SQ of 09-z2

[Marcin_gps]

Why don't you share your thoughts about the latest version of XXHE? Personally I prefer z1.

[Telstar]

Why don't you share your thoughts about the latest version of XXHE? Personally I prefer z1.

Why?

[PeterSt]

On my part nothing has explicitly changed !
Which is no guarantee it didn't ...

[Marcin_gps]

Why don't you share your thoughts about the latest version of XXHE? Personally I prefer z1.

Why?

Maybe because z2 crashes / stops playback very often :D

[PeterSt]

Only if you perceive the sound of a crashing airplane or something, this can be related to SQ, don't you think ?
hehe

[ivo]

z2 = nice piece of code. pretty stable here at my rig. Fabulous space and 3D stereo is coming out of my speakers. Thanks.

Ivo

[arvind]

Hi Guys,

Ever since Adaptive Mode, I have been using it & had always found it better than the other modes. With 9z-1, I thought let me try out the other modes, for comparison & I was surprised to find the Special Mode works by far better, in my set up, of course. The bass is tighter, the mids are cleaner, top end is about the same.
I presume it would be the same in 9z-2 too as Peter hasnt indicated any change made in this area.

Arvind

[pedal]

Hi Guys,

Ever since Adaptive Mode, I have been using it & had always found it better than the other modes. With 9z-1, I thought let me try out the other modes, for comparison & I was surprised to find the Special Mode works by far better, in my set up, of course. The bass is tighter, the mids are cleaner, top end is about the same.
I presume it would be the same in 9z-2 too as Peter hasnt indicated any change made in this area.

Arvind

SQ of Adaptive Mode vs. Special Mode is very much depending on your setting of soundcards latency. With 09-z1 and 09-z2, Special Mode sounds best with lowest possible latency. With Adaptive Mode latency should be increased to something like 1024.

This is the general consensus here, as far as I have understood. (Or, you might say that with the myriad of parameters, setups and user preferences of XX and its users, the only consensus is that there is no consensus). 

[PeterSt]

Well, at least I agree with it. I must say though that at finding Adaptive better (some weeks back), it is time to hop back to Special again, and see which of both I now prefer.

[PeterSt]

Allright. Three evenings of playing further, and I must say that Special Mode is definitely better. But as far as I could judge in this time period, it takes some special settings you won't be able to perform at this moment ;

All starts with my own "desire" to play with 32/352.8 (or 384), which I myself (with my not the fastest PC) can't get done properly with the perceived necessary low latency without glitches here and there. Of course, this is not all that much different from when we startet out with the ultra low latency playback, but while (I guess not only me) accepted those glitches with the positive tradeoff of better sound, by now I am used to the good sound anyway, and the glitches now disturb. Oh, it can be without glitches (or merely small ticks), but it is so fragile that it depends on the mood of the PC or whatever it is. and I don't want to run into this.

The acceptable low latency for me is 64 samples at a buffer size of 512, and with the 32/352.8 playback this depends on the mood of the PC.

Before proceeding, first what I think of the sound with those settings :
For "normal" music it is hard to distinguish from Adaptive, Q1=1 and a buffer size of 2048. This firstly means that the warmth and emotion from this latter setting can be achieved with the ultra low latency just the same. Notice though that I dedicate this to the Phasure NOS1 in its latest setup for a large part, because before this wasn't the case (Special @ low latency was too sterile IMO). However, the in fact outrageous difference emerges at playing the better ambient stuff, like Shulman, Ott and even Jens Glad. The difference with this better ambient music is the high transient information which is in there, generally coming from unnatural (high transient) synthesizers. So, this is about the "strange" sweeping sounds, which are not there at all with Adaptive Mode - or are excessivley there with Special Mode.

Thinking further, I guess this is what Adaptive Mode is giving its better warmth, BUT -and it luckily always goes like this- which will be the case when not all is 100% right. In other words, when the higher transients are not processed too well, they are smeared, and credit to the overall sound quality. This would happen with Special Mode and the not 100% operation. With Adaptive Mode the high transients won't be fed to the DAC in the first place (well, that is what I think of it), and smearing is not in order. But high transients neither !
So keep in mind, to me it seems that it is now highly dependend on the DAC (and possibly more), and never mind the details which must be at sub-micro level, the net effect is huge.

Now, the way I got it working glitchlessly, may be of special interest for Marcin.

