XXHighEnd

Hi Peter and all

I have a strange problem. When i tick double and upsampling my extrernal processor locks at 88.2khz but the sound is distorted.
I thought that it was a driver problem, until i used foobar with wasapi plugin and it works as it should. Any ideas???
i'm using the latest version of XXHE

Hi,

I have a question which may look unrelated at first glance, but it very much is :
How can you tell that the output from the plugin is 88K2 ? Most probably you can't, meaning that it is possible that the OS resamples for you (and *then* the output to the processor is 88K2, but in a unintended way).

To me something like the above would be the only explanation *unless* I am doing something wrong with the upsampling (but as you probably know this is finished for quite some months now, and I don't hear anything about it (including from myself :)).

A test which might give an indication, is checking whether Foobar output is in Exclusive Mode (if you know how to do it).
In the mean time I will investigate Foobar myself; wanted to do that anyway. :whistle:

Peter

I think i understand why you asking that.

I'm sure that os doesn't resamples because in audio advanced properties (drivers settings) if i set for example 96khz i still get 88khz at spdif when i press the play button.
If the driver do something to xxhe output then i should have seen 96khz.
I 've done the same test with foobar. Wasapi output, secret rabbit resampler at 88.2 and i set the audio properties at 96khz. When i press the play button i get 88khz at the dac but the sound is normal. I tried with different drivers but i got the same. So i believe it is something wrong with xx upsampling.

... which would be true if it didn't work here or with others either, but it just does ...
However, might you be in doubt, tick the AA checkbox, which would give your SSRC resampling :yes: (loading takes long then).

Also, depending on the type of file you use, unticking Enable Crack Detect is a good thing to test with in this case (supposed yo played DTS or something like that).

Anyway, I just checked out Foobar, and indeed I can't find anything wrong with it (at normal 44K1 play and upsampling to 88K2 via PPHS resampling). So it wouldn't be in that area. I noticed something else though ...

I took these 30 minutes to checkout SQ of Foobar now, and at noticing some remarkable differences (not in favour of Foobar hehe) I started setting the parameters concerned. This pointed me at the minimum buffer size of 100ms, while XXHighEnd uses 1ms - 270 ms depending on combinations of settings. Now :
The Foobar buffer in Playback - Output does not act the same as the buffer in XXHighEnd I refer to BUT at using the Foobar buffer at e.g. 100ms for sure you would feed the audio chain with something which is much more "at ease" than the XX buffer at in fact any setting.
To work as stressless as possible in XXHighEnd you should set the Q1 slider to a higher value (but too high and it doesn't work anymore (skips and things). And I see from your sig that you use a Q1 of -4, which is the most stressful setting (but generally with the best sound !) and which for sure not everybody can use ! (ticks, distortion).

Looking at the input buffer at the soundcard, I couldn't find differences between Foobar and XXHighEnd; both need 96 (+ 64) samples of buffer in order to play 88K2/16 (I didn't try 88K2/24 btw). Both can use 48 (+64) samples to play 44K1/16. So at that end both are as "lean" which btw (to me) is obvious for WASAPI playback (at that end all is the same when talking about achieveable latency).


Can you do something with this information ?
Please keep in mind that I'd do anything to solve this for you, because if "some" WASAPI player (that's what it comes to) is able to work with a soundcard, there's really no reason XX shouldn't do it too. We must only find out what causes it not to, ok ?


PS: There's so many things involved that it's easy to forget one; It just slipped into my mind that XXHighEnd would be upsampling to 24 bits (actually 32 bits) ALWAYS as long as the soundcard supports it *and* you told XX that the soundcard supports it. So note that when you'd upsample in Foobar to 88K2/16 there's twice as less data going through opposed to XX which will output 88K2/32. When you want to test this, set your DAC = to 88K2 (or 96) / 16 bits (or make sure you play Foobar at 88K2/32 (padded)).

PPS: I hope I make sense to you ... :heat:

thanks for your time Peter.

I had almost  (double and triple) check everything from the above. I'm a computer expert and i tried almost everything before i post.
If it is working for you then it must be something with soundmax driver. I know that soundmax resamples 44.1 to 48 when transmitting to spdif and for that reason i want to sent 88khz by software, to play untouched after resample. I would give it a try with another soundcard and i'll tell you if it works for me.

With soundmax i tried 88.2/16 bit and i set the dac option to 96/16 at xx. So this is something that i must check. Thanks for the info.

Anyway the difference is huge between xx and foobar. i love xx for the sound and the only reason to use foobar is just for testing :).

Yeah, but although that might help you, it isn't the solution to something which just should work from the beginning. Most important of course is if Foobar behaves the same in all aspects (about your resampling to 48K first maybe). Because if not then it's apples and oranges.
Please keep me posted !

hmmm...
when i set foobar resampler to 48khz i get 48 at the dac and sound its ok. if i change to 88/32 i get an error of unsupported format.
at 88/24 no problem at all.

At xx when i set dac is 96khz 24 bit, i got an error message that the dac supports only 88200 / 32.
when i set dac is 96 /32 and i press play it says that the device does not support exlusive mode.
I think that this is the "problem". For me, i guess the only solution is to resample without changing the bit depth at xx. Is that possible??

Ps: i run the dac test and it supports only 2|88200|16 ,no 24 no 32

Ok finally after some testing i find out that my sound device supports only 16 bit in xx. The dac test says that in exlusive mode i can use only 16bit (in foobar i can play 24 bit. strange.). So when i try to upsample the xx output is 24 bit and i hear this distortion.
If i set the dac is 96/24 i get a message at playback that xxengine is stop working so i can only set 96/16. So at this mode the upsampling shouldn't be allowed, or you must fix it to play at 16 bit.
I also notice that if i set my dac is to 96/24 or 32, i cant select the 24 bit (untested) box.

You are right. This is wrong in XX. On the other hand, I am 99.99 % sure that being able to set it to 24 bits won't change a thing. But :


... I hear you saying this too of course.

I will fix the error of not being able to set the DAC to 24 bits (in your given example) and then we must see further. But then again :


This should tell me enough. Or IOW, when Foobar allows something in this area, it resamples first. Maybe you remember that at some stage I quit allowing resampling by the OS because it just would be for the worse. Thus, XX doesn't allow this now, while WASAPI actually does allow it.

But to be sure (for me) : do you have information that your soundcard/DAC supports 16 bits only ? because if so indeed, we just might wonder how Foobar seems to manage ...  :dntknw:

My card in theory supports 24 bit. I can confirm that by using the test tones of the windows driver. it plays 44.1/48/96 at both 16 and 24 bit with directsound. i'm not sure if the /24 bit modes are supported with exclusive mode. XXHE says no.At the begining  i used foobar, to confirm that it plays, it works but now i'm not so sure if this is exclusive mode. On the other hand if i measure pci latency with dpc, while playing foobar at 48 or 88 either at 16 or 24 bit, i get very low values which shows (almost the some in both circumstancies) that it is exclusive mode.
with direct sound output plugin i get almost 3x latencies.

I have been pulling my hair what the difference could be, and I came to something ...

Because the whole Exclusive Mode thing is so buggy I applied quite some tweaks. One of them is contained in the 96.dat file which resides in the XXHighEnd folder. Now :

Keep your own 96.dat safe, and put the below one in your current XX folder. It "untweaks", and should normalize things far more to the Foobar situation.
Please note that the normal behaviour is that the nearest possible samplerate and bit depth is selected when the one needed is not there. It would be normal if Foobar does just that, and keep in mind that XX does *not* do that (because it would imply a conversion).

The first thing you might do is run the DAC Test and see whether there is any difference.

Do not forget for this XX version (0.9v-3) that you replaced this 96.dat file, because it won't allow you other things;
When it helps, I can make it dynamically.

Peter

I replaced the 96.dat file and the results of dac test are the same.

