MP2 (MPEG I, audio layer 2) support?

I use a Pure Bug to save from DAB broadcasts in the MP2 (MPEG I, audio layer 2) codec. Is there a possibility of full support for this codec in a future firmware update? I know MP3 decoders should be capable of handling MP2 (and even MP1), but many don’t. I have only just got my 8GB Fuze, and am looking forward to loading up and playing back a host of FLAC files, but stable playback of MP2 would also be really handy.

If I have jumped the gun and the Fuze already handles MP2s (possibly renamed as MP3s), then apologies for wasting this message.

O.K., so I have now tried copying an MP2 file to the Fuze. Well, with its original suffix, theFuze does not even report its existence. With the suffix renemed to MP3, the Fuze reports it, but fils to play it, switching straight to the next file. I do hope MP2 compatibility will be provided via a future firmware update.

Oh, and in case somebody is thinking of suggesting trancoding to mp3 or ogg, I want to keep what limited audio quality the orignal MP2s have to offer. Transcoding to an alternative lossy codec must necessarily reduce that limited quality further.

Why don’t you recode the MP2s to FLAC?

That is indeed my (hopefully temporary) solution, but I would rather not bulk up a 128 or 192kbps mp2 to something like a 500kbps FLAC if I could avoid it by the implementation of MP2 decoding into the Fuze firmware.

If the quality is limited anyway, you might try transcoding to Ogg with -q6 or -q7 (192-224kbs). In my experience, Ogg is much better at preserving subjective quality than MP3 due to its encoding algorithm and the quality-based bandwith allocation. Of course this depends on the nature of the audio tracks in question, frequency spectrum, used instruments etc.

I did test transcoding 128kbs MP3 to Ogg -q6 once, and didn’t perceive any quality loss. For my old DAP (Samsung YP-U1 512MB), I had to transcode my -q6/-q7 library down to -q2 or -q3 to fit any sensible amount of songs on it. Using my Koss PortaPro headphones, the step between -q2 and -3 was quite noticeable, but any higher bitrates didn’t give much improvement - although this was probably a limit of the YP-U1s audio output. But -q3 was still pretty good, even after doing the Ogg-to-Ogg conversion.

Just give it a try and listen to the original MP2 and transcoded Ogg, it’s quite possible that you won’t perceive any noticeable difference. Trust your ears. Take into account where you listen to the tracks and what kind of headphone you use - this may make any quality loss totally secondary.

Well, I eventually gave up on trying to get the Fuze to handle mp2s, (still can’t understand why a codec so widely used on digital broadcasting is not incorporated in to the Fuze), and am happily using a little Archos 1 Vision for mp2 playback. The codec is not listed in the documantation for that device, but it plays my mp2s beautifully. The Fuze is now devoted mainly to FLAC playback.

The problem with transcoding from MP2 to OGG is that the power consumption drastically increases. Here is a good paper on the subject. As you can see on the third page the energy consumption of OGG is approximately 40% greater than energy consumption of MP2.

@doccolinni wrote:
The problem with transcoding from MP2 to OGG is that the power consumption drastically increases. Here is a good paper on the subject. As you can see on the third page the energy consumption of OGG is approximately 40% greater than energy consumption of MP2.

You can’t actually make conclusions like this because the power consumption depends on the individual codecs used and how they’re optimized for an individual player.  You need to look up which codecs they used and how they optimized them and then take that into account.  Of course you can’t do that because that paper is stupid and doesn’t say what they compared.  So their results are useless.  

Even ignoring that, they’re really weird numbers.  MPC and MP2 are very similar computationally (MPC is actually derived from MP2), but they claim its much slower?  Obviously one of their codecs wasn’t well optimized.  MP1 and MP2 are also much faster then MP3  (you can skip the IMDCT stage in mp3 which takes about 25% of the decode time), but they only show a small difference.  I think these guys just didn’t know what the heck they were doing.  Which is probably why it looks like they couldn’t get that paper published.    

---- 

If you want actual useful results for various formats on the Fuze, I suggest these from rockbox:

http://www.rockbox.org/wiki/CodecPerformanceComparison#Sansa_Fuze_v1_ARM922T

These are more interesting since theres a lot of CPUs compared, and you can look up the codecs and how they are optimized. 

