Wave or Flac

Because, there is no information (stuff) needed to be put back as it is not removed. Simplistically when expanding FLAC out to WAV, the information  is being put back into its inefficient encoding types.. Such as a WAV file of stereo PCM samples.

The information stays the same, the encoding of the info varies. We are converting the encoding not the information. You can do this encoding translation infinitely and no information will ever be changed (Lost or corrupted)

Simon

Oh i see, is this why it actually doesn't use a lot of CPU to decompress the file. Its not really having to do an awful lot.

An 'awful lot' is a relative term, but yes, decoding FLAC to WAV stereo PCM can be done very efficiently and quickly in code.

The bottle necks can be more linked to I/O bandwidth/throughput on small UPnP/NAS devices rather than the CPU load per se. But that's another matter...

Simon

I'm new to this stuff & as an experiment have ripped about 10 CD's to both FLAC & WAV.  Some are 24bit HDCD, others are regular 16/44.  I have yet to sit down for a serious listen, but will be doing so.  The next step will be to transcode the FLAC on the fly to WAV & check that out.  

But as WAV seems to give me all the metdata I need & my NAS is running at 15% full, its purely academic.   

Outside of the different technicalities between FLAC v WAV  (agreed exemplary explanations from Simon)  surely its about SQ. If you do hear a difference between the formats & you prefer WAV but are concerned that it compromises your available HD capacity, then surely this indicates you need more capacity.

when you rip a cd to mp3 you loose information is in it the same with flac it`s not better to rip straight to wave and to leave it as it is without converting again

Mike absolutely... But if you transcode to WAV away from your Naim equipment, your equipment has no way of knowing what format the information was before it was encoded as PCM in the UPnP media stream. Having more efficient encodings of the PCM helps with storage and speed of incremental backups.. 

But I think the key point is whether you prefer the sonic signature of Naim playing WAV, FLAC or no preference.. The advantage of current UPnP technology makes this user preference as simple as ticking a box in a config. It doesn't have to be determined by what format your thousands of CDs are encoded as.

Simon

PS when you transcode the info on the fly look at a Wireshark trace of the PCM whether originated by WAV or FLAC.. Tell me what differences to the sample data you see 

I'd expect there to be none, if the transcoding application is doing its job correctly.

Mike. At the beginning i did a lot of tests trying out flac v wav v flac transcoded to wav. Wav sounded a lttle more natural compared to flac, but ive never been able to tell a difference between wav and flac trancoded to wav.

Mind you not saying that others cant tell a difference, but i certainly cant.

Graeme

Thanks Simon your explanations have really helped. Now where is db's batch converter. More time at the PC :-(

Simon's answers are erudite as always (besides its always good to hear from a fellow fan of early Kraftwerk).

However, the following made a bit more sense to my baffled daughter: i think this is all analogous to maths notation so

wav would be: 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2 x 2

flac would be: 2 exp 10 

and mp3 would be: 1000 ish

the wav and flac contain exactly the same data but the flac is more efficiently stored whereas the mp3 contains an approximation that cant be reverse engineered back to the raw data? 

......in reality it's far more cunning than this but you get the drift ?

....oh and to my ears Wav definitely sounds better than flac on an nds. i followed simon's advice and use Asset for rock and electronic though i prefer minimserver for classical Metadata.

meni48,

There is a lot of science that can be bought to bear on WAV vs FLAC, but ultimately it comes down to what you hear in the context of your system.

A number of years ago I built a Windoze based CMP2 system and used a HiFace to get the sound out into my nDAC. In that system I heard no difference between FLAC and WAV.

The problem with the system for me was that I had stripped out ALL extraneous functionality from the operating system, including networking. This meant that I couldn't use remote storage, and that the PC was akin to an electronic turntable. A year later I bought a Naim NS01.

With the NS01 I could immediately pick between FLAC and WAV, and WAV was better.

Now the NS01 pulls the files from my NAS using SMB, not upnp, and so this is not directly transferable to your situation, but I would encourage you to test all the variations, including zero compression flac.

M

To my knowledge, MP3 is lossy format and FLAC is in lossless same with WAV, AIFF & ALAC, therefore there is no data loss when transcoding between the 4 lossless formats. Flac is compressed (meaning zip file), we only have to unzip it to become WAV or AIFF. 

I have not compared Wav with Flac because all my albums are in Aiff. However, I do hear a subtle difference (for the better) in Aiff over Alac format. 

Mikapoh, ALAC is not that dissimilar to FLAC, and to all intent and purpose the default PCM encoding in AIFF is the same as WAV other than the sample data least and most significant digits are reversed.

