• 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!

A tale of three DACs: Bricasti Design M3, M1 SE, M3h Mdx

misureaudio

Member
Joined
Jul 29, 2019
Messages
90
Likes
108
Location
Mantova, Italy
Some spectral estimates are presented here, about three Bricasti DACs.

Two M3 DACs are examined first, an early (March 2019) non Mdx sample and a very recent (presumably first half 2022) M3h Mdx.

A M1SE DAC, with last-revision (May 2019) non Mdx digital board is further examined.

Tests are made with a National Instruments PXI-1033 having two PXI-446x Dynamic Signal Analyzers. Those are connected to a host PC, which is equipped with LabVIEW software and custom made acquisition and analysis tools. A SFDR (Spurious Free Dynamic Range) of 118dB is declared for the PXI-446x series DSAs.

The DAC under test is used as generator, driven by ASIO USB.

Foobar2000 is the audio player software. Appropriate plugins are installed to allow DSD playback in DoP mode when required, via ASIO USB.

The ASIO driver is the TUSB 1.6. (A newer one exists but it showed a problematic behaviour, so I decided to stick to the former version).

Test files containing the multitone stimulus signal are generated by Matlab and saved in flac format when PCM.

Multitone signal is calculated with an attenuation coefficient of -0.25 dB with respect to the maximum allowable digital value. (What you see in the spectral estimates reports the level of the single tone components, with obvious attenuation, while the signal, in the time domain, reaches peak values near the allowable maximum).

Alternatively the same multitone signal is converted in DSD128 using the TEAC editor, to test the DSD capabilities of the DACs.

Tests are made with a 4Ms-long acquisition buffer, allowing 20 passes to calculate a RMS average. Entire test with multitone stimulus, sampled at 192kHz-24bit, lasts about 7min 40secs.

4-th order Blackman-Harris windows are used in calculating the spectral estimates.

Here the report about Bricasti M3 non Mdx.

Firmware version 1.03. S/N 1012

Bricasti M3 non Mdx: PCM decoding, linear reconstruction filter:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[PCM192]-[4Ms]-[TUSB]-[LinFlt]-[RMS avg lin]-OK.png

Bricasti M3 non Mdx: PCM decoding, minimum phase reconstruction filter, Sync Fine mode:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[PCM192]-[4Ms]-[TUSB]-[MinFlt]-[RMS avg lin]-OK.png

Bricasti M3 non Mdx: PCM decoding, minimum phase reconstruction filter, Sync Wide mode:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[PCM192]-[4Ms]-[TUSB]-[MinFlt]-[SyncWIDE]-[RMS...png


Bricasti M3 non Mdx: DSD decoding, DSD2PCM conversion from DoP DSD128 signal:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[DSD128]-[4Ms]-[TUSB]-[DSD2PCM]-[RMS avg lin]-OK.png


Bricasti M3 non Mdx: DSD decoding, NDSD Native DSD mode from DoP DSD128 signal:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[DSD128]-[4Ms]-[TUSB]-[NDSD]-[RMS avg lin]-OK.png


Bricasti M3 non Mdx, 1kHz, 0dBFs signal, 4 passes RMS average, Sync Fine mode:

[1kHz]-[44100-24]-[180s]-[0dB]-[RD]-[PCM44.1]-[4Ms]-[TUSB]-[MinFlt]-[SyncFINE]-[RMS avg lin]-OK.png


Bricasti M3 non Mdx, 1kHz, 0dBFs signal, 4 passes RMS average, Sync Wide mode:

[1kHz]-[44100-24]-[180s]-[0dB]-[RD]-[PCM44.1]-[4Ms]-[TUSB]-[MinFlt]-[SyncWIDE]-[RMS avg lin]-OK.png


We report here, instead, about the M3h Mdx.

Firmware version: 1.08. (S/N not reported: borrowed sample, serial not printed on the device)

Bricasti M3h Mdx: PCM decoding, minimum phase reconstruction filter:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[PCM192]-[4Ms]-[TUSB Mdx]-[MinFlt]-[RMS avg li...png

Bricasti M3h Mdx: DSD decoding, DSD2PCM conversion from DoP DSD128 signal:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[PCM192]-[4Ms]-[TUSB Mdx]-[MinFlt]-[RMS avg li...png


Bricasti M3h Mdx: DSD decoding, NDSD native conversion from DoP DSD128 signal:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[DSD128]-[4Ms]-[TUSB Mdx]-[NDSD]-[RMS avg lin]...png


Bricasti M3h Mdx: PCM decoding, 1kHz, 0dBFs signal:

Bricasti M3 - 44.1kHz - 24bit - TUSB Mdx - 4Ms - [1kHz 0dB signal] - RMS avg lin.png

(In the new Mdx configuration, there no more switchable options (Sync Fine/Sync Wide)

Bricasti M3h Mdx, same multitone signal at a level of -90dBFs:

MultiTone[37]-[192000-24]-[600s]-[-90dB]-[RD]-[PCM192]-[4Ms]-[TUSB Mdx]-[MinFlt]-[PK Hold]-OK.png
 

