Quad Anti Imaging

[PeterSt]

I don't think people have been using Anti Imaging Upsampling very much at the time it was introduced (then called Anti Aliasing, AA). I myself did though (for two months or so in a row), and the fact that I could last for two months, well, tells something (to me at least ). Then Arc Prediction was introduced, and people rave about that, including me of course. Notice that either means was created for the filterless DAC, and only Arc Prediction showed that on OS DACs it does its positive job just the same.

As a sad result of introducing Arc Prediction, Anti Imaging stopped working unnoticed. You could choose it, but it just did nothing. I know, this is a really bad thing, especially for those who took the effort in comparing etc.; Now :

All is up and running again, and btw in a far more robust way than it ever worked before. The penalty is that it won't work gapless for Attended Playback anymore (or at this moment), but alas.

After testing (bug solving) AI yesterday, I just left it on, and I must say that I immediately perceived the same darkness and quietness I recall from using it for the very first time back then. This should mean that -at least for my filterless DAC- it could be a good thing afterall when the higher frequencies are just cut off after 20KHz. But keep in mind : this is a ringing filtering means.

My point for today is : this time I used this mode with Engine#4 and Adaptive Mode (soundcard at 128, Q1 = 1), and I am sure under the impression that somehow this is a "working" combination. So, the darkness is there, but the details seem even more (this was no different back then), but there's also the freshness, and *that* was not there before.

So, you might try it as well. It probably won't work at all on the OS (thus filtering) DACs, because opposed to Arc Prediction it merely has downsides (ringing) and no benefits for OS DACs. Anyway, I seem to like it afterall, or at least it is very different again.

Peter

[manisandher]

Peter, I can't get QAI to work! Having pressed play, I have waited for minutes with absolutely nothing happening. Is this my poor Atom processor showing its weakness?

Mani.

[PeterSt]

Mani - I doubt if "nothing" really is nothing. What does your cpu tell you ?

[manisandher]

Yes, I've got to be more patient with this Atom PC. It takes the processor, running at full pelt, quite a while to get all the AI pre-processing done.

In any event, Quad Anti Imaging sounds interesting... dry, dark and 'weighty'. It could sound very good on some systems...

However, I've just compared a 16/44.1 upsampled with QAI and then with QAP against the same native 24/192. There are no two ways about it; QAP sounds closer to the native 24/192 than QAI. The ambience is missing with QAI.

Mani.

[PeterSt]

Thank you for the feedback (and tests) Mani.

Btw, I now realize that for less powerful processors the way AI currently works is quite unworkable. So, I will try to reinstall the "quick start", only for the first track though, meaning that Unattended will remain "not gapless" for this.

Peter

[earflappin]

I agree with you Mani - QAP sounds better than QAI to my ears as well. 

Peter, two questions:

(1) I can run special mode at 32 buffer and Q1=3 samples and with my Lynx card buffer at 32.  Is there any reason why adaptive should sound better this special mode?

(2) I'm getting an out of memory alarm when I try to play a 6-10min file in QAP mode even though I have 4GB memory and I have the split file set at 3000.

Thanks,
David

[PeterSt]

Is there any reason why adaptive should sound better this special mode?

David, a similar question would be "why does Engine#3 sounds worse than Engine#4". It is DIFFERENT, and often different is enough to let it sound better (hey, that is 50% chance without me thinking about it; with thinking it could increase a bit ).

I have 4GB memory and I have the split file set at 3000.

Oops, that is not how it works. The Split Size must be seen as a kind of gross size, looking at the file size on disk. Thus, files of 200MB exist easily, and it is *that* size to fill in. Your 3000 suggests files of 3GB.
After your amount has been read from disk, the file will expand (up to 8 times) in memory.

I think this is clear enough ?
Peter

[earflappin]

Ah...not really....

Let's say my file is 200mb - 8x that would be 1.6Gb....so what's my problem?  I have 4GB RAM?  What setting do I need to change and to what value to not get the memory insufficient error?

Thanks.

[PeterSt]

Ah ... your first problem would be the 2GB limit for one process (which is there in the setup I use ).
Subtract the OS from that (600MB) and something like 170MB (Split File size) should go. I'm sorry, I didn't think of this (2GB) before.

[PeterSt]

The ambience is missing with QAI.

Djeezz, are you right about that !
I was just testing the AI "quick start" way of working (btw, works again), and tried it on a Joe Pass ... I didn't know what I was hearing !
So, completely true, and apparently *the* virtue of Arc Prediction !

I'm completely objective here, and just started using AI again after it was reinstalled (0.9y-7), BUT used it at the same time KS Adaptive Mode worked for the first time (also 0.9y-7). So :
If you don't know better (me), it seems that QAI with KS Adaptive Mode is better than QAP with Special Mode. But once you hop over to QAP/KS Adaptive ... well ... you know better.

Another small leap for ... blahblah
Peter

[PeterSt]

Ah ... your first problem would be the 2GB limit for one process (which is there in the setup I use ).
Subtract the OS from that (600MB) and something like 170MB (Split File size) should go. I'm sorry, I didn't think of this (2GB) before.

Btw, I guess by now I'm up to creating a 64 bit version for the audio engines. But you can bet it will sound "different".
But are you up to it ... or shall I better wait for a month or so.
Yes, I guess so.

haha

[earflappin]

Sounds grreat.  Bring it on....