When do measurements correlate with subjective impressions

[Audiophile Neuroscience]

12 hours ago, cat6man said:

 

I have to disagree with much of the above.

 

1.  i have spent 30+ years simulating complex digital telecommunications systems and, while often causing much consternation among my colleagues over this time (everyone like easy answers), i have found that MOST simulations oversimplify the problem and answer a question or problem, but via oversimplification or by not incorporating something (known or unknown), they do not solve THE question or problem.

do you have a realistic jitter spectrum?  what is the PDF (probability density function of the jitter)? what is the reconstruction filter?  it is technically obvious from the math (don't panic, not included here) that the combination and interaction of the jitter with the reconstruction filter are intimately related, so how is this handled?  some filters are not, in fact, even linear as some have lookup tables for filter coefficients that are data sequence dependent.

 

you may use your program to see if your particular waveform and jitter stimulus is/isn't audible to an individual (and there is value in that) but abstracting that to DACs in general is highly questionable.

 

2.  "Jitter is a fairly simple concept"?  sorry, but that just isn't correct.  jitter has a spectrum, it has a pdf, it may have various forms of correlation, the clock may pick up noise of various sorts which could be random, impulsive, RFI, etc.

 

i will try out your program, assuming it runs under WINE, as i'm curious to see what i can/cannot hear but i know enough about simulations to be suspicious of claiming relevance of a level below which jitter not not matter.

 

let's try a gedanken experiment.  assume i have a 50ps rms jitter clock and there is an impulsive noise (power line spike for example......A/C is pretty crappy) impacting that low level signal 1% of the time and causing the clock off instantaneously by 10x the RMS jitter (500ps).  i'm guessing the average RMS jitter measurement would not change at all but that the DAC would see a 10x timing offset 410 times a second (i.e. 1% of the time)...........maybe that could be audible?  i don't know but it certainly seems possible and something that would likely (?) not be measured with averaging turned on.

 

[technical analogy--feel free to skip]

an example i know intimately from simulation:  in a 4G LTE system, high speed data is controlled by a 'scheduler' that assigns time slots at the mobile and base station to different users.  in addition to assigning time slots to each mobile user, the scheduler must assign from a limited set of control channel slots in order to tell each mobile when its data time slot is coming up.  if there are not enough control channel time slots available to serve all the mobile users at a specific time, that mobile user is temporarily blocked and cannot send or receive data.

the metric commonly used in simulations was average control channel utilization, and it was thought that keeping that value below 70-80% on average was sufficient for good system operation.  however, simulation of many mobile data users with many different data usage profiles showed interesting results.  even with say 50% control channel utilization, the blocking could be as high as 5-20% as the distribution of the number of instantanous users could be highly skewed.  therefore, the solution turned out to be much more complex than expected as the tail of the distribution of #users needing control channel slots dominated overall performance and the average utilization was essentially irrelevant.

just a little non-audio example to show you where i'm coming from.

there is always a simple answer to complex problems, but it will often be wrong.

unfortunately, many people (and upper management) prefer a simple answer to a complex "well it depends" answer.

 

as always, YMMV

 

IMHO a refreshing way of looking at things.

 

It has been said  "to the simple everything appears simple" which is a pejorative way of perhaps explaining why most/many of us just want answers expediently and especially in the neat form of a black and white test result .

 

A better way of expressing it may be " Everything should be made as simple as possible, but not simpler".

[opus101]

The trouble with statements like that is that they're either meaningless or else self-contradictory when examined. If its not possible to make something any simpler, why the prohibition against it? So I think the word 'possible' needs replacing with something like 'reasonably practicable' to rescue it. Or alternatively, simply don't examine it at all.😉

[Audiophile Neuroscience]

I think " Everything should be made as simple as possible, but not simpler" (Einstein? but could be apocryphal) is a great goal but of course there is "reasonably practicable". In either case it is difficult to agree (or know?) when the goal has been reached. happy to continue discussing elsewhere but not here as I think it would move off topic.

Cheers

David

[jabbr]

9 hours ago, opus101 said:

The trouble with statements like that is that they're either meaningless or else self-contradictory when examined.

Occam’s Razor

 

Apropos to what we are doing is that things like Spice capture the components but not the parasitics. 3D SIV software is better but expensive and needs the actual board layout which is not available to third parties, 

 

Regarding “injecting” jitter into a file, this makes some big assumptions because jitter is a time phenomenon and files are not... in the same way that SPICE makes assumptions that board layout is not significant, the in accuracies of Meeks relate to assumptions. The other saying is “Don’t ASSUME ...” 😉

[pkane2001]

1 hour ago, jabbr said:

Regarding “injecting” jitter into a file, this makes some big assumptions because jitter is a time phenomenon and files are not... 😉

 

A PCM file is voltage sampled over time... so it's not a time phenomenon?

 

[jabbr]

1 hour ago, pkane2001 said:

 

A PCM file is voltage sampled over time... so it's not a time phenomenon?

 

Well it’s jitter that’s first baked and then frozen, so it might taste just like fresh jitter but might not, depending ...

[pkane2001]

1 hour ago, jabbr said:

Well it’s jitter that’s first baked and then frozen, so it might taste just like fresh jitter but might not, depending ...

 

I prefer my jitter fried...

 

[cat6man]

fyi for linux folks.............i installed DISTORT under Crossover (supported version of wine) but i could not get all parts of it working (didn't see plots, for example).  installation automatically installed .NET framework 4.5

 

when i installed .NET framework 4.6.2, i started to see the plots and was able to save a WAV file after going through distortion

[cat6man]

17 hours ago, opus101 said:

The trouble with statements like that is that they're either meaningless or else self-contradictory when examined. If its not possible to make something any simpler, why the prohibition against it? So I think the word 'possible' needs replacing with something like 'reasonably practicable' to rescue it. Or alternatively, simply don't examine it at all.😉

 

a counter example to show how wrong this is:

another telecommunications example (sorry)

in a data system, i want to measure the data throughput when downloading from the web.

i can model all of the downlink and show its capacity from the physical layer, frame sizes, signal to noise ratio, etc.

 

however, the downlink rate depends on tcp/ip (and RLC as well on a cellular wireless network) which is a bi-directional protocol.  if i have lots of errors or lots of latency in the uplink, it will throttle the downlink.

so simplifying the simulation is certainly possible and will give you an answer, but the wrong one in my opinion. (just ask anyone with a cable modem with a oversubscribed shared uplink!)

 

so, it is absurd to say that oversimplification is meaningless or self-contradictory.

oversimplification is rampant, both intentional (e.g. the real world is too complex so we have to, or my boss wants a yes/no answer so i have to) and unintentional (we're all a lot smarter today than we were before, and likely to be a lot smarter in the future)

 

by the way, i have no intention of starting (or participating in) a food fight or philosophical discussion.

i'd like to see this thread stay technical and help solve problems.

 

as always, YMMV

[cat6man]

and to lighten the mood, check this out.

particularly relevant part starts at 2:18

 

 

[Superdad]

1 hour ago, cat6man said:

and to lighten the mood, check this out.

particularly relevant part starts at 2:18

 

Yes, Placebo Domingo is one of my favorite opera singers. He fakes me out every time. 

[opus101]

2 hours ago, cat6man said:

so, it is absurd to say that oversimplification is meaningless or self-contradictory.

 

 

Seems then you misunderstood my point. I wasn't in any way trying to say there's no such thing as oversimplification (I am, like you convinced there definitely is), I was merely pointing out the internal self-contradiction in the expression.

