• AnyStream is having some DRM issues currently, Netflix is not available in HD for the time being.
    Situations like this will always happen with AnyStream: streaming providers are continuously improving their countermeasures while we try to catch up, it's an ongoing cat-and-mouse game. Please be patient and don't flood our support or forum with requests, we are working on it 24/7 to get it resolved. Thank you.

Bit exact audio in Reclock without any resampling

You are also making me repeat myself like an old broken record...:D
Using my RME FF400 and setting the clock to 46080 (25fps) and 48048 (23.976fps).
Note that James said that the KS mode is bugged, so perhaps that's the reason it hasn't worked with you...
alright, that might explain...coz I was getting glitches every few mins w/ b43 & Waveout.

last time I tried your own DLL w/ b42 in 23.976fps@48.000Hz it seemed to work a lot better than that...I could easily see 15 mins w/o any glitch.

yet, when you see how much the refresh rate constantly oscillates in mVR or beliyaal's MPC builds, I've got a hard time believing you could watch a 2H movie w/o any audio "catch up" glitch...for the reason I gave above : when Reclock sees 48.000Hz, that's not the real refresh rate...that's what DirectX detects and there's a pretty big error margin.

comparing the old resampler from 1.7b4 and your new builds was night and day....but some slight changes would require to output to TotalRecorder and make THD/SNR measurements w/ WaveSpectra, using a 1KHz 0dBFS test tone or pink noise :

Ws1kamp.gif
 
Last edited:
alright, that might explain...coz I was getting glitches every few mins w/ b43 & Waveout.

last time I tried your own DLL w/ b42 in 23.976fps@48.000Hz it seemed to work a lot better than that...I could easily see 15 mins w/o any glitch.

yet, when you see how much the refresh rate constantly oscillates in mVR or beliyaal's MPC builds, I've got a hard time believing you could watch a 2H movie w/o any audio "catch up" glitch...for the reason I gave above : when Reclock sees 48.000Hz, that's not the real refresh rate...that's what DirectX detects and there's a pretty big error margin.

comparing the old resampler from 1.7b4 and your new builds was night and day....but some slight changes would require to output to TotalRecorder and make THD/SNR measurements w/ WaveSpectra, using a 1KHz 0dBFS test tone or pink noise :

Hey LeePerry

If I run MPC-HC with Beliyaals EVR custom renderer, I get an audio/video glitch about 1 or twice per hour. (With or without Reclock.) It's otherwise perfectly smooth. I wonder if it is related to what you just mentioned...


And is that you on the ASUS Xonar forum? I got a Xonar HDAV Deluxe for the 8 channel analogue output (with RCA jacks). I'm really hating the control panel...
 
alright, that might explain...coz I was getting glitches every few mins w/ b43 & Waveout.
Waveout is working fine, the problem only affects KS. So, your soundcard simply should not support the custom clocks.;)

yet, when you see how much the refresh rate constantly oscillates in mVR or beliyaal's MPC builds, I've got a hard time believing you could watch a 2H movie w/o any audio "catch up" glitch...
As I said, I haven't yet watched a 2 hour movie, but you have to consider that the error in the computing of the refresh rate in madVR or MPC-HC EVR-CP should be a lot higher than the errors in the soundcard and PC system clocks. You should not consider the values given by them has a reference, they're just an aproximation.

comparing the old resampler from 1.7b4 and your new builds was night and day....but some slight changes would require to output to TotalRecorder and make THD/SNR measurements w/ WaveSpectra, using a 1KHz 0dBFS test tone or pink noise :
I did not understand what you meant with this phrase.:confused:
 
