9z-5-2: Wrong pitch with 48khz at x4

[AlainGr]

I normally leave the output sample rate at 4x, but when I have a song is initially sampled at 48khz, the pitch is not right.

It only happens at 4x (I can't go higher).

Thanks.

Alain

[PeterSt]

Ehm  ... 9z-5-2 pitch problem with 16/48 combined with 4X upsampling ??

Amnesia ?

[AlainGr]

Amnesia... ? Is it a keyword to trigger something ? ;-)

[AlainGr]

Ah ! I got it now... Sorry for this... Sorry for this (I have to write it twice) ;-)

Alain 

[PeterSt]

Hey Alain - Small problem; I can't copy this behaviour. It just works ok.

Is that 24 bits you are having there ?
Do you use Arc Prediction ?

Thanks,
Peter

[AlainGr]

Hi Peter,

The song is natively at 16 bits 48 khz... I tried with and without Arc Prediction. My DAC is working with 24 bits (not 32)...

Thanks,

Alain

[PeterSt]

Still no dice.
A bit of a problem is that I have no means to test this sound wise. But I've looked in the program, and all what is done for 48000 is done for 44100 exactly the same (unless my searching capabilities are not the best). Meaning : if this happens with 48000 it should happen with 44100 just the same.
... which of course is not happening in your situation.

All I can think of is that 16/48000 hardly will be a normal audio format, and whatever it is you have there, it will be a non-common rip. This means that something could be wrong in the header, and that *this* is not covered for correctly by the software - somewhere.
On this matter, you could send me a track of this.

What also -theoretically- can be the case, is that somehow your DAC is not able to do this right. Now, however, my story really must be the other way around from above : playing 16/48 will be the most normal to a DAC (because this is video standard).

Now I think of it : I will be able to test it without sound, by measuring the track time. I mean, I do have a 24 bit device (HiFace) but no DAC for behind it. But I feel it is better if I first have an example track. Ok ?

Thanks,
Peter

[AlainGr]

Hi Peter, I have sent the "suspicious" file ;-) I downloaded this file again and though I bought the whole album one year ago, this song seems to be the only one recorded at this weird format (16/48)...

When I reduce de X factor from 4 to 2 (x 2), everything is OK... Maybe it is related to either the USB -> SPDIF I have (Audiophilleo 2) or my DAC (Bryston BDA-1) ?

I forgot to say that this is not of real importance for me. I just thought it would be useful to you, but if it is related to my USB -> SPDIF or my DAC, there is not much to do...

Thanks anyway,

Alain

[PeterSt]

Thank you Alain. But we really must hop over to plan B (IOW, all still ok here).

Ok, if this is not of big importance for you, I better let it be for now.
Not for later ...

Thank you for your help !
Peter

[AlainGr]

Hi Peter,

I just downloaded an album today (Rodrigo y Gabriela and C.U.B.A: Area 52). It is 24/48... And I use KS, adaptative, 2.5ms, SFS 50 (I am waiting for Lenovo memory banks), Q1 = 7. I have Windows 7 pro SP1, 64 bit, 2 GB of RAM... CPU is Intel G620, 2.6 ghz, 2 cores.

When I use the multiplier (x 4) to get 192 khz, the pitch is slower. At x 2, the pitch is OK...

I don't have a clue. I have tested this with and without Minimize OS. I have tried it with WASAPI and the same symptoms are happening...

It was doing this on my laptop, now on my recent PC... I did try on my main PC and no problem... But the difference is that the laptop and the new PC are connected to my audio chain, while my PC is standalone...

Alain 

[PeterSt]

Yo Alain,

It was doing this on my laptop, now on my recent PC... I did try on my main PC and no problem... But the difference is that the laptop and the new PC are connected to my audio chain, while my PC is standalone...

I had already read this twice before I saw the (my) emphasis. So ...

WHAT ?

Now I'm floored.

The difference should be your DAC (??).
But this is not 100% true, because the difference also will be that 24 bit output. So, in-PC that is 32 bit I suppose ? (please check the DAC Needs setting on your main PC)

It now also seems clear that it is not bit depth related ... am I correct ? so, both 16 and 24 bit files don't work.
Btw, earlier in the topic I said that 16/48 is a rare format, but it is not at all of course. This is just what is ripped from DVDs (which in there will be 16/44 although in lossy "MP3" format).

Please confirm the few things from this post, and when you do, I guess I *have* to find some bug somewhere. As you will remember, I already checked everything, but I must do it again. It's a bit stupid that I can't test it. Or ... hmm ... I think I can with the analyser (and the HiFace I have, which allows for the DAC Needs 24 bit setting).

Peter

 

[AlainGr]

Peter,

To prevent any confusion:
- Lenovo is the desktop connected to my audio system.
- Asus is my general purpose desktop. It is more powerful than Lenovo.

Yes, Internal DAC has to get 32 bits on Asus.

As for the audio system (Lenovo), outboard DAC has to get 24 bit...

I forgot to mention that I got an Audiophilleo2 USB to SPDIF converter on my audio system... :-( I have corrected my signature a few minutes ago...

I ran some tests...

Lenovo: first, I indicated to XXHE that I would use the integrated soundcard. It works at 24 and 32 bit, but sampling rate is limited to 96khz. Not useful since x4 upsamble 48khz to 192khz.

Lenovo: I switch to the audio system.

Lenovo -> USB -> Audiophilleo (24 bit) -> SPDIF -> Bryston DAC (24 bit). I select "DAC is: 24 bit/192 khz". As for the  "At 16 bit, the DAC needs", I have to choose 24 bit. (pitch is then slower at x4).

Now I turn to my general purpose desktop...

Asus is a little confusing: I can either select "DAC is: 24/192" or "DAC is: 32/192". As for the "At 16 bit, the DAC needs", It has to be 32 bit (pitch is ok at x4). 

I could connect Asus to the audio system to check if it will behave the same as Lenovo... I can do it if it can be of some help ?

As for this situation, there is no rush. I am only worried that I could forget what is happening at x4 for 48khz songs ;-)

I started computer programming at 30 for 15 years. I still do some programming from time to time, but it is not as stressful as it used to be. I know how it can be...

Regards,

Alain

[PeterSt]

I could connect Asus to the audio system to check if it will behave the same as Lenovo... I can do it if it can be of some help ?

Alain, if that is easy for you - Yes please.
(and the same problem is to be expected)

[AlainGr]

Hi Peter,

It took me a while (...) to do the test with my ASUS computer, but I did it this morning and... Yes, the pitch has the same symptoms as with the 2 other PCs (HP laptop and Lenovo desktop).

Since I did not experiment this problem with my ASUS soundcard, my guess would be that with a 24 bit DAC (or with the Audiophilleo), there could be something wrong happening.

If you need more precision, please feel free to ask ?

Regards,

Alain

PS: I changed the title, since I changed the XXHE version when it came available...

[PeterSt]

Ok Alain, thanks. So I *have* so find something eh ?

[AlainGr]

Mmmm ? Not sure I understand your answer ?

Oop... It seems that by replying to your post, the title is back to 9z-5-2...

[PeterSt]

About my earlier search for the culprit and that I can't find anything. I can't imagine that I find something to be there (wrongly) now, but something *has* to be wrong.
That's all ...

[AlainGr]

Ah ok 

I tested different parameters, but...

I would say that it does not bother me, but it seems more music is available in either 24/44.1 or 24/48...

At least I keep on buying CDs I slowed down on 24/192 downloads... Too much... Cr.. ?

Regards

[PeterSt]

Alain,

I have just been finishing the Release Notes to 0.9z-7 (which does not mean 0.9z-7 is ready to go out ) and ...

I ran into a most peculiar unresolved ToDo. So, of course your little issue is in there as well, but here I'm talking about a coincidental DAC Needs = 24 test I did only two months back with the NOS1 (which it appears to be able to do) but with 48KHz ...

TE PITCH IS TOO LOW.



Now, I never saw the relation to your issue, but strolling through the (solved/unsolved) ToDo's I suddenly saw the combination. But wait, not too fast ...