[pkane2001]

5 hours ago, cat6man said:

fyi for linux folks.............i installed DISTORT under Crossover (supported version of wine) but i could not get all parts of it working (didn't see plots, for example).  installation automatically installed .NET framework 4.5

 

when i installed .NET framework 4.6.2, i started to see the plots and was able to save a WAV file after going through distortion

 

WINE 5 worked for me under Ubuntu 18. It installed the correct .NET framework automatically with DISTORT and all functions seemed to work after that.

[cat6man]

14 hours ago, pkane2001 said:

 

WINE 5 worked for me under Ubuntu 18. It installed the correct .NET framework automatically with DISTORT and all functions seemed to work after that.

 

thanks.........i was still running WINE 4 when I had to change the .NET framework.

[Audiophile Neuroscience]

On 3/8/2020 at 11:42 AM, Superdad said:

 

Yes, Placebo Domingo is one of my favorite opera singers. He fakes me out every time. 

 

Yes, and his evil twin Nocebo Damingo !

[ferenc]

On 2/28/2020 at 8:01 PM, Shimei said:

 

Likewise, has anyone ever sat above a dozen violins on the balcony of an indoor concerto? No electronics, that is, no speakers etc. Just percussion, horns, stringed instruments and pure vocals. Or an Opera where the singer turns and faces another direction? 

 

Has anyone ever heard a man made system that could reproduce the reality of that kinda soundstage? The "all that exists here now" or "is" in the moment? Has any technology "achieved" or "surpassed" that kinda standard?

 

Are we living in the Matrix? 🤣

 

In all seriousness, doesn't the reference or standard bearer become the very source in which everything else is measured? If so, then very observer/listener themselves may become subject to distortion or not perfectly "reflecting" back the source. 

 

Actually to enhance our senses and to educate our brains, regarding recordings and mics and music production technology, I organized the so called My Reel Club events, where you can go to the studio and sit just meters away from the musicians, instruments while the recording is made live and without any post-production, processing, practically only a mixing session on the recorded tracks, if we use more than two mics.

 

It is very different experience than being on a concert and sitting far from the stage. The studios rather small, there is usually no way to squeeze more than 20-30 people so it is comparable to the situation when the musicians would visit you at home and play just for you.

 

After the recording, mixing completed, you get the files (in DSD256) and can listen at home. Usually we measure the loudness and dynamics as well of the event so you can set your volume up to the same level if you want. 

 

This way you can understand not only what a direct instrument sound means in the studio, but the whole production chain as well, how many different ways a post processed commercial recording is changing the original instrument sound, dynamics, tonality, etc. 

 

Let me share a sample, it is 3:39 min long ( 800 MB size) from the Gabor Varga Quartet, cool jazz music recorded in December. There is no editing or any kind of digital processing in this track, it is only mixed using a large SSL analog console. The peak of this track was 92-93 dB, roughly 4m from the drums in the middle of the audience.

 

https://we.tl/t-V2UIAnM6L0

 

 

[audiobomber]

A couple of excerpts regarding measurements:

the "Power" that sites like ASR are gaining is really concerning and detrimental to the end sound quality IME.

The trouble with the guys on ASR is that they ONLY judge a products performance via "simplistic" technical measurements and IME this does a massive disservice to the general HiFi community - especially those who are scared to listen for themselves...

 

As far as I'm concerned, in the case above, if you want "Best measurements", then buy the cheaper unit and be Happy and "Laugh" at the rest of us  but DON"T dare say the new design sounds worst because in some areas it has poorer measured performance...

You will be surprised how many (it not all) designer I know will privately agree that standard "static" measurements have little refection on sound quality (try as they might to measure and "quantify" sound quality) - but just as in so many walks of life are too afraid to have there voices counted... 

 

https://www.diyaudio.com/forums/digital-line-level/331363-ak4499eq-dac-25.html#post6130649

 

JohnW is John Westlake, a renowned designer. He developed my last two DAC's and both punched above their weights, as do all his products. 

https://en.wikipedia.org/wiki/John_A._Westlake

 

 

[manueljenkin]

Intellectual meme.

