• 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). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Measuring RME ADI-2 Pro with QuantAsylum QA401

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#41
During Jan's last visit, we used the APx525 to measure the Autoranger- and the measurement limitation was the AP. What dedicated measurement systems like the AP or dScope get you is measurement efficiency and versatility, not necessarily better performance.
That's very interesting! I know I have read about the autoranger in the past but forgotten about it.
Do you have any links to info about it?

The combo might not quite be able to match the Apx555 but still......
 
Last edited:

Miska

Active Member
Joined
Feb 20, 2019
Messages
137
Likes
120
#44
Aahh so it's inadequate to look at the inaudible stuff ;) as I said I'm not interested in the inaudible.
You sound like the people manufacturing NOS DACs for playing 44.1k content in NOS way...

For some strange reason all modern converter chips still sport digital filters. And I strangely suspect people designing the filters, modulators and the chips overall also look outside 20 kHz. I'm doing same stuff, without need/pressure to cram it on a $10 chip though.

Anyway, each to their own.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#45
You sound like the people manufacturing NOS DACs for playing 44.1k content in NOS way...

For some strange reason all modern converter chips still sport digital filters. And I strangely suspect people designing the filters, modulators and the chips overall also look outside 20 kHz. I'm doing same stuff, without need/pressure to cram it on a $10 chip though.

Anyway, each to their own.
Not in the slightest. However there is a point where things become fit for purpose. The same reason I don't use silver mains plugs Cooled to absolute zero. Copper at room temp is fit for purpose.

I also want see justification when I see remarks such as yours above that a spuria at - 105dB at 44 kHz or 88kHz at - 120dB implying it has an audible impact.

As I said, let's not disrupt this thread. If you want to create another thread where you can demonstrate audibility of such things then I will gladly contribute.
 
Last edited:

Miska

Active Member
Joined
Feb 20, 2019
Messages
137
Likes
120
#46
I also want see justification when I see remarks such as yours above that a spuria at - 105dB at 44 kHz or 88kHz at - 120dB implying it has an audible impact.
Or -50 dB at 352.8 kHz? You have not yet explained at which point it matters and which not? Why the performance is not allowed to be "too good" at no extra cost?

So this is also perfectly fine, NOS R2R DAC running at 44.1 kHz with transformer on the output as reconstruction filter?

Metrum-Musette-1k-441-graph.png


As I said, let's not disrupt this thread. If you want to create another thread where you can demonstrate audibility of such things then I will gladly contribute.
You are the one who wants to discuss audibility limit numbers. I'm looking at this from technical correctness point of view.

To stay on topic, you can also disable digital filters from the ADI-2 DAC side, if you think that OOB images don't matter.
 

Miska

Active Member
Joined
Feb 20, 2019
Messages
137
Likes
120
#47
By the way, for measuring with ADI-2 you may also want to keep in mind that the ADC chip's anti-alias filter has 85 dB attenuation...
 

QAMatt

Member
Manufacturer
Joined
Mar 1, 2019
Messages
18
Likes
45
#48
Thanks Matt

I think the thd + noise plot sort of masks the issue.

Try plots of something like 1kHz at - 60dB, - 20dB and 0dB. Maybe 32 or 64 k fft. Observe the rise in the noise floor. My tests showed it to be significant.

Don't get me wrong, value for money is great, but this issue was a deal breaker for me I'm afraid.

It depends on usage, it could be fine for others requirements
Hi March, OK, understood. Take a look at the video below. It's stepping through levels, 5 dB increase each time, up to -1 dBFS. Also, keep your eye on the N+D in the upper left. That measures everything outside the fundamental in the indicated bandwidth (20 to 20 KHz). So you should be able to sanity check the readings at -61, -21 and -1 dBFS (close to your requests)

As noted, there's is a few dB of rise to the noise floor, but it's pretty much in line with what AKM shows for the part. And anyone should be able to replicate this on their QA401 too.