Start Audio Device Analysis Exclusive Mode support
04E1FD0C

    Supported : 2|16|44100|176400|4
Not supported : 2|24|44100|264600|6
Not supported : 2|32|44100|352800|8
    Supported : 2|16|48000|192000|4
Not supported : 2|24|48000|288000|6
Not supported : 2|32|48000|384000|8
    Supported : 2|16|88200|352800|4
Not supported : 2|24|88200|529200|6
Not supported : 2|32|88200|705600|8
    Supported : 2|16|96000|384000|4
Not supported : 2|24|96000|576000|6
Not supported : 2|32|96000|768000|8
    Supported : 2|16|176400|705600|4
Not supported : 2|24|176400|1058400|6
Not supported : 2|32|176400|1411200|8
    Supported : 2|16|192000|768000|4
Not supported : 2|24|192000|1152000|6
Not supported : 2|32|192000|1536000|8
Not supported : 2|16|352800|1411200|4
Not supported : 2|24|352800|2116800|6
Not supported : 2|32|352800|2822400|8
Not supported : 2|16|384000|1536000|4
Not supported : 2|24|384000|2304000|6
Not supported : 2|32|384000|3072000|8

if i set dac is 96/24 (double & upsampling checked) and press the play button XXEngine3.exe hangs.

Thanks for trying (here it works with that "other" 96.dat at 16 as well as 24 bits).

There is something I don't understand;
It would be my idea that when the soundcard (towards XX) rejects the 24 bit modes, you would not be able to even start playing 88K2/24 (actually 88K2/32 which would happen at Doubling/Upsampling). However, you told (first post) that the 88K2 lamp on your processor lits, but the sound is distorted. This tells me that you must have selected a DAC Is setting with 16 bits. Can this be true ?


JFYI : I don't think it is reliable to check this by looking at the latencies. Best would be (99% sure) to try to switch of the sound by means of the loudspeaker icon in the taskbar tray (rightclick - Open Mixer) and the icons and sliders in there. Here this works okay (Foobar), but then I feed everything with data of which I know it can be handled without conversion.


PS: Don't forget to set back the original 96.dat.

Peter you are right. I forgot to mention something that i did while trying to fix my problem.

I have already told you that with the latest version as it is, when i hit the play button (dac is 96/16 double & upsampling)
i get xxengine.exe stop working and no sound at all.

So i tried with an older file of xxengine.exe (version 09s0). This was the engine i used with my old soundcard and upsampling was ok.
With this version if i tick upsampling and double  and press play my dac locks at 44.1 which is propably normal because this old file dont have the code to cooperate with the latest version. here is the strange thing. if i tick AA also then i get 88.2 and the distorted output.
It was an expirement for me to try to solve the problem without bothering you. sorry that i forgot to mention that. I didn't remember it after so many changes i had made to make it play.


At foobar with 88.2 and 24 bit if i try to change the volume nothing happens. So it is exlusive mode. I dont know if the output is 24 bit or 16 and i cant think something to verify that.

Ok, so now I understand better.


But this was with the before sent 96.dat, right ?
Or is it also with the one that normally belongs there ?


Please check whether your earlier post is correct, because that tells about xxengine3 stop working at 24 bits ... :scratching: :


:)

The right is that when i check upsample & double i get no sound and xxengine stop working.
The some with the new 96.dat.

The above posts are wrong because i had the older xxengine in the folder.

Ps If i check only double it says that the dac doesnt accept the data sequence (2|32|88200)

Oh help, I am getting lost on this one. :wacko:



These both do not combine for me. Double is exactly the same as Double/Upsample regarding the "setting" of the soundcard/DAC. So, if you are correct about both above, I should be able to find something really, but I don't think it can be correct ...


Would you please be so kind to

a. install XX in its current consistent version (i.e. no mixture with other XXEngine3 versions etc., and use the normal 96.dat)
b. act as if you didn't post anything about this yet, and explain as accurate as possible what does not work, and which works in Foobar WASAPI.

I really want to (and will) help you, but currently I am lost a bit. :sorry:


PS: Your idea of using an old XX version in order to try, is actually quite good. But, you must use consistent versions always (I mean : as how they can be downloaded), and when such an old version works it really could be useful to know which one. But please don't do that right now because I first must know the exact situation at your side from the current version.