[pkane2001]

Just came across this video by two researches, Steve Temme, and Sean Olive, on correlating objective measurements (THD, IMD, NCD) and subjective preferences for a number of headphones. This is a discussion of an AES published paper results covering this topic:

 

 

 

The published paper: http://www.aes.org/e-lib/browse.cfm?elib=14232

 

[John Dyson]

On 2/20/2020 at 10:32 AM, jabbr said:

You are really trying to get me to install Windows aren’t you?

Or, bring up a tempoary version under VMware or under one of the emulators built-in on Linux (using gnome.)  In emergencies, I tend to use VMware instead of the built-in tools, but is limited to non-commercial use.   It is easy to run windows under an emulator...  I normally do not do it -- and use a laptop for Windows.

I forget if audio works under VMware, but seem to remember that it just might.

John

 

[lucretius]

On 7/14/2020 at 10:24 AM, John Dyson said:

Or, bring up a tempoary version under VMware or under one of the emulators built-in on Linux (using gnome.)  In emergencies, I tend to use VMware instead of the built-in tools, but is limited to non-commercial use.   It is easy to run windows under an emulator...  I normally do not do it -- and use a laptop for Windows.

I forget if audio works under VMware, but seem to remember that it just might.

John

 

 

Yes, audio works under VMware but there are limitations -- the 2 limitations I can think of off hand are:

1) You are limited by the capabilities of the "virtual" sound device. (You can forget about playing DSD!)

2) I cannot get exclusive mode to work inside a VM. This creates a lot of problems (e.g. changing sample rates and bit depths, etc.).

 

 

 

[lucretius]

2 hours ago, lucretius said:

 

Yes, audio works under VMware but there are limitations -- the 2 limitations I can think of off hand are:

1) You are limited by the capabilities of the "virtual" sound device. (You can forget about playing DSD!)

2) I cannot get exclusive mode to work inside a VM. This creates a lot of problems (e.g. changing sample rates and bit depths, etc.).

 

 

 

 

In theory, one should be able to connect a USB DAC directly to a VMware VM.  I had very limited success with that -- it only worked for lower sample rates and would sometimes cut-out -- it choked on high sample rates.

[pkane2001]

2 hours ago, lucretius said:

 

Yes, audio works under VMware but there are limitations -- the 2 limitations I can think of off hand are:

1) You are limited by the capabilities of the "virtual" sound device. (You can forget about playing DSD!)

2) I cannot get exclusive mode to work inside a VM. This creates a lot of problems (e.g. changing sample rates and bit depths, etc.).

 

 

 

 

I use VMWare with a USB DAC connected to the host computer (iMac in my case). No problems! When connecting the DAC to the Mac, I get a prompt from VMWare asking if I want to connect it to MacOS or Windows. Connecting it to Windows makes it work just fine with WASAPI and ASIO drivers installed on the VM. That's how I do most of my development.

 

[lucretius]

1 minute ago, pkane2001 said:

 

I use VMWare with a USB DAC connected to the host computer (iMac in my case). No problems! When connecting the DAC to the Mac, I get a prompt from VMWare asking if I want to connect it to MacOS or Windows. Connecting it to Windows makes it work just fine with WASAPI and ASIO drivers installed on the VM. That's how I do most of my development.

 

 

My experience was with a Linux guest. I don't have any Windows guests. Nonetheless, I am glad to hear it will work with a Windows guest  Does it work in exclusive mode?

[pkane2001]

3 minutes ago, lucretius said:

 

My experience was with a Linux guest. I don't have any Windows guests. Nonetheless, I am glad to hear it will work with a Windows guest  Does it work in exclusive mode?

 

Yes. The USB device is mapped directly to VMWare, MacOS doesn't see it or control it in this case. The native DAC driver inside Windows VM sees this as a proper type USB device connected to Windows. I've even done firmware updates this way on one of the DACs 😱