• 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!

MSB DACs measurements

Matias

Master Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
5,029
Likes
10,796
Location
São Paulo, Brazil
I wonder if @amirm 's test would confirm these impressive numbers that MSB has once shared about its R2R DACs. Comments?

Select-90db-Graph.jpg
 
Last edited:
OP
Matias

Matias

Master Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
5,029
Likes
10,796
Location
São Paulo, Brazil

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
2,690
Likes
6,013
Location
Berlin, Germany
It may not be a fake plot but it is very misleading and important information is not present, making the whole thing a deception:
1) Note the dB scale is dBV (dB relative to 1Vrms), that means we do not know what the 0dBFS level corrsponds to.
2) A large FFT size is used, 1024*1024 (1048576) points, which means the noise floor of each frequency bin is really low because the bins are so narrow with 48kHz sample rate. The true noise floor (rms value, wideband 20...20kHz) is larger by factor of more than 700 (57.2dB).

"28.5 effective bits" is completely impossible and they know that...

EDIT: MSB describe their measurement setup here (and still makes 28.5 bits impossible) : http://www.msbtech.com/products/measurements.php?Page=dacSelect
(note: you must click the stop button of your browser immediately after loading, otherwise the page will redirect to MSB's main page. They don't want us to see this, for some reason).
 
Last edited:
OP
Matias

Matias

Master Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
5,029
Likes
10,796
Location
São Paulo, Brazil
Their website before home page printed below.

msb.JPG
 

dc655321

Major Contributor
Joined
Mar 4, 2018
Messages
1,597
Likes
2,235
I'm sure it's all very impressive, but:
a) their Select DAC is $85,000, before "upgrades"
b) even with ~60dB of FFT gain, one cannot get around physics
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,522
Likes
37,050
Their website before home page printed below.

View attachment 34298
Nothing wrong with this, it all would work. I've done something similar to show a few sigma delta DACs are in fact putting out an accurate signal to the 23rd bit and not far off the 24th bit. Even though I don't have any gear with 24 bits of resolution. What I did was feed the digitally reduced signal to the DAC, and feed it into a microphone preamp with 60 db of gain. The microphone preamps analog circuit will amplify noise along with signal while adding very little of its own noise. So having boosted the signal well above the noise floor of my ADC I take the recorded result and reduce it digitally until the test signal is back at its original low value. This pushes the noise floor of the DAC (which I amplified above the noise floor of my ADC) down proportionally. This lets me recreate the signal and with FFT gain actually accurately measure signals down near -150 db even though I don't have any gear with that low a noise level themselves. This is what they have done in the above described setup only using better gear than I have.

Where they are playing fast and lose with definitions is in the annotations to the graph they show. Let us look at the Select DAC which is the green line. One of the issues is with a 1 meg FFT you'll get 57 db of extra depth looking into noise over the bandwidth they are examining. Meaning the squiggly line of noise they show at about -189 db were it white noise would equate to an approximate noise level for that bandwidth of -132 db which is excellent. But isn't a dynamic range of 173 dbFS. You see how they are getting this, they show SINAD of 83.44 for the -90 db signal. Adding 90 db and 83 db for 173 db. Something is off with their SINAD measurement however, as what they are showing with the FFT gain considered isn't 83 db SINAD. The distortion and power supply noise might be at this level, but they must be ignoring wideband noise. My guess is they are only totaling up THD and power supply noise while ignoring wideband noise. That might get you close along with the dbV reference KSTR noted upthread. The Select DAC can output 12 volts which is nearly another 22 db. Perhaps they are assuming that, but for all we know the output noise rises 22 db as well. So they are trying to deceive a bit here.

You also would think about output impedance. Resistance or impedance creates a floor of noise you can't get below. For 24 khz bandwidth as in these graphs even a 1 ohm output impedance means thermal and other noise will be -154 dbV. If output impedance is 10 ohms that would be -144 dbV.

Their gear does appear to have very low output noise at low levels, but their THD of -90 db with high signal levels and their IMD doesn't look all that impressive. Good enough to be inaudible, but lots of gear will match that part for a tiny fraction of the cost.
 