What I never saw in your posts in this topic is HOW MUCH the pitch is too low. So, imagine my logic that you could only be talking about half of the speed or so.
But what I see here with this issue on my side is a 1000Hz tone coming out at 918Hz. And I was flabbergasted, because it really looks like the clock suddenly runs too slow. Or at least I have no other explanation, while this actually is no explanation either. So in ad-hoc fashion I am typing this :

1000 / ~918 = 1.0893
48000 / 44100 = 1.0884

Tadaaa ....

But now the question : is this what you perceive indeed, or is it half the speed like I always thought ?

Assuming you too perceive that ~10% off only I sure can now mimic your situation. And, it is 24 bits related. Both seen from different angles, and in both (for me as thought completelt separated situations) I couldn't see what was going on. Your case has been an "oh well", but my case is really flabbergasting me. And of course I (now) recall to have been through all the code and found nothing. Surely, with my situation at hand I can really start trying things, but since I know that the setting of the sample rate should be out of my hands because the OS is doing that through the header data in the file, I wonder - could it be an OS bug ?
(most probably not, and most probably I will find a bug on my side now).

But, is this your situation ?

Peter

[AlainGr]

Hi Peter,

You are right, I never mentionned by about what percentage the pitch was lower than normal, but it sure was not half of the speed. Maybe the x4 and x2 thing got confusing. I would say that a 10% decrease in speed is about it.

Thanks !

Alain 

[AlainGr]

It still behaves like this. Like you mentioned before, it could be 10% slower. When I upsample to x2 for the right pitch.

Thanks,

Alain

[PeterSt]

Call me crazy, but this should have been solved somewhere in 0.9z-7. Although, I briefly searched the Release Notes, but couldn't find anything.

Are you sure this hasn't been solved in between versions ? I am pretty sure we talked about this.
Also, this is about the back and forth working/not working of all around your DAC. Up to that situation which needed Play twice because the first (after initial playback) produced an error.

I am getting crazy of this and I guess when you can't come up with an explicit "never worked" I will give up on this. Anyway I am pretty 100% sure that some version contains Release Notes about this and I ask you politely to look that up (and when found, try that version and when it helps, try to find the version which made it stop working again).
This looks a little bacl and white, but it is a difficult problem anyway, and without clue where again it stopped working - and with the notice with all of the problems which happened in between around this - ...

So please Alain, you must help me here.
Peter


PS: And let me softly say it once again ... PM and stuff about problems is really not going to help. I use one source to keep track of problems, and this is the forum. Maybe you can find things about this in emails or PMs ? I really can't with 150 "unstructured" pages in there.

[PeterSt]

After all, here : XXHighEnd Model 0.9z-7-4 (alows for more PA situations + new SQ setting). But notice the heading under which it is listen "Prior to ...".

Peter

[AlainGr]

After all, here : XXHighEnd Model 0.9z-7-4 (alows for more PA situations + new SQ setting). But notice the heading under which it is listen "Prior to ...".

Peter

Peter,

Though you indicated me from where to start, I chose to start from 9-6 up to 9-7-5... WASAPI (to do it faster). Unless there would be a difference about this situation with KS, they all behave the same... 10% difference (about) between x2 and x4...

If you wish I can do the same again with KS ?

Thanks.

Alain

[PeterSt]

Ahum.

The solution for KS is different from WASAPI, plus in my comment it states "WASAPI not tested at this time but isn't used by anyone anyway". Haha, oh boy.
So I am sorry, but yes this matters !

IIRC what I wrote yesterday it is not necesssary to start with 0.9z-6. Somewhere 0.9z-7 it will be that I declared it solved.

Btw Alain, it will be a good idea to tell me again how you test this today, because quite a few scenario's exist. This may be different from how this topic started. Okay ?