To keep in mind, this is about the difference where ticks and glitches are audible regularly, and where bringing up XXHighEnd (from Unattended) will show a lot of crackling on one side, and totaly nothing the matter on the other. Well, this is the other side :

I still have 64 samples of latency, and the buffer size is still set to 512 (btw, I didn't try lower, but I guess lower is possible now). The output is still 32/352.8;
Where the Phasue NOS1 is full of tricks, I applied one of them, and this is letting the driver run at 88200. In the mean time though, the DAC runs at 705.6, which thus is twice as fast as the "with glitches" situation.
Of course everybody is confused now, because the "resolution" output of the DAC is only 352.8 in this situation. But this is about my applied trick (see the DAC thread) at achieving better jitter figures (theoretically 6dB max). Anyway ...

Here we go again : it is the driver which lets things choke, because at twice the output, but letting the driver run at half the speed (with still net twice the output), all is infinitely better. And this includes SQ too - apart from the joy of "being glitchless".
And better not ask about my tricks ... I'm full of them. Haha.

I think normal audiophools ends up in bed late when things sound really good, but since I never play music after dinnner, in my case dinner is just a little late. Like yesterday, 11:30pm.

If I am right and not on dope, this is just pure extasy. I must emphasize again though, that there is music beyond normal music. This music -as far as I can tell- only expresses in special ambient. And this is the most rare. If I have 20 of such albums it will be many. It is not just synthesizer music, because all what is commonly known (Vangelis, JM Jarre, Schulze, Eno, 50 more) doesn't have it. At all. It's the ones I mentioned. Add Shpongle to it. It takes digital synths with sufficient bit depth (the early Vangelis days used 8 bits, to name something), and a high amount of intelligence to make this music special. Notice that even the good old Yello doesn't come close.

I want to ask Roy - who I dedicate to be our modern ambient specialist - whether he knows more, the above mentioned in mind. It may be hard to recognize if you don't have the sound yourself, but since it was Roy who put me to the mentioned artists in the first place (some year ago), maybe by now he knows more of them. Good to know : Carbon Based Lifeforms doesn't cut it either. Each 60 minute trip without using is welcome, and when you are not sure, mention it anyway and I will sort it out. Thank you !

So far,
Peter

[Marcin_gps]

Hello Peter,

Long time no speak :D I've been busy and enjoying music - at last! What's worth mentioning: I gave up the idea of second, dedicated audio server, it was far from convenience. So, I decided to switch back to my regular, fast PC and install my Cantatis@ESI sound card inside. What the hell? Now it sounds better here and that's without ANY OS tweaks and undervolting/underclocking. You may say, that I've gone crazy, but now that I think about it, it makes sense. Why? Because I discovered the great ESI drivers/firmware on my old PC, which allowed me to run smooth playback with 1 sample latency. After going back to my new PC, superiority of motherboard and PSU quality and it seems that motherboard and PSU brought new, better quality - and I'm not talking about small differences. So folks, quality motherboard and PSU! This is the base of good pc audio. Of course, that is if your system depends on PCI/PCI-E or even USB (without external PSU) digital transport, so basically 95% of all available interfaces.

Peter, unfortunately I can't set fixed sample rate in my drivers control panel now, because it uses auto rate switching. However, it wasn't possible before AFAIR.

Cheers,
Marcin

PS
Somehow my activation code works on my new PC as well, yeaah. I thought, you were gonna kill me for asking again, but noooo 

[PeterSt]

I guess one day you wil be definite and sure about all Marcin. But the best is that you listened to music. Out of al I think I know what the difference is between ever trying to get things well, and listening to things well done.

unfortunately I can't set fixed sample rate in my drivers control panel now, because it uses auto rate switching. However, it wasn't possible before AFAIR.

I am not sure why you say "unfortunately", but I am used to "auto", and as far as I am concerned, it should be like that. But I had it the other way around one time after a reinstall of the ESI drivers. Nothing else to do than reinstall them again and all was good. But good = "auto".

or ?

Peter

[Marcin_gps]

I was referring to your 'trick'. I can't imagine how could I set the driver to run at half of the speed, but I'd get my hands cut off if someone told me how to set it x2, because I can't play 48kHz and multiplier sample rates (49.152 clock instead of 22.5792 on my Cantatis). Peter, tell us how did you do it?

PS
Yes, I'm able to set the playback to 768kHz and it works, but 'in slow motion'. Isn't the driver suppose to run at double speed instead or am I confusing sth?

[PeterSt]

Ah ... no, you can't do it. It is XXHighEnd doing it internally, and next it coorporates with the DAC I have here.
So this is only possible with XXHighEnd/Phasure NOS1.

The driver can't go higher than 192. But as you can imagine ... when I can go higher anyway for the net result, I can also lower the driver (and have a higher net result).

Full of tricks ...

[Marcin_gps]

After a while I got tired of all that details coming from Special mode and ultra-low latency. I gave it up and went back to Adaptive and 4096 samples. And the most important thing - player priority: low, thread priority: nothing. This gives the best sound imho. Thread priority at realtime is definately more detailed, but if your system is highly transparent, it will give you fatigue instead of joy. I prefer fun

[PeterSt]

Allright. Three evenings of playing further, and I must say that Special Mode is definitely better.

I went back to Adaptive Mode again after a couple of days. Indeed it is so that too much detail can make you restless or without peace or something, but I don't think it is that with me. Adaptive just feels better. It doesn't make me *look* for details or other technical stuff. It makes me listen to music.

Thread priority at realtime is definately more detailed

Yes, I can imagine that. Lowering the Thread priority would be following the same line.
But (for others too) be careful you won't receive glitches at track boundaries from lowering the Thread Priority.
I will try it too ...

[Marcin_gps]

Try it, try it...
As long as Player Priority stays at low, there are no glitches. I'm glad that am not the only one here changing the settings like socks :D

[ivo]

I can agree with that Special now is better than Adaptive. Adaptive is kind of smeared, but Special is more precise and concrete. Bass is tighter, mids are closer and highs are a bit more sparkling. Downside is that Special needs more CPU as we go with lower Q1 values. The lowest I can go is Q1=-2, but then my E6300 rocks around 30%.

As I have found that HiFace uses 2048 sample buffer then I set Special mode to 2048 and tune Q1=0. It gives me 16 samples of latency between XX and the Hiface and I am happy with it for now.

Ivo

[wushuliu]

I can agree with that Special now is better than Adaptive. Adaptive is kind of smeared, but Special is more precise and concrete. Bass is tighter, mids are closer and highs are a bit more sparkling. Downside is that Special needs more CPU as we go with lower Q1 values. The lowest I can go is Q1=-2, but then my E6300 rocks around 30%.

As I have found that HiFace uses 2048 sample buffer then I set Special mode to 2048 and tune Q1=0. It gives me 16 samples of latency between XX and the Hiface and I am happy with it for now.

Ivo

I am new to XXHE and am still learning about the settings. The first few days I played around in Adaptive Mode trying out different thigs. Yesterday I switched to Special. I had to keep Q = 1, any higher and there was distortion. However, the sound quality was as you described - more precise. But I did not find it analytical or detailed, just cleaner, maybe even slightly, very slightly rolled off at high frequency. The sound is not as wide as I heard in Adaptive, but there is more depth. To me it sounds like there is a reduction in background noise. It is the closest to vinyl 'analog' I've heard so far in HE, or any music program. Amazing.

[Marcin_gps]

Folks, it seems that all of my previous combinations was WRONG. I tried different priorities to find the best sound, but neglected the split size setting. I thought that setting it high would be the best, well, it's NOT! Try it yourself. Quick comparison between 50 and 2000 MB would give you the 'image'. Small value is speed, speed and detail and nothing fatiguing about it, while 2000MB seems warmer at first, but it's just dull pulp where everything is hazy. I wonder how much is to discover...

[PeterSt]

Yeah, ain't it nice ...

I wonder how much is to discover...

I must say that I knew this, but it is a bit of a theory, and completely dependend on how your PC (mobo) is sticked together. I mean, there's always stuff like higher memory addresses which need to be addressed indirectly ... memory banks which don't work the same (but a first always does), so indeed I always stick to a low amount. But it's hardly science ...

So to be clear : this sure won't count the same for everybody !

What a hobby.

[Marcin_gps]

Another remark from my side:
I prefer Scheme-3 only with upsamling. Plain 16/44 sounds best with No-Scheme

[ivo]

OK, after getting some knowledge from Peter about the buffers i did some listening tests with latest XX in KS mode. All tests are unattended with Dev buffer = 2048 (original Hiface). Did use Chesky demo disc (24/96 files).

1) Engine4, KS:Adaptive mode, Q1=1
Sound is wide, individual instruments are kind of playing each at its own location if looked at the stereo image.

2) Engine4, KS:Adaptive mode, Q1=30
Sound is wide, individual instruments are kind of playing each at its own location if looked at the stereo image.
Compared to 1) it is duller. Bass and voices have a little like dullness.

