MQN testing/experimentation thread

[jkeny]

While waiting for the full set of results files to upload in google drive - here's the two sets of data files from the above MQN test
https://www.dropbox.com/sh/q82z5qc1bgvqkqe/6O6kPR37q_

There are only 3 files in each set that are of interest - open one of them in Veusz & the plot will be generated.
The names each begin with:
- load_iqhfspectrum - this is the HF plots I have mostly posted here
- load_iqreport - this will give a plot o low frequency timing errors plotted over time rather than frequency
- load_iqspectrum - this will give a plot of LF timing errors but in parts per billion i.e nanoseconds (yes 1000,000,000th of a second)

The other files are CSV files of the data which have been generated by the analysis program which is run on the recorded output file. They can be opened in a spreadsheet & show timing & errors but understanding the details requires understanding the program, I believe.

[jkeny]

Why mqn sd ? I think sq is the best and others report sw as being good, would be interesting to know how these fair in comparison and whether sq does have better timing.

Oh, I haven't kept up with the latest favourite, Gordon - I just picked what I last heard was the best.
I will do a recording of sq & sw now but I expect the plotted differences to be even more subtle than MQN Vs Foobar

Also, remember all these tests are being done with a bog-standard laptop, not an tweaked audio PC

Edit: I don't see a 1024-sq version just sd,sb,sw(all dated 6th Nov) - did you mean sb?

[jkeny]

I've done the recording & plots but they are not worth posting up as they show no real differences - as I suspected
But remember, I'm using a bog standard laptop, I also didn't have a Ciunas DAC available for this one so straight from laptop analogue outs.
I'm sure the subtlety of any differences that might be seen on the plots will be masked by the less than stellar hardware I'm using.

What would be interesting would be a plot of the before & after clock changes that are being talked about on the MQN thread.
I can't do that with this AMD CPU laptop.

Maybe with the Intel NUC but all this takes time - hence the call for volunteers!!

[jkeny]

If people consider this testing worthwhile then let's make it a group effort!!

It doesn't have to be restricted to MQN testing - this technique can reveal timing errors in playback resulting from hardware, PS, registry tweaks, cables, etc. It really has great value - reporting timing errors in microseconds without expensive measuring equipment. We just need to verify it's veracity & this can only be done with a body of tests across different hardware (& the harder problem - software).

So who's on for this group effort??

What is needed:
Those who wish to do recordings
- a recording device - this could be a line in on a soundcard or an external recorder?
- a bit of time & effort in playing back a 3 minute file (I can supply these) & recording it's output
- processing this in an analysis program which takes seconds

There is also a need for volunteers to look at plots to try to spot differences/trends - what's needed:
- a download of plotting program & the files to be plotted
- a bit of time running through some of the various plots which can be done


There is a further need for some C programmers to try to tease out the large spikes that are present on the plots. I can assist in this but haven't the time to analyse the program. It's not a big deal as the plots around these big spikes can be ignored but still it would be great to get these spikes out of the picture. I have the sources for these programs - there's only 2 - a file generation program & an analysis program.

So what do you think guys? This would be a great joint effort which may, if successful, give us a great timing differences handle on what we are doing with tweaks, PS, etc.

It would also garner wider interest & attention for this group & for this forum :)

[jkeny]

OK, as promised here's the link to the results folder https://drive.google.com/folderview?id= ... sp=sharing

It contains the following folders:
- Sanity Check - results of running the input file directly into the analysis program

- Zoom recordings (recording to the Zoom recorder):
- direct from laptop (16/44)
- through Ciunas DAC (16/44 & some 24/96)

[Diapason]

Thanks John, I'll try to have a look at this tomorrow if the opportunity arises.

[jkeny]

Thanks John, I'll try to have a look at this tomorrow if the opportunity arises.

Great!

[jkeny]

I said that there were no tweaks to the laptop but I forgot about turning off the high-precision timer in Device Manager
So all the previous plots were done with timer off.
Now here's a Foobar pair (direct from laptop) - one with timer turned off & one timer on

high precision timer on



high precision timer off


Am I imagining it or are the fluctuations less severe with HPet off?

[jkeny]

BTW, a good 2010 paper on "Mechanisms of Sound Localization in Mammals" http://physrev.physiology.org/content/9 ... l.pdf+html

Edit: It appears that the modern thinking on lateral sound localisation is that timing (ITD) is not restricted to lower frequencies & loudness/intensity (ILD or IID) only operational at higher frequencies. This dual model has been superseded & ITD now is used at HF also.

Also a nice explanation for how we localise sound partiulalry in the vertical plane

Cues for sound localization. A: spectral analysis as a monaural cue for sound localization in the vertical plane. Interaction of a broadband sound with the head and, primarily, the outer ear alters the effective spectrum of the sound impinging on the eardrum in a manner dependent on the location of the sound source in the vertical plane. Most prominently, the central notch (black arrows) in the effective spectrum of the sound shifts to higher frequencies when the sound source is shifted from below (−45°, green) to above the horizon (+45°, red). The spectra shown are digitally computed from 1-s white noise stimuli (cutoff frequency, 44.1 kHz) convoluted with KEMAR head-related impulse responses and averaged over 50 repetitions. B: interaural time differences (ITD): the difference in the arrival time (Δt) of a sound wave (gray lines) at the two ears is used to localize a sound source in the horizontal plane. For frequencies below ∼2 kHz, information in the fine structure of the sounds is available for ITD processing. C: interaural level differences (ILD): for frequencies higher than ∼2 kHz, the shadowing effect of the head creates increasingly sizable differences in the intensity of the sounds at the two ears (ΔI) that are utilized for sound localization in the horizontal plane.

The underlined text was what I was working from plus other reports but current research says this is incorrect

[LowOrbit]

John

I'll happily join in the testing, not sure if I'll get a chance this week though.

Mark