TAK Support

[bhobba]

Hi Peter and All

Am unsure exactly what section to post this under but this seemed it was the most appropriate even though it is not about sound quality per se.

At present until I can get XXHighend working I am using Foobar.  But I am not in that much of a hurry because I am simply listening via some cr*ppy speakers connected to my computer until I get my main rig set up.

Of recent times I have been doing some experimentation on lossless compression and found Flac is far form the most efficient around.  I am currently using it because XXHighend supports it but others are much more efficient.  I tried monkeys audio and while it was great its performance on high res material was not that great.  Then I came across TAK:
http://wiki.hydrogenaudio.org/index.php?title=TAK

My experimentation shows it encodes faster than Flac 8 even at the insane setting I use.  And it is supposed to decode fast as well but that is hard to tell on my machine because I have an I7 and foobar doesn't really give you an indication how hard it is working - I can say however my CPU usage is practically zero.  On 44.1 material monkey audio has a slight advantage (but it is very small).  However on high res material TAK is way better - something like 20% better than monkey audio and even better again than flac. 

I know that disk space isn't that expensive these days but it is still not free and I don't really feel like wasting it.  Is there any chance of XXHighend supporting TAK in the future?  What do others think?

I know that TAK is not widely used and that is an issue.  Even though musepac leaves mp3 for dead I recently converted my musepac files to mp3 because it looks like it is going thew way of the do do.  For example squeezbox claims it supports it but closer examination showed it only supported earlier versions - later ones simply do not work.  When I asked about it I was told musepac is quit uncommon and they won't be updating their player for such a small number of people.  Pity that the better option lost ground.  I hope the same does not happen to TAK but my gut tells me it probably will.

Thanks
Bill

[PeterSt]

G'day there Bill,

First off, it wouldn't be all *that* much trouble to implement it, but it should be worth it. And I have doubts on that one ...
So let's see ...

Your finding hat TAK encodes faster than FLAC-8 will be true, but also is not of much value. This is because there is no use case in using FLAC-8 opposed to FLAC-5. You can almost nothing, and it takes ages longer. Solution : use FLAC-5 (not 6, not 4, just 5 which is the default).
Also, any comparison with Monkeys Audio is not much useful (I know, you didn't either, but mentioned it still) because it takes ages to decode.

Then, TAK compresses some 2% or so better than FLAC. Do we really bother about that ? of course, on 1 TB disk this allows you to store some 40-50 albums more, but it is still 2% only. This won't save your life once the disk is full up to that level; you'll buy a new one in either case.

The real "culprit" is about the decoding speed. I read that this is "at FLAC levels", which I would say too if I were the second best. But, it is 20% more slow, or something in that area. But careful now ...

It may have occurred to you (no, maybe not you I think, but others then) that decoding a complete album via XXHighEnd takes 5-8 seconds or so. This does not compare at all to the official lists with the relative respective numbers. For example, Monkey may take 2 minutes for it easily, and even the FLAC site talks about over 5 minutes for FLAC itself, although this was done on a PII 333MHz whatever by now old machine. The (relative) figures aren't correct, and this is caused by the way XXHighEnd is able to process things internally, which happens to go so with FLAC, and if it were for me, with FLAC only. I'll admit it is a kind of coincidence, but I myself am not even able to normally copy a file faster than FLAC converts it including the reading of the source file and writing of the destination file. So this is why e.g. Monkey becomes 20 times or whatever more slow. The same will happen to TAK (or at least I wouldn't know why not).

So, all together I don't see much virtue in doing it, which btw is a bad thing by itself for the developer of TAK. On this matter, compression is a sport and requires a lot of smartness plus knowledge. Thus, reasoning like I just did doesn't honour that much, does it ?
But still ...

Remember, it can be done. But the above incorporated it's always a thing of "don't I have other priorities ?".

Peter

[bhobba]

Hi Peter and All

Yes I agree that the the extra 2% or so you gain on low res recordings simply is not worth it.  However that is not the case for hi res recordings - it is something like 30% or more:
http://www.hydrogenaudio.org/forums/lofiversion/index.php/t64933.html

Flac at 8 - 46.2mb - TAK - 31.94mb - thats over a 30% saving which is certainly worthwhile.  I think the conclusion reached in the above link is the right one:
'Well, so it looks that FLAC isn't really well-suited at the moment for low-passed files, thus for 96 and 192 KHz files. Considering the results, I think it will be better for me to stick with FLAC for 16/44 files, but to switch to another format for 24/96 and 24/192 files. I don't like the idea of handling two different lossless formats in my collection, but the gap is huge, and I don't see any reason for losing almost 350 MB of disk space for a single album.'

However right now there is not much hi res material about so it doesn't look like its something worthwhile to do anything about now.  However as time goes by that will change. 

I think the bottom line here is its something not worthwhile doing now but when high res becomes the norm it would probably be worthwhile looking at it again.  I also have to say it looks a like a simple tweak to FLAC to handle hi res material differently would solve the problem anyway - with 30% or more gains possible I don't think the Flac developers will let that go by.

TAK decodes in close to the same time as Flac  That was one of its design goals:
http://flac.sourceforge.net/comparison.html

Still it is a bit longer.  What I suspect however is by the time hi res formats are the norm we will have a lot faster CPU's so the slight increase will be of zero importance.

Just by the by I always recode my flac at 8.  You are right - you usually don't gain much but it doesn't take long and the way I look at it why waste disk space.
 
Thanks
Bill

[PeterSt]

Allright Bill - 30% *is* a lot of difference. And I do have a couple of 100 hires albums hanging around.

Maybe the decoding speed of TAK is also faster for hires; I always had the feeling that FLAC is very slow on those (I know, they are much bigger too, but still).

I will look into it !
Peter

[bhobba]

Hi Peter and All

Thanks for looking into it - it is really appreciated.  I believe hi res recordings are the way of the future and 30% is worthwhile.  But I don't think its something of any real urgency.  I suspect doing it during a few spare moments now and then is all that is required.

Thanks
Bill