Waveout is working fine, the problem only affects KS. So, your soundcard simply should not support the custom clocks.
I was outputting bit-perfect from Reclock, and relied on KMixer/my M-Audio drivers to resample 48048@96Khz.
I did not understand what you meant with this phrase.:confused:
well AFAIK, to benchmark the efficiency of resampling algorithms you have to measure their THD/SNR rates.
And is that you on the ASUS Xonar forum? I got a Xonar HDAV Deluxe for the 8 channel analogue output (with RCA jacks). I'm really hating the control panel...
yeah, I'm in contact w/ the Asus drivers engineers...so they'd leave 44.1/48 through ASIO/KS untouched and NOT resampled...they only leave 96/192Khz untouched at this point for some reason :rolleyes:

we need an "auto" samplerate selection in the drivers.

and I'm also nagging M-Audio because the same drivers package gives KS on Vista SP1, but on XP SP3 it's dead silent....their techsupport guy told me "it's working!" but he was using DS :rolleyes:

ah, quite a struggle I tell you :bang:
 
Last edited:
I
yeah, I'm in contact w/ the Asus drivers engineers...so they leave 44.1/48 through ASIO/KS untouched and NOT resampled...they only leave 96/192Khz untouched for some reason :rolleyes:

we need an "auto" samplerate selection in the drivers.


I'm trying to as well.

PLEASE tell them to stop the stupid card upmixing channels all the time (it will affect the Essence ST with the extension card, but maybe not yours).

I want to leave it set on 5.1, and when I play a stereo track, I want 2 speakers making sound....not all 6. I only want upmixing when I tell it to (like picking Dolby Prologic 2 for instance). Not all the time.

Currently I have to fiddle with the control panel EVERY time I change type of media (5.1/stereo). There is no way to turn off upmixing, other than setting the Analog Out to 2 channel....even then I'm unsure if the upmixing algorithm is disabled, or merely upmixing 2 channels to 2 channels.

And yes. I'm onto them for an "as original" setting for Sample Rate, I'm not religiously opposed to resampling, but the damn card should be able to pass a signal through without molesting it (channels and sample rate).

But I'm stuck with the email answer guy after a "screenshot".
 
ah well, hopefully Reclock will output 96/192Khz at some point...so at least the damn drivers won't screw around w/ Reclock's output in KS

and it will upsample 44.1/48KHz to 192KHz...by design :rolleyes:

Emmanuel Piat told me that a very accurate resampling algo(like the voxengo r8brain PRO stuff) should be fully transparent...and anyway this STX card has killer components(I will add some LME49720), so maybe I will enjoy it after all!

as I previously said w/ Reclock, I'd rather have resampling on a killer sound card than full bit-perfect pitchshift on a lousy soundcard w/ JRC2068 op-amps :p

OK I've created a 16/48 1KHz 0dFS test tone and made WaveSpectra screenshots for THD/THD+N....lemme set up Total Recorder through Reclock in KS and I will post benchmarks of the Reclock resamplers.

 
Last edited:
I was outputting bit-perfect from Reclock, and relied on KMixer/my M-Audio drivers to resample 48048@96Khz.
You are still not understanding completelly how this method works...
When using this method, the audio is not at a sample rate of 48048, is still at a sample rate of 48000.
Let me try an analogy to see if you understand it better.

Let's consider a book with 110 pages, and two different readers: reader A which can read 10 pages/min and reader B that reads 11 pages/min.
Now imagine that they are reading the book to an audience and are receiving a new 110 pages book each 10min. What would happen?
Reader A:
Will always miss 10 pages of each book.
Reader B:
Will always read the entire books, never missing a page.

Now imagine that you rewrite the 110 page books to only 100 page books (how, I will let it to your imagination;)).
What would happen?
Reader A:
Will always read the entire books, never missing a page.
Reader B:
Will always read the entire books, never missing a page, and will be silent for almost a minute.

Now imagine that you only have access to reader A to entertain your audience. You have to give him only 100page books, because the audience will not like to loose part of the books, and also will not like to wait in silence until a new book is delivered.
That's the typical soundcard. What reclock does is rewriting the books to keep the audience satisfied.

