• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Measuring DACs

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
203
Likes
497
Location
Stockholm, Sweden
Hi!
I currently have a QuantAsylum QA-401 Audio Analyzer. I've been using it for testing my current project, a headphone amplifier. I've made a lot of equipment around it to be able to measure lower THD with a super low THD signal generator and a twin-T notch filter. But here's my problem. I want to measure DACs. On my list is to see if different power supplies have any effect on my newly bought Topping D50 DAC. I've seen some proper measurements from some of the people on this forum, like @March Audio doing it with the QA-401.

So here's my question. Because I can't use the QA-401 output, I need a software signal generator or downloadable files. Any tips on where and what files I need to characterise a DAC would be much appreciated. I've tried some generators and downloaded files, but get different results. What is needed is a small How-To for DAC measurements. Anyone up for writing a short guide?
 

gvl

Major Contributor
Joined
Mar 16, 2018
Messages
3,427
Likes
3,981
Location
SoCal
I glanced over the QA-401 info and I think there was an option to mirror the digital signal to the default audio device in Windows, that looks promising although I'd think the signal will go through the Windows mixer with all the consequences.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
March will be best to answer some of your questions as he has owned the QA-401.

He has taken a shine to this software:
http://www.virtins.com/multi-instrument.shtml

You can get a 21 day free trial. I'd suggest downloading and studying the manual carefully before starting the trial.

I didn't ever get it to work with more than one device in Windows to my satisfaction. You'll have to have two instances of it open to do it anyway. I had better luck using two computers running an ADC and DAC with Virtins on both ends. But maybe I just never figured it out well enough.

Now I can point you to making your own test files. Audacity or Adobe audition are capable of it. Or I can share some I've made for myself. Let me address those in another post.

You also can use RMAA for some basic stuff.
http://audio.rightmark.org/products/rmaa.shtml

I have a few quibbles about what it does, but its fine for some purposes. I think it would run with your QA. Worst case you can generate and playback a test file you record with the QA, and then the RMAA software will take that file recorded and do the analysis.
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
Alright for what I've put together for myself, I have a high level signal test file that runs a bit over 5 minutes. And a low level test file of a bit over 6 minutes.

Most of my high level signals are .1 db from clipping. I set output levels so the ADC is close to max, but no clipping.

I have a -4 db white noise test which I'll explain in a bit.
Then a 2hz to 20 khz sweep over nearly 60 seconds.
I then have a twin tone sweep with tones seperated by 1 khz which is an IMD sweep.
I have a 1 khz tone at -6 dbfs, -12 dbfs, and -18 dbfs. Lets me see both basic THD, and if the THD drops rapidly with the lower levels.
Then a -6 dbFS 18 khz + 19 khz tone also followed by one 6 db lower and 12 db lower in level. Lets me check IMD.
A 12 khz (which is a quarter sample rate of the 48 khz I use for most testing) at .1 dbFS. To check for jitter sidebands.
A 1 khz at -.1 dbFS signal and 18+19 khz at .1 dbFS level.
Finally a -60.0 dbFS 1 khz tone. I of course have a three seconds of silence between these, and do first right and then left offset in time so I can check crosstalk of all these.

Next is the low level test in the following post.
 
Last edited:
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
203
Likes
497
Location
Stockholm, Sweden
Alright for what I've put together for myself, I have a high level signal test file that runs a bit over 5 minutes. And a low level test file of a bit over 6 minutes.

Most of my high level signals are .1 db from clipping. I set output levels so the ADC is close to max, but no clipping.

I have a -4 db white noise test which I'll explain in a bit.
Then a 2hz to 20 khz sweep over nearly 60 seconds.
I then have a twin tone sweep with tones seperated by 2 khz which is an IMD sweep.
I have a 1 khz tone at .6 dbfs, -12 dbfs, and -18 dbfs. Lets me see both basic THD, and if the THD drops rapidly with the lower levels.
Then a -6 dbFS 18 khz + 19 khz tone also followed by one 6 db lower and 12 db lower in level. Lets me check IMD.
A 12 khz (which is a quarter sample rate of the 48 khz I use for most testing) at .1 dbFS. To check for jitter sidebands.
A 1 khz at -.1 dbFS signal and 18+19 khz at .1 dbFS level.
Finally a -60.0 dbFS 1 khz tone. I of course have a three seconds of silence between these, and do first right and then left offset in time so I can check crosstalk of all these.

Next is the low level test in the following post.