PS. You can read THD+N from any point off that graph that pops up at the end. For example, for -5 dBV = -11 dBFS output you can see the N+D @ -106.5 dBV which is -112.5 dBFS. That gives ratio THD+N of (-11 - -112.5) => THD+N = -101 dB


https://www.screencast.com/t/zWSUK7mA0oLK
 

QAMatt

Member
Manufacturer
Joined
Mar 1, 2019
Messages
18
Likes
45
#49
https://www.audiosciencereview.com/forum/index.php?threads/audio-measurement-gear.113/post-62396

https://www.audiosciencereview.com/forum/index.php?threads/audio-measurement-gear.113/post-62413

You can look here and see details showing the noise floor modulation at higher signals. Especially above -10 db. The THD+N also shows a rise at those levels.
Hi Blumlein, I don't have a Meridan to try, but looking at the plot marked "-6 dB at 1 KHz" I can try to recreate it in loopback

1) Picture of setup looked single ended, so that was used
2) Adjust output of QA401 to achieve same measured input (979mVrms versus 984mVrms)
3) Adjust relative to indicate the peak at -6 dBr
4) Offset freq slightly to give same skirt witdh
4) Apply 5 average (original measurement used 50)

In QA401 loopback, N+D measures -101 dBr, THD measures -101 dB, and THD+N = -94.6 dB.

In the Meridian measurement at the same level, N+D measures -92, THD measures -95, and THD+N = -86.1.

What conclusions do you draw from this?

BTW, at -60, the N+D measures -115 dBr, which is 15 dB worse that at -6 dBr, but that is in line with the plot shown earlier which is also in line with AKM plots.
 

Attachments

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#50
Hi March, OK, understood. Take a look at the video below. It's stepping through levels, 5 dB increase each time, up to -1 dBFS. Also, keep your eye on the N+D in the upper left. That measures everything outside the fundamental in the indicated bandwidth (20 to 20 KHz). So you should be able to sanity check the readings at -61, -21 and -1 dBFS (close to your requests)

As noted, there's is a few dB of rise to the noise floor, but it's pretty much in line with what AKM shows for the part. And anyone should be able to replicate this on their QA401 too.

PS. You can read THD+N from any point off that graph that pops up at the end. For example, for -5 dBV = -11 dBFS output you can see the N+D @ -106.5 dBV which is -112.5 dBFS. That gives ratio THD+N of (-11 - -112.5) => THD+N = -101 dB


https://www.screencast.com/t/zWSUK7mA0oLK
Hi Matt

Thanks again but the video doesn't match the behaviour I have seen and as @Earfonia has. It s very interesting and it would be great to get to the bottom of the issue.

I tried many different sources with my 401 and they all exhibited a 20 to 30 dB rise in noise floor with min to max input signal level when measured on the 401. Ignore THD and THD + N for the moment, the addition of distortion into the mix just obscures the issue.

If you look at the Earphonia measurements
https://www.audiosciencereview.com/...2-pro-with-quantasylum-qa401.6877/post-152943
you see with input dropping from 0dBV to - 6dBV to - 12dBV sees the noise floor drop from around (32k fft) - 122dBV to - 130dB to - 138dBV respectively.

So that's an noise floor increase of about 16dB with signal increase of 12dB.

This is exactly the sort of thing I would see. We can be confident this isn't the ADI as measurements with other devices don't show it.

Have their been any circuit revisions in newer units that might account for this?
 
Last edited:

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
6,352
Likes
5,941
#51
Is it possible the QA ADC is sensitive to noise above the nyquist frequency at which it is working? I've at times seen other ADCs that were. Noise above nyquist would elevate in band noise floors. Say at higher levels an analog output of the dut has some harmonics beyond nyquist or sigma delta ultrasonic noise and this raises the noise floor below nyquist of the QA. If the QA doesn't have the noise floor modulated in loopback and does in testing external devices then what is different to cause this issue?
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#52
Is it possible the QA ADC is sensitive to noise above the nyquist frequency at which it is working? I've at times seen other ADCs that were. Noise above nyquist would elevate in band noise floors. Say at higher levels an analog output of the dut has some harmonics beyond nyquist or sigma delta ultrasonic noise and this raises the noise floor below nyquist of the QA. If the QA doesn't have the noise floor modulated in loopback and does in testing external devices then what is different to cause this issue?
Unfortunately I dont have the unit anymore to check, but my recollection is it still had the problem in loopback.
 