Now imagine that you have access to 151 readers to entertain your audience, with reading rates from 50 pages/min to 200 pages/min. IF you give them books with no less than 50 and no more than 200 pages, the audience would be always satisfied, because they could hear the original books without any silence between.
That's how my soundcard works, so reclock can bypass the book rewriting and still keep the audience satisfied.

What you have done was somethink like this: the reader A only read 100 pages of the full 110 to another guy who would translate it to the audience, hoping that he would tell them the full book. That's not possible, because he's also not getting the full book...
If you don't have access to the reader B, you will have to relly always on book rewritting.

Note: I intentionally left the concept of pitch shifting outside of the analogy, because that's the more complicated part in understanding all this. It's more complicated for us to understand the concepts that use "time", because it's not a material thing.
 
well, one registry entry makes sure that resampling is disabled...and the other one sets the output sample rate :confused:

I've tried 32KHz and the sound was sloooooow, and 96KHz and it was twice too fast.

we disable resampling, and output bit-perfect pitchshifted audio ?!

like when you output 1:1 from Reclock(w/ your modded DLL), and then set your clock to 48048Hz(for 23.976@48 ) in the RME drivers ?!

I was outputting bit-perfect pitchshifted 48048Hz to KMixer via WaveOut, and KMixer was resampling to 96KHz.
 
Last edited:
Yes.
Setting BypassResampler to 1 is the same as using my modded dll. Soon I will edit my first post with these changes and remove the modded dll, because there is no more need for it.

Setting ForceSampleRate to 46080 or 48048, is the same as setting these values in RME drivers control panel. With these registry keys, I don't need to use the RME drivers anymore to switch the clock.

@James,
Thanks for the new version. KS is working flawlessly too. Soon I will download Windows7 RC1 to give it a try and then I will let you know if it's also working fine with WASAPI... By the way, in the changelog you still kept the ForceSampleRate in kHz, when it's Hz.
What about the auto mode for the ForceSampleRate? It's the only thing that I still miss, so I don't need to be messing with the registry anytime I want to switch from a source type...;)
 
I was outputting bit-perfect pitchshifted 48048Hz to KMixer via WaveOut, and it was resampling to 96KHz.
No. As I said, you were outputting bit-perfect not-pitchshifted 48000Hz to KMixer via WaveOut, and it was resampling to 96KHz.

The pitch-shifting is performed by the soundcard when converting from digital to analog using a different clock.
From my analogy above. The book will always have 110 pages, you need to switch the reader, and you cannot to.;)
 
No. As I said, you were outputting bit-perfect not-pitchshifted 48000Hz to KMixer via WaveOut, and it was resampling to 96KHz.

The pitch-shifting is performed by the soundcard when converting from digital to analog using a different clock.
well, I was using the option to bypass the resampler, 32KHz made the sound very slow, and 96KHz twice too fast...I do believe that my audio was indeed pitchshifted to 48048Hz :doh:
 
well, I was using the option to bypass the resampler, 32KHz made the sound very slow, and 96KHz twice too fast...I do believe that my audio was indeed pitchshifted to 48048Hz :doh:
So it seems that when you say that you can hear 48000 vs 48048 pitch difference it's just a suggestion because you know that there is a small difference...;)

Setting it to 32000 or 96000 does not prove that you can set it to 48048, because the former values are standard values that all soundcards accept. You should try with 40000 or 80000.
If the soundcard accepts clocks not standard like the above, you will clearly notice the pitch shift. That's what happens with my FF400. If there is no pitch shifting, your soundcard does not support other than standard clocks.
 
