Exporting at 192kHz

Thanks for the reply, Joe

When it comes to audio files, I fully agree with you. However, with a software instrument there is, theoretically, some logic behind the idea that an increased sample rate would result in an audible difference, as 192khz  might allow the sound-engine to process overtones more than two octaves higher than the standard 44.1.

As I have Standard, as opposed to Pro, I can’t check for myself - this is why I was hoping someone might have done a comparison.

Piano doesn't have overtones two octaves higher than 44.1 KHz.

There is a reason why mastering is done at 192khz.

48k is perfectly enough for time resolution considering the piano spectrum on the whole. I have Pro and I hear absolutely no difference between 48k, 96k and 192k. I do see increased CPU usage, but when there's no benefit, why even go there?

Not everywhere and not by everyone. Also when majority of mastering plugins are using oversampling, it's not even really necessary... Only for archival purposes, I'd assume, but even then 96k suffices.

Also, an interesting read:

https://xiph.org/~xiphmont/demo/neil-young.html

Of course I do, time is what sample rate discretizes. I've studied DSP in college.

In any case, almost any input signal can be discretized at 48k and recreated back into analog realm perfectly, with today's precise DACs. Even those that have a lot of high frequency content, like cymbals etc.

If there's stuff happening between two samples at 48k, you're dealing with a signal that has harmonics above 24 kHz in its spectrum. This is not the case with piano. So, regarding time resolution, 48k is enough, as I said, and higher sample rates bring no real benefits, as far as I'm concerned. I am not sure what you meant by "not accounting for time resolution". That is EXACTLY what sample rate accounts for.

Why don't you stop being vague then and say what you mean? I am pretty sure I know what I'm talking about.

http://www.skylondaworks.com/sc_resol.htm

Time resolution. By this description, it is pretty much the same thing as sample rate. They even mention Nyquist's theorem there.

By your own words even you're not an "expert in the field". So really... stop being vague and say what you mean, then?

192 KHz may be imperceptible when played 'straight-up' as compared to a similar 44.1 KHz recording, but the 'richer' recording allows for more post-processing, for anyone who wants to do such stuff in their DAW.

As a 'visual' analogy, think of the digital photography world:  36 MP files don't look any different than 6 MP files when printed as 4"x5" photos, but 36 MP files allow a large opportunity to crop sections of the frame, yet still print those crops without pixelization.  Similarly, while the eye can't perceive the subtleties in 24 bit or even 16 bit color-spaces as compared to 8 bit or 12 bit color, the eye can certainly see the excessive posterization that happens when you try to do too much 'massaging' of the 8- or 12-bit image with exposure adjustments.

In summary, more information allows more manipulation of the information, while still having imperceptibly smooth results.

- David

So, it’s a no, then?

Thanks for all the replies, folks!

The thing is, Pianoteq is not going to generate more piano harmonics when you work at 192k, because they don't exist in the real thing either.

I never said 192k is a marketing thing. However that it can be a liability on great majority of playback devices, it can, and the article I linked to proves that.

Some time ago, apparently for version 6.0, I tried to find what kind of winnings gives a high number of discretizations. I exported the sound in two versions and found some differences. It is obvious that no extra frequencies are added. The timbre does not change. But with a greater sampling frequency, the processor certainly makes more calculations. Those. more calculations per unit time are obtained. Accordingly, it can be assumed that a larger number of calculations can lead to a more subtle result. It can be assumed that the greater the difference, the greater the calculation - i.e. The differences will be more noticeable with a long sound(When differences can accumulate). Which, in principle, I managed to notice. It is interesting that afterwards, after reconverting the WAV format 192000Hz and 44100Hz in MP3, these differences were preserved. The difference is small - some deterioration in the integrity of the picture in the interaction of overtones. Since then, Pianoteq updated the engine responsible for the spatiality of the room and I do not know how things are with this now.

Hi, I have experienced - and still do - very pronounced differences with playback in different sample rates, RedBook CD vs.HighRes SDM vs. DSD, reproduceably, beyond any placebo effects, despite of my old ears.

But at the same time the other arguments are just as valid - no inherent differences. And I have a good fully rational explanation for this paradox, which gives a good scope for consideration and decisions to be based on it. 

My personal attitude is that a sample rate about 44.1 doesn't matter much but the bit depth 16/24 can make a huge audible difference. But this is not the whole story.   

All SIGMA-DELTA DACs (most we use) are a HW/SW package, and their firmware do intensive processing on their SDM input (like 44.1KhZ/16bit) to achieve a DSD-like bitstream conversion to be able to make an analog physical output. The resources of the DAC HW is mostly very limited in comparison to an 8-core computer system, so their processors, algorithms, filters are also inherently limited. Most of them maintain a sweet input format, which would be processed both effectively and with the best quality while other inputs might get slightly deteriorated.

