Measurements of MiniDSP 2x4HD

[March Audio]

@BE718
Well I get the same results with Audacity and Wave Spectra. Here are some measures of an actual device. The cyan line is an 18i20 which is running 77 ppm fast. The gold line is the same signal once the speed has been corrected within about 10 ppb more or less. The second image has a pink line which is a synthetically created file 77 ppm fast. Hann window, 32 K fft. Again it only matters close in. In the second image you see how they all come together 200 hz either side of the 12 khz tone. 77 ppm fast btw is offset by .924 hz. You see the point at which -120 db crosses the graph is very close to where your .8 hz offset occurred at roughly 75 or 80 hz left and right.

Now trying to get back on topic. I originally brought all of this up because you showed the SMSL as having less close in phase noise than the miniDSP. Yet the amount of the difference or most of it could be nothing more than speed differences in the clocks of the two DACs vs the clock in the QA. Or it could be a real difference in close in timing.

If you still have both devices an easy quick check to see if that is some of it would be to look at the 12 khz waveform in Audacity to see if one is smooth and one lumpy which would indicate such a speed difference.

View attachment 11505
View attachment 11506

View attachment 11507

Nonetheless it still doesnt make sense to have zero leakage The variations you show above seem more in tune with what I would expect, and yes you are correct in saying that some of the difference between the SMSL and minidsp could be due to this frequency difference. Unfortunately I dont have the SMSL anymore to confirm.

What this does make clear is that we should be using much higher FFT resolutionsto minimise the issue.

262k point FFT min max

 

[March Audio]

The dsp functionality of the 2x4hd is its reason for being..as interesting as the measurements may be.

Well yes, I agree to an extent, however ultimately that gets output through its DAC which appears to be toward the lower end of performance. I also want to test the impact of its enforced re-sampling to 48kHz or 96kHz.

Ultimately I found for my DSP speaker project it didnt have the processing power to deliver sufficiently high resolution FIR filtering and regardless of that I subjectively found it wasnt as good as I expected (all usual subjective comment qualifiers apply ). Hence I went to a PC running Accourate.

These units are exceptionally easy to use however and would probably still outperform a passive crossover IMO.

 

[March Audio]

How about a measurement using the Analog input, since that is how I'm driving it right now.

I will try and get round to that soon

Perhaps I could also do a loopback audio recording to compare to the original.

 

[Blumlein 88]

Nonetheless it still doesnt make sense to have zero leakage The variations you show above seem more in tune with what I would expect, and yes you are correct in saying that some of the difference between the SMSL and minidsp could be due to this frequency difference. Unfortunately I dont have the SMSL anymore to confirm.

What this does make clear is that we should be using much higher FFT resolutionsto minimise the issue.

262k point FFT min max

I agree on the higher FFTs for this purpose. Most of your other measurements aren't hurt by sticking with 32 k FFT. Close in phase noise is however at least for the 100-200 hz each side of the test tone. Highest I can conveniently go with free software is 128 FFT. I'm not competent enough with Matlab or Octave to use that for the FFT, but those who are can do one as large as they wish. 1 million point FFT is enough for some really fine resolution.

Also with smaller sized FFT's switching to a Blackman_Harris window can be helpful vs Hann.

 

[March Audio]

Have a play with Arta and kaiser window

http://www.artalabs.hr

 

[Digital_Thor]

Hello everyone

Here in Denmark - I have listened to two different versions of minidsp - OpenDRC and C-DSP. They both have the same problem with noise, so that you need to use an amplifier with input gain to attenuate the line level, so that the noise floor becomes acceptable. It's a clear hiss and ac-noise. Everything has been tried, from using different amps, wall sockets, linefilters and so on.
But with my DCN28, noise is gone - even with 110dB horn speakers.
http://groundsound.com/dcn28.php

Thank you for a great homepage, that I now hope to enjoy even more.

 

[Wombat]

I get the same result(no noise) with a DCX2496 plus DEQ2496. Altec 8o2 CDs.

$US600(for both)new vs $US5500+ for the DCN.

 

[stunta]

I get the same result(no noise) with a DCX2496 plus DEQ2496. Altec 8o2 CDs.

$US600(for both)new vs $US5500+ for the DCN.

Is the DCX2496 for crossover and DEQ2496 for EQ? Do you use these for anything else?

What I am looking for is a single-box unit with a competent USB DAC and digital crossover functionality with XLR outputs - for a simple 2.1 setup like I have, this would be ideal. I can continue to do EQ in Roon.

I could use my existing USB DAC into the DCX2496, but it seems pointless/sub-optimal to be DA, AD and then DA again. There may not be any audible difference with this path, but the inefficiency bothers me, not to mention the additional wiring and equipment.

 

[Wombat]

Is the DCX2496 for crossover and DEQ2496 for EQ? Do you use these for anything else?

What I am looking for is a single-box unit with a competent USB DAC and digital crossover functionality with XLR outputs - for a simple 2.1 setup like I have, this would be ideal. I can continue to do EQ in Roon.

I could use my existing USB DAC into the DCX2496, but it seems pointless/sub-optimal to be DA, AD and then DA again. There may not be any audible difference with this path, but the inefficiency bothers me, not to mention the additional wiring and equipment.

DCX: crossover with 3 analog inputs(one suitable as a digital stereo AES/EBU input) and 6 analog outputs, all XLR. NO USB.

DEQ: Equaliser /Real Time Analyiser: analog XLR input and output, TRS aux. out and S/PDIF in and out plus RTA mic. in. NO USB

 

[March Audio]

Just been having a quick look at the minidsp re-sampling as it converts all inputs to 48kHz or 96kHz for processing.

I input a sweep from 10Hz to 22kHz and looped it back (through USB) to see if there were any obvious SRC issues. All other DSP processing is turned off. Note this is not going through AD / DA conversion. A colourmap display is great for this as it makes the low level spuria stand out.

Original signal 44kHz - note the clean background

Loopback - Has been converted to 48kHz. The images are around -130 dB so it looks worse than its audible impact. Interesting reflection abouve about 18kHz.

48kHz input - clean

Output, no SRC required. Noise floor increases to about -146dB from around -164dB but no spuria/images

Adobe Audition conversion to 48kHz - Perfectly clean and no increase in noise floor. Much better than the miniDSP.

Video of minidsp 44 to 48kHz sweep

 

[March Audio]

Just for fun the file below contains 3 tracks for comparison, the original, one of a TI ADC recording the output of a Chord Mojo and the other of the MiniDSP recording the Mojo. Can you identify any differences or which is which?

http://alanmarch.hopto.org:5000/sharing/myDXpuTLe

 

[Blumlein 88]

So further down you've got some aliasing there. Pretty interesting images especially the video. That one where the reflection as the sweep nears 20 khz with one coming the other way and then going away is common for gear with half band filters.

 

[March Audio]

So further down you've got some aliasing there. Pretty interesting images especially the video. That one where the reflection as the sweep nears 20 khz with one coming the other way and then going away is common for gear with half band filters.

Yes its interesting. Can we hear this? I dont know, but I have been sus about this aspect of the minidsp for a while. All I can say is that the SRC is definitely sub optimal compared to say the results of Adobe Audition.

EDIT: I hade the colourmap setting wrong on the Audition plot above - updated below, you can see some faint images.

 

[Blumlein 88]

Yes its interesting. Can we hear this? I dont know, but I have been sus about this aspect of the minidsp for a while. All you can say is that the SRC is sub optimal compared to the the results of Adobe Audition.

I doubt it audible, but yes sub-standard. Though you might be surprised. I found I could hear the difference in ABX of Arnie's jangling keys file which might be a decade old. He had recorded jangling keys which have substantial output into the 30 khz plus range at various sample rates. I could score near perfect out of 20 on it. It turned out to be the resampling software he used. I resampled it with Sox and was doing no better than guessing.

There also was the well done test of sample rates using music school students and concurrent recording. They couldn't hear 44.1 vs 88.2 khz recordings. But did hear 44.1 downsampling of 88.2 khz. So again the resampling they used. I'm thinking it was a fairly high quality resampler they used too. Maybe even Izotope.

 

[March Audio]

Have you ever come across this site?

http://src.infinitewave.ca/

 

[Blumlein 88]

Have you ever come across this site?

http://src.infinitewave.ca/

Yes, I've seen those images. Sox, Izotope and maybe one or two others stand well above any other resamplers I am aware of. What is sad is many DAWs use built in resampling for various purposes and some of those are pretty substandard. Tests of them like those at SRC show horrid looking results.

 

[March Audio]

My understanding is that the minidsp uses the in built SRC block of the 21483 DSP.

http://www.analog.com/media/en/tech...sheets/ADSP-21483_21486_21487_21488_21489.pdf

http://www.analog.com/media/en/technical-documentation/data-sheets/AD1896.pdf

Asynchronous Sample Rate Converter (SRC)
The asynchronous sample rate converter contains four SRC blocks and is the same core as that used in the AD1896 192 kHz stereo asynchronous sample rate converter and provides up to 128 dB SNR. The SRC block is used to perform synchronous or asynchronous sample rate conversion across independent stereo channels, without using internal processor resources. The four SRC blocks can also be configured to operate together to convert multichannel audio data without phase mismatches. Finally, the SRC can be used to clean up audio data from jittery clock sources such as the S/PDIF receiver.

 

[Blumlein 88]

Just for fun the file below contains 3 tracks for comparison, the original, one of a TI ADC recording the output of a Chord Mojo and the other of the MiniDSP recording the Mojo. Can you identify any differences or which is which?

http://alanmarch.hopto.org:5000/sharing/myDXpuTLe

At first I thought I could. Then I realized there was a tiny time mismatch. A tiny bit between all three, but I only noticed the larger one for the 1st and 3rd. I put them in Audacity without looking at much, but did align a transient in the first few seconds. After that I couldn't tell a difference. I didn't get super intense on trying it, but listening late at night all the same to me. Some forensic analysis might turn up which is which, but just listening nope not tonight anyway.

So my system sucks, my ears are cloth instead of gold, or well you know the spiel.........

 

[March Audio]

Re-reading the miniDSP plugin specs it sounds like the 2x4HD re-samples only to 96kHz (unlike some other miniDSP products) so here is 44.1kHz to 96kHz loopback recording.

https://www.minidsp.com/images/documents/Product Brief-2x4 HD1 plugin.pdf

 

[March Audio]

Using JRiver to perform the SRC to 96kHz prior to input into the miniDSP. I think the lesson here is perform the SRC to 96kHz first to avoid the miniDSPs somewhat sub optimal conversion!