@James,
Thanks for the new version. KS is working flawlessly too. Soon I will download Windows7 RC1 to give it a try and then I will let you know if it's also working fine with WASAPI... By the way, in the changelog you still kept the ForceSampleRate in kHz, when it's Hz.
Argh. :bang:

What about the auto mode for the ForceSampleRate? It's the only thing that I still miss, so I don't need to be messing with the registry anytime I want to switch from a source type...;)
Maybe in another life... ;)
 
Setting it to 32000 or 96000 does not prove that you can set it to 48048, because the former values are standard values that all soundcards accept. You should try with 40000 or 80000.
any sample rate works, 10KHz, 20KHz, 50KHz, 60 KHz...I've tried it w/ my 1KHz test pattern...it's being resampled to the current sample rate(by whatever KMixer or my M-Audio drivers, dunno)
leeperry,
please try this resampler.dll and let me know how it sounds when resampling, comparing it with the resampler.dll that ships with reclock...
View attachment 7294
so you've lowered the output volume?

I can't benchmark Reclock in regular adaptative mode, coz it constantly resamples to different rates.

I'm forced to check "slave to audio clock", and Total Recorder/VAC don't support KS...and because Reclock doesn't support ASIO, I'm forced to use WaveOut(w/ 23.976fps@48Hz).

so that's the original file :

7bf64d34912241.gif

that's w/ the stock resampler :

a6a3c534912245.gif

that's w/ your new "lower volume" resampler :

7e5d4e34912248.gif

hopefully I'll be able to add measurements for 96/192KHz resampling soon, and I'm willing to bet that the THD values will drastically lower :rock:
 
Last edited:
Maybe in another life
same goes for 96/192KHz resampling? jirotoh.gif

it's strongly frowned upon in studio to resample a 48KHz DAT tape to 44.1 CDDA :
http://209.85.229.132/search?q=cach.../1.php+resample+48KHz@44.1&cd=1&hl=en&ct=clnk

The 44.1kHz to 48kHz resampling project so far has proven to be a little more difficult, namely, I believe because of the large interpolation and decimation factors and long filters required.

resampling at 99.9% is as lossy as could possibly get...actually, you can't get anymore inaccurate than that(well maybe 99.999999999%) :(

but well, I believe I've done my best to prove you that the THD distortion thanks to 200/400% resampling would simply become inaudible....I understand Reclock is not an audiophile tool, but 2 lines of additional code might come a very long way :rock:

just an option for double/quad samplerate in the settings, and a line of code in the output reclock component that says "if checkbox enabled, double/quad multiply samplerate"? :bowdown:
 
Last edited:
any sample rate works, 10KHz, 20KHz, 50KHz, 60 KHz...
But do you hear any difference in the audio picth? If you always hear the same pitch it's not accepting the sampling rates...

so you've lowered the output volume?
Yes. When resampling between sampling rates not a multiple of each other could appear some clipping in the resampled data, so the volume should be lowered to avoid it. These kind of conversions, 48000-> 50000 and 48000->47952, are very prone to it. I don't know if reclock is using any internal decrease of the volume to give some headroom for the resampling. If it's not, it can be outputting clipping sound in some parts, when performing resampling, and that could be the reason of the bad sound quality that you are referring to. If that's the reason, doubling the sampling ratio as you are requesting will not solve it.;)
For that test version I have simply multiplied by 0.8 the original sound, but if James confirms that no internal volume reduction is performed internally, I will try to calculate the correct value that should be used to avoid the clipping and upload a new source for resampler.dll... James?
 
1)well yes, the pitch changes depending on the samplerate I set in the registry. It works as expected....easily audible w/ a 1 KHz test tone ;)

2)the biggest quality KILLER is 99.9% resampling for sure...if it were 200/400% the distortion figures would lower a long way :)

any resampler on the market would fail w/ 99.9% ratio :eek:

we could also look into gain compensation afterwards indeed, something like 99% volume?
 
Last edited:
For that test version I have simply multiplied by 0.8 the original sound, but if James confirms that no internal volume reduction is performed internally, I will try to calculate the correct value that should be used to avoid the clipping and upload a new source for resampler.dll... James?
No, there is no volume reduction.
 
No, there is no volume reduction.
So I will look into this and update resampler.dll source accordingly to avoid clipping. Do you agree, or do you prefer to do it yourself?
 
Back
Top