0.9y-4 - w00t

[manisandher]

With some nice Jamaican inspiration I just thought of a way to uncompress compressed albums ...

Well, there's nothing like Jamaican inspiration Considering the prevalence and level of compression nowadays, this would be really

... if 6dB would be needed to extend the peaks, I don't know of any album that has that headroom

No, to understand why they apply 6dB attenuation to 16-bit material in the BADA, you need to understand the 3 HDCD 16-bit amplitude encoding options. To avaoid going too OT, I will give a brief overview in a new thread.

In any event, none of this has anything to do with Arc Predicition, other than the fact that using AP on 16-bit material will switch off the 6dB attenuation in the BADA.

Mani.

[PeterSt]

In any event, none of this has anything to do with Arc Predicition

It has though ...
This needs the "knowledge" of how the 16 bits domain operates, and there is no way peaks can extend 6dB without attenuating the whole lot 6dB first. Thus, from your other topic (thanks !) :

allows peak levels up to 6 dB above standard full scale level (+6 dBfs) on HDCD 16-bit recordings without generating “overs

can not be done at all, unless you believe in ghosts. Just for the 16 bit domain (mind that !) this would mean : drop the least significant bit hence attenuate with 6dB, and then allow the soft limited peaks (on purpose with HDCD) to extend to -0dBfs.
+6dBfs does *not* exist. It can't.

This all sure is related to A.P. because the limited peaks have to be extended (and will automatically !), while it first needs the headroom in order to be able to that (which latter currently does *not* happen).

What it comes down to is that A.P. does not need HDCD (knowledge) in order to work so anyway. The limited peaks are sufficient knowledgde to extend them (which currently *does* happen, until -0dBfs is reached).

I don't have a HDCD decoder, but you will see in the next version (and if I understand all correctly ) that you won't need HDCD decoding in order to "decode" it anyway. But, you will also see that non-HDCD encoded matertial will be worked out as good as HDCD, if only soft (or hard) limiting has been applied. And this is true for most of the albums (see my before post before I'm repeating everything of it). One thing : it will be *better* than HDCD decoding, because HDCD will loose one bit at least, while I think it is two. A.P. will not loose enything, if it's only *non* HDCD material.

Notice that this only counts for the Peak Extension, and not for HDCD limiting (worthless IMO) and certainly not low level extension (which just destroys the original sound).

We'll see.
Peter

[manisandher]

In any event, none of this has anything to do with Arc Predicition

It has though ...

Sorry, I didn't make myself very clear, did I? I was referring to the 6dB increase in volume that BADA users are experiencing. This isn't really down to AP reconstructing lost peaks. As you said, this would be pretty incredible...

Mani.

[manisandher]

... the samples (sample rate) didn't allow the proper peak, and that peak is now added (restored). However, when the track has been mastered at (or near) peak level already, that level is extended a little. But, it can't extend, so what happens instead is that the already cut peak gets widened a bit, and while your meters may not trip on a few samples of maximum value, it may on a few samples more.

Notice that things could really improve for SQ again if I first drop the volume a bit and *then* apply QAP.

I totally get this. And if it beats HDCD's Peak Extension, then great. By how much are you hoping to drop the volume before allowing AP to kick in?

Mani.

[PeterSt]

As how I think of it now, I will do it the other way around : just calculated the peak values (no matter they will become + dBfs) and then lower the volume accordingly.

But as you will notice once 0.9y-5 is there, much much more is going on.

[manisandher]

Firstly... Happy New Year to everyone!

Now, I had some time this morning and I really wanted to verify Peter's comment below:

I don't have a HDCD decoder, but you will see in the next version (and if I understand all correctly ) that you won't need HDCD decoding in order to "decode" it anyway.

I chose an HDCD track which I know has been mastered with the HDCD options engaged (from the Jacques Loussier Trio playing Satie).

With no upsampling and vol at 0 on XXHE, the HDCD indicator on my DAC lit up, as expected. The sound was very warm and lush... listenable, but not very exciting or foot-tapping.

I then set the vol in XXHE to -6dB, keeping everything else the same. My DAC automatically disengaged its HDCD processing (the LSB had changed after all) and increased its output by 6dB (all as expected). I expected to hear a decrease in SQ... BUT this just wasn't the case. The sound was actually clearer!!! More tuneful. Everything started sounding more real.

Then on to QAP with vol still at -6dB in XXHE (but making all the necessary changes elsewhere for my DAC to accept 24/176.4K). And there was yet another step up in SQ!

But the killer was when I then increased the vol in XXHE to -0dB (and decreased my analogue volume control by as close to the same amount as I could). Now, this sounded very, very good... substantially better than when I started out.

The only thing I can really conclude from all this is that how attenuation is applied is absolutely critical.

Also, don’t worry about buying HDCD material and not being able to decode it properly – it sounds wonderful played on 0.9y-4, with or without upsampling.

Mani.

[PeterSt]

But the killer was when I then increased the vol in XXHE to -0dB (and decreased my analogue volume control by as close to the same amount as I could). Now, this sounded very, very good... substantially better than when I started out.

Assuming this has been the only change compared to the last described situation before, the least I can say is that analogue volume works differently from digital volume. So, it hardly can be said that the digital volume degrades in this situation (-6dB), but the additional analogue volume may just do the trick. It can work the other way around just the same and notice this depends largely on the amount of attenuation needed in the first place.

Also to keep in mind (for those trying) : something like a TVC is quite unlinear (regarding freq. response), so in that case it is easy to find a setting that sounds quite different from another.

Thanks for sharing Mani. We keep on learning (and improving from there !).
Peter


PS: I only now realize that where I said yesterday "I love the sound from Kernel Streaming" (W7) it is also the (QAP) "peak extension" engaged now which may do some tricks. I didn't A-B, but I applied it together with hopping over to W7 and completely forgot about it because of W7 itself (but at first I thought the peak extension didn't workout at all because of the WASAPI sound of W7 (again, can be personal)).

[bernharddiner]

Hi all,

my first post here 

I hope I won´t confuse more as I can contribute by now.

Thinking about my overall very positive experiences with the y-4 (which is at the same time my 1st XXHE version) I thought first that I should collect at least some 8 ... 10 listening sessions to make me familiar with XXHE. But after approx. 50% of those I have the feeling that I should share with you following:

1. my current favourite setting is NO OS in XXHE at all
2. followed by the 2x AP
3. and finally the 4x AP

This ranking seems to remain independent from the other settings.


- The #1 above delivers in my case (my individual taste & setup with a Weiss DAC2) at most the feeling of an analogue/real acoustic atmosphere.
- #2 seems to be more dynamic/accurate, with tighter impulses in bass, but seems to loose transients and esp. the feeling for the space in comparison to #1. It seems to drift in the direction of "digital" - impressive but virtual
- #3 finally goes along the path of #2, but significant further down the road and I can enjoy it only with an inverted phase - why ever!?

Nevertheless: XXHE, despite the lack of comfort in comparison with esp. MediaMonkey (with iTouch), Jriver and Foobar - is for me currently in terms of SQ the most enjoyable way to listen music reproductions; in addition to the good old analogue way.

It represents for me at that moment also the hope, that crossing the stream to the "Mac side" is not an obligation, when you´re searching for the SQ optimum.

Thank you Peter, so far!
Please proceed that way and don´t forget the "good upsampler DACs" - even if my impressions might change later on!?

Cheers

Josko


PS: My OS is Vista on a dedicated music notebook via Firewire to the DAC.

[manisandher]

Hi Josko,

Thanks for sharing your experiences. Actually, your order of preference is exactly what I would expect from someone using an oversampling DAC (which your Weiss is, along with the 99.99% of all other DACs on the planet).

What you're experiencing with AP (I think) is your DAC doing unnecessary oversampling, once this has already been done in XXHE. The point here is that with 4 x AP (I like to call it QAP), no more oversampling really needs to be performed - everything 'nasty' is well outside the audio band. However, delta-sigma DACs have to continue oversampling in order to work. Many multibit DACs also continue oversampling with 4fs material... BUT my contention is that they don't have to.

Although I've never heard it, 8 x AP should be even better... played on a NOS DAC... that can handle 24/352.8K.

In any event, I think the one thing we all agree on is that XXHE is a great player... Welcome to the club!

Mani.

[bernharddiner]

Thank you Mani,

considering my "theoretical expectations" I am/was not really surprised and even less disappointed (with the DAC2  ) + share your opinions so far as I can follow ...

Finally XXHE is an additional attractive ball in the audio game and the playground has never been so broad as these days.

Apart of the good old analogue world: obviously the period has aready started where software will dominate more and more the setups and the related reachable results.

Josko

btw: Chapeau! You took a legendary name

[manisandher]

btw: Chapeau! You took a legendary name

For the record, I want all our German friends to know that I do not drive an Opel Manta.

Mani (not short for Manfred)

[bernharddiner]

... even I am not a German - sorry!

To become serious again: during the comparison btw. XXHE and the known Win alernatives (JRiver, Foobar, MM) I have got the impression that XXHE depends less on formats (in my case FLAC<=>AiFF). With other words: by switching from FLAC to AIFF I observed with the classical software greater improvements in SQ (or less rel. difference in SQ) as with XXHE; but don´t have any clue, why.

Josko

[Per]

....by switching from FLAC to AIFF I observed with the classical software greater improvements in SQ (or less rel. difference in SQ) as with XXHE; but don´t have any clue, why.

As far as I know XXHE convert everything to WAV before playing it. That will explain your observation.

Per

[bernharddiner]

Sounds reasonable. If I remember well, I have seen something written about that.

thank you Per!