing zero's vs (more-or-less) full-scale output would be a preferable and more realistic method. Many DACs will have special things in place in case of no output to prevent things like pops during the transition to playing sound. They may influence the noise floor.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
39,430
Likes
179,341
Location
Seattle Area
That can be interpreted in many ways of course. No signal doesn't really make sense anyway, because that's never how you would listen to it.
Well, it is what he stated. If you look at his no signal graph, you see that there is nothing at 1 kHz so there was nothing there. Even at the lowest -144 dBFS of a 24-bit signal, you would have had a tall spike, going well past the noise floor of -180 dB. So there is no ambiguity on what he meant there.

Whether this is a valid test is another topic. There are methods for example to examine the spectrum of noise floor to see if it changes with different signal values. That is said to be more important than whether it moves a bit as a whole or not.
 

JSmith

Major Contributor
Joined
Feb 8, 2021
Messages
3,233
Likes
7,355
Location
Algol Perseus
As a result, the comments Rob makes about how low the noise floor is, i.e. -180 dB, is useless. That is NOT the actual noise floor of the DAC.
RW caught being deceptive and misleading again on Chord's behalf.

Nice review... and solidly clears up some points perpetuated by RW and followers.

@TabCam & @PcChip ... what say you now? ;)


JSmith
 

voodooless

Master Contributor
Forum Donor
Joined
Jun 16, 2020
Messages
5,666
Likes
9,316
Location
Netherlands
Well, it is what he stated. If you look at his no signal graph, you see that there is nothing at 1 kHz so there was nothing there. Even at the lowest -144 dBFS of a 24-bit signal, you would have had a tall spike, going well past the noise floor of -180 dB. So there is no ambiguity on what he meant there.
Yeah, but what does that mean, "no signal"? Does that mean the DAC is not fed any data, or does it mean that it was fed with zeros?
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
39,430
Likes
179,341
Location
Seattle Area
Yeah, but what does that mean, "no signal"? Does that mean the DAC is not fed any data, or does it mean that it was fed with zeros?
As I explained, it means the cable is connected but the DAC is not asked to output anything. So you are seeing the idle noise of the DAC compared to when it is producing near max signal. Any valid PCM value would have shown a spike in the "no signal" case at 1 kHz and we see none of that.
 

Astoneroad

Senior Member
Forum Donor
Joined
Feb 16, 2022
Messages
486
Likes
912
Location
Cave in the desert
Congrats! Once again, you have seemed to "Struck a Chord". /s
(I love multi-layer puns:cool:)
Thanks again Amir for answering John Lennon's call to "Gimme Some Truth". I can just feel the "dischord" rising from the noise floor as the "For Sale" ads hit all the usual venues. When I finished reading your review, I was reminded of Carl Sagan's "Pale Blue Dot" warning... "Our planet is a lonely speck in the great enveloping cosmic dark. In our obscurity, in all this vastness, there is no hint that help will come from elsewhere to save us from ourselves." So Amir, you are amongst the intrepid skeptics of Lennon & Sagan... of "sound & fury"... but unlike MacBeth's warning... the truth does signify something... however fleeting. Thanks again for your pitbull like pursuit of objective facts.
 
Top Bottom