OP
Matias

Matias

Master Contributor
Forum Donor
Joined
Jan 1, 2019
Messages
5,029
Likes
10,796
Location
São Paulo, Brazil
About wide band noise, they published this as well.

images (8).jpeg
 

RayDunzl

Grand Contributor
Central Scrutinizer
Joined
Mar 9, 2016
Messages
13,198
Likes
16,981
Location
Riverview FL
Wow...

-200dB Measurement System Noise.

1569283105788.png


That's beyond 32bit.

?
 

FrantzM

Major Contributor
Forum Donor
Joined
Mar 12, 2016
Messages
4,337
Likes
7,729
These "measurements" will likely disappear in the next few days. High End Audio is taking notice of ASR. The people who frequent this site have the knowledge to call their BS.
Thanks guys...
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,522
Likes
37,050
Wow...

-200dB Measurement System Noise.

View attachment 34302

That's beyond 32bit.

?
Well -200 db with a 1 meg FFT is only -143 db for the full bandwidth. So almost 24 bit. And remember in the diagram of the measurement gear setup they were using a 22 bit ADC (plus analog gain to push noise above the noise floor of the ADC).
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,522
Likes
37,050
Here is a measurement of the March Audio DAC I did. Signal is -60 db. Only a 128 k FFT. The line of noise would drop another 9 db if it were 1 meg FFT putting it close to -170 db. And if I used the method MSB used I'd state this DAC has about 148 dbFS of dynamic range. With a 1 meg FFT I might also lower that by a few more db.

1569287369267.png
 

YSC

Major Contributor
Joined
Dec 31, 2019
Messages
3,194
Likes
2,570
Here is a measurement of the March Audio DAC I did. Signal is -60 db. Only a 128 k FFT. The line of noise would drop another 9 db if it were 1 meg FFT putting it close to -170 db. And if I used the method MSB used I'd state this DAC has about 148 dbFS of dynamic range. With a 1 meg FFT I might also lower that by a few more db.

View attachment 34310
This is really a shame given their price! as a poor working class I can't afford their dac but this is like Ferrari claiming their car can do 0-60 if tyre have no grip limit! R2R are known to be measuring worse than DS dacs, and I believe those who would buy one at this price don't need them to cheat like this and will pay with a properly measured result with say 90-100db SNR
 

WolfX-700

Addicted to Fun and Learning
Forum Donor
Joined
Jul 26, 2019
Messages
537
Likes
4,508
Location
China
Today someone asked me about the MSB measurement result.To explain this problem, I partially reproduced the measurement.

According to the measurement block diagram released by MSB, they use -30dbFS signal and -60db DAC attenuation, then use a +70db LNA to amplify the measurement, and shift the vertical axis to keep the signal level at -90db.At the same time, it is known that the MSB uses 1MFFT.

The following is the measurement result when I use a DAC that is less than 300 US dollars, without LNA, and directly use 1MFFT:

90.jpg
 

YSC

Major Contributor
Joined
Dec 31, 2019
Messages
3,194
Likes
2,570
Today someone asked me about the MSB measurement result.To explain this problem, I partially reproduced the measurement.

According to the measurement block diagram released by MSB, they use -30dbFS signal and -60db DAC attenuation, then use a +70db LNA to amplify the measurement, and shift the vertical axis to keep the signal level at -90db.At the same time, it is known that the MSB uses 1MFFT.

The following is the measurement result when I use a DAC that is less than 300 US dollars, without LNA, and directly use 1MFFT:

View attachment 122128
Are there any link to the MSB results for a quick comparison?
 

conman

Member
Joined
Jan 11, 2022
Messages
20
Likes
38
I know this is an old thread, but I stumbled across these independent measurements on their Select DAC.
 

Goodsound

Member
Joined
May 21, 2022
Messages
8
Likes
7
I know too, this is an old thread, but I stumbled across these independent measurements on their The Analog DAC. Just for completion:
35BEC7A9-FE6E-4BD4-A3B4-FD2ACC92D434.jpeg

The text description added: measured Jitter of 161ps on all inputs. Seems not bad for 2013.
 
Top Bottom