My DSD-capable DAC can reproduce the same music markedly best when played back the DSD recording format. And interestingly, it gets not that hot while playing back. I was surprised, how a good playback SW (HQPlayer) can reproduce classic CDs when letting it to process and convert it to DSD before output to the DAC.

So I use Pianoteq with the foo_dsd_asio ASIO proxy (will be displayed as a normal ASIO driver) which converts the PT output to DSD format with  the Bitrate of 352800 and that makes a much more livid musical experience. There is still a little audible difference with different Audio sample rate, but when I chose 1764000 (for
then the CPU would limit a cracking-free 5ms latency playback (for some reason 176400 is the maximum where foo_dsd_asio my DAC do make the automatic conversion). I can imagine, the higher internal bitrate, the less quantization errors in the outputs. But even these are giving much less variations than the above DSD conversion itself. I would say, that quality improvement with DSD playback does make these subtle quality variations just percievable at all.

I would say the audio improvement I could achieve this way with DSD was in the range of what we usually get with a new PTQ new version number. (An this was additionally, outside of Pianoteq - but which Pianoteq supports ).

I would suggest if You have a DSD-capable DAC it is absolutely worth trying. I am also very curious about Your experience, because this all is very system-specific.

(My system is nothing of a high end: DAC Light Harmonic GOV2 Infinity, the first version, now no more than some 300$ , Sennheiser HD650 or Dynaudio boxes, i7 4 Ys. old desktop PC with WIN10...)

Thank you but before i look at this can you confirm if you started with Pianteq running at internal sample rate of 192khz please?

Thanks for this effort Mr. scherbakov

Well, these mp3 tails sound different.

I listened to these .wav files at least a dozen times using my RME Babyface Pro to swap playback from 44,1 to 192 KHz. Used both speakers and headphones. The 192 KHz file sounds smoother, fuller and more natural to my ears from beginning to end. No contest.

Perhaps that could be the PianoTeq software and/or the BabyFace conversion. I am thinking there is no sample rate conversion going on inside PianoTeq and the 192KHz mode naturally has more refined processing/mapping. But these are both black boxes so who knows.

Maybe Modartt should provide some sort of trial version of PianoTeq Pro so users can noodle around with the 192KHz mode.

I'm with you. Just one experience I had recently:

At the moment I'm using my Intel N4200 Laptop (fanless), which has a subpar Audiocodec ALC255 with just 44100 and 48000 Hz PCM.

On my some years older Laptop with its better ALC283 I always used 48000 Hz with 64 samples. The ALC255 needs a minimum of 128 samples to work stable at this rate. Believe it or not - with 128 samples the system feels more sluggish, than with 64 samples. I don't like that and don't want to adapt.

So I decided to downsize to 44100 Hz to be able to use 64 samples again. That's my setup to test the new v6.2.x releases at the moment. Latency excellent again. Pianoteq dito.

But the point is: With 44100 Hz it sounds different to 48000 Hz!

If "better" or "worse" is in the ear of the listener, but that it sounds different at all is funny. From Nyquist's point of view, it shouldn't - but it does. You might say "placebo!", but as long as a Laptop is a blackbox it must be allowed to say "difference!"

cheers

What I can't quite work out is why there's so much of a difference in the dynamics of the two examples (44.1 in green, 192 in blue). Note for example the tail being visible for at least 3 more seconds in the 192khz version (The green line disappearing around 33 sec leaving just the blue line (192khz version) visible till just past the 36 second mark):

Is there maybe some random number generation in the modelling of the sound production in Pianoteq so you wouldn't be able to get the same result twice ?

Well if that's the case doesn't that make direct comparisons in the minutiae of the audible difference between these high sample-rates near impossible ?!

One explanation (apart from Pianoteq modeling being better at 192 kHz internal) is the quality of up and downsampling on the DAC used for listening. Maybe some modern chips work better at 192 kHz and downsample to lower sample rate with some errors added.

One site of interest is this free site having the same recording in a variety of styles, formats and sample rates: http://www.2l.no/hires/

If you hear a better result from these recordings at 192kHz than lower, maybe the DAC is to blame...or you have exceptional ears.

Interestingly I came across a Sound On Sound article a few days ago that linked to a white paper written by someone well above my mental capacity (This guy: http://www.lavryengineering.com/about.html ) that actually states the counter argument that we'd be worse off at 192khz using current technologies: http://lavryengineering.com/pdfs/lavry-...theory.pdf

I generated 2 files for comparison 44100 and 192000 Hz. I used Reaper, Pianoteq Steingraeber (Recording 3 by default) and binauralizer from Noisemakers:

44 100 Hz

192 000 Hz

video:

https://youtu.be/159USGOD5hk

If you listen to a comparison from a device that can switch the mode of the audio card to the resolution of the file, then you can hear some differences. (for example Audirvana or Neutron player for Android(Yes, yes .. my android supports 192,000 Hz.))

What differences do you find?

Upd^( I want to note that Pianoteq, while exporting 192 000 hz, was in the regime of 48 000 hz.)