WMA again

[Marsdaddy]

Just to let you know...

I've started work on a fixed point WMA decoder. It's essentially a conversion of the WMA decoder from the ffmpeg project. I intend to get it working stand alone and "float free" before probably submitting it for others to incorporate into rockbox.

That's the plan anyway. Don't hold your breath, I get very little free time, etc. etc. I reserve the right to get bored and give up!

Seriously though, lack of WMA bugs me in rockbox so I am going to give this a serious shot. I've already made some progress and will try to keep you posted...

[Llorean]

Score one point for those who've been saying "If someone wants it badly enough, they'll start working on it."

But seriously man, if you do get a fixed point decoder working, that's one step closer to making a lot of non-me people very happy. (I'm satisfied with vorbis.)

[bk]

Cool. Consider contributing your code back to ffmpeg (or maintain it as a separate project), because I'm sure Rockbox isn't the only project that could use a free fixed-point WMA decoder.

[Mr. Brownstone]

Cool! Keep us posted, however it goes.

[bbad]

why you should do something thats already done - http://forums.rockbox.org/index.php?topic=1258.msg11398#msg11398

[Marsdaddy]

why you should do something thats already done - http://forums.rockbox.org/index.php?topic=1258.msg11398#msg11398

Well, yes I have looked at that but there were a couple of issues:-

(1) It looks like it might have been based on an old version on the floating point decoder as there are some bugs in there that I've had to fix in the version I got

(2) Err... it's littered with floats still, all over the code  . I can see there has been some work on fixed point, but it's very much a work in progress.

[MU4L]

Excellent news, Marsdaddy! I wish you good luck!

Please keep us updated  ,


MU4L

[Marsdaddy]

Thanks for all your interest!

Just to let you know I'm making pretty good progress, I currently have a hybrid decoder which is mainly integer based. A couple of tough problems to overcome and I should be there...

Watch this space.

[MU4L]

*trembles with excitement*

I'm just one of many helpless chaps who has zilch in the way of coding knowledge/experience, but who has a fair few WMA's in their collection (folly of youth and all that  ), so to see some actual progress on a WMA decoder is awesome.

Perhaps (at last) I can finally enjoy the whole of my collection within the Rockbox Firmware rather than having to shutdown and re-boot into the vastly inferior iRiver FW,


Kudos, and good luck (again),


ManchesterUnited4Life

[Llorean]

Man, I'd personally rather just spend the few days reencoding from source rather than have to switch back and forth to play my music.

[MU4L]

Man, I'd personally rather just spend the few days reencoding from source rather than have to switch back and forth to play my music.

Well I have re-encoded what I can (from source, of course) to either -q6 ogg, or 192+KBpS LAME VBR mp3.

What I have left are 106 WMA files (out of a collection of around 7,000) that I no longer own either the cassette, or CD's I created thereof, with which to re-encode.

In any case, WMA is the preferred codec of a quite a few people (whether by choice, or by unintentional ignorance), which I'm sure is why the progress of this decoder will be followed with anticipation.

It would also open up new doors for Rockbox, in terms of prospective users, who may have been put off by the intensive process of re-encoding ...




Either way, I wish Marsdaddy the best in overcoming the problems he/she may be encountering

[Llorean]

Yeah, I'm definitely looking forward to WMA support, just so there's no disadvantage (in audio capability) to using Rockbox with most targets.

Though I'm curious, you say some use it by choice. As far as I can tell, most double blind listening tests have had WMA come up a decent amount lower than LAME MP3 and OGG in a lot of bitrate categories, if not almost all of them. Since there's no question that MP3 is more widely supported, what are the benefits of WMA?

[MU4L]

Though I'm curious, you say some use it by choice. As far as I can tell, most double blind listening tests have had WMA come up a decent amount lower than LAME MP3 and OGG in a lot of bitrate categories, if not almost all of them. Since there's no question that MP3 is more widely supported, what are the benefits of WMA?

Just before I begin, I'd like to say that you're preaching to the converted (even that's not to say I wasn't to begin with, though  ), Llorean.



To be fair, the layman doesn't have the equipment to carry out *accurate* double-blind/ABX testing, nor are they willing to. According to them, if there's a minor discrepancy (in terms of SQ), then that's what it is - minor. Why exert any more effort, if, at the end of it all, the result is a minor amelioration? This situation is even more appropriate when talking in contemporary terms, of the majority of AAC/iTunes/iPod users. 

Another example; in the early days when we were dealing with MP3 players that held mega- rather than giga- bytes of storage, and before OGG and LAME were widely supported on portable units, a balance was required between having decent SQ output and small filesize. With stock headphones, the difference between 128KBpS MP3, and 64KBpS WMA (as was often compared by Microsoft), was negligible. The 'choice' then, was simple. It might not be the case today, but the saying remains true, in that old habits die hard. Even if people *did* later realise the difference, it is far easier to click on a drop-down box in WMP (to increase Bitrate), than to go digging around for new codecs/rippers.


Personally, I've been using MP3 (and MP3 players) since '98, first experiencing WMA encoding in 2001. I encoded quite a bit into WMA, since it allowed for far more music onto my 32, 64 and 128MB MPMan players. I did toy around with MP3 transcoding, but even with crappy stock headphones, the difference was massive, and ruined the point of listening to music in the first place.

When I bought my iRiver iHP-120 + EX71's back in February '04, I was finally able to hear the drastic difference in sound between WMA and MP3/OGG. OGG was a new experience to me at the time, and being one of the last of a dying breed, I actually got of my proverbial back side, and did some research.

From that day to this I've not ripped to WMA, and have re-encoded the CD's I still own (in the files and bitrates aforementioned). FLAC and WAVpack are relatively new to me, so I have no relevant files. I also don't have the appropriate equipment to enjoy FLAC and WAVpack (being a poor student) ...


MU4L,

P.S. I see I haven't actually answered your question  - sorry, I guess you'll have to ask a die-hard WMA user 

[Llorean]

Well, I'm not arguing against WMA by any means. If there are valid reasons for using it more power to those who do.

I definitely understand the "I don't own earphones good enough to hear the difference between 128mp3 and 64wma" argument.

As to the double blind thing, I guess I was just making the assumption that the "don't know better" category encompassed anyone who didn't research their audio format choice up front. 

[Marsdaddy]

Well, as far as this debate goes, I feel that whatever you think about WMA, if Rockbox supported it that would help some who are cautious about using Rockbox to make the switch.

As far as my work on the decoder goes - I now have a fixed point decoder . Here's what is left to do:-

1) Test it on a wider variety of WMAs - I've only used a very small number and I know there are code paths in there that haven't been tried, and will almost certainly fail.

2) Optimise it further and tidy it up

3) Incorporate it into Rockbox - I've started to look at this and think that I will tackle this myself.

I reckon that it'll be a couple of works before there could be a version of Rockbox with WMA suport. Not sure what the protocol is for getting my changes into CVS.