0.9u-6 early thoughts

[PeterSt]

Now, objectlively, I have even more issues with it. As I've mentioned before, aliasing has a definite effect below 22.05KHz.

Sorry, but I couldn't help myself - have a look at the following graphs.

Ok, I wanted to put this in a separate topic, but actually it is 100% related, so here goes :

Mani, I don't know about your signal processing skills, but personally I don't understand what I'm seeing, looking at your plots.
Remember what I said elsewhere : "I don't like to apply an AA filter, because it would roll of the highs" (similar). Mind you, this thinking of mine is taken from what can be read all over the place;

Your plots show (and mine obviously will too) that the roll of is already there *without* an AA filter. Now what ? was everybody looking only after the AA filter had been applied ? Or IOW, what is actually going on here ?

a. there's a problem in the (XX) software;
b. the DAC isn't capable to cope with the high frequencies at the higher resolution (that's what it would come to) (for others : keep in mind that this goes over the analogue output of the Fireface, with or without loop back cable (the loopback it can be done internally);
c. the aliases influence the phase of the audible side of the mirror, and therewith attenuate;
d. the method of measuring is faulty;
e. this is just normal signal processing theory stupid, get your books !

Please keep in mind that I look at this all from the context of knowing what the program does. So unless a. above is applicable, I can't think of why a sequence of samples coincidently implying a higher frequency, would decrease in volume (that's what the plots show) other than c.
I don't change that volume, nor is there any arithmetic that implies it ...

[manisandher]

Hi Peter,

I'm assuming c)

In any event, here are the results with Foobar/ASIO (running 24/192 shared mode). There is obviously an AA filter being applied.

Mani.

[manisandher]

I think it's time to brush the dust off those old undergrad books on our shelves... or at least on my shelf.

Mani.

[PeterSt]

g. http://www.epanorama.net/documents/audio/spdif.html

uhohh


Thank you for the Foobar test Mani.
Now I wonder what happens at double/upsampling with "DAC is" at a 16 bit setting. If this looks normal, then

1. It is most probably the SPDIF limit;
2. Foobar (SRC or whatever it is you were using) doesn't utilize the bits (only uprates the sample frequency).

Btw, I am not asking you to try this, but can't do it myself at this moment (tonight I can).


Oh, when things still don't look normal, I'd try ADAT instead of SPDIF.


PS: I was Googling for an hour or so on the SPDIF matter, but couldn't find a clear answer. Keep in mind that "we" are using SPDIF over Firewire (400 in my case) and that things may matter there.

PPS: I got the idea here : http://www.avsforum.com/avs-vb/showpost.php?p=13353422&postcount=897 which was singing in my head the whole day because I couldn't believe that, while on the other hand, that (MS) guy should know. Suddenly the combination with our topic here popped into my head ...

[manisandher]

It's my lunch break, so I can do this quickly... with pleasure.

Find below the two curves. Just be aware that I'm using my system 2) at the moment - i.e. no external dac and no spdif.

I have no idea what Foobar's SRC does with the bit depth, though in shared mode, Vista pads it to 24 bits, no? (I have it set to 24/192).

Mani.

[manisandher]

I'm in the same camp as Mani on not using the upsampling option. When activated, the precision looses a little something. Mellower and not as articulate.

Thanks Russ. I was beginning to think I'd lost it for a minute...

I totally concur with everything you say.

The only thing that doesn't make sense to me is this: why does quad upsampling sound as good as it actually does. No, IMO, it doesn't sound as good as no upsampling... but it still sounds pretty damn good. If you read the literature on aliasing, it is supposed to sound awful, even in moderate amounts...

Mani.

[hybride]

I'm in the same camp as Mani on not using the upsampling option. When activated, the precision looses a little something.

I don't get this. Upsampling means more precise reproduction of the original waveform. That precision looses a little with Upsampling cannot be true in theory. When you hear the opposite effect, there must be another reason.         