OK peter sorry for the misunderstanding. I will try to be as more accurate as i can.
forget about the above posts.

Using the latest version 09v-3 as it is after download (i don't know if any registry value must be deleted before i use it)
with the new 96.dat file.

If i set dac is 96/24 (engine 3, double & upsampling checked) and play a wav, i get endpoint create failed > the dac does not accept the byte sequence, and engine exe doesnt start. The same with only double checked.

If i set dac is 96/16 (engine 3, double & upsampling checked) and play a wav, my external processor locks at 88.2 but the output is distorted.
nothing hangs in this case.

If you want any other information please ask.

Now lets go to foobar.

After a clean installation, and using the necessery plugins ( wasapi out, and a resampler), 
i check as output wasapi, buffer 100 and output format 24bit. I set the dsp resampler at 88.2 and play a wave.
My processor locks at 88.2 and the output is ok. the same happens, if i set as output format 16 bit.

I think that if upsampling in xx was at 16 bit, everything would work fine for me. for some reason there is a bug in xx or in the driver that doesn't allow 24 bit with it.

thanks again for the help.

Ok, thank you very much. One small (but important) thing though :


Do you mean the one I uploaded yesterday, or the one received with the normal download ?

the one you propose to download yesterday.

Ok, here's some thinkering ...

I looked at Foobar again, and I found a difference. Foobar will not output in 32 bits. That is, I receive an error from that setting. Now, XXHighEnd ONLY outputs in 32 bits (hence not in 24). Why ? well, because

a. my Fireface soundcard (/DAC) won't operate when fed with 24 bits
b. I didn't hear from even a single person that he couldn't use 32 bits (per the way I do it), or the other way around, that 24 bits was *needed*.

Ad a.
Somehow it must be my lacking knowledge that I can't (or couldn't at the time) get this going. I conclude this by
1. Foobar throws an error for 32 bits hence the 24 bit setting really must be different;
2. Foobar plays those 24 bits through the same soundcard/DAC which I (in XX) can't get going.

Ad b.
You might be the first.
Also, how could one ever tell without the comparison means, which now exists (Foobar). Thus :
More people might be suffering from this, and think it is their driver.

Despite the above, and assuming it is true that your sound device really needs 24 bits (and can't use 32), it still will be rather "impossible" for me to create something that works, because I can't test it as long as it just doesn't work (kind of your situation). Also, Foobar not allowing the 32 bits looks fishy to me the least. Why ? well, because the error thrown doesn't come from Foobar but from WASAPI, and it is just the same as the one you receive when you receive the Endpoint Create Failed message. Now here we have a problem, because this implies that both (XXHighEnd / Foobar) work mutual exclusive the other way around, and that can't be ... Added to this, is that I of course know what I'm doing, so without doubt I am outputting 32 bits. And this implies that Foobar is wrong, but I can't tell in what.

The latter isn't much important, weren't it that we both try to use Foobar to work out why XX doesn't play in the same situation. So, it *is* important.
Btw, sadly, I can check the bits used only up to 24, so I can't check whether Foobar by accident has things twisted (plays at 32 when set at 24 and the other way around (which then indeed doesn't work, like with me)).

Important : Might you check things yourself sometime again, always quit Foobar after a change in the bit depth area, and restart it. So, restart playback isn't enough, and you really can't see the difference (I can with some software I have).


What I will do next, is trying to find a way to let my DAC work at 24 bits, or prove otherwise that each soundcard / DAC needs 32 once more than 16 are in order. 8)

Now I think of it ... there is another difference :

You would be using Vista SP1 (which is W2008), while I use SP0. And, since all (audio stuff) is so buggy in there, *and* Peter from Foobar says that really SP1 is needed it might just simply come to this :

I couldn't get the 24 bits going because of a bug in Vista, and now I just could. And, because I couldn't get it going, the way it tries it in the current version could be just wrong (remember, I couldn't test it because it just never worked -> chicken-egg thing).


