WAV vs. FLAC - Should Naim aim for no difference?

The latter I think. FLAC seems like the solution to a problem that didn't exist. Hard disk real estate isn't expensive and after that it's all garnish if musical enjoyment is the priority. 

Naim (unlike Linn) have stated that the best results are acheived with uncompressed files - WAV - and have stated the reason why they think ths is. The serving and ripping hardware they recommend rip to this format so why not design it to sound the best with this in mind. If Naim think they can't design the NDS to handle all formats universally well then so be it. I'd rather them design with one prime format in mind and make the best of it rather than compromise. I'm surprised that the difference was that marked in the demos though. Was the same server providing both the WAV and Flac files ?

James

I don't see why this difference, if it is real, should be accepted as inevitable at all. Linn engineers, quoted on this thread on the naim forum say that in the Linn DS "if we measure the power rails that feed the audio clock and the DAC we see no evidence of any processor related disturbances. There is no measurable difference (down to a noise floor measured in microvolts) between FLAC and WAV".

So a difference is certainly not inevitable; if there is one it just reflects on the design. Naim were late coming to the FLAC party; Linn were into FLAC from the beginning. Maybe Linn have had longer to get it right.

FLAC is designed to be computationally trivial to decode, well within the capabilities of Mickey Mouse processors such as you find in £50 streamers and cheap players. If thousands of pounds worth of fancy streamer are in any way taxed by decoding it I would ask for my money back.

Does an NDS feeding an nDAC have the same problem? 

There should be no difference between WAV and FLAC .... if there is then something is broken, as far as I'm concerned. As I understand it the NDS will have no sonic difference between lossless formats, as the optical separation between renderer and DAC should kill the noise. I hope this is the case as my music collection is not in WAV format, so I'm only interested in something that plays FLAC/ALAC/AIFF or a format I can tag successful with no loss of quality.  

UOJames, when listening to the NDS (preproduction) there were clearly differences between WAV and FLAC. