How to demonstrate MQA's leaky filter

[GoldenOne]

In a 2018 RMAF talk by Chris Connaker, MQA handwaved away evidence collected by Archimago that their filter design was leaky, by implying he was involved in the industry and was biased. And that because Chris wouldn't reveal Archimago's real name, the information couldn't be trusted.

Chris quite rightly said that it didn't matter who Archimago was, anyone could quite easily demonstrate the issue themselves, and this post will show how to do just that.

 

1) Download and install the "VB-Hifi cable" software.

This is a bit-perfect virtual audio cable which we will use. (They also have a standard "VB-Cable" software, this is not the same and is not bitperfect).

Link: https://vb-audio.com/Cable/index.htm
Direct download link is about halfway down the page.

 

2) Install "Audacity", "Tidal", and "Spek"
https://www.audacityteam.org/download/
https://tidal.com/

http://spek.cc/
 

3) Configure the cable

In windows, right click the sound icon on your taskbar, and choose "Open Sound Settings"

 

 

Then go to "Sound Control Panel" on the right.
Find "Hi-Fi Cable Input" on the list, right click and choose "Properties"

Next, click the 'advanced' tab, and set the cable to 24 bit 88200hz.
If you know that the track you wish to test is 48khz, then choose 96000 instead. But most will be 44.1khz and unfold to 88.2khz.

 

Repeat step 3 for the hifi cable OUTPUT in the "Recording" tab of the sound control panel. Both must be 24 bit 88200hz.


4) Configure Audacity
Open up Audacity, and make sure that the following are set correctly:

1: Windows WASAPI is selected
2: Input device is "Hi-Fi cable output"
3: Project rate is 88200

 

 

5) Configure Tidal
In the settings section for Tidal, make sure that streaming audio quality is set to "Master"
Then make sure that the output device is set to "Hi-Fi Cable Input"
Finally, click the "More settings" text next to 'sound output', and configure as shown

 

6) Record audio

In Audacity, hit the record button, and then in Tidal, play a "Master" track. Let it record for as long as you desire.

7) Export
In Audacity, hit the yellow "stop" button, and then go to File->Export->Export WAV
Save the file somewhere accessible.

8) Analyze
Open up 'Spek', and drag your WAV file into it. (Alternatively can use any FFT or Spectrum analyzer)
You will see the spectrum view of your file, and will see that there is >22.05khz content present.

 

 

Congrats, now you know that it wasn't Archimago's bias, but indeed a genuine concern with MQA's upsampling filter.

[GoldenOne]

34 minutes ago, PeterSt said:

OK, what do I miss ... ?

 

You play a track that is supposed to be 88.2 after unfolding and you show context above 22.05 ?

Nothing wrong with that.

 

So I must have missed some data point here.

Thanks.

Sorry, should have clarified in the original post that this is intended  to be done using a track that came from a 44.1khz master. (Roon will tell you the original master of an MQA file or track on Tidal so you can check if it was ever really high-res or not, most MQA releases came from 44.1khz).

This test could also be done to demonstrate leaky filterer in the 'renderer' by recording the analog output of a renderer tricked into using it's MQA filter on a normal file using Mansr's tool. https://code.videolan.org/mansr/mqa/-/tree/master (I wasn't able to actually get that working, but I had my own track published in MQA so used that instead).


MQA Core Decoder (with HQP sinc-M in blue for comparison, both digital recordings):



MQA Renderer Filter in Blue, ifi GTO filter in white (both analog recordings):



HQPlayer sinc-L TDPF to show device limitations in the above comparison (analog recording):