XXHighEnd

All,

I am the most sorry to say, but I again found that the ever named "compiler bug" is still in there. By now I don't dare to tell anymore where it exactly exhibits, but at least I now found a normal frequency in the audio band. Well, sort of - 20KHz.

My Dirac Pulses (which are on another frequency outside of the audio band) were solved last time, but now I coincidentally ran into this frequency. And with this exact 20KHz as one of my test signals I assume there may be many more.

And so I can't advise other than switch Peak Extension On again because then it doesn't exhibit (what ? distortion).

And you know what (here comes a :teasing:) ?
I am a step further again on SQ but I started to hear some kind of nastyness. A strange sibilance in the highest frequencies. And now I think it can well have been this. I play with PE On right at this moment, and ... shivers !

What I did not try myself (say, fedup with it) is letting this loose on Windows 8. So, no no, don't reinstall W8 again. But when someone is around who has it readily available on a Dual Boot or something, please try this. It can well be that it solves a few things there ...

Sorry for this !
Peter

So for SQ Peak Extension should be switched on or off?

Hi Jud,

Peak Extension ON is my understanding :)

Alain

Yes, should be On.

I'm still using W8 in my office. Switching Peak Extension to 'ON' has exactly the effect that you'd expect from previous PE experience. The edge is taken off the sound and it comes across as more balanced, and less annoying. It's actually quite nice. Unfortunately I don't have W7 on dual boot on this PC to compare, but I could live with this sound (in my office at least).

Could it be that W8 is actually more transparent than W7, and what we've been hearing are the side effects of compiler bug issues with PE='OFF'? Seems a long shot to me, but maybe worth investigating further. It's prompted me to sort W8 out on my main system again. I do have W7/W8 dual boot there, but for some reason, my W8 seems to have become corrupted - I think by physically disconnecting the HDD whilst in Min OS mode.

Will report back once I've had a chance to compare, but probably won't be for a while.

Mani.

Hey Mani - that is interesting I will try it tonight. If it is possible to lose the "edge" in W8 that would solve two problems in one for me.

Paul

Peter, is this also true of my Vista version 0.9z-8f?

Yes Charlie.

This is slightly embarrassing - I just checked and found that I have been listening to W8 and W7 with PE on without realising it. I have been concentrating on all other settings so much that I missed it. No problem though.

Paul

I believe the following two statements are correct.  Can you confirm?

This compiler error bug doesn't manifest:

• 1)  if the interpolation slider is set to 1x.
• 2)  if no interpolation method is selected. (bit perfect samples are repeated)

=====

What about if L or AI are selected?

Charlie,

Without hooking up the analyser and so by NOT 100% guarantee :

•1)  if the interpolation slider is set to 1x.
I don't think that will matter (problem still occurs).

•2)  if no interpolation method is selected. (bit perfect samples are repeated)

From what I last found, even that exhibits the problem, although this is not what I had in mind. I could also say : for my 20KHz test signal I tried to find one situation which worked, but I did not find any. Please notice though that with the NOS1 all combinations double because there's the "Is NOS1" mode and "normal" and this all can make a difference.


Same story, although I sure did not try Linear. AI is more difficult to test/compare because it only upsamples to 192KHz max.

There's just something messing up with the Volume and where -somehow- the base "factor" of 1.000000 does not turn out like that at all. Also, this is actually a known phenomenon and something like 1.000000 actualy is 1.000001 or something like that. And then for insiders : The maximum (think 16 bit domain) is 1 x 32767. And you know what ? 1.000001 x 32767 = more than 32767 and this means that it will become negative (-32767.032).
Now, everybody will say "come on, this is solvable !" and it is. For example by turning the standard 1.000000 into 1.001. But now the volume will not be "losless" anymore and when I sneaked that in as a solution a couple of versions back, people complained about SQ ...

Peter

Peter, another question about PE and HDCD...

Since 0.9z-8-2, I've been drawn more and more to HDCD recordings. I've said before that there's a 'sweetness' to them that I really like. I think my system is finally resolving enough to distinguish between good and bad ADCs used during recording/mastering. There's a lot of variability between the sound of various recordings of course, but HDCDs sound consistently good.

