HQPlayer and the NOS1

[PeterSt]

Mani,

Please remember, someone challenged me to try this, and without that I never would have done it. In this realm, the last thing I would do is expressing about it, but since it should be regarded as advice I now have to, after what I have seen ...

None of the filters make a pulse train of a fed pulse train; all is pure sines. This means transients are killed.

THD+N in the audio band is worse up to around 16KHz (not much but still). Beyond that - but up to the stop band - HQPlayer is better. Notice that this is by design and that 16Khz is "holy" for me (I have talked about this before).

I did not listen at all. I could have before measuring but I wanted to save my windows in the first place. This appeared justified because when "anomalies" occur, sync is lost and static occurs. Notice that this was with the 2-Wire setting (which by itself seems to be OK) and the default buffer setting; after changing that to 100ms I did not see it anymore, but the fact this can happen scares me (no preamp here).

Beyond the audio band (audible / influending or not) things were only worse. By itself this is to be expected to some degree (thinking noise shaping bases, of which I am not sure this is a base at all but it can be, thinking about Miska's general technology).

If it weren't for the getting out of sync (creating static at close to -0dBFS) I would have listened for sure. However, knowing how pulse trains implying sines sound (will buzz) I wasn't tempted much and I am not sure whether that's a good thing (now I have no bad expression about SQ) or just not fair. So, I keep on saying what I always said : if "we" (and that includes me in the end) find a better playback means than XXHighEnd, we should use it. So I leave that up to you.

I have one other remark which I can't reason out much;
All filters behave exactly the same in the stop band, which is the same behavior as Arc Prediction : Heavy imaging is in order, meaning that behind the Nyquist mirror the energy is only 3dB down to the original signal. This is what I personally don't like from Arc Prediction, no matter those frequencies don't play much (between 20Khz and 22.05Khz), so I was surprised to see that here just the same. One (for me) peculiar exception : the IIR filter; that does not show this at all.

All 'n all, and especially with the heart I have for Miska, I don't understand it.
Remember, all seen through an NOS DAC which doesn't do a thing herself; I have no idea what Miska can see ...
But, knowing he has a Mytek which I own too, I will later repeat the measurements through that.

Regards,
Peter

[manisandher]

Hey Peter, thanks for looking into this so thoroughly. Your findings reflect what I'm hearing, I think. With the settings I've described, HQPlayer sounds more laid back than XX. And the 'reverb effect' isn't present. I've been assuming that the latter is an AP anomaly that I'm hearing. Maybe it isn't. As I said, it's actually very, very pleasant and really gives the music a 'recorded live' feeling.

I've stated before that AP sounds much better than the 'Polynomial' filters in HQPlayer. These are just awful, and I agree with Miska's recommendation not to use them.

But I still think it's nice that NOS1 owners have something else to try and play around with. I mean, there just aren't enough parameters in XX to keep one occupied

Mani.

[PeterSt]

LOL!

[Jud]

But I still think it's nice that NOS1 owners have something else to try and play around with. I mean, there just aren't enough parameters in XX to keep one occupied

Mani.

Hah!  But as you say it is interesting to hear just how profoundly filtering and dither changes alter what we hear.

Peter, what are your thoughts on dither?  Does Arc Prediction incorporate it?

[PeterSt]

Jud,

No dither anywhere; against everything and all I just don't like the principle, not matter the theories seem good to me. But additionally :

When I tried to implement it I couldn't measure it at the outputs. So, I operated in the blind and can use the theories only. And the down side (for me) : we'll throw out one precious bit because of it.

So notice, all the nice plots we see (like from Miska for example) are "digital" plots showing the theory only (like a signal still being there at -140 or whatever dB). This, while for the NOS1 the signal drops into the noise at the 18th bit. So, useless (and no other DAC will do any better).
But also : the other day I "prooved" that while resolvement is possible up to that 18th bit, 24 bits still do their work for a nice sine around the noise level. I don't know where that post is hanging out, but I wrote about it somewhere here.

Peter

[Jud]

Not meant to argue with anything you said (I don't have the technical knowledge in the first place): I do think it's been demonstrated that people can "hear into the noise," though as you say no DAC will provide 24-bit resolution.  Still seems nice that in terms of internal mathematical calculations at least, things are correct down to that level.

And recalling some of the discussions about how noise may not be the worst thing as long as it is relatively random, this seems like something analogous to dither in a way.

[PeterSt]

Because I followed a link from somewhere else, that ending up in this topic - and after reading it all through again, I must say now and in this topic that I have been sitting on something for quite a few months, and see that I can nicely tell about it here.
It is almost exclusively for Mani ...