Theres no MP2 results there, but I just tried on my Fuze V1 and I get 26 MHz for a 128k file, almost exactly the same as MPC.  As expected, thats a lot faster then MP3/Vorbis/WMA, so you are correct about not wanting to transcode the files.  However, since I think the Fuze OF gets the best battery life for either WMA or MP3, that might be your best bet.

[quote]You can’t actually make conclusions like this because the power consumption depends on the individual codecs used and how they’re optimized for an individual player.  You need to look up which codecs they used and how they optimized them and then take that into account.  Of course you can’t do that because that paper is stupid and doesn’t say what they compared.  So their results are useless.  

Even ignoring that, they’re really weird numbers.  MPC and MP2 are very similar computationally (MPC is actually derived from MP2), but they claim its much slower?  Obviously one of their codecs wasn’t well optimized.  MP1 and MP2 are also much faster then MP3  (you can skip the IMDCT stage in mp3 which takes about 25% of the decode time), but they only show a small difference.  I think these guys just didn’t know what the heck they were doing.  Which is probably why it looks like they couldn’t get that paper published.[/quote]


Sure, the guys from Tunghai University didn’t know what the heck they’re doing. What do you mean they couldn’t get their paper published? :expressionless: I don’t understand.

Also, MP2 and MP3 only showing a small difference? Are you nuts? The increase in power consumption between MP2 and MP3 is almost 20%, I wouldn’t call that small. Also there was no MP1 in their tests, MP2-1, MP2-2 and MP2-3 are all MP2 files with different headers.

If you don’t understand what and how they were doing it doesn’t automatically mean that they didn’t understand what they were doing, they’ve explained it in detail in section 3.1 Experiment Environment and 3.2 Experiment Setup.

Also relevant are these test results, although they are for Sansa Clip. But whatever player you used, such a huge difference between MP3 and OGG cannot be just a random fluke of the player. As you say MP2 definitely is less power consuming than MP3 (and based on that paper by at least 20%), so imagine the difference between MP2 and OGG if the difference between MP3 and OGG is so huge.

Message Edited by doccolinni on 2010-04-19 10:04 PM

Message Edited by doccolinni on 2010-04-19 10:05 PM

@doccolinni wrote:

 

Sure, the guys from Tunghai University didn’t know what the heck they’re doing.

 

 Correct.

doccolinni wrote: 

 

What do you mean they couldn’t get their paper published? :-| I don’t understand.

  

 The version you linked had no journal information and was not hosted by a journal, so I assumed it was unpublished.  Looking at your link, thats a conference proceeding (which is an electronically generated document that is not viewed by a human being before going onto the website), not a publication (which is a document that is edited and reviewed by a human being for content before being accepted).  So i believe I am correct.  They were not able to get these results published.

Presumably they were not published because either the editors or the authors realized they were ■■■■. 

doccolinni wrote: 

 

Also, MP2 and MP3 only showing a small difference? Are you nuts? The increase in power consumption between MP2 and MP3 is almost 20%, I wouldn’t call that small. Also there was no MP1 in their tests, MP2-1, MP2-2 and MP2-3 are all MP2 files with different headers.

 

 Compared to the difference they should have got, yes, it is small.  And really, comparing different MP2 headers is even weirder then what I thought they were doing.  Why would the header have any impact at all?  Does the author realize how silly that comparison is?  If he doesn’t realize what the header does, then what exactly does he know about MP2?  Anything at all?

When you read a scientific paper, you’re trusting the writer to know what hes talking about, unless like me you’re willing to go out and repeat his experiments.  If you read something thats completely dumb and suggests the author doesn’t know what hes talking about, should you really trust him?  I wouldn’t. 

doccolinni wrote: 

 

If you don’t understand what and how they were doing it doesn’t automatically mean that they didn’t understand what they were doing, they’ve explained it in detail in section 3.1 Experiment Environment and 3.2 Experiment Setup.

 

 No the fact that they don’t realize you need to say how they decoded the files is why I know they don’t know what they’re doing.  The experimental information they give is completely useless.  The useful information is withheld.  Either the author is intentionally being malicious, or the author is incompetent.  I assume he is incompetent rather then malicious.    

doccolinni wrote: 

 

Also relevant are these test results, although they are for Sansa Clip. 