Miska

Active Member
Joined
Feb 20, 2019
Messages
137
Likes
120
#53
Is it possible the QA ADC is sensitive to noise above the nyquist frequency at which it is working? I've at times seen other ADCs that were. Noise above nyquist would elevate in band noise floors. Say at higher levels an analog output of the dut has some harmonics beyond nyquist or sigma delta ultrasonic noise and this raises the noise floor below nyquist of the QA. If the QA doesn't have the noise floor modulated in loopback and does in testing external devices then what is different to cause this issue?
Yes, exactly. This is what I've been trying to explain. Since DACs commonly have images around for example 352.8 kHz (when running at 44.1k), and ADCs created for music recording don't necessarily have very strong anti-alias filters, you may get aliasing back to audio band.

If you measure something like ADI-2 through loopback, the DAC and ADC are running at same rate and thus images and aliases neatly stack up so it is harder to detect. When you run the two at different rates (DAC at 44.1k) and ADC at 48k-base rate then you begin to see also aliases more clearly.

Aliasing of random noise of course doesn't stack up, so that creates elevated noise floor.
 

restorer-john

Major Contributor
Joined
Mar 1, 2018
Messages
2,255
Likes
3,479
Location
Gold Coast, Queensland, Australia
#54
If the QA doesn't have the noise floor modulated in loopback and does in testing external devices then what is different to cause this issue?
Poor filtering in the 'external devices' I would expect. Maybe the QA needs the equivalent of an AES-17 filter to really get rid of the HF crap and limit its bandwidth for 'normal' measurements.
 

Miska

Active Member
Joined
Feb 20, 2019
Messages
137
Likes
120
#55
For tricky cases like class-D amps, where extra filtering may be needed, I have external analog filter:
http://www.prismsound.com/test_measure/products_subs/filter/products_lpf.php

With the dScope, when measuring DACs, such is not needed because the noise levels are not so high. But with class-D amps it can be different, also because reconstruction filters in class-D amps are much simpler.

I don't know how good digital AA-filter the QA has. But for example ADI-2 may need some extra help for some cases with it's 85 dB digital filter.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#56
Yes, exactly. This is what I've been trying to explain. Since DACs commonly have images around for example 352.8 kHz (when running at 44.1k), and ADCs created for music recording don't necessarily have very strong anti-alias filters, you may get aliasing back to audio band.

If you measure something like ADI-2 through loopback, the DAC and ADC are running at same rate and thus images and aliases neatly stack up so it is harder to detect. When you run the two at different rates (DAC at 44.1k) and ADC at 48k-base rate then you begin to see also aliases more clearly.

Aliasing of random noise of course doesn't stack up, so that creates elevated noise floor.
We understand this, you dont need to explain it ;), and we have discussed it elsewhere. Again its down to appropriate input filtering (whether its a speaker amp or ADC input buffer amp) as out of band noise can come from any source such as external RF. DACs are not the only source of OOB noise. OP amps can rectify and give additional noise and DC off sets. You cant predict what conditions an amp may encounter, so you have to deal with it by appropriate input filtering.

This issue of the significantly rising noise floor on the 401 I havent seen in any of the audio interfaces I have tested. Its of course possible the 401 may not have adequate filtering.

Anyone want to draw the input circuit :)

1552210293700.png
 
Last edited:

QAMatt

Member
Manufacturer
Joined
Mar 1, 2019
Messages
18
Likes
45
#57
Hi Matt

Thanks again but the video doesn't match the behaviour I have seen and as @Earfonia has. It s very interesting and it would be great to get to the bottom of the issue.

I tried many different sources with my 401 and they all exhibited a 20 to 30 dB rise in noise floor with min to max input signal level when measured on the 401. Ignore THD and THD + N for the moment, the addition of distortion into the mix just obscures the issue.

If you look at the Earphonia measurements
https://www.audiosciencereview.com/...2-pro-with-quantasylum-qa401.6877/post-152943
you see with input dropping from 0dBV to - 6dBV to - 12dBV sees the noise floor drop from around (32k fft) - 122dBV to - 130dB to - 138dBV respectively.