[manisandher]

My limited understanding of upsampling is that it has a single potential benefit: it allows the use of a less-steep anti-aliasing filter to be used. This is turn reduces ringing effects in the audible range.

I do not understand how upsampling can create more information than is recorded onto the CD. The only technique that I'm aware of that helps with increasing resolution is dithering...

I kind of remember Bob Katz and others doing some tests on this. They found that in an ideal world, an anti-aliasing filter would have a low pass band starting no lower than 50 KHz or so. In which case double, but especially quad upsampling might make sense.

But what's happening here is that we are getting a load of (I assume) nasty digital artifacts above 22.05KHz, interfering with our beloved music. This cannot be good. But I don't understand why it doesn't sound worse than it actually does...

... maybe my hearing isn't good enough to hear the mess that going on up there

Mani.

[hybride]

Can it be so that the audible effect of artifacts and distortion with NOS (without filtering) is heared and interpretated as more details? In other words; that what we think to hear is not what we really hear.
 

[manisandher]

Do you really want to upset the NOS crowd???

Mani.

[hybride]

  no!, because i also like to listen to my diy (Monica) NOS dac. Sounds very nice. But....

My Stello 220mkII Dac is connected to 100% clean power (Kemp powerstation).
The Stello can be easy switched from NOS to OS.
The audible differences I hear are more in agreement with the theory, wich i suppose as the truth;
OS (192khz):  More subtile, a little bit less dynamics,  much more microdetails in sense of dying-out sound wich makes the sound image more complete and enjoyable . 
NOS:   A bit raw, more dynamic sound.  Very good musicality but ‘not complete’


Moderator's note for future reference :
The "NOS" referred to above, is unrelated to the official "Non OverSampling" phenomenon. The Stello 220MKII is a Delta-Sigma Oversampling DAC.

[PeterSt]

Can it be so that the audible effect of artifacts and distortion with NOS (without filtering) is heared and interpretated as more details? In other words; that what we think to hear is not what we really hear.

So far the discussion is not about NOS, but sure the lacking filtering in there comes down to similar stuff (/ perceivement ?). Otoh, I don't think it is good to bring that into the mix, because it will be different for inherent anomalies and (listening) results.

Then, I think it is important not to mix up with unrelated things as to what the alias at the other side of the mirror might bring for audible anomalies on one side, and what obviously happens on the left side of the mirror which SHOWS anomalies. It will be 100% sure that what shows there, is audible too.
Since we left our books covered with dust, we could wonder how the right side of the mirror (which makes perfect sence that that aliasing  happens) influences the left side *if* it is that what is going on. And I am not so sure about that yet.
Probably right now we are a bunch of stupids with to few knowledge to talk about this seriously, and IMO the only reason why it could be interesting even to the knowledgdeable, is the possible situation that an AA filter not only removes the aliasing, but *also* compensates for the known highs roll of implication, and just nobody saw that the roll of is already there when the aliasing is just left alone ...

With above fuzzy lines I only want to say this : I don't think we should try to work out (or "find out") the obvious, for those who just know. We'd only make oursellves more stupid to those who do know.
On this matter it is my idea to ask a person of which I'm 100% sure he knows ... Bruno Putzeys. Why him ? well, because he has been all over into this stuff, *and* we actually know him. He might not recognize it, but it was me asking for the highest grade Hypexes as I think are around by now, and hybride, you might know him more personally while I only talked indirectly to him (via MM-Audio). Send him an email, direct him to this topic, and ask politely ...
Btw, for those who don't know him : he once DIY created an amplifyer taking DSD for input ...

