0.9y-4 - w00t

[GerardA]

Just curious: Bitperfect, first 16 bits?

Does Arc calculate the extra bits from 16 to 24/32 or are they padded with zeroes.
Because only then it makes sense that double can do so much...

[SeVeReD]

happy because from what I've heard, lower sample rate

Uhm ?

Furthermore, I don't know. The only thing I know is that IIRC 64 always worked for me at 24/96 but I always set it to 96 for safety. At this moment I don't know, because I don't have the FF connected to the music PC anymore.

Safety?! yikes  Anyway, I said lower was better because when I've left it at 128 sample (after listening to 24/96 and forgetting to set it back) and listening to regular 44.1 in the past it did not sound as good as when set back to 48 samples ... so, just figured lower sample rate was better.  Anyway, running it at 64 samples seems to work well and FF settings says it's a 88.2 signal it's sending to dac... so, alls well so far, nothing falling apart.

If I've understood peter right, the Qs interact with the dac you have ... so everyone may be on their own with Q settings?

...I think (think ! people preferring Q2/3/4/5 = 0 (not Q1 !) have the ability to sort out distortion better, opposed to people who like the non-zero settings...

Did you just accuse me of preferring to listen to music with distortion over cleaner SQ?  Thems fighting words. hehe Anyway I've stuck with 4 26 26 0 0 all day today and I hear no reason to switch; to me it's sounding very realistic, and I use the term realism because the changes I'm hearing with 9y-4 are just so vast and different from other versions.

[PeterSt]

[oh, found this this morning. Forgot to post it yesterday. Sorry.]

If Fidelio stays alone (with his wife of course ), he should try to listen to this with an NOS DAC.

My wife loves listening, she plays bass in a nice punk rock band so she is cool with loud music, and sensitive to SQ.

What DAC are we talking here? Can you give me an example of a NOS DAC that is nice?

For this Arc Upsampling it shouldn't matter much, but to do really good it should be 24/192. 24/384 preferrable.

The DAC Mani has may be in the leage, and otherwise I don't really know.
To do really well, these are the prerequisites :

- NOS
- Filterless
- 24/192

May you not know, my own DAC does comply (and this is in the end where I made the Arc Upsampling for), and it will be commercially available.
I don't want this to sound like a commercial though, and of course I made the Arc Upsampling available to you all in the first place. XXHighEnd is one, and Phasure NOS1 is two. Them together is what I just listened to for a couple of hours, and it may be just that other step into what we may judge as normal in a couple of years.

But what is normal these days ?

[Calibrator]

Woot indeed !

My immediate neighbours must really hate each new release as I'm sure they must get sick of listening to the collection of reference tracks I play in the evaluation process .. haha.

I think I was the second to download this but refrained from posting until I had a couple of days to fully comprehend what I am now hearing. This extended listening period also negates the effects from Mr. Placebo

There have been a number of milestone releases in the development of XXHE during the last two years .... this is one of them!

I'm not sure I still fully understand what has altered, but it's noticeably for the better, and has reset, yet again, my level of acceptance for music reproduction.

There is a level of transient and dynamics apparent now that, upon reflection, simply wasn't there before or was masked. I was listening to some Flim & The BB's albums yesterday, at levels I rarely use, and when "Jazz Patrol" started up, from their "Further Adventures" album, I nearly jumped out of my skin when a sudden crash of cymbals sounded. Amazing stuff. Further listening just reinforced this new found clarity, and I found myself jumping from favourite album to favourite album, relishing in it all.

If I had an intravenous drip inserted I would still be in my chair listening, but I suppose one has to eat and drink sometimes .. hehe

I have XXHE now configued for Arc Prediction using Doubleing, and this is due to a technical restriction of my pre-processor ( Integra DHC 9.9 ). If I feed it a 176.4 signal the pre-pro disables the onboard parametric equalizer which I use for taming room modes. The DAC's in the pre-pro are Burr Brown PCM1796's , and are Oversampling in nature ( http://focus.ti.com/docs/prod/folders/print/pcm1796.html )

Even though the DAC's are OS, the signals being fed by XXHE are of significantly better character in nature, and that translates to an audible benefit that's quite apparent. I can only but wonder what something like the  NOS box that Peter has created would reveal.

Peter ... is this Arc Prediction your proprietary idea or something you found in a white paper somewhere and have adapted, as I've found no reference on the net about it at all ? Does the technique lend itself to specifying a set output characteristic such as 24bit/96Khz to fully exploit the capabilities of our DAC's, rather than, what I gather currently, is a integer double or quadrupling of the sample frequency ?

Back to my listening chair for now.

Well done Peter .. again

Cheers,

Russ

BTW .. I'll get onto the bug report later

[PeterSt]

Just curious: Bitperfect, first 16 bits?

Does Arc calculate the extra bits from 16 to 24/32 or are they padded with zeroes.
Because only then it makes sense that double can do so much...

No, not bit perfect, because that can't exist with Upsampling. However, to stay in the "bit perfect" realm, the original samples are still there (which does not happen with OS/Filtering), and only the additionally needed samples (for the new sample rate) are injected.

As many bits are really used as the DAC permits. So if it's 20, then 20. If it's 24, then 24. However, calculation is done in "32 bit float", and after that decimated to the 24 bits (etc.), BUT which is done by the DAC itself, unless it is a 24 bit-only device, then it happens in the software.

Because only then it makes sense that double can do so much...

Not really. The bit depth matters for the accuracy, but it is really the sample rate and my means I perform the Upsampling, that makes this happen.

I hope it is clear a little (bit ).
Peter

[GerardA]

It is a little clearer now, I think.
So in theory it would also be possible without upsampling, just recalculating the extra bits so the shape of the wave gets more 'natural'? (leaving out the peaks between two samples of course).
Makes me think of making a picture by connecting the dots on the paper. 4 dots can become a rectangle, but if you get more dots it appears it is supposed to be a circle. You are then calculating the perfect circle.
No idea what sinx will do with that!

[Fidelio]

The DAC Mani has may be in the leage, and otherwise I don't really know.

?

To do really well, these are the prerequisites :

- NOS
- Filterless
- 24/192

But no commercial DAC I have found boasts these parametres. Or am I wrong?

[manisandher]

But no commercial DAC I have found boasts these parametres. Or am I wrong?

Hence why so many of us wait with baited breath for the commercial release of the NOS1 DAC... me included... even though I might already have a DAC that is 'in the league'.

These are very exciting times!

Mani.

[PeterSt]

Did you just accuse me of preferring to listen to music with distortion over cleaner SQ?  Thems fighting words. hehe Anyway I've stuck with 4 26 26 0 0 all day today and I hear no reason to switch; to me it's sounding very realistic, and I use the term realism because the changes I'm hearing with 9y-4 are just so vast and different from other versions.

No, but I did try to accuse you of using "sample rate" (the 64/96/128 subject) where it should be "buffer length in samples". I failed on that I guess.

On the distortion matter, I think this subject is more interesting and 5am deep so to say, than what's normally accepted. I mean, if you'd for example claim that NOS sounds better, any techie on the area would accuse you of liking distortion. He would even mean well, but would be sure you just like distortion (just not saying that you are completely deaf and can't recognize distortion). This is at the most rough level, and IMHO alredy completely untrue because where added distortion is one thing on the negative side, there's the all so wrong filtering on the other side which brings another type - and even more distortion (see longer post), but *that* our techie won't see, because his maths are correct and proven by the analyser (equipment).

What we talk about here (the Q2/Q3) is at such a super micro level that it can't be measured by known means (but my own software), and theoretically it should be so that 0/0/0/0 is the best. However, when we think about e.g. harshness, we all know how it can be solved : adding more distortion (read : filter it, smear it). Thus, liking something like 30/30/0/0 better, may indicate that something else is to be filtered. This is no crime, and when it works out for the better, why not. However, it is never to forget that when indeed it is so that a higher setting adds distortion (all still being bit perfect !), there might be a cause somewhere else, and of course it is always good to find the source of it.
All 'n all, it still can be so that when someone hears the added distortion as an added something, that may disturb so much by itself, it better not be there, and leave you with that other (base) distortion.

I could blahblah further about how people's brains and stomache etc. work upon/with audio, but I think in the end it is just so.

Lastly, it is not said at all that a higher Q2 etc. setting adds distortion. This is only what I expect from theories. It just as well can cancel out distortion, referring to jitter. And let's not forget, I made it like it works for a reason, and it is all based on everybody's experiences, and my perception of what kind of things may work out for the better. Also, there's xx% distortion in digital audio anyway, so why not attempt to cancel it out.

I hope that by now at least it is clear that Q2/3/4/5 are no explicit means to create "a" sound, nor that having them all on 0 is the absolute best. Nevertheless if I must judge something for real merits, I *have* to set them all to 0 (not Q1) because I know they otherwise influence, while without influence seems more honest.

Don't make a big deal of it, and just set them to something which is good for you !

[PeterSt]

Peter ... is this Arc Prediction your proprietary idea or something you found in a white paper somewhere and have adapted, as I've found no reference on the net about it at all ? Does the technique lend itself to specifying a set output characteristic such as 24bit/96Khz to fully exploit the capabilities of our DAC's, rather than, what I gather currently, is a integer double or quadrupling of the sample frequency ?

It is completely my own idea, and the kind of explicit necessity to outbetter what a DAC normally does, while the DAC under construction just has left out all this "fuzz".
So, at knowing a little what SRC's, digital and analogue filters and all do, I *had* to outbetter it, because otherwise the whole DAC project would have been without sense.

In brief, the technique is a combination of all what is needed to actually reconstruct the analogue wave, and which comprise of very basic elements in signal processing. The only thing I left out are the solutions from the schoolbooks, which had become objectives of their own (like : "let it ring, and the wave will reconstruct automatically"). There's two general parts in it :

1. Reconstruct the wave as good as possible, which is different per (upsampled) rate;
2. Reason out why something is 100% good, while it still looks wrong.

Mind you, I have been staring to simulation graphs for two weeks on how to do it (think of some 80 hours of staring, which is no joke), next I have been trying for two weeks to achieve something I could see as possible for further improvement, next - when that at last worked visually (in the simulation) - it sounded as sh*t, and next I needed two days to reason out how the in my eyes not yet 100% solution was 100% afterall.
So, as I said earlier, the waves cannot be resonstructed 100% looking at the graphs, but they sure just are afterall after applying some good math. In the end it comes down to, say, two different waves being able to output the exact same net result. One of those is the real analogue wave how it ever was, and the other is the Arc Upsampled representative of it. As with more things, I leave it to other experts as a challenge how this ever can be so. But, they may never get to the beginning of it all, because all begins with the prediction of the samples itself which can exist in one form only : the good one.

In order to answer your question more directly, the higher the upsampling, the better it gets. However, this does not say that the higher the upsampling the better you will perceive it. This is because the base (or objective) of it all is removing harmonic distortion of the stepped digital wave, while at some point other forces are there which prevent from making it better, or be able to measure it. Still the theory is so that the more upsampling the better, but (thus) with the foot note that you can just as well stop at some point at doing it.
Going from 44.1 to 88.2 clearly is not enough, because the HD is fairly above the capabilities of my DAC (but here you go already ... *my* DAC ...).
Going from 44.1 to 176.4 is as good as the dreaded ringing filtering does it, which may be a measure for some. And, I don't have another measure because a. I can't measure beyond 96000 for the "half of nyquist" (equipment limitation) and b. I have no means to output 352.8 from the PC (half of Nyquist would be 176.4 then), which is necessary in the case I would try to go beyond Quad upsampling (Octral ?).

All 'n all Arc Prediction puts the sample exactly where they would have been at a higher (A/D) sample rate, but it uses the representative (wave) model for it.
And this is what you hear.

Now, as said earlier, I must sit back and stare again, in order to reason out how it can be that for OS/Filtering DACs it works too (or at least for some as it seems). But I really can't imagine it works as good. This seems "mathematically" impossible. But we'll see.

Peter

[PeterSt]

It is a little clearer now, I think.
So in theory it would also be possible without upsampling, just recalculating the extra bits so the shape of the wave gets more 'natural'? (leaving out the peaks between two samples of course).
Makes me think of making a picture by connecting the dots on the paper. 4 dots can become a rectangle, but if you get more dots it appears it is supposed to be a circle. You are then calculating the perfect circle.
No idea what sinx will do with that!

Haha, no, it is just the other way around. The extra bits do (about) nothing, because they are only about the amplitudes of the waves (frequencies). So, this is accuracy in the amplitude domain (also quite important of course). It is only possible with upsampling, because only then you have the "opportunity" to capture the shape (but think in peaks only for your conveniency) better. With extra bit depth only, you can only let rise or drop existing peaks and this can't be done on its own (unless all goes in an even fashion, but then there wouldn't be any change accept general volume change).

Your imagination of the picture and 4 dots is okay. But this is not the actual problem. Draw that rectangle, and give it four corners. Imagine those corners are the sample points. Note : In order to really do this, you'd need infinit rise time which does not exist. Now shift the time domain a little, meaning that your A/D converter just starts a fraction later, which is 4mm on you paper to the right. What do you get for the shape of your original rectangle ? Keep in mind, your rectangle stays where it is (the singer started singing), only the A/D starts a litter later.

[AUDIODIDAKT]

There is more going on than what I currently can reason out 100%;
This is related to hires material not sounding as good as this upsampled 16/44.1; I won't say that this is caused by all hires being upsampled (by the same wrong means) from 16/44.1, but something is going on with "native" e.g. 24/192 material which makes it hide those on stage sounds and everything. Personally I don't think we can end up at "but Arc Prediction is wrong afterall", just because it does so right for such a long time (5 weeks in my case). Notice the difference with "faked high detail", which is very easy to create by just removing some (lower) mid frequency output. Result : "Ah ! you can hear her spitting in the microphone now !!". Yeah, right. But this won't last for long, because it is "created" by removing something else first. It won't last for 5 weeks.
So, it intrigues me that it is just sheer high resolution I perceive from this Arc Upsampling, while native 24/192 (or 24/96) already shows explicit lower resolution. So, "as good" would already be strange (but according to my own applied theories it can be done), but how "worse" falls into place ? ...

This still remains odd, yes, don't like the HiRez SQ either.

Strange!

[arvind]

LIVE & ALIVE, that's the only way I can describe 0.9y-4. Fantastic SQ, Peter you have done a remarkable job.
This is without having an NOS/Filterless DAC, as yet. Borrowing one in a few days to hear the difference.

Arvind

[PeterSt]

Thank you Arvind. I was hoping for that, especially for you (being especially dissatisfied with the later version(s)).

I'm already very interested in the results from your comparison to come.
Peter

[leifchristensen]

hello P
is there any upside soundwize in upgrading from 9y3a to 4 if I cannot turn off the upsamling?
best
Leif
Norway