MQN testing/experimentation thread

[jkeny]

A lot of this seems to be involved with the underlying mechanisms of "auditory scene" creation. This is how we isolate/follow one conversation from among many or from a background of noise. Also how we follow a particular musical strand & can jump from one to another.
These areas in psychoacoustics aren't yet fully understood/explained so it's not surprising that we are also struggling to define what's going on.

[jkeny]

Mark,
As you are investigating Audacity, here's something for you (& maybe others to try) - using Audacity, not for measurement for for introducing a known value of noise into a music stream to test what affect it has on the perception of music playback.

Here's what I suggest to investigate the effect of very low noise on music:
Generate a noise track (pink or white), using this feature in Audacity - then low pass filter it (another feature) to 10KHz, 5KHz or whatever low freq desired.
You can now merge & render this noise track with a music track to check the audibility of it's effect.

This might be interesting to see if we are on the right track?

[jesuscheung]

just had a simple idea.

1. play a song with MQn
2. record the song from DAC line out.
3. play the recorded song with MQn again.
4. record it again....

repeat it like 10 times.

if the version of MQn has something wrong, let say it compresses sound width by 1%.

so if you play same song 10 times, the width will reduced by >10%

this jitter will be very easily audible. now, sbgk can go on and fix this jitter!

i know it is not totally automation, but it feels practical.

[DaveF]

Here's one possible way to measure the noise on audio signals and to examine if different MQN builds are really affecting it.
http://www.aaronia.com/products/spectru ... BBV-NF-35/
In the link here is an LNA (Low Noise Amplifier) that will work from 1Hz up 30MHz. It has a gain of 35db so you can see a lot of what's going on in a signal that otherwise would not be seen even with a pretty decent scope. I've used it before.
It would have be done on slow signals such as I2S or SPDIF or perhaps on the signals going into the DAC. Think USB would be too fast for the bandwidth of this device.
Anyway what you'd have to do is generate a mere second or two worth of audio samples in MQN and repeat them all back to back continously but the trick would be to generate a trigger signal that your scope (with the LNA attached) could lock onto. That is the really difficult bit, piece of cake for an FPGA to do but on a PC, I'm not so sure it can be done. With this repeating pattern of samples you can zoom in on the scope to your hearts content and look for tones in the noise.

I have to say I'm pretty sceptical that different MQN builds are changing the noise out on the DAC but this might be one way.

[sbgk]

Here's one possible way to measure the noise on audio signals and to examine if different MQN builds are really affecting it.
http://www.aaronia.com/products/spectru ... BBV-NF-35/
In the link here is an LNA (Low Noise Amplifier) that will work from 1Hz up 30MHz. It has a gain of 35db so you can see a lot of what's going on in a signal that otherwise would not be seen even with a pretty decent scope. I've used it before.
It would have be done on slow signals such as I2S or SPDIF or perhaps on the signals going into the DAC. Think USB would be too fast for the bandwidth of this device.
Anyway what you'd have to do is generate a mere second or two worth of audio samples in MQN and repeat them all back to back continously but the trick would be to generate a trigger signal that your scope (with the LNA attached) could lock onto. That is the really difficult bit, piece of cake for an FPGA to do but on a PC, I'm not so sure it can be done. With this repeating pattern of samples you can zoom in on the scope to your hearts content and look for tones in the noise.

I have to say I'm pretty sceptical that different MQN builds are changing the noise out on the DAC but this might be one way.

wish there wasn't any difference.

the only question you need to answer is do different o/s and different players all sound the same ? Most people's experience is that they don't. That should then lead to further questions.

[DaveF]

Here's one possible way to measure the noise on audio signals and to examine if different MQN builds are really affecting it.
http://www.aaronia.com/products/spectru ... BBV-NF-35/
In the link here is an LNA (Low Noise Amplifier) that will work from 1Hz up 30MHz. It has a gain of 35db so you can see a lot of what's going on in a signal that otherwise would not be seen even with a pretty decent scope. I've used it before.
It would have be done on slow signals such as I2S or SPDIF or perhaps on the signals going into the DAC. Think USB would be too fast for the bandwidth of this device.
Anyway what you'd have to do is generate a mere second or two worth of audio samples in MQN and repeat them all back to back continously but the trick would be to generate a trigger signal that your scope (with the LNA attached) could lock onto. That is the really difficult bit, piece of cake for an FPGA to do but on a PC, I'm not so sure it can be done. With this repeating pattern of samples you can zoom in on the scope to your hearts content and look for tones in the noise.

I have to say I'm pretty sceptical that different MQN builds are changing the noise out on the DAC but this might be one way.

wish there wasn't any difference.

the only question you need to answer is do different o/s and different players all sound the same ? Most people's experience is that they don't. That should then lead to further questions.

The difficult bit for me is that everyone is using motherboards of different types, layouts, PCB design techniques so that even if MQN reduces the noise floor in some instances it probably would have no affect at all in another situation with another board. But we dont know the nature of this noise. I think the method I suggested above is a very valid. Here at work, if we were to look at low noise this is probably how we'd do it.

A pity that we have to use a PC at all as its far too clunky and bloated for the simple task of sending audio bits to a DAC.

[sbgk]

Here's one possible way to measure the noise on audio signals and to examine if different MQN builds are really affecting it.
http://www.aaronia.com/products/spectru ... BBV-NF-35/
In the link here is an LNA (Low Noise Amplifier) that will work from 1Hz up 30MHz. It has a gain of 35db so you can see a lot of what's going on in a signal that otherwise would not be seen even with a pretty decent scope. I've used it before.
It would have be done on slow signals such as I2S or SPDIF or perhaps on the signals going into the DAC. Think USB would be too fast for the bandwidth of this device.
Anyway what you'd have to do is generate a mere second or two worth of audio samples in MQN and repeat them all back to back continously but the trick would be to generate a trigger signal that your scope (with the LNA attached) could lock onto. That is the really difficult bit, piece of cake for an FPGA to do but on a PC, I'm not so sure it can be done. With this repeating pattern of samples you can zoom in on the scope to your hearts content and look for tones in the noise.

I have to say I'm pretty sceptical that different MQN builds are changing the noise out on the DAC but this might be one way.

wish there wasn't any difference.

the only question you need to answer is do different o/s and different players all sound the same ? Most people's experience is that they don't. That should then lead to further questions.

The difficult bit for me is that everyone is using motherboards of different types, layouts, PCB design techniques so that even if MQN reduces the noise floor in some instances it probably would have no affect at all in another situation with another board. But we dont know the nature of this noise. I think the method I suggested above is a very valid. Here at work, if we were to look at low noise this is probably how we'd do it.

A pity that we have to use a PC at all as its far too clunky and bloated for the simple task of sending audio bits to a DAC.

Haven't had anyone that's tried it say there isn't a difference

In theory it's a simple task, but the implementation is not good enough to fool the ear.

Don't think I have the equipment to follow your testing method.

[jkeny]

Interesting, Dave - pity it's a grand - costly for one experiment.
Do you know of anywhere it could be hired from?

I believe from all our discussions we are left with two possibilities to explain audible differences between MQN (given that the datastream is still bit perfect) - noise or timing (or both combined).
My rudimentary attempts at timing measurements using IQ-Test didn't show up anything obvious but I'm not fully convinced about the validity of the test (or maybe how I was using it). It's still an area that might be worth revisiting, particularly now that MQN can handle high-res.
So noise is the other possibility unless you have another suggestion?

One interesting piece of info - just saw on XXHighend http://www.phasure.com/index.php?topic=2994.0;all that PeterSt is claiming the upgrade to his NOS1 DAC is now immune to anything happening on the PC (including software, yes his own software XXHighend is redundant apart from the filters) including USB cables. If true, then the details of the upgrade are of interest. Not surprisingly the upgrade details are somewhat sketchy but from what I can gather one interesting change which is possibly of significance is that the USB receiver clock (normally 12MHz) in the NOS1A is slaved to the audio clock(s). The other piece he states (although English is not his first language, so he's difficult to understand) is that his unique way of doing galvanic isolation introduces no additional jitter. His measurements suggest that the very low & steady jitter figures are the evidence of this immunity

[DaveF]

Here's another LNA that's half the cost. Less gain but probably enough.
http://www.aaronia.com/products/spectru ... BBV-NF-25/

I'll see if I can get a contact number for a rep dealing with lab equipment.

On the subject of bit perfectness, have we test results showing that it is and how was it done? (At the DAC inputs)

[jkeny]

Here's another LNA that's half the cost. Less gain but probably enough.
http://www.aaronia.com/products/spectru ... BBV-NF-25/

Ah, right - I didn't look through their range - will do now

I'll see if I can get a contact number for a rep dealing with lab equipment.

Very cool, thanks

On the subject of bit perfectness, have we test results showing that it is and how was it done? (At the DAC inputs)

I don't think we have explicitly tested bit-perfectness as such - really it's a given really considering what Gordon's software is doing. Maybe Gordon has checked this. But you are correct that any testing should establish this fundamental first. It's relatively easy to do this - play a HDCD file through MQN into a HDCD compatible DAC which has a HDCD indicator. If the HDCD indicator lights we have a bit perfect stream (the HDCD indicator is a single bit indicator, AFAIR)