Attachments

  • MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[DSD128]-[4Ms]-[TUSB Mdx]-[NDSD]-[RMS avg lin]...png
    MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[DSD128]-[4Ms]-[TUSB Mdx]-[NDSD]-[RMS avg lin]...png
    28.5 KB · Views: 64
Last edited:
OP
misureaudio

misureaudio

Member
Joined
Jul 29, 2019
Messages
90
Likes
108
Location
Mantova, Italy
Bricasti M1SE non Mdx

Firmware version 1.51. S/N 0955

PCM conversion, minimum phase reconstruction filter Minimum0:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[PCM192]-[4Ms]-[TUSB]-[RMS avg lin]-OK.png

DSD to PCM conversion, reconstruction filter DSD2:

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[DSD128]-[4Ms]-[TUSB]-[DSD2PCM]-[RMS avg lin]-OK.png

Bricasti M1SE non Mdx, 1kHz, 0dBFs:
Bricasti M1SE - 44.1kHz - 24bit - TUSB Mdx - 4Ms - [1kHz 0dB signal] - RMS avg lin.png

Bricasti M1SE , multitone, -90dBFs:

MultiTone[37]-[192000-24]-[600s]-[-90dB]-[RD]-[PCM192]-[4Ms]-[TUSB Mdx]-[MinFlt]-[PK Hold]-OK.png
 
Last edited:

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
3,334
Likes
3,278
Location
.de
Interesting. Looks like M3 has a bit of a jitter rejection issue and cleans up almost entirely in native DSD/DoP (odd, why would that be?). M3h is cleaner overall but some difference between PCM and native DSD remains.

M1SE is probably cheaper, higher odd-order harmonics and multitone distortion rising towards the low end (which suggests some sort of coupling capacitor issue or perhaps something that generates a negative supply internally).
 
OP
misureaudio

misureaudio

Member
Joined
Jul 29, 2019
Messages
90
Likes
108
Location
Mantova, Italy
Interesting. Looks like M3 has a bit of a jitter rejection issue and cleans up almost entirely in native DSD/DoP (odd, why would that be?). M3h is cleaner overall but some difference between PCM and native DSD remains.

M1SE is probably cheaper, higher odd-order harmonics and multitone distortion rising towards the low end (which suggests some sort of coupling capacitor issue or perhaps something that generates a negative supply internally).
M1SE is about 50% higher in price than M3. Both share identical DSP Motherboard. In the earlier non Mdx M3 the DSP motherboard is 'M1 digital'. I don't know if the two mbs are identical revision level: something could change in relevant details as the adopted reference clocks. The M1SE is completely dual mono as it goes to analog stages. There are respectively three and two linear power supplies in the M1SE and M3. In the M1SE you have a power supply for each analog output board.

Interestingly, the NDSD section is exclusive of the M3. The M1SE, instead, offers just DSD to PCM conversion, with a total of fifteen reconstruction filters available for the user to select.

What we see in the 1kHz test on the M3 involves difference components with the exact double of the mains in Italy (50Hz*2).
 
OP
misureaudio

misureaudio

Member
Joined
Jul 29, 2019
Messages
90
Likes
108
Location
Mantova, Italy
Ifi iDSD Signature fw v.5.3c GTO filter - For reference

Multitone (DSD256)

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD] - [DSD256]-[4Ms]-[RMS avg lin - 20].png

Multitone (PCM 192)

MultiTone[37]-[192000-24]-[600s]-[-0.25dB]-[RD]-[4Ms]-[RMS avg lin - 20].png


1kHz (PCM192)

[1kHz]-[0dB]-[44.1-24]-[4Ms]-[RMS avg lin - 4].png


Muiltitone -90dB (PCM192)

MultiTone[37]-[192000-24]-[600s]-[-90.00dB]-[RD]-[4Ms]-[RMS avg lin - 20].png
 
OP
misureaudio

misureaudio

Member
Joined
Jul 29, 2019
Messages
90
Likes
108
Location
Mantova, Italy
Wavelength Proton DAC 24/96 v1.12b 12/06/2010 S/N 28-0194 - For reference

Multitone -0.25dBFS fs=96kHz

MultiTone[34]-[96000-24]-[600s]-[-0.25dB]-[RD]-[ASIO4ALL]-[4Ms].png


1kHz 0dBFS PCM 44.1 kHz
1kHz -0dB - [44.1-24]-[2ch]-[ASIO4ALL]-[4Ms]-[NoGrid].png

Multitone -90 dBFS (DSD64 -> PCM 44.1kHz via Foobar2k)

MultiTone[37]-[192000-24]-[600s]-[-90dB]-[RD] - [192 kHz]_2.8M-[ASIO4ALL]-[44.1-24]-[4Ms].png

DUNN test

ST-DUNN-11025Hz+229.6875Hz - [ASIO4ALL]-[4Ms].png
 
OP
misureaudio

misureaudio

Member
Joined
Jul 29, 2019
Messages
90
Likes
108
Location
Mantova, Italy
M1SE reassessed with Cosmos ADC/Multitone Analyzer
 