Simon

Martin

Are you saying there is a clear difference between these two application in sound quality terms. This again baffles me if they are sending the same information to the player

Graeme

I've got cloth ears, but can still hear the difference and much prefer streaming WAV. I believe the difference is because the streamer has less work to do to get the PCM stream out of WAV than it does with FLAC. I now rip to FLAC and get the UnitiServe to transcode for playback.

So FLAC is interpolated. That's what I suspected and it potentially explains why I can hear a difference.

Graeme, Martin is referring to the metadata as opposed to the sound of the media in his post. 

For whatever reason Martin appears to like how Minimserver maps the classical music metadata for UPnP (the tags, labels and descriptions) as opposed to Asset.

personally I believe this can all be handled when creating the metadata in the first place, but Martin will have his reasons.

BTW I can hear (subtle) sound differences between UPnP app/OS combinations, but I will leave that for another thread as it has nothing to do with FLAC and WAV

Harry, can you explain how you mean FLAC is interpolated? it works with integer samples like WAV as opposed to floating point.

Simon

If you're a software developer and writing a FLAC encoder, it should be pretty easy to test it:  convert test files from WAV to FLAC and then back to WAV again.  If the 'before' and 'after' WAV files are the same then you're in business.

The only possible SQ difference would be because of different levels of digital noise in the rendering device (e.g. NDS) - although this is hypothetical.  From personal experience I'm highly skeptical about people who claim to be able to hear the difference between the two formats.  However, I can believe in "FLAC anxiety" brought on by a need to worry about things.

I backed into the FLAC vs WAV difference by accident. I couldn't understand why some of my HiRes recordings had a slight tendency to sound thin and phasey. I put it down to the recording quality (which in the vast majority of cases has more to do with reproduction than any possible listening chain issue) but converting them to WAV and playing back without conversion did seem to improve matters. 

If it is necessary to subject a file to interpolation, however processer friendly and however accurate the result, we're back to the kinds of processes that happen in CD playback that streaming is supposed to banish. To me this kind of defeats the object, or at least a part of it. If WAVs were impossible to tag or took up ten plus times more space than FLACs and HDD storage was very expensive (as opposed to the cheapest it's ever been and going down still) I could better understand a need. And if you don't compress FLACs and therefore don't need to reassemble them and add bits back in, they are apparently bigger than the corresponding WAV? That's useful then.

FLAC doesn't use interpolation - a decompressed FLAC file is identical to the original.

Agreed, but it's still got to be decompressed. Bits that are no longer in the compressed version must be  inserted. Maybe I'm misusing "interpolation", perhaps it should be "interpretation". The results must be accurate for playback to be possible and as you say, if you compare before compression and after compression there is no difference. But the process of doing this may potentially impact on sound quality and perhaps this is why I prefer to listen to WAVs played back? Don't know.

Harry

IMHO that is exactly what is showing in revealing systems and people sensitive to these small differences. The extra processing required for rebuilding the original PCM-stream from FLAC, can create extra noise that becomes noticeable.

i'm involved in another software player and there it shows that much of the things a processor has to do in getting the bits ready for playback, will impact sound quality, e.g also the way it communicates with memory and the order in which it retrieves, copies and flushes processor caches and driver IO-area.

There is no bits-are-bits Utopia.

Aleg, Harry indeed there is no difference in the sample data conveyed in FLAC, AIF or WAV. The sample data however is encoded or perhaps I should say stored differently in each format.

The renderer has to work to extract the stereo samples from the various encodings. The amount of work is greater with FLAC and ALAC vs WAV or AIF for a given set of  samples. 

In the world of micro controllers and processing electronics work can be proportional to electrical noise created. Electrical noise modulates ground planes, supply lines and radiates through EMI. Digital clocks are hyper sensitive to noise, and sensitive analogue stages are very susceptible  to noise.

Therefore it stands to reason that when decoding FLAC more noise is created than with WAV, and perhaps some of us hear this noise negatively.

I do for example, which is why I transcode. FLAC to WAV, thereby reducing the amount of work my Naim renderer has to do.

As Aleg intimates there is a world of difference between being 'bit perfect' and 'bit are bits' from the quality of the audio we hear.

Simon

Hi Lars,

Phil at Naim has recommended transcoding FLAC to WAV by the UnitiServe. The rationale is indeed the reduced processing by the streamer. This does contradict the manual, but I can't remember how Phil justified the discrepancy. I suppose it's better to reduce processing wherever possible so you set it to native if all your files are WAV. My non-Naim rips are mostly FLAC so I transcode. 

Keith