• 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!

Which DAC would you choose?

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
Here are some measurements of three DACs I have on hand. This is a 2k FFT spectrogram. In the below view the background goes to light gray at -100 db FS. Just above that it becomes faintly blue then purple then red then white as level goes up.

The measures are a full scale single tone sweep 0-20 khz. And a twin tone full scale sweep with the tones 1000 hz apart or an IMD sweep.

For completeness all these DACs are very linear to -120 db, the Jtest is very clean for two and quite good for one. No big issues.

DAC A B C .png


All the above are pretty similar. Let us assume my speakers are max 105 db SPL in room. Since all the above DACs are very nearly 100 % clean at - 100 db FS will they sound different? DAC B does show just a trace of 2nd and 3rd harmonic at upper frequencies. -98.5 db and 97 db respectively. There is just a trace of IMD sidebands at upper frequencies at like -99 db.

So will they all sound the same, sound interchangeable? I mean there are no spurious tones, USB bleed thru or any other kind of distortion poking above -100 db effectively. What could possibly sound different about them?

Now let us look a little further into things. This is the same spectrogram except I have moved the the spectrogram floor to - 130 db FS.

DAC A B C 130.png


Peering deeper we see some differences now. DAC B has all the harmonics though at a very low level. After the 3rd harmonic everything is -110 db FS or lower. Will they matter at all?

The other DACs have only 2nd, 3rd and a little 4th harmonic. Those are at - 110 db and less.

DAC B has a constant tone near 16 khz. It isn't USB noise as it is actually at 16,438 hz at -123 db FS. It only occurs when connected to my Macbook. Will that be a problem? Should I not use it with a Macbook?

Also you see aliasing/imaging (it is some of both) where the harmonics turn back downward in the spectrogram. They are gone below -130 db by 20 khz and lower. Will that matter?

Why is this in psycho-acoustics? I wanted to get some perspective about how good is good enough. I'll take the best performance I can get, but should I only do that if it is cheap past a certain point? And what is that point in measured performance? Where to draw the line?
 

Attachments

  • DAC A B C 130.png
    DAC A B C 130.png
    782.2 KB · Views: 171
Last edited:

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,201
Likes
16,983
Location
Riverview FL
Here's a quick (5 seconds) log chirp in-room... JBL LSR 308 and Cheezewoofers.

I think I see 3rd harmonic from about 300 to 3000Hz, and 60hz room noise, 34Hz room resonance.

upload_2018-4-11_3-15-7.png
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
Here's a quick (5 seconds) log chirp in-room... JBL LSR 308 and Cheezewoofers.

I think I see 3rd harmonic from about 300 to 3000Hz, and 60hz room noise, 34Hz room resonance.

View attachment 11986

I don't see any 3rd harmonic there. You have a parallel frequency which is odd. Harmonics will diverge at an angle to the test signal, and I don't see that in yours. Also you might try dropping the window size to 2048 or 1024. It will catch things that happen in a shorter time period that way. You'd be surprised in room how that can gloss over short events due to averaging out so many samples in the FFT. Your using .68 seconds for each capture while you'd be using only .04 seconds at 2048 on the window size.

Here is an example. Twin tone sweep. The upper one is 2048 window size and lower one is 32K. On a sweep the 32 k requires so many samples it misses some of the signal that is there. On a static tone it would be fine, but shorter FFTs can work better on sweeps. Same range and gain on both of these only the window size is different. This will be even more so on a fast moving 5 second sweep.
DAC A two FFT sizes.png
 
Last edited:

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,201
Likes
16,983
Location
Riverview FL
I don't see any 3rd harmonic there. You have a parallel frequency which is odd.

The displays are different.

I have a log chirp, and log vertical scale. That makes the multiples appear as parallel lines.

Yours appears to be linear chip and linear vertical scale.

Smaller window does tighten up the harmonic a bit, but loses the detail in the lows, on a log sweep.

upload_2018-4-11_4-35-13.png
 
Last edited:

Arnold Krueger

Active Member
Joined
Oct 10, 2017
Messages
160
Likes
83
I don't see any 3rd harmonic there. You have a parallel frequency which is odd. Harmonics will diverge at an angle to the test signal, and I don't see that in yours. Also you might try dropping the window size to 2048 or 1024. It will catch things that happen in a shorter time period that way. You'd be surprised in room how that can gloss over short events due to averaging out so many samples in the FFT. Your using .68 seconds for each capture while you'd be using only .04 seconds at 2048 on the window size.

Here is an example. Twin tone sweep. The upper one is 2048 window size and lower one is 32K. On a sweep the 32 k requires so many samples it misses some of the signal that is there. On a static tone it would be fine, but shorter FFTs can work better on sweeps. Same range and gain on both of these only the window size is different. This will be even more so on a fast moving 5 second sweep.
View attachment 11987


The little thin blue line going up at a higher rate than the main chirp is a digital artifact - related to quantization error. One of those things that proper dithering makes go away. Seen and heard it many time in my experiments.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,201
Likes
16,983
Location
Riverview FL
I still think mine is third harmonic, which the JBL LSR 308 has in its 8" woofer, and the MartinLogan has less of, in its panel.

Showing 3rd and just a hint of 5th here on the JBL, and lack of third in the midrange on the ML.

upload_2018-4-11_12-15-22.png
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
I still think mine is third harmonic, which the JBL LSR 308 has in its 8" woofer, and the MartinLogan has less of, in its panel.

Showing 3rd and just a hint of 5th here on the JBL, and lack of third in the midrange on the ML.

View attachment 11993


You are correct Ray. I didn't notice you used a log scale and log sweep. I use linear because I find it easier to spot diverging faint lines than parallel faint lines if you get down in the noise. So a little 3rd and 5th and less on the panels. So your level of 3rd is what maybe .5% there. So if the DAC adds about .001% to it then it likely wouldn't be heard would it?
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
The little thin blue line going up at a higher rate than the main chirp is a digital artifact - related to quantization error. One of those things that proper dithering makes go away. Seen and heard it many time in my experiments.

I don't think so in this case Arnold. I am using linear sweeps and linear graphing. So harmonics go up at twice or thrice and so on the steepness of the sweep. Here is an FFT of DAC A where you see a 5031 hz tone with the cursor set to 15092 hz or the 3rd harmonic. Which is showing at - 103.4 db. This RMS'd with the 2nd is about 2 db better than the listed spec for this device.

18i20 3rd harmonic FFT.png
 
Last edited:
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
So no one is going to choose a DAC?
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,201
Likes
16,983
Location
Riverview FL
So no one is going to choose a DAC?

I'm unqualified for such duty.

I'm trying to fault my poorly measuring miniDSP 2x4HD as a practical matter, and haven't been able to quite yet.
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
What's the price?
Without any other information, I would take DAC A... Higher noise floor, but so low... Aaaaand, it is the first in the list:cool:

Two of them were $350ish and one goes for $1500.
 

Theo

Active Member
Joined
Mar 31, 2018
Messages
288
Likes
180
Do I think?:rolleyes:
Well, seriously, how can I tell? Can't see any $ sign in the FFT...:D The only thing that I can tell at the moment is that it is the first in the list!
 

Theo

Active Member
Joined
Mar 31, 2018
Messages
288
Likes
180
Maybe if you play "money", we may see a difference in the FFT?
 

Arnold Krueger

Active Member
Joined
Oct 10, 2017
Messages
160
Likes
83
I don't think so in this case Arnold. I am using linear sweeps and linear graphing. So harmonics go up at twice or thrice and so on the steepness of the sweep. Here is an FFT of DAC A where you see a 5031 hz tone with the cursor set to 15092 hz or the 3rd harmonic. Which is showing at - 103.4 db. This RMS'd with the 2nd is about 2 db better than the listed spec for this device.

View attachment 12000


I agree. I tried to duplicate the plot in question using the tools at hand. I think this is a fair approximation, subject to me not taking the time to go overboard. Its simply third order nonlinear distortion, in this case about 1%
1 pct third 24192.png
 
Last edited:

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
Looks like you will never disclose the answer until someone made a decision. Then I choose C since it is less noisy than A and less distortion than B.

Since I am not a believer of $$$ = quality and I am not a self-proclaimed golden ears anyway I have nothing to lose if C is the cheapest or from a brand with bad reputation.
 
OP
Blumlein 88

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
DAC A is the Focusrite 18i20. 8 channels for $400.

DAC B is the Antelope audio Zen Tour. The ADC is much better than the DAC. 4 channels for $1700.

DAC C is the Focusrite Forte. 2 channels for $400 before being discontinued.
 

bennetng

Major Contributor
Joined
Nov 15, 2017
Messages
1,634
Likes
1,692
Thanks for fulfilling my curiosity.

BTW does the Forte provide switchable I/O sensitivity? I used RME Multiface II in a restoration studio several years ago and that switch doesn't have any SNR penalty which means I can for example, set it to -10dBV to test a budget smartphone with some 0.5Vrms tiny output.
http://www.rme-audio.de/en/products/multiface_2.php

PS: it is an external PCIE interface

Despite being an old and discontinued product with poorer specs than the adi-2, Multiface II has much more resasonable I/O layout, and the ADI-2 "FS" discussion really made me worry about the future of this company, it looks like it is a beginning of being "evil".
 
Top Bottom