Last August when I was in LA, I have been speaking to Prof. Keith Johnson for something like 30 minutes. Let me tell right away that this conversation was not so easy because something surprising was happening : Mr Johnson got old. Very old.

I have been asking him the questions to which the answers always lacked (of you / us) like the genuine 24 bit ADC. Well, this is what I recall from it :
Yes, but very difficult to implement with sample and hold in analog and always going back and timing and ...
And what *I* got from it was a Sigma Delta in analog. Couldn't make more of it (or make it better) because of the sample and hold was about some feedback and maybe professors talk in a way which needs college for several years, which I lack.

Whether he wouldn't be interested in reproduction of the Pacific Microsonics (II) - No. But ... No.
Maybe .... NO.
Oh.
And then after all some explanation of parts not being available anymore - the story we know.
But see here a sort of out of date "behavior" which I could not revert - suggesting the money in my pocket.

Ringing.
The story had to deviate somewhat, but about non-ringing filters.
CAN NOT EXIST.
Yea, well, but ...
CAN'T.
... I have that working.
Not sure what the next reply was, but it went something like "that's what everybody is saying, until I look at the THD and see the distortion".
But ...
NO.

Such a pitty and such a shame while I had this super chance.
There was a guy standing next to me, who later hunted me down, suggesting to send him the output of Arc Prediction by email so he could check it and see.
I have almost (or more) tears in my eyes now, because it really can not work. Not anymore.

Mani, I should have told you right away, but a vacation followed this and next a speaker to tune for a show and, well, since then I have been thinking about it. But seeing this topic accidentally made me tell about it after all.

Done.


PS: And then to think this was after a seminiar Phasure organized with the preprogrammed contributors as you may have read in a topic elsewhere (California Audio Show). Some showed up as planned, others did not, but somehow Prof. Johnson appeared there without being planned so. I had been sitting there for an hour during the seminar itself, thinking what to ask and how afterwards. I'm still a bit shakey.

[manisandher]

Hey Peter, thanks for saying it explicitly. But I kind of already surmised your conversation about non-ringing filters with 'Prof' KOJ from what you posted at the time (I think you referred to him as 'the most famous recording engineer of the lot', or something like that). So no real surprise for me there.

As for the PMII. Well, there's so much 'secrecy' about it that it makes you wonder why. I mean, it's an old machine now and there's no intellectual capital to protect. So maybe the continued secrecy is to conceal some 'untruths' that were told by the PM team at the time. Maybe one of these was that it was a genuine R2R ladder ADC (as explicitly stated by Michael Ritter in at least one article)??? In any event, it was and still is a good ADC, however it actually works.

I'm confused by this though:

I have almost (or more) tears in my eyes now, because it really can not work. Not anymore.

What are you referring to here that "can not work"?

Mani.

[PeterSt]

Mani, let's say : To create a most decent ADC once again, based upon the knowledge from then which is still very current if only the key person wanted to cooporate which he seemingly can't "anymore";
I too will at one time have become too old to have the idea I want to change the world.

Something like that ...
Peter

[AlainGr]

Strange... Each time I wanted to change the world, the opposite happened...



Alain

[manisandher]

Mani, let's say : To create a most decent ADC once again, based upon the knowledge from then which is still very current if only the key person wanted to cooporate which he seemingly can't "anymore"

Ah, OK. Understood. Yep, it's always sad when someone brilliant loses their mojo.

But is this "knowledge from then which is still very current" really only in the head of KOJ? Surely there must be others who have it too, no? And if there really aren't, then might it be possible to recreate the knowledge by taking something like a PMII, studying its ADC chip and reverse-engineering it?

Mani.

[CoenP]

Why not forward engeneer? Linear ADC has been done for ages.  Silicon is much much more capable than ten years ago. Maybe we can do true 20 bit acurracy in a real ladder adc at 384 KHz. Caveat: you will probably have to design a soc to meet to the specs. This is very high speed ultra precision stuff. Now that still isn't well suited for experiments to say the least.

Technically the delta sigma dacs will still run circles around it, are they really that bad? Theory says that when properly bandpass filtered and decimated, there will be no difference in sampled signal. The delta sigma noise is gone, in either case the signal has to be filtered (hence ringing if done steep) to avoid aliasing.

It would be usefull to have a true ladder adc'd signal to assess the true merits of the ladder concept. Anyway it is my optionion that the ladder has the edge in practical playback, where (multibit) delta sigma the edge has in adc.

dSD/DXD is a different story though, now that would be interesting. Jitter issues deluxe, but no mathematical conversion necessary only filtering...

Lets move on!

Regards,  coen