Regards,
Peter

[AlainGr]

I did the same test for each version the same:
- Decompress all in a XXHIGHEND folder on the OS disk, always emptied between versions
- Reboot between each version
- As you stated (in a PM that I started - I know, I know) that the OCX files (from setup.exe) do not require to be reinstalled between versions, I did not change them.
- Adjust minimally (WASAPI, 24/192, 24 bit [not 32, my dac does not recognize 32 bit), peack extension, arc prediction, x2 and x4 [for each test].
- I play at x1 (48khz)
- I play at x2 (96)
- I play at x4 (192)
- reboot
- clear the XXHIGHEND version to leave place for the next version.
- As for those which had an update, it was done first without, then with the replacement files (like the xxengine3 that was added for 9-7-4).
- All on the same PC (music PC) of course...
- 1ms clock resolution.
- contiguous memory
- None of them activated (...)
- All in demo mode (no minimize OS)
- SFS at 12

I wondered if it was necessary to kill xxengine3 between each upsampling but since the pitch was changing, I guessed it was not necessary.

Alain

[AlainGr]

Peter (good morning)

In one of your answers about this, you mention that the routine for 48khz is the same as with 44.1 and that it shoud have the same behavior. Maybe I could copy this song and downsample it to 44.1 just to see if this will behave the same ?

Alain

[PeterSt]

Alain, help !!

You didn't tell what went wrong and where !
Unless of course you mean that it (what upsampling ?) never plays right. And since I don't (and never did) expect *that*, I don't get it.

- Adjust minimally (WASAPI, 24/192, 24 bit

I don't understand this either; WASAPI ? Didn't we "agree" that you should test with KS ?

And then lastly, this was about finding out out where it was right and wrong again. You say nothing about that though. And it feels wrong to have my own conclusion about it ...

In one of your answers about this, you mention that the routine for 48khz is the same as with 44.1 and that it shoud have the same behavior.

Probably I said something like that, but then in some context now lacking here, and *thus* I don't understand this either.

Okay, it could be so that I don't understand anything much today !

??

[AlainGr]

Alain, help !!

You didn't tell what went wrong and where !
Unless of course you mean that it (what upsampling ?) never plays right. And since I don't (and never did) expect *that*, I don't get it.

I hope I understand your question correctly: when I started using XXHIGHEND, I was not using upsampling. After reading here in the forum, I decided to use it. At this point and from the moment I had this song, the version I was using was 9z-5-2. Then I discovered about this. I never saw it change throughout the following versions, though I admit I did not test it with each of them.

I don't understand this either; WASAPI ? Didn't we "agree" that you should test with KS ?

That's true. I would have guessed though that if WASAPI is a part of the code you do not really touch, it would be simpler to diagnose and correct, since WASAPI is also problematic about this... But yes, I will redo the tests in KS mode this evening...

And then lastly, this was about finding out out where it was right and wrong again. You say nothing about that though. And it feels wrong to have my own conclusion about it ...

I never said it was ok Peter, because it never happened...

Probably I said something like that, but then in some context now lacking here, and *thus* I don't understand this either.

The following quote is taken from a post you did in this very thread:

Still no dice.
A bit of a problem is that I have no means to test this sound wise. But I've looked in the program, and all what is done for 48000 is done for 44100 exactly the same (unless my searching capabilities are not the best). Meaning : if this happens with 48000 it should happen with 44100 just the same.
... which of course is not happening in your situation.

All I can think of is that 16/48000 hardly will be a normal audio format, and whatever it is you have there, it will be a non-common rip. This means that something could be wrong in the header, and that *this* is not covered for correctly by the software - somewhere.
On this matter, you could send me a track of this.

What also -theoretically- can be the case, is that somehow your DAC is not able to do this right. Now, however, my story really must be the other way around from above : playing 16/48 will be the most normal to a DAC (because this is video standard).

Now I think of it : I will be able to test it without sound, by measuring the track time. I mean, I do have a 24 bit device (HiFace) but no DAC for behind it. But I feel it is better if I first have an example track. Ok ?

Thanks,
Peter

Okay, it could be so that I don't understand anything much today !

??