Attachments

  • 002-SQ100Hz-4M_32AVG_96kHz_R-M1SE_MONO2L.jpg
    002-SQ100Hz-4M_32AVG_96kHz_R-M1SE_MONO2L.jpg
    112.7 KB · Views: 42
  • 002-SQ100Hz-4M_32AVG_96kHz_L-M1SE_MONO2L.jpg
    002-SQ100Hz-4M_32AVG_96kHz_L-M1SE_MONO2L.jpg
    111.9 KB · Views: 43
  • 002-SIN1000Hz-4M_32AVG_96kHz_R-M1SE_MONO2L.jpg
    002-SIN1000Hz-4M_32AVG_96kHz_R-M1SE_MONO2L.jpg
    97.6 KB · Views: 43
  • 002-SIN1000Hz-4M_32AVG_96kHz_L-M1SE_MONO2L.jpg
    002-SIN1000Hz-4M_32AVG_96kHz_L-M1SE_MONO2L.jpg
    98.2 KB · Views: 38
  • 002-MT32AP-4M_32AVG_96kHz_R-M1SE_MONO2L.jpg
    002-MT32AP-4M_32AVG_96kHz_R-M1SE_MONO2L.jpg
    120.9 KB · Views: 43
  • 002-MT32AP-4M_32AVG_96kHz_L-M1SE_MONO2L.jpg
    002-MT32AP-4M_32AVG_96kHz_L-M1SE_MONO2L.jpg
    121 KB · Views: 42
OP
misureaudio

misureaudio

Member
Joined
Jul 29, 2019
Messages
90
Likes
108
Location
Mantova, Italy
Topping D90MQA reassessed with Cosmos ADC/Multitone Analyzer
 

Attachments

  • 001-MT32AP-4M_32AVG_96kHz_L-D90MQA_MONO2L.jpg
    001-MT32AP-4M_32AVG_96kHz_L-D90MQA_MONO2L.jpg
    122.5 KB · Views: 39
  • 001-MT32AP-4M_32AVG_96kHz_R-D90MQA_MONO2L.jpg
    001-MT32AP-4M_32AVG_96kHz_R-D90MQA_MONO2L.jpg
    122.7 KB · Views: 41
  • 001-SIN1000Hz-4M_32AVG_96kHz_L-D90MQA_MONO2L.jpg
    001-SIN1000Hz-4M_32AVG_96kHz_L-D90MQA_MONO2L.jpg
    96.9 KB · Views: 35
  • 001-SIN1000Hz-4M_32AVG_96kHz_R-D90MQA_MONO2L.jpg
    001-SIN1000Hz-4M_32AVG_96kHz_R-D90MQA_MONO2L.jpg
    96.6 KB · Views: 40
  • 001-SQ100Hz-4M_32AVG_96kHz_L-D90MQA_MONO2L.jpg
    001-SQ100Hz-4M_32AVG_96kHz_L-D90MQA_MONO2L.jpg
    113.8 KB · Views: 32
  • 001-SQ100Hz-4M_32AVG_96kHz_R-D90MQA_MONO2L.jpg
    001-SQ100Hz-4M_32AVG_96kHz_R-D90MQA_MONO2L.jpg
    113.6 KB · Views: 38
OP
misureaudio

misureaudio

Member
Joined
Jul 29, 2019
Messages
90
Likes
108
Location
Mantova, Italy
So can we rank these from best to worse?
Initially I was interested in a comparison between the two measurement protocols, (mine with a multitone signal generated with Matlab and analyzed with a National Instruments DSA) (PXI 446x series), and with a recently acquired E1DA Cosmos ADC, driven by Multitone analyzer. The two protocols slightly differ but are well consistent. So Cosmos ADC is a perfect alternative to the much older NI analyzer and my Matlab/LabVIEW software.

Having the two DACs in my lab it was ok to reassess both of them with Cosmos ADC.

I concentrate on the datum of SFDR, Spurious Free Dynamic Range, which differs so much between D90MQA and M1SE.
(https://www.ni.com/en/support/docum...ined--spurious-free-dynamic-range--sfdr-.html)

D90MQA confirms the high quality of its implementation, ending with an estimate better than 105 dB SFDR.

M1SE shows here a very low value of SFDR, just a bit better than 55 dB, probably due to low frequency jitter that is able to induce a frequency modulation of the test signal.

It looks like the problem worsens at higher signal frequencies.

Analysis window is a quite long 4 Ms.

Multitone Analyzer is able to detect and display the content of the frequency bins that are contiguous to the center frequencies of the test signal.

The spread of the near-the-center components is probably low enough to be inaudible, due to to the differential frequency threshold of human ear.

However this is not a good sign, if confirmed, because what we see in spectral estimates derives from modulations of instantaneous frequency of the signal.

An investigation in time domain probably would require an analysis tool like Huang-Hilbert Transform.

Every idea about this is welcome.
 
Last edited:
Top Bottom