This information is great! All the info you can provide is very much appreciated. It's good if the measurements from us guys who don't have an AP are using the same tests. Makes it easier to compare. And that's why I love this forum. Just the facts ma'am!
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
The low level signal I set things up differently. I'll have the DAC at max output unless the prior test has shown distortion is too high. Then I may lower output a few decibels to whatever resulted in clean output. Quite a few DACs have the distortion grow a lot between max and -3 db levels.

If the ADC has gain, I'll turn it up as much as I can. The idea is this boost the noise floor of the DAC above the noise floor of the ADC so I can see what the real noise floor of the DAC is without it being swamped by the ADC noise. One of my ADCs has 20 db gain and that is usually enough, but another has over 40 db of gain. Now gain can and will add some noise, but in good devices that usually isn't more than 3 or 4 db. Still getting you closer to seeing the low level results of the DAC under test.

My low level test signal starts with 1 khz at precisely -60 dbFS. None of my low level signals are higher than this. Next I do a single tone and twin tone sweep with the signals at - 60 dbFS. Then I do my low level linearity tests which are a bit different than how Amir does them, but some other testing people do something similar. What I do would be DNL (differential linearity) and Amir does INL(integral linearity). I use a 12 khz tone at 60.21 dbFS and then follow dropping the level by 1 bit (-6.02 db) each until the 24th bit level is reached. The INL version is corrupted somewhat by noise and this one is somewhat immune. The noise is there, but most modern sigma delta based DACs put out near perfect linearity even if it is hidden in the noise of the device. I prefer to treat the inherent linearity and the noise separately.

So when I've recorded this test file and used some gain, the recorded levels are actually incorrect. If I used 40 db of gain the 1 khz reference tone will be at -20 dbFS in the recording. I look at whatever this reference tone is and adjust it in Audacity so it is -60.00 dbFS. I adjust the whole file this amount. So in the example if it were -20 dbFS I'd reduce the file level by 40 db and everything is where it should be.

So without quite yet getting into all the gritty details I can take these two files and using Audacity and/or WaveSpectra determine all of your normal specs and more. It isn't quite automated, but for a home hobbiest it isn't too bad. Plus I can simply store these recorded test files to look at later.
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
About the white noise I recorded at - 4 db.......it can be useful to run the high level file again, but let the recording ADC record results at a higher rate like running my file to the DAC at 48 khz and letting the ADC record at 192 khz. This lets you see the response of the anti-imaging filter of the DAC.

There are some other tricks for finding aliasing or imaging using spectrograms.

Here are a couple examples with explanations below.
Tact AP2 vs TCIT Toslink noise.png

In the above I have recorded 48 khz white noise at 192 khz. You get to see the shape of the imaging filter by how it drops off above 24 khz.
Jurgen sweep tact 48.png


Here I've the white noise in the blue line, and a high level sweep in red (caught at about 19 khz), and in green the base noise floor. You see the white noise raises the noise floor above 24 khz a bit vs the green noise floor of silence. You see the high level 19 khz tone leaks thru the anti-imaging filter some to show up at 29 khz. Note this mirrors around the nyquist frequency of 24 khz. 24- 19 is 5 khz. And 24 +5 is 29 khz where the mirror image shows up.

The blip in red at about 11 khz is an alias in the ADC I used for this. It is always there. It is some oddity in the digital filtering of this ADC. It mirrors around 30, 500 hz. 30.5-19=11.5 khz which is where the red blip is. It shouldn't be there, and in this case isn't in the DAC.

So I run the high level file and record. The high level file and record at 192 khz the second time. Then the low level file. I can go back and learn many things about the results after the fact.
 
Last edited:

Thomas savage

Grand Contributor
The Watchman
Forum Donor
Joined
Feb 24, 2016
Messages
10,260
Likes
16,298
Location
uk, taunton
It would be great to see more guys publishing measurements here and Absolutely brilliant if they were synchronised to a degree in terms of methodology and instrumentation.

A Great source of learning that will develop greater understanding within our community.
 
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
203
Likes
497
Location
Stockholm, Sweden
So I started to connect up my measurements using the Topping D50 DAC connected to my QuantAsylum QA-401 audio analyzer. But there's a problem. It seems to be a ground loop. Here's the measurement of the audio out from the D50 connected to the analyzer.

USB ground loop 1.png


These measurements are without any stimulus signal, just connected to check the noise floor. So there are high peaks at 750Hz, 1kHz and 1.5kHz. So obviously something is wrong. So I unplug the USB cable to the DAC, and this is the result.