Alcool, drugs and electric trains do not mix well 

Alain

[PeterSt]

Okay Alain. Retry from my side :

You HAVE to be more explicit about -I'd say- EVERYTHING. To me it looks totally impossible that up to today I apparently don't understand the problem. I myself did not read back in this topic which also feels a little useless because not everyting said may be in the topic. But I am pretty sure this was about my idea of CHANGING the sample rate. Again, please look back. And this is totally different from only playing at 4x which again is different from a pitch being off 10% and the speed being double - as how I started out responding in this topic. Again, *I* did not read back.

Do notice now that I never in my life changed something about some "absolute 4x" not being OK with 48KHz. I was just never told that, possibly similar to you actually not telling a problem in your previous longer post. Actually it is worse, because you never in the whole topic told a thing, or it has to be the confirmation from (finally ?) two days back that you had a pitch problem instread of double speed or something. This I had to guess too you know. But I do that because things don't fit.

So, am I complaining ? no, not much. But I do have to tell you that we won't be able to solve problems like this.

The solution for KS is different from WASAPI, plus in my comment it states "WASAPI not tested at this time but isn't used by anyone anyway". Haha, oh boy.
So I am sorry, but yes this matters !

You may not like me by now, but to me it seems that you tested all in WASAPI again (why ?) while you should have done so with KS after my post I mentioned this in. Possibly I am as unclear as you are though.

I wondered if it was necessary to kill xxengine3 between each upsampling but since the pitch was changing, I guessed it was not necessary.

What ? what pitch CHANGING ? Didn't I just conclude from your words that nothing was CHANGING ?
Be explicit Alain. Far more explicit.

Below you see the files you sent me to test with. You know, I solved the problem. What problem ? THAT I know don't know. So I solved a problem, but not yours. Can I still use these files for my testing of whatever is your problem ?

Do not repeat that test again. Please spend your time on explaining 100% clearly about what the problem is. I *still* don't know you know ...

haha

Peter


PS: Apologies if I sound a bit like hitting the keyboard hard. Well, I do. Can't stand that I still don't get it !

PPS: At reading back this post before posting, I suddenly see a clue (maybe) :
- You test at e.g. 2x
- You test at 4x and hear the pitch CHANGE.
Correct ? If yes, only confirm this. I don't need anything else then. Ok ?

[PeterSt]

One more thing to add, for your better undertanding (hopefully) :

Suppose you say, Yes Peter, of course that is the problem !
Well, then you have to know that there is no logic in this for me, because I see no way it can happen. What can (and did) happen is that the pitch is wrong at going from 44.1 based to 48 based.

So yes, if your problem is as I think you try to tell it to me (all the time), it is easy to see when reading back. But first I'd have to know it because otherwise I would never guess it.
Never.

A PS again : It sure is good that I can regard this as a new problem (which has been there forevere). This is far more easy than finding code which changed and causes it.
If this is your problem now (change from 2x to 4x and the pitch changes). And before I can't mimic this problem : it can be your DAC which is not able to cope with the bandwidth. Try Foobar please.

[PeterSt]

So there we are ...

That first file in the list I showed is a 96KHz (the other one is a 44.1Khz). Not 48KHz. Should not matter by itself I guess, if you only KNOW that there is no way you will have been able to play this file at 4x. Can we agree ? Ok, we can.
Next, I just played it (with KS !) at 1x, 2x and 4x and at 32 bits *and* your used 24 bits, and the 4x with 24 bits is two times  slow. Could be 220% but somethig like that.

This isn't 10% off you know.

So, tell me that there is a problem somewhere and maybe I will find one. Haha.
But is it your problem ?

So I found myself an 48KHz and you know what ? 10% too slow at 4x and 24 bits.

One thing left : I never can have used your 96KHz file to test that 10%. I did solve something with it, but not this of course.
But as you see, I get (to ) you anyway !

Well, finally. Now solve it.
No further response needed. Yea, later please, when I announce it solved in some Release Note.

[AlainGr]

Peter,

Maybe I am confusing, but you are going too fast ! I am at work, so can't verify certain things here... We're on a different time zone It is 13H24 here... When I will be at home, I will make the proper verifications, then come back to you.

Alain

[AlainGr]

Peter,

Well I at least can try to "un" confuse 2 things.