So what settings should be used? I've tried 3 different settings, with totally different results:

1. PE and HDCD on. Very laid back, smooth sound. Easy on the ear, but lacking a bit of sparkle.

2. PE off, HDCD on. Too bright and too forward. Slightly aggressive leading edges. (Might this be due to VC issues?)

3. PE on, HDCD off. The best sounding. Very smooth but with the sparkle intact. Maybe just a tad too 'holographic' sounding.

So, 3 is my favourite. But I'm worried that I'm missing out on the correct HDCD decoding.

Any thoughts?

Mani.

Mani,

The solution to this will be a very different, fresh and unexpected one. I will announce this in the Release Notes to 0.9z-8-3.
Prepared to be shocked. The most shocked ...

Peter

PS: And no, this is not about the "tweak" you already know about. Hahahaha.

Hi Mani,

Maybe irrelevant as we are about to be shocked  :o, but there is one combination you didn't mention which is both PE and HDCD 'OFF'. This is my personal favorite, as to my ears the sparkle is there without leaning to aggressive.

Cheers,
Diede

Hmmm... it's complicated...

Peter says we need PE 'on' in order for a truly lossless volume control. Certainly, with PE 'off' the sound is more lively and has more 'sparkle' - but is this the dreaded 'fake detail' that we can all get caught out by? Don't know.

As for HDCD, well it depends on how the Pacific Microsonics machine was set during the mastering. There are two HDCD processes; either, both or neither of which could have been engaged during the HDCD mastering. I'm not sure if the HDCD plugin that Peter uses (I'm assuming HDCD.exe) is intelligent enough to do the job of figuring out which of these three possibilities has been used. In any event, if either or both of the HDCD processes have been used, then HDCD should be 'on' in XX. Replaying these HDCD-encoded files without correct HDCD decoding just sounds wrong to me... all IMHO of course.


There are only a few things coming from you that would truly shock me:

1. AI is shown to be better Arc Prediction. (Nah, never.)
2. The VC is shown to be always lossy and you have started to advise people to use a preamp. Hmmm, could be possible. I still wonder if a totally transparent preamp (maybe passive) exists, in which case XX could be set permanently to the PCM1704s' ideal output level (-9dB?).
3. XX is no longer a true 'memory' player. Again, could be possible. Low SFSs are now working very well. Maybe the ideal is an SFS of zero?

Ha, this is fun. Anyone else with any ideas of how we are going to be "the most shocked"?

Mani.

No takers? OK, here's another one from me...

4. XX now converts everything to DSD. If this were the case, I really would be "the most shocked".

Mani.

Well, we've had hints of laptops and wifi....

XXHE broadcasting DSD over wifi for Win8, Mac, and Linux laptops?

And the splitter bar doesn't move?     :swoon:

I tested this last night and for HDCD recordings "3" is so much better than the others. Silly of me I never deactivated HDCD before.

But also this setting sounds *louder*. I'm not sure if it's really louder, or it's because of increased dynamics.

Stanley

I have the odd experience that the detrimental effect of HDCD "on" is also audible with non-HDCD material.
When cofiguring a new XXHE version this is usually the first feature I disable.

regards, Coen

Maybe an interesting sort of side-step :

With the notice that I disregard PE as a really influencing factor, except for the very subject of this topic (meaning : just count this out), I noticed this, and I notice it for a longer time now :

Q sound (think Amused to Death, Immaculate Collection (Madonna) and some Sting album) ... does not sound right anymore these days. Remember, this is severely manipulated (phase) stuff.
I don't need to be right on this, but I just noticed it and though that perhaps it is now very much audible when too much has been changed to the sound. Of course this is nothing much different from severe "DSP'd" mixing and such, but it sort of goes along with superb sounding albums from era's where no manipulation was done.

Maybe the HDCD manipulation (with actually no real idea about the severety of it) also is too much of it.

I'd have to add that I never A/B'd HDCD albums with HDCD decoding On/Off, BUT I always find it the most noticeable that HDCD decoded (so ON) sounds so much more silky. Of course this can resemble this :


... which can be about the less sparkle just making it more smooth (silky in my words). I just never compared.

It has been said before, it is so, so (SO) difficult to differentiate freshness from distortion that we all can be on a wrong track for really long times (and I sure don't exclude myself from that). But at least for me, slowly, I start to recognize from other properties of the sound what is the "real" thing in this realm.
I could say ("state") that less sparkle never is good, if - and only if the more sparkle can be sustained as OK throughout your whole collection (within reasonable limits).

Notice the word (or phenomenon) "sparkle" here, which actually is a very good description of some more of these properties involved; for those with the experience, compare to Windows 8. No single way I would dedicate that any sparkle as such. Hard, to the point, sharp ... those would be descriptions for W8 and no way I mean these as negatives (to get the meaning of the words how I mean them). But sparkle ? sparkle contains colours, warmth, maybe even love - and most important as how I see it : colours in music notes. Call it timbre. So, even piano notes can contain sparkle as long as the harmonics are sufficiently shown.

I think you can well say that the more sparkle the system exhibits, the better you will hear colours in cymbals. So, I think the both go together. With W8 there is no sparkle and cymbals show the high frequencies only (too few colour) while at the same time I'd call that hard and pinpointed. But sparkle ? never. Sparkle sings. Sparkle makes inherently (??) too close high frequencies like with Enigma sing instead of being grey. Just like cymbals must sing, and the more they do the better you see the sizes of them.

And to (attempt to) complete it : when things are to "hard" (W8) frequencies get too high compared to how they should be. With too much filtering (just the word, no other meaning), things get dull. When too high frequencies are made dull, they become more pleasing to listen to. But not better. And once you are used to the higher frequencies, your brain can't cope with the blanketed sound.

Sparkle is in the middle of this all ...

Peter

Well, aren't we all ready for the mad house by now.
So, no way there's strange or buggy code or anything involved. However, other processes run and since everything matters this can matter too. To explain this, think like this :

You play an album and use Normalized Volume. In this case it wasn't Normalized yet, so that has to happen first. Now *that's* a process of some "size" that won't happen the next time you play this album. And these things matter. Some more, some less, and depending on what it is what's being done.
HDCD decoding - just examining whether it has to happen - is also such a process, though tiny compared to my example.

So, don't lift off to the madhouse yet.
But things do get quite crazy these days, and I'm afraid it won't get less ...

Peter

I have the same impression. Something is going on.

Joachim

Yea, but you were already banned to ...
well, somewhere far away. Or ?

hehe

So BTW - and my apologies in advance if this is something that has been explained repeatedly before or is in a tooltip I should have paid attention to: Does "Q" have anything whatever to do with the Q of a filter?

Also: Might be interesting for folks to look at the iZotope SRC thread over at Computer Audiophile.  Has some interesting stuff, and speaking of "sparkle," I just lately made a comment in that thread that talks about how we are trying to describe (in that case, and partially also with XXHE) digital filter effects so inadequately with language from back in the analog days when all we knew about changing was frequency response and a little about time ("transients" and such).

No Jud, Q just stands for Quality. It originates from special and (still) secret WASAPI settings, while for Kernel Streaming it's nothing more than a part of the Buffer Sizes setting.

Thanks,
Peter

Cool, we three, Mani, Coen and me are then ready for the madhouse ............. .

Never had time for a serious A/B comparison, but now …… .

The release 0.9z-7.5 has the best “sound staging”, “tonal balance” and “coherence” with the right size of the instruments.
(see a previous note by Stefano too - http://www.phasure.com/index.php?topic=2490.msg25926#msg25926)

On my system I get the best results with the memory management “mixed” (better sound staging).

I’m in a concert by Chick Corea tomorrow and then I can compare XXH with the real Live.

Joachim

Don't apply Phase Alignment???

Well, consider me totally shocked.

Mani.

PS. Bert, I want my money back on those caps you put into my amps ;-)

PPS. Will try 09z-8-3 with Peter's 'ultimate' settings (and no PA  :o) when I get back home this weekend.

Very glad To READ that .
BTW I am the first downloader of the new version :-)

s