So that's an noise floor increase of about 16dB with signal increase of 12dB.

This is exactly the sort of thing I would see. We can be confident this isn't the ADI as measurements with other devices don't show it.

Have their been any circuit revisions in newer units that might account for this?
We have two knowns here: That a correctly functioning QA401 doesn't exhibit this, and that a correctly functioning RME ADI-2 doesn't exhibit this.

We also have two unknowns here: It's not known if the QA401 used in this test or the RME used in this test are functioning correctly nor do we know all the details of the test connection, the XLR construction, etc.

If someone has a QA401 that is showing the rise in loopback mode, I'd consider the unit defective and we'd repair it. It would be very easy to detect this in loopback. There could be an instance where application of high DC to the input could have degraded the input opamp??

@Earfonia. can you make some loopback measurements on your QA401? And then we can repeat on whatever DAC you have access to. Let me know if you can and I'll paste the test procedure.
 

QAMatt

Member
Manufacturer
Joined
Mar 1, 2019
Messages
18
Likes
45
#58
Is it possible the QA ADC is sensitive to noise above the nyquist frequency at which it is working? I've at times seen other ADCs that were. Noise above nyquist would elevate in band noise floors. Say at higher levels an analog output of the dut has some harmonics beyond nyquist or sigma delta ultrasonic noise and this raises the noise floor below nyquist of the QA. If the QA doesn't have the noise floor modulated in loopback and does in testing external devices then what is different to cause this issue?
Potentially, but it'd have to be really, really bad. If someone has a QA401 they can replicate this test pretty easily: Set the QA401 to 48Ksps, short L- and connect an external sig gen (quality not important) to the L+. Set the external siggen to 0 dBV and 49 KHz. You should see an 1 KHz aliased product show up at -95 dBV. And then repeat at 97 KHz. You should again see a product show up again at 1 KHz at around -105 dBV. And then try 193 KHz.

In a table:

49 KHz -> 95 dB of rejection
97 KHz -> 105 dB of rejection
193 KHz -> 105 dB of rejection

So, the noise level above >50 KHz or so would have be pretty bad to make a difference. But as we move through the eval, you are right, it's something to keep in mind.
 

March Audio

Major Contributor
Manufacturer
Patreon Donor
Joined
Mar 1, 2016
Messages
3,811
Likes
3,503
Location
Perth Western Australia
#59
We have two knowns here: That a correctly functioning QA401 doesn't exhibit this, and that a correctly functioning RME ADI-2 doesn't exhibit this.

We also have two unknowns here: It's not known if the QA401 used in this test or the RME used in this test are functioning correctly nor do we know all the details of the test connection, the XLR construction, etc.

If someone has a QA401 that is showing the rise in loopback mode, I'd consider the unit defective and we'd repair it. It would be very easy to detect this in loopback. There could be an instance where application of high DC to the input could have degraded the input opamp??

@Earfonia. can you make some loopback measurements on your QA401? And then we can repeat on whatever DAC you have access to. Let me know if you can and I'll paste the test procedure.
Hi Matt, absolutely, we need to work through the possibilities.

@Earfonia connections looked good, balanced split to the BNCs. My connections would have been same, or SE split.

My recollection is that the issue was there in loopback with straight bnc to bnc connection.

High DC offset doesn't seem likely as the issue was observed with multiple different DAC. sources

I'm afraid I can't help much as I don't have the unit anymore, but hopefully @Earfonia can assist
 

QAMatt

Member
Manufacturer
Joined
Mar 1, 2019
Messages
18
Likes
45
#60
High DC offset doesn't seem likely as the issue was observed with multiple different DAC. sources ...but hopefully @Earfonia can assist
Hi March, oops, sorry, I meant that if the input is exposed to a high DC value (eg trying to measure noise on 48V supply) it can degrade the performance while still working, seemingly normally. The guys building tube amps have +250V grids, and one user accidentally touched the QA401 input to +250V. The isolation worked as expected and the PC was fine, and the logic continued working. But the analog side of the QA401 couldn't be revived.
 
Top Bottom