 

Those tests show a 72% difference.  Your link shows an 18% difference.  How can you disagree with me that your link cannot be applied to the Fuze when its off by several fold!  You probably couldn’t have guessed a number and been closer then those “measurements”.

doccolinni wrote: 

 

But whatever player you used, such a huge difference between MP3 and OGG cannot be just a random fluke of the player. As you say MP2 definitely is less power consuming than MP3 (and based on that paper by at least 20%), so imagine the difference between MP2 and OGG if the difference between MP3 and OGG is so huge. 

 

 

Why should I imagine it?  I already measured it and told you what it is.  You have the information you want.  Stop speculating and use it.

And the differences are not a fluke.  The one in the paper is due to the paper being dumb and wrong.  The one with the Sandisk firmware is due to a bug in Tremor that will hopefully be corrected in future firmware revisions.  The correct results I gave you are due to my results being correct :slight_smile:  So three widely different numbers, three completely different reasons for them.   

I double checked that piece to make sure I wasn’t missing something.  In every single chart, Mp2 uses less power then AC3.  Every single one.  The conclusion:

Experiment results show that playing music encoded by AC3 consumes the

least amount of electricity, but the sound quality of AC3 is the

worst.

 

So it seems I’ve read the paper better then the authors who are apparently unaware of their own results!

I’m also a little annoyed to hear them complain about the quality of AC3 without specifying that they’re using ffmpeg’s encoder, which doesn’t actually have any of the usual bits that make an encoder sound good.  Its just a simple shell of an encoder meant to produce a compliant bitstream, not good quality, so naturally it sounds quite bad.  That doesn’t mean AC3 sounds bad, it just means they picked bad software.   

For what it’s worth: RockBox supports mp2, at least the samples I found here http://www.squee.eclipse.co.uk/samples/

played quite well…

Many thanks for the Rockbox tip. I will investigate.

First, It depends on how a perticular codec is supported weather or not it will use more power to decode or not, Higher bitrate will take more power to decode then lower(flac vs ogg for example) but thats a different topic.

Vorbis(ogg) on some players KILLS batt life, on others like the fuze, you loose at most an hour or 2 of batt life(my testing the WORST I loose is 1.5hr vs using MP3 files)

Vorbis using aotuv’s latest encoder could likely give you audible transparency at q-2 or q-3, only rare tracks take higher when using the latist aoTuV encoder, q-6 is considered transparent on all but a few “killer tracks” (tracks designed to cause problems)

Note MPC uses less power to decode then mp3, but isnt currently supported by fuze or any other sansa player either.

My advice if you want to carry around your pod casts or whatever that by default come in MP2 format, convert to vorbis and toss them on your fuze, They will use less power then the FLAC your already using and you will be able to fit more of them on the player.

I use vorbis because no other codec I have found has the quality to bitrate that vorbis encoded with aoTuV’s encoders offers, and Im sorry but carrying around a few flac’s on a portible player is silly, it eats batt life, the players also tend to have a lower snr then a good set of IEM’s or cans(the cans would likely also have pretty high resistance so u would need an amp) when it comes down to it, Ogg just makes sense for these players, it decodes fast, offers excelent size to quality ratio and has plenty of encoders to play with(dbpoweramp FTW tho)

Oh and this isnt a one off with only loosing a little bat life, my samsung yp-u3 lost around 1 hr moving to ogg from mp3.

Ogg can be decoded MANY ways, there are a few common ways, but there are also alot of less common decoders written to be optimized for specific player hardware, samsung and sansa and cowan and archos tend to do a decent job on newer players when it comes to their decoders being well optimized to use as little power as possible and still maintain excellent sound quality.

alot of cheaper players use…far less optimal vorbis and/or flac support that ■■■■■ even more batt life, I have seen some cheap asian players that had their batt life cut in HALF by using anything but mp3(not kidding, its really that bad) then again i have seen some asian players that had BETTER support for non-mp3 formats then they do for mp3, for mainstream players like the fuze/clip/exct the diffrance in batt life shouldnt be very large these days, I know Im not gonna go back to mp3 over ogg for my 8gb fuze(with 8gb card) I like being able to fit lots of music and audio books onto my player without having to resort to my stuff sounding like it was recorded in a toilet bowl :stuck_out_tongue: