Bug with FLAC ?
Posted by: mige0 on 27 August 2011
If there isn't any at the moment, anybody to know when this possibly to be awaited or any suggestions for a workaround (Qnap NAS, Twonky)?
Michael
I haven't noticed any SQ issue when playing FLACs off USB stick (I don't have a NAS yet and WMP doesn;t reocgnise flacs so can't stream)
I had loads of artefacts on a sample track I downloaded. I downloaded it again in FLAC, in WAV but the artefacts are present on those too and in the same or very similar position in the track so I assume it is the recording. I certainly wasnt as low quality as MP3 (assume you mean quite low rate since you don;t specify)
Have you tried playing the track(s) on another player.
Could the track(s)be upsampled?
Coldn't belive myself first time, thought it could have been a burn in issue.
Converting FLAC to WAV - stored at the same NAS - auditioning - outch ! - there definitely is something wrong with FLAC playing at the UnitiCute.
Dynamics are compressed also loss in image and stage compared to WAV.
I'll go and try the USB too - didn't check until now
Michael
Simon
So the tracks are OK when converted /trancoded to WAV?
The issue is less audible when playing from USB (with exactly the same files !)
Also - there seems to be a difference between WLAN and LAN. Playing the NAS over LAN - which I just did try - there is more or less the same sonic difference between WAV and FLAC like playing from USB (more subtile).
Looks to me to be a buffering problem - which is kind a paradox, as FLAC should be less data transmission.
Michael
If it was a buffering problem, wouldn't there be dropouts or stuttering in playback, rather than a loss of SQ?
Anybody knowing if there is any check for bit accurate reproduction (Error correction = 0) ?
Michael
Sounds broken to me.
So the DAC interpolates missing data? Sounds unlikely.
Simon
Does converting to WAV get them back on track?
Mige0 -
Much has been written about the subtle improvements that are possible when going from FLAC to WAV streaming using Naim network players. But most of the speculation I've read is that the improvement is a simple matter of reducing the processing load on the player by not having to do uncompression, and thus reducing the amount of noise the processor generates. I do not see this as a bug with FLAC, but instead as an optimization using WAV.
I rip all of my music to FLAC (level 0), and have my UPnP server convert on-the-fly to WAV. But if I could go back and do it all over again, I would have used dbpoweramp to rip my CD's using its new "uncompressed" FLAC option.
If you have access to dbpoweramp, you might want to try comparing WAV to uncompressed FLAC. I would bet you will hear next to no difference, but it would interesting to hear your results.
Good luck.
Hook
PS - Sorry in advance if you are referring to something completely different here. I own an NDX -- and not a Uniti or UQ -- so I do not know if these two products have something wrong with their FLAC replay. It is definitely not the case with the NDX. Mrs. Hook has shown me that she hear a difference between streaming FLAC and WAV (we did a blind test), but I confess that I cannot! On the other hand, she could hear no difference between WAV and FLAC converted to WAV on-the-fly. I figure it can only help, and cannot hurt, to do this uncompression work on the UPnP server, as opposed to my renderer (so I've left this server option for on-the-fly conversion to WAV turned on).
Simon
Thank you both. It never occurred to me to convert FLAC to WAV but it's hardly a chore and worth investigating.
thanks for all your input
Sure I could convert to WAV but thats roughly a 3 times more HDD space - and - there are wiggles with the different versions of WAV (including tags or not).
If I take the effort and cost to convert whole HDD, I would only do it if SW-up-sampling to 96k will prove for better reproduction compared to 44kHz
@ hook
well - if a gear does not handle FLAC if it is sold as to being able to handle FLAC as its meant to be = eg. lossless (!) - at least I would prefer to call this a bug - but I do understand and respect you trying to be polite.
As my (pretty quick) checks did indicate that there are differences regarding the choice of input too (LAN / WLAN / USB) it may be worth to look into the issue for NAIM.
Yes its been dbpoweramp I used for ripping to FLAC (guess its been default level 5, not sure).
Subsequent converting to WAV was done with SoundConverter
@ Hook & Simon-in-Suffolk
"transcoding on the fly" looks interesting - will have a check if I can find more about ASSET and how to implement on my Qnap NAS - any advice from your side?
Michael
Perhaps another Naim type solution is to use an NDX to transcode to s/pdif and feed it into an nDAC?
Yeah, seems that compression levels other than zero in FLAC is what NAIM can't handle.
Just switched to 96k for a test 44.1k versus 96k and here the same thing apples.
Thanks for pointing me to the right direction !
FLAC 96K converted in default Level 5 sounds (slightly) off, but FLAC 96K at level 0 sounds OK to me - virtually on par as WAV 96k.
So it seems I have to up-sample to 96k and use FLAC level zero to bring out the best of this NAIM gear at considerable disk space consumption.
Michael
Simon
Alternatively get a liitle WHS machine, put asset on it, and have the flexibility of a full blown uPNP server with access to one or more NAS's . It what I' do and it works so well I forget about it's there.
If you do this you find that a full blown uPNP server like asset provide useful options such as it can transcode differently to different clients, ie you can tell asset to transcode to mp3 for the clients in the conservatory, garden and kitchen via wifi, and transcode to a full uncompressed stream to the ndx in the lounge, from all from your FLACs. Perfect.
Simon
On the other hand - having jumped into that things and discovered SQ difference where non should be - I up-sampled to 96k - just out of curiosity and - hey - thats *really* worth to do for me.
So for today I ended up to go this route:
to SW up-sample to 96k as NAIM seems to be way more happy with this format and stay with FLAC (compression level zero!) to stay within reasons regarding disk space.
Michael
You might be interested to know on the NDAC and NDX Naim up samples to 768kHz (from memory ) but they don't interpolate. Of course by up sampling you will have probably created some arithmetic rounding errors and created some spectrum noise. But that is clearly sounding preferable for you. What algorithm are you using for up sampling and to what data word resolution ? Failing that what program are you using?
Simon
If the hypothesis is that it is the extra processing required to decode FLAC that is responsible for an NDX sounding poorer with FLAC as opposed to WAV, has anyone tested whether switching wireless encryption off has a similar effect?