USB ground loop no USB cable plugged in.png


The peaks vanishes and all the high-frequency noise as well. So I tried to change USB cables, with and without ferrite beads. But no changes. I cobbled together a low noise 5-volt power supply that can deliver 5 volts, 1 ampere with the noise of 4 µVrms. No difference. I've also tried to use both powered and unpowered USB hubs in between in different configurations, but all with the same results. If I leave the audio cable from the DAC connected to the meter, but disconnect the DAC USB cable, the noise disappears, and the noise floor goes down to a reasonable level. So it's definitely a ground loop. But I can't get my head around it. According to QuantAsylum, The PC interface is isolated from the analog IO. But here's the kicker, if I take the Topping D50, put the ground from the RCA output connectors directly to one of the BNC connectors on the audio analyzer, the problem disappears. And yes, I have tried with the computer both on mains and battery. No difference.

I will try to connect the DAC to my headphone amplifier and then measure the amp instead. It's my test mule, so it's been attached to the QA-401 many times, so I know how it should look.

Does Anyone get an idea how to fix the noise? I will take the D50 apart tomorrow, just to snoop around a bit. But something is strange.
 

gvl

Major Contributor
Joined
Mar 16, 2018
Messages
3,427
Likes
3,981
Location
SoCal
QA401 is isolated but I'll wager $1 D50 is not, so PC noise may find its way on the analog outputs. I don't get the part where you can eliminate the noise by connecting RCA ground to BNC ground, isn't RCA ground already connected to the QA401's BNC ground when the D50's outputs are connected to QA401 inputs with cables? Bad interconnects?
 
OP
jackenhack

jackenhack

Active Member
Forum Donor
Joined
Oct 25, 2018
Messages
203
Likes
497
Location
Stockholm, Sweden
QA401 is isolated but I'll wager $1 D50 is not, so PC noise may find its way on the analog outputs. I don't get the part where you can eliminate the noise by connecting RCA ground to BNC ground, isn't RCA ground already connected to the QA401's BNC ground when the D50's outputs are connected to QA401 inputs with cables? Bad interconnects?

Exactly my thought! I've tried four different cables. The ground is connected via the RCA connectors, but still that problem. I have tried with both of the RCA connectors, but the same problem. I can't make heads or tails of why. There should be sufficient grounding by one or two RCA cables! But it's something strange going on with the USB connector on the D50.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
Can't the D50 be powered by the wall wort or simply by USB cable? And have you tried it both ways on that?
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
So I started to connect up my measurements using the Topping D50 DAC connected to my QuantAsylum QA-401 audio analyzer. But there's a problem. It seems to be a ground loop. Here's the measurement of the audio out from the D50 connected to the analyzer.

View attachment 17393

These measurements are without any stimulus signal, just connected to check the noise floor. So there are high peaks at 750Hz, 1kHz and 1.5kHz. So obviously something is wrong. So I unplug the USB cable to the DAC, and this is the result.

View attachment 17394

The peaks vanishes and all the high-frequency noise as well. So I tried to change USB cables, with and without ferrite beads. But no changes. I cobbled together a low noise 5-volt power supply that can deliver 5 volts, 1 ampere with the noise of 4 µVrms. No difference. I've also tried to use both powered and unpowered USB hubs in between in different configurations, but all with the same results. If I leave the audio cable from the DAC connected to the meter, but disconnect the DAC USB cable, the noise disappears, and the noise floor goes down to a reasonable level. So it's definitely a ground loop. But I can't get my head around it. According to QuantAsylum, The PC interface is isolated from the analog IO. But here's the kicker, if I take the Topping D50, put the ground from the RCA output connectors directly to one of the BNC connectors on the audio analyzer, the problem disappears. And yes, I have tried with the computer both on mains and battery. No difference.

I will try to connect the DAC to my headphone amplifier and then measure the amp instead. It's my test mule, so it's been attached to the QA-401 many times, so I know how it should look.

Does Anyone get an idea how to fix the noise? I will take the D50 apart tomorrow, just to snoop around a bit. But something is strange.

Taking measurements using the same PC as connected to the DUT is problematic. If I do so I use an isolator such as the Intona which usually solves the problem.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,280
Location
Gold Coast, Queensland, Australia
Turn off your WiFi/Bluetooth on the PC? (Disable the adapter)

I get all sorts of issues from close proximity WiFi adapters on FFTs.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
Yeah, I always use a separate computer for ADC and DAC. Might try that as March suggests. I usually plug both PCs into the same heavy extension cord to reduce the chance of loops. Sometimes it helps to have one on battery and one powered from the wall. Sometimes both from battery.

If you don't have a second PC, you might try connecting to a phone or tablet via USB.
 
Top Bottom