Maybe you already told it, but what happens if *you* try the 32bits in Foobar ?

Peter i have 2 things to propose. I don't know how difficult is it to accomplish.

First of all you can try with your onboard soundcard, to work at 24 bit. All this cr*py soundcards behaves the same.
They resamples 44.1 input to 48, before send the signal to spdif. If you choose 44.1 as output they convert again to this samplerate.
So we have 2 convertions and this is what i want to keep out of, with upsampling to 88.

The second is if possible, to modify the code of xxhe, so that if someone choose 96/16 or 192/16 (at dac is) the upsampling is kept at 16 bit depth or to have the option of 16 or 24.

Anyway i appreciate your help and your hard work with this program.
For me even at 44>48>44 khz xxhe sounds much better than foobar.
It is the winner by far :)

I get an error. i can't use 32 bit BUT for me vista sp0 have bad impact to many other things including speed, latencies etc.
I can't use q=-4 with sp0. I recommend to anyone to use 2008 or vista sp1. especially all the users that have a good soundcard with native 44.1. Give it a try, i'm sure you gona like it.

Ps: vista sp1 is not exactly w2008. It is difficult for a simple user to set up 2008 but for power users is quite easy and much faster than vista.

That is exactly what XX is doing by that settings. Earlier I checked whether those 16 bits are really 16 bits (and not more by accident), but it is OK. This is why it is so strange that it doesn't work for you (combined with that it works in Foobar of course -> Foobar does those things correct as well; I checked).

And thanks for your nice words. We'll keep in touch.
Peter

Peter i think i found something...

Upsampling at 16 bit is ok. I roll back the drivers of the soundcard and now i get normal sound until q=1. at 0 i hear clicks and pops
and below 0 i get the same kind of distortion i had with the old driver. i' m little tired to test right now, but it seems that the previous driver had smaller input latency than this one. So i suppose that the only problem for now is why exclusive mode rejects 24 bit in xx.

Another question i would like to ask you, is if xx take care of the clipping that sometimes creates the upsampler in some recordings.
What is the difference between double & upsampling to double, upsampling and AA?

Ah, good ! so the (for me) most strange thing has been solved. :)


I don't know what you mean ... With the AA checkbox unticked, XX can't create clipping. The music data may already clip from itself of course (this happens in more "recordings" than you like).
If you think there is a problem somewhere in this area, please start a new topic for it, ok ?

Double is just changing the sampling rate (to double the source value) without touching the data.
Upsampling is also changing the sampling rate (double or quattro), but now the additional samples are calculated; without upsampling the additional samples are repeats from the previous.
AA = Anti Alias is a more official way of upsampling but which changes the data all over to filter out aliases which always emerge because of upsampling (because Nyquist gets disobayed then).

Thus, first you tell what the sample rate must be (double or quattro the source) and then you tell how to output that (upsampled or not, and if yes with or without AA filter).

OK peter thanks again for your help.

Is it possible that the distortion i get at -1, -2 etc is from the output buffer of xx?

I tried to explain this here (http://www.phasure.com/index.php?topic=544.msg3892#msg3892) :


You could say that the Q1 slider towards the lower levels, brings XX closer to the DAC. It influences the DAC more, which is not necessarily better, but in general (I mean for most people) yes, it is better.
It is not said at all that one is able to play at -4 because this highly depends on the latency of the system, but also on the playing software itself (if you followed the development of XX you've seen that at certain versions suddenly people couldn't play at e.g. -1 anymore and had to use a higher number).
Concluded, it is not strange at all when one can't reach -4 because things get distorted, BUT you could say that if your system can't reach -4 your system isn't optimal (and mind "system", which is a very wide phenomenon). The same counts for the upper level of Q1; it is not said that you can play without distortion at the topmost levels, which depends on many things but mainly the DAC itself.

The Q1 is a very fragile thing, and with some experience you can hear that at the lower levels - when too low - violins start to sound digital. This is no distortion yet, but a too high influence on the DAC and you could say it gets nerveous. Now, at interpreting these things one must be very very careful not to draw the wrong conclusions, because the too digital violin IMO is just the better representation of something which is digital from the beginning (the WAV file). You can hear this by sensing that all receives better detail at the lower Q1 levels, and you could say that as long as it is not about violins, it is just for the better. Mind you, this all is so fragile that I certainly wouldn't bother to slide up the Q1 when a violin comes by, but I'd know that violin can sound a bit better. On another note, I already play the Q1 somewhat higher lately (4) which has a reason of course (and I can play -4 without any problem).

Interesting for you might be the fact that the Q1 values of -2 and -3 are a kind of special to the sense of that those values don't workout in a consistent way. Hard to explain, but thinking in terms of the time something takes (is ordered to take) this time varies at those Q1 settings. More at -3 than at -2. But, since there's an averaging aspect in there (think in terms of how long something takes on average) the effect will be that those -2 and -3 values incur for even more stress than -4 (-4 is a consistent working value again). Thus, when your system coincidentally is in the critical range, it might happen (theoretically that is) that -2 and -3 can't play without distortion, but -4 can.

Lastly remember, in general we can say that the latency used is 1ms, and if something holds up for longer than that, you can't play (without distortion). For that matter, the 100ms as the minimum from Foobar could be interpreted as 100 times more headroom for other tasks in the system.
Btw, before you ask, the 1ms XX uses, has a purpose.

This has to do with the output latency, i mean 1ms is the same for all Q1 settings?
Or it has to do with data output, for example lower q1 sends, (how to say) more "accurate" signal to dac.

Another question here has to do with phase. When i hear at -4 region if i don't tick invert phase the sound is very unconfortable.
At +2 or +4 the difference is not so huge, I'm not sure what is better. Is there any explanation for this?


I think "digital" is not the right word. I saw in another post someone else use the some word, but for my ears i have a feeling of more analog,detailed and comfortable playback. I dont know if my tube amp has any influence in this. I want to hear some seperated instruments sounds to verify that. With music by lowering Q1 i get more pleasant sound.

I hope not to bother you with all these questions :)

The 1ms may vary, and it is not about sending a more accurate signal (I don't think that would be possible). The way I do it influences the DAC though, and it is the DAC that can be more accurate or less. :secret:


Hahaha, no, not one that I can think of. It is true though, and it is the exact reason the Invert checkbox was created in the first place. It *is* part of the process though, and where you can imagine that there's a scale of being in phase and out of phase - which really is about time alignment - you can imagine that there is one best setting. There's one thing I can tell you : an amount of jitter doesn't behave the same in the high frequencies as it does in the low frequencies (generally, when you have it right for the high frequencies it will be wrong for the low frequencies).
Also, because all is software-incurred :yes:, while the base for this all can be better and worse, in the earlier versions of XXHighEnd with the Q1 slider, that slider had much more influence on the sound than it has today. The principle of Q1 didn't change, but the software in the base did ...

I hope you're not an engineer, because then you wouldn't believe your own ears ...
:)

Yes, it will. But since you're comfortable at Q1 = -4 and for that matter probably didn't reach your limit, you might try the Volume slider at -6dB. There's some indirect trick working in there as well ... :evil:
(don't go further than -24dB when your DAC plays with 16 bits)

I ask all this questions just to understand how to handle Q1 and what to expect from this everytime it changes.
For sure many things changes and it's difficult to understand what is better. I understand its also hardware depended, by means that it behaves different as result, with everyones dac, soundcard etc. Also i understand that those settings, behaves different with other recordings. But for sure is less difficult to have a target and try to adjust Q1, to achive your perpose and this is the reason for all the questions.

I f i understand well, by doing this i remove 1 bit (or 2 V=-12) of info that contains noise and distortion and is very difficult to hear anyway.
It will be a good idea to set dac is to 44/18 bit, and what it will be the result of this action for this 1 bit?

That is correct.

For the program, the 24, 20 and 18 bit settings are the same. This is more for yourselves and to understand what you are setting (because 18 and 20 bits DACs exist).