Now, since I actually don't even have those books (yeah, you knew that) there is much more going on which I found well over two years ago, and for which I don't have explanations. In brief this is about the whole audible spectrum being FULL of aliases, if you only know how to look at it, and how to incur for it. I can tell you, only nos DACs show this best, and actually it is quite unbelievable we can hear music through that mess. This too is about the Nyquist frequency being a mirror, and "colliding" parts of the mirrored frequencies being able to create the exact same SPL as originating frequencies elsewhere. Believe it ot not, but e.g. a 16Khz (somewhere around there) tone can create an as high volume at 20Hz. Mind you, no spurs of 20Hz are in the original tone in this case.
This concluded for now : the nos DAC is as bad as can be (for matters you can't guess and which I did not explain), BUT WE LIKE IT.


All 'n all (again, for now), assuming I can't hear the anomalies on the other side of the mirror, the only thing I see is a roll of of the highs (but careful, because you can see a drop a 2KHz already) and an anomaly at the low end. If this indeed would be all that's the matter it hardly harms sound but for more dullness (and what comes from that of course -> should be less detail). Mind you, apart from the thingy at the low end, it's rather lineair.

Last thoughts : if this is a phase impeeding thing, I'd say that we can't capture it with a *digital* loop back. But I can't believe this is phase related ...
The strange things I found I talked about in the above, are in the pure digital domain also ...

And oh, let's keep in mind in advance of things : theoretically it can be that what we see is not real. Not so much because RME software would be wrong, but because the representation of it all actually also needs an AA filter. How this can be made consistent with software applying an AA filter (like most probably Foobar and its used plugin does), and all looking good, is another thing.

Peter


PS: This didn't incorporate hybride's last reply, above.

[PeterSt]

One other thing to think about :

Since this is a topic that makes me brabble around anyway, take an MP3 if you dare (take the higher bitrates though).
So far I did not spend much attention about it (too much ), but you may find yourself finding it *better*. Maybe the pinpointing is less, but for sure the stage is wider and deeper, the bass is better ... Now what ?
And yes, I even have theories for it why it can be better, and 0.9t (and u for that matter) is a bit based on this theory. Mind you, a theory I had to come up with afterwards, so nothing I could reason out in advance.

What does this tell ? well

a. do never think that an e.g. wider sound stage is better, because how can an MP3 be better ?
b. the fact that, say, noise is added, not necesserily means we perceive it as worse. It should be worse though for theories.

Anyway, this kind of explains why we also can perceive the nos DAC to be better, although there's some more to that.
For those who know : I've been always referring to the Kodac DCS whatever camera for comparison. There is *NO* camera showing its resolution better than that one. There is NO camera more difficult to deal with at making photo's and avoiding the things it just can't do, like a roof with overlayed stone cover (sorry I don't know the name in english) which will show moire all over. Even with trees you must be careful. And you know what ? there is just no AA filter in there. It is the only camera which allows to see the 14MB picture at full size, and it wil show no degradation on sharpness. Also it is the only camera that will not need sharpness appliance ...

With these things in mind, it is more easy to see why we can perceive things better, while actually they are not, or why we *should* perceive things better when we only comply to the rules.
This is a great deal of XX ...

[GerardA]

Hi Peter,

Do you know the site src.infinitewave.ca?
They show interesting testpictures of a lot of upsampling algorithms, most of them look very bad.
Even the ones that are reported by some to sound good look awfull.
I was hoping you could make one that sounds good and looks good, but that must be a very difficult challenge!
Maybe you can do the same tests they did, or find some clue what the competition is doing.

[hybride]

Since this is a topic that makes me brabble around anyway, take an MP3 if you dare (take the higher bitrates though).

Well this topic is about sound quality.. 

I agree that 'what sounds good' not always has to be a 100% agreement to theory. 
If it was, our hobby would have ended after Philips made his first cd-player  .
But 'impressive' mostly gets boring through the time. Although, i think that kind of phenomenons keeps the highend market alive.
Don't you agree that its more about the natural level of sound with stable precision in more complex parts of music (like orchestra's).
It is very hard to achieve real natural sound. 1 'wrong' capacitor in a circuit can screw things up.
Mp3's may sound impressive, but will i.m.o never fit in the framework of highend.