3) Engine4, KS:Special mode, Q1=1 (16 samples)
Sound is not so wide, it is a bit narrower than with Adaptive. It is a bit thinner, but more towards listener. Individual instruments are now more in a concert rather than each at its own location. Voices are clearer and closer to listener.

4) Engine4, KS:Special mode, Q1=30 (2032 samples)
The same like 3), but seems like getting more full, however maybe too much...

5) Engine4, KS:Special mode, Q1=7 (the min value where I got 2048 samples for XX buffer)
Choose to stick with this, as it seemed to me the most i like.

Anyway, to me it seems like special mode is sounding like altogether, but the interesting thing is that in special mode i start loosing the feeling of a stereo. I mean in Adaptive mode the stereo effect is easier to hear. However, in special, the sound is deeper and I guess highs are better pronounced.
Actually the testing was done to understand how the XX buffer length impacts the SQ. To my understanding in Adaptive mode, shorter buffer is better, but in Special just the opposite.

One last question to Peter: In Adaptive mode when Q1=1 and Q1=30, how many samples are these values?

Ivo

[PeterSt]

One last question to Peter:

That will be the day !

With Q1 = 1 it will be the Dev Buffer Size. Anything higher for Q1 will have "some" higher amount, but it is nothing to work with or do something with. But it's also not meant to do something with. But think something like each higher Q1 value adds 10% or so, and its not even linear.
With a higher Dev Buf Size setting you can reach 40000 easily, but I think I limited the upper range to 1 second, because at least XP USB devices can't go over 1 second (which would be 44100 if you play 16/44.1).


Ivo, I must admit, it must be rather tough for someone like you -who just wants to know it for the better results- to get the answers from someone like me who doesn't want to expose everything and all. But I'm doing my best to make it as clear as possible and with the honest figures, so *if* you are working with them, that it makes sense (it would be very wrong to let you work with different figures from reality).

Regards,
Peter

[ivo]

Thanks Peter!

I just want to understand what I am turning I mean what this control like Q1 actually does. That is my nature, I am an enigineer.

Ivo

[Marcin_gps]

I went further and decreased the split size value to minimum, that is 12 MB and have to tell you that it is even better now. Beside the fact that the speed is great, the stereo went '3D', bass is tighter and more powerful. How could I be so wrong before. You have no idea (or maybe you do :D) how much time I've spend on fine-tuning the settings in XXHE But it's good -best sound I've ever heard from my system 

Peter, do you think it can improve further? I mean XXHE capabilities.

[PeterSt]

Hey Marcin. Again thanks for sharing this all ... something good may come from it indeed.

Peter, do you think it can improve further? I mean XXHE capabilities.

Well, as I said a few days ago, I already had the feeling/idea that using a specific "layer" of the memory would allow for improvement;
At this moment I can't be sure whether it is exactly *that*, you prooved to be the cause of better SQ, or that it's just resource related, but eaither reason should theoretically be able to lead to better SQ. On this matter it better be the memory, because the resource related thing will be much more tough to control *and* will differ per user (PC).

I'll start a bit of experimenting in your directions and see what I can make of it myself.
In the mean time, I can only encourage others to do the same, once again referring to the major importance of equal views on this, making it "law" automatically.


Peter

[boleary]

First impression of split file size of 12 v. 60: 12 is brighter everywhere but also sounds more thin, more like Engine 3, no kernel streaming. 60 sounds less thin with a fuller mid range. Again first, very quick impression--was leaving the house for work. Will try again this evening.

[Marcin_gps]

Yes it's thiner, but try the buffer size at 4096. This should make things better or not :D Maybe you will like the sound of lower values, it all depends on your system.
I wonder if I can have both - details and rich, powerful sound.

[boleary]

Gotta live with compromises....I guess. After playing several files many times the "best" sound seems to come with different split-file sizes. Generally, female vocals and violins sound better with higher sizes, 120 to 200 (Patty Griffin doesn't sound bad at 4200, though 200 is preferred!). Bass guitar and piano like lower values, 12 (Oscar Peterson, Night Train,24/96) to 60 (Bass Face Trio plays Gerschwin, 16/44). The age and manner of recording might be part of the reason the Oscar Peterson sounds so good at 12.....don't really know.

Another generlization may be that at lower volumes 12 is probably the best size.

 

[Marcin_gps]

I play with PeakExtend on. I also lowered the buffer size to 1024 samples and found it richer. 4096 is great, but I think you need very powerful amp to get enough kick. So where am I? It seems that I made a circle and  went back to my initial best settings, only with different split size