MQN testing/experimentation thread

[DaveF]

That would be an interesting check, though I do think this "bit perfect" measure is nothing to get too hung up on.

For me, its quite an important check to make and a critical one.

If the output of any electrical system has changed then something back along the chain of components has been altered to cause this affect. Be it either in the analog, digital or software domains. One of these has changed.

If the differences in sound are real then the output of the Amp or Dac has changed. I dont think anyone could disagree with that.

So lets work our way back up the chain to see where and what can be tested.

The easiest to test is the input to the dac as we have direct access to the digital data going in.
Choose 2 or 3 versions of MQN that people believe sound different. Run them and each time look at the data going into the DAC. If all 3 match, then each version of MQN is bit equivalent which implies that everything else back up the chain is doing the same thing for each version of MQN. If they are different then the data is being altered even if unintentional.

If all versions of MQN are bit equivalent and people still believe that they sound different then the next area to check is analog output of the DAC and into the amp. This is a little more difficult but still doable.

[Aleg]

DaveF

I think this is an oversimplification, based on a viewpoint that digital is just 'bits'.

You assume that it is just 'data' that is entering a DAC, but that's not true on a physical level.
On a physical level it is a electrical signal wholy in the analog domain, with all types of noise riding on the signal, with EMI/RFI influences, with common mode disturbances, with reflections due to impedance mismatches, causing timing effects on the signals, creating disturbances on sensitive micro-electronics.

Despite all this several layers up in the OSI-model the signal is still interpreted properly into logical 1's and 0's, but all the influences on the physical level have not just disappeared. They have done something unless the engineers have made sure they have no effect, which they can't do for everything without causing other effects.

There is no 'ideal digital' logical transmission.

Cheers

Aleg

[DaveF]

I think this is an oversimplification, based on a viewpoint that digital is just 'bits'.

You assume that it is just 'data' that is entering a DAC, but that's not true on a physical level.

No I don't assume anything of the sort.

What I proposed above is to determine if the content of the digital samples is being altered by the software as a result of whatever algorithm the coder is using. Its a valid argument and test. It hasnt been done and therefore it still remains as one possibility for differences heard.

On a physical level it is a electrical signal wholy in the analog domain, with all types of noise riding on the signal, with EMI/RFI influences, with common mode disturbances, with reflections due to impedance mismatches, causing timing effects on the signals, creating disturbances on sensitive micro-electronics.

Yes I know.

Do you believe that MQN software changes which is several layers above other layers is changing the electrical noise, EMI or impedence within the electronics?

In the vast array of embedded systems I've had experience in, never has software changed anything electrically unless its job was to do just that via the hardware below it.

But without knowing the exact hardware below, its difficult to say for sure.

[LowOrbit]

That would be an interesting check, though I do think this "bit perfect" measure is nothing to get too hung up on.

For me, its quite an important check to make and a critical one.

If the output of any electrical system has changed then something back along the chain of components has been altered to cause this affect. Be it either in the analog, digital or software domains. One of these has changed.

If the differences in sound are real then the output of the Amp or Dac has changed. I dont think anyone could disagree with that.

So lets work our way back up the chain to see where and what can be tested.

quote]

From a logical and investigative point of view I agree.

My point is that bit perfect refers only to there being or not being any intentional manipulation of the samples (DSP, resampling etc). Any unintentional changes would be random (relative to the sample stream) and perceived as noise (distortion). I've seen or heard nothing which indicates and intentional changes are being made within MQn.

I don't believe we should be thinking beyond the PC interface level because people have differing dacs and systems and are reporting fairly coherent differences.

[LowOrbit]

My take on what MQn is doing - by whatever mechanism, intentional or otherwise - is that thi8s software is enabling the computer to output a sample stream that is "cleaner"/better aligned and is therefore allowing the DAC an easier job in converting this to an analogue signal. That cleaner signal results in less masking of detail.

The challenge is figuring out how it's enabling that, given - as Dave has rightly pointed out - the player is abstracted several layers up the process foodchain in the PC.

It would be good to try and list all the processes/dependencies inside the computer that are involved in delivering that sample stream to the outside world.

[Aleg]

No I don't assume anything of the sort.

Phew, relieved to hear that :-)

...
Do you believe that MQN software changes which is several layers above other layers is changing the electrical noise, EMI or impedence within the electronics?
....

No/yes, I don't expect it to reduce what is already there, but it might just be not adding any additional disturbance by reducing the amount of CPU and RAM-processing. General purpose PC's are not designed to be low noise devices. It is hardly relevant for ordinary usage.

Cheers

Aleg

[sbgk]

did anyone try the rax test 128 version in the test folder, might give you an insight into where the noise is being generated.

[DaveF]

No/yes, I don't expect it to reduce what is already there, but it might just be not adding any additional disturbance by reducing the amount of CPU and RAM-processing. General purpose PC's are not designed to be low noise devices. It is hardly relevant for ordinary usage.

This would largely depend on the layout of the PCB. Even if there was reduced activity on an external memory bus, it wont mean that less noise here will couple onto other components related to audio. It might through poor PCB layout and how close they were together. Since many people report the same differences, yet are running different PCs/laptops, some with external dac etc, I find it highly implausible that a reduction in noise is key here.
If we were all running a fixed software/hardware plaform then maybe.

If the processor was being asking to do something very mathematically intensive such as FFT calculations with lots of continuous accesses to memory then a power/noise fluctuations would be observed alright.

[sbgk]

If the processor was being asking to do something very mathematically intensive such as FFT calculations with lots of continuous accesses to memory then a power/noise fluctuations would be observed alright.

Whats your view on streaming data ? I thought it would involve lots of continuous reads/writes from/to memory.

Anyway, listen to the rax test 128 version, it gives an audible queue as to when/where the noise is being generated.

[LowOrbit]

If the processor was being asking to do something very mathematically intensive such as FFT calculations with lots of continuous accesses to memory then a power/noise fluctuations would be observed alright.

Whats your view on streaming data ? I thought it would involve lots of continuous reads/writes from/to memory.

Anyway, listen to the rax test 128 version, it gives an audible queue as to when/where the noise is being generated.

I just had a quick listen. Didn't tell me much except noise is generated at - what I would gather - is buffer refresh. Would you class that as an exaggerated example or simply what happens when you get it really wrong?

Mark