• 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!

Review and Measurements of Mytek Brooklyn DAC

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,388
Location
Seattle Area
Hello everyone. Courtesy of a friend of the forum :), I am in possession of boatload of new hardware to measure/test. I am starting with the well regarded Mytek Brooklyn DAC measurements.

Brooklyn-Dac-for-web.png


This is a $2,000 DAC with support for MQA and DSD, balanced output, and multiple inputs (Toslink, S/PDIF, AES/EBU). It even has an analog input!

The Mytek Brooklyn has a very nice OLED display, volume control, remote and headphone output. Unit is self powered with its own AC input/internal power supply. The box is the size of a paperback book, albeit in square format. It is heavy enough to feel solid and stay put.

Driver installation was a breeze and all worked the first time. For some reason I cannot get into any of the menus. The four buttons in the front that are supposed to do that, do nothing in my loaned input. Maybe there is a lock function some place.

As usual, I threw the 24-bit, 12 Khz, 48 Khz J-test signal at it. Results show comparison to iFi iDAC 2 (around $350) and Behringer UMC204HD ($70):

J-test.png


As we see, all three show very clean output. The Behringer noise floor is lower but so is its output level. If compensated, it would look like the other two.

Next up is a test of harmonic distortion using 7 Khz tone. For the Mytek, I used a 6.7 Khz tone so that its output is shifted to the left, allowing simultaneous comparison. Here, I am comparing it to the Behringer:

7 Khz.png


Sadly we see the same problem we had seen with iFi iDAC with pretty high second harmonic distortion. The Behringer on the other hand only has a third harmonic which it shares with the Mytek Brooklyn just the same. Remarkable how the $70 Behringer DAC is able to have such low level of harmonic distortion.

Now to a problem that cost me a lot of time. When I first tested the Mytek with the same 7 Khz tone as the others, it generated a ton more distortion products than what is seen above. After some investigation, I realized that the distortions were due to overflow in the signal processing/filtering in the Mytek Brooklyn. See this comparison of the 6.7 Khz tone at -1.0 db to 7 Khz at 0.0 db:

0db clipping.png


Notice all of those spikes with 0 db signal. The levels of each spike may seem low but what we hear is their combined energy which will be quite a bit higher. I will do some listening tests later but for now, it is disappointing to see that they do not have enough headroom in their math to avoid this overflow. Neither iFi DAC nor Behringer have this problem.

Given the fact that a lot of music today is compressed with levels a hair below or at 0 db, I expect this to be a real nuisance.

Another note: the levels out of the box was some 8 to 9 db higher than other DACs I tested. The manual says there is a -4db adjustment that can be made by changing a jumper. This not being my box, I did not attempt to do that. But suffice it to say, subjective testing will present the user with much higher level than other DACs, potentially resulting a much more positive impression than would exist otherwise.

Summary:
This seems like a well implemented DAC from physical point of view. The box, display, copious inputs, etc. are all nice. Alas, when it comes to analog performance out of the DAC, it shows the level of maturity that exists out there where a $70 DAC from Behringer outperforms it in distortion product. The clipping at 0 db FS is also disappointing.

As usual, I very much welcome comments, inputs, feedback, corrections from manufacturers or members alike. It is easy to make mistakes in conducting such measurements and I am happy to acknowledge any errors in them.

EDIT: I received more information from the designer of Mytek Brooklyn on the issues I found above as reported in this post: http://www.audiosciencereview.com/f...-of-mytek-brooklyn-dac.1828/page-3#post-47017
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
Try sending that Mytek the white noise at -4 db, -3 db and -1 db with wideband plots. You know send it the signal as a 48 khz version and show the output over 192 khz.

That -4 db noise signal just every so often has a peak level that gets close to clipping. The reason it wasn't used at a higher level I suspect. For some DACs I can set a -.1 db sine wave which is fine, and even then that - 4 db noise signal will clip a bit much. So this might be a way to see more about what is going here. You of course need to do that for the other DACs to get a good comparison between them.

There is a thread over on CA asking if the UMC204 HD outperformed the iFi DAC. I suppose now we need one there that asks the same of the Mytek Brooklyn.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,202
Likes
16,982
Location
Riverview FL
Try sending that Mytek the white noise

When I look at music files up close, the 44k or 48k samples can look pretty ragged.

So, sometimes I've resampled to 384k to smooth things out a bit to see what is likely happening after that low rate signal is smoothed out by the filter after the conversion in a DAC. I don't know if that is entirely legal or not, but here's an example.

Method of madness:

Take a full-scale White Noise 48kHz (last track below is the original), copy and paste it 5 times, so each track starts out the same.