1. This thread is about ONE song that, played alone, does have a -10% pitch variation when I upsample it at x4, as opposed to x2 or nothing. This problem has been consistently here since 9z-5-2, that happened to be the first time I discovered this. Though I admit that I did not test this song with each and every version that followed, but each time I did, the same was happening. That is why I followed your suggestion to try to redo each version up to the actual. I did, but with WASAPI. I will do it with KS this evening.

2. There was also a situation where 2 songs, each with a different format (song 1 = 16/44.1, song 2 = 24/96) were behaving strangely (this is the one I have to check when I will be back home this evening). But I don't recall creating a thread about this. Maybe I did that through PM 

About problem 1: I did this test with JRiver (it did not create any problem with upsampling). I will do also with foobar as requested.

Alain
PS: You know like I know that these are the kind of situations where it's like looking for a needle in a haystack. My life does not depend on this (I learned that last summer). Ok ?

[PeterSt]

Alain, by now and since my last post, please, when you get home spend time with the family instead of this. I have found a sufficient amount of problems to solve, and "undoubtedly" (ehm) yours is amongst it.

Maybe I can tell you : it is only praiseworthy that you really are the only one coming up with this.

What I saw will be easy to find and solve. No need for Foolbar anymore either. Really. And hey, if you dare post again about this I will delete that post(s) anyway. Does that help ?

Peter (reaaaaly pushing the keyboard keys softley this time)

[AlainGr]

And hey, if you dare post again about this I will delete that post(s) anyway. Does that help ?

Don't tell me you want this through PM ? I am so tired of using PMs
 
Alain

[PeterSt]

Hey Alain - you are mean. You sent a PM and I can't delete PMs. Ok, I can, but what's the use.

Ok, solved the problem but didn't listen to it. All was a matter of forgetting one small = somewhere. Yep, that's how it works in C.

Might you want to try it (and don't you dare to do not - actually you should be playing 48KHz files for the rest of your life from now on ... okay, kidding), paste the below (XXEngine3.exe) over your current 0.9z-8-1a folder (or other 0.9z-8 folder, but not nothing else !!).

And hey, thanks for being persistent !
Peter

[AlainGr]

Hey Alain - you are mean. You sent a PM and I can't delete PMs. Ok, I can, but what's the use.

Ok, solved the problem but didn't listen to it. All was a matter of forgetting one small = somewhere. Yep, that's how it works in C.

Might you want to try it (and don't you dare to do not - actually you should be playing 48KHz files for the rest of your life from now on ... okay, kidding), paste the below (XXEngine3.exe) over your current 0.9z-8-1a folder (or other 0.9z-8 folder, but not nothing else !!).

And hey, thanks for being persistent !
Peter

  Hey... Thanks for your patience Peter I am eager to go back home I know how those small things can be stealthy when you look for them. They say that a good programmer is also a good debugger 

WoooooHoooo 

Alain

[PeterSt]

Haha, I'd say a good programmer doesn't need to be a good debugger.

Anyway debugging is more difficult than good programming (time wise).
My opinion of course.

[AlainGr]

Haha, I'd say a good programmer doesn't need to be a good debugger.

Anyway debugging is more difficult than good programming (time wise).
My opinion of course.

Also true, but nobody's perfect 

[AlainGr]

Hey Alain - you are mean. You sent a PM and I can't delete PMs. Ok, I can, but what's the use.

Ok, solved the problem but didn't listen to it. All was a matter of forgetting one small = somewhere. Yep, that's how it works in C.

Might you want to try it (and don't you dare to do not - actually you should be playing 48KHz files for the rest of your life from now on ... okay, kidding), paste the below (XXEngine3.exe) over your current 0.9z-8-1a folder (or other 0.9z-8 folder, but not nothing else !!).

And hey, thanks for being persistent !
Peter

I came back home, downloaded the file and... (drums rolling)... Tada ! It works ! I does !!!!

Now I am sad... What to find next, except good music with SSQ (Super SQ)... 

Regards,

AHA - A Happy Alain