Amplify by -4, -3, -2, -1, and 0 -- to make your suggested files at 48kHz.

Smooth out the result of those with resampling to 384kHz, where the idea is that the upsampled version is closer to what the analog output would look like (if i'm wrong, let me know).

Then hunt down the clipping, of which there is plenty.

upload_2017-7-31_1-27-49.png


Maybe this isn't valid since the DAC output filter is a low-pass, or maybe not since the source bits are already bandwidth restricted. I don't know. Neither does Donny.

PS: My DAC claims it has 3.5dB headroom for those nasties. It also runs everything coming in on the S/PDIF at 211kHz. How it accomplishes all that, I don't know.

Reading this, briefly, I suppose they attenuate the incoming digital signal by 3.5dB:
https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-2-digital-processing
 
Last edited:

Jinjuku

Major Contributor
Forum Donor
Joined
Feb 28, 2016
Messages
1,278
Likes
1,180
Could you get a snapshot of the balanced output on the Behringer?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
Might be good to do a twin tone IMD test of the Mytek too.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
When I look at music files up close, the 44k or 48k samples can look pretty ragged.

So, sometimes I've resampled to 384k to smooth things out a bit to see what is likely happening after that low rate signal is smoothed out by the filter after the conversion in a DAC. I don't know if that is entirely legal or not, but here's an example.

Method of madness:

Take a full-scale White Noise 48kHz (last track below is the original), copy and paste it 5 times, so each track starts out the same.

Amplify by -4, -3, -2, -1, and 0 -- to make your suggested files at 48kHz.

Smooth out the result of those with resampling to 384kHz, where the idea is that the upsampled version is closer to what the analog output would look like (if i'm wrong, let me know).

Then hunt down the clipping, of which there is plenty.

View attachment 8022

Maybe this isn't valid since the DAC output filter is a low-pass, or maybe not since the source bits are already bandwidth restricted. I don't know. Neither does Donny.

PS: My DAC claims it has 3.5dB headroom for those nasties. It also runs everything coming in on the S/PDIF at 211kHz. How it accomplishes all that, I don't know.

Reading this, briefly, I suppose they attenuate the incoming digital signal by 3.5dB:
https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-2-digital-processing

Your examples illustrate why the white noise needs to be at - 4 db, and also why a sine wave even -.1 db will be fine while with all the same level settings the -4 db white noise will cause clipping every once in awhile. I ran into this testing DACs. I used a 440 hz tone to set levels as close to max in the ADC as possible. I used 440 because it was an irrational fraction of any of the common sample rates. I found you could then run the - 4 db white noise and get just a few instances of clipping over 15 seconds. So I started using the white noise to set levels instead. I suppose white noise maximizes the chances of intersample overs. Maybe that is how Benchmark came to the decision to have 3.5 db headroom.

I think your up sampling to see the waveform better is probably just fine. If someone has Adobe Audition it draws the actual waveform instead of just connecting dots with the sample values. Your upsampling would do a similar thing for us Audacity Cowboys.

I assume you know the -4 db white noise signal is an idea I stole from Stereophile who was told about it by Juergen Reis of MBL. I am thinking he picked -4 db for all the same reasons.

My thinking to test the Mytek that way was to see if it is clipping from intersample overs when it shouldn't be or if it clips more easily than it should as you raise the level of the noise above -4db in steps.
 
Last edited:

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,497
Location
Monument, CO
I thought the crest factor of pure white (which may or may not be Gaussian) noise was greater than 4 dB? I thought it was more like 10 dB (1-sigma), and when I was testing data converters we typically applied noise at -10 dB FS to avoid clipping. Been a while since I looked it up... The numbers I have memorized are not applicable to audio testing. Note two uncorrelated sine waves at the same amplitude when added result in a 6 dB increase in voltage.
 
Last edited:

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,202
Likes
16,982
Location
Riverview FL
Your examples illustrate why the white noise needs to be at - 4 db

New observation:

One minute of Audacity White Noise at 24/48 needed attenuation of -6.3dB to eliminate full-range samples after resampling to 24/384k.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
Took a little while to look at this too. Here are 10 seconds of white noise at -4, -5 and -6 db. These are upsampled to 384 khz rates. I don't know how close to gaussian pure white noise the generator in Audacity gets. I did find at -6 db if I generate say an hour of it then you get places with minutes that have no clipping and get areas with a bit of clipping. So a full pure noise signal may indeed have 10 db crest factor. Also in each of the levels below there is a 5 db difference in the nominal level and the RMS level.

Upsampled white noise waves.png


I also wonder if Juergen Reis wanted some clipping to create a flat band white noise level above the sample rate. I'll have to try these with real signals, and see what happens at -4 thru -6 db. Below is what the FFT looks like. Red is -4 db, blue is -5 db and green is -6 db. This only resampled to 192 khz as Wavespectra doesn't work at 384 khz. Note to the left the value of the cursor which I placed at 24 khz for each of the plots.

wavespectra noise plot.png
 
Last edited:

Wayne

Active Member
Forum Donor
Joined
May 26, 2017
Messages
172
Likes
46
Location
Los Angeles, CA
Trying to get a handle on the above tests, as I have no experience in this area.

As usual, I threw the 24-bit, 12 Khz, 48 Khz J-test signal at it.

You are inputting two signals, a 12 Khz and a 48 Khz? The 48 Khz is to check for aliasing frequencies. Is the 12Khz to look for harmonics? Is there a reason why 12 Khz was chosen?

Next up is a test of harmonic distortion using 7 Khz tone.

OK, the 7 Kz is for harmonics, why this frequency? Is the "volume" higher than in the above test?

Thanks,
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,388
Location
Seattle Area
You are inputting two signals, a 12 Khz and a 48 Khz? The 48 Khz is to check for aliasing frequencies. Is the 12Khz to look for harmonics? Is there a reason why 12 Khz was chosen?
No, no. It is a 12 Khz tone at the *sampling rate* of 48 Khz. The tone has its low order bits toggled which over links such as S/PDIF can aggravate jitter. For USB bus it doesn't do that but does cause bit changes/activity in the DAC which may show something.

Outside of that 12 Khz at full amplitude is a high frequency signal which brings out internal sources of jitter and voltage modulation which show up as other spikes.

It can show harmonic distortion also but I usually don't go high enough to show them. The first harmonic would be at 24 Khz. To show harmonics better, I have been using the 7 Khz tone so that its second and third harmonics are still within audible band.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,388
Location
Seattle Area
And oh, 12 Khz was used because it divides the sampling rate exactly by 4. This allows the the tone to be precisely generated as to not need added dither/noise t conform to those PCM samples. That in turn allows the measurement noise floor be lower.
 

DonH56

Master Contributor
Technical Expert
Forum Donor
Joined
Mar 15, 2016
Messages
7,835
Likes
16,497
Location
Monument, CO
48 kHz is the base sampling rate for the audio in most videos these days. And 48/4 = 12 whilst 44.1/4 = 11.025 -- 12 is easier to type. ;)

Most data converter testing does not use even submultiples of the sampling rate because it causes multiple binning of harmonics, making it more difficult to gauge performance. I use the scheme from the IEEE Standard (1241) to calculate suitable test frequencies.
 

Wayne

Active Member
Forum Donor
Joined
May 26, 2017
Messages
172
Likes
46
Location
Los Angeles, CA

I think I am beginning to "see the light." In the ASR forum I have seen many statements that basically say "jitter" in most audio devices is not an issue (ref Jitter Stew, etc.). So based on the methodology of simulating "jitter" as used in the J-test signal; the above tests demonstrate that "jitter" is not an issue in the three devices tested.

Is this an accurate interpretation of the above test results?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
48 kHz is the base sampling rate for the audio in most videos these days. And 48/4 = 12 whilst 44.1/4 = 11.025 -- 12 is easier to type. ;)

Most data converter testing does not use even submultiples of the sampling rate because it causes multiple binning of harmonics, making it more difficult to gauge performance. I use the scheme from the IEEE Standard (1241) to calculate suitable test frequencies.

Where can we see that standard? I assume this is similar to why some test CDs used 997 hz instead of 1 khz so it would exercise more bits more variably than a quickly repeating frequency.
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,202
Likes
16,982
Location
Riverview FL

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
997 is a prime number.
It is, but I don't think that was why it was used. I seem to remember it exercised 14 of 16 bits before repeating in either 44,100 hz or 48,000 hz sampling while 1000 hz only exercised something like 5 bits in 48 khz and maybe only 7 in 44.1 khz. Or maybe all primes have similar properties in digitally sampled systems.

Maybe we should have used a sample rate of 44,101 and 48,017. Or maybe I should market that standard with some wrinkles as a competitor to MQA to fix what is wrong with digital audio.
 
OP
amirm

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,388
Location
Seattle Area
Also, is there a reason for choosing a sampling rate of 48 instead of 44.1 (the CD standard)?
Either can, and are used. I use 48 Khz because I have a C program that generates it and it is set for 48 Khz. And as Don said, it is the standard for video and recently a few high-resolution downloads.

Either way, if there are jitter components, they show up with J-test signal in either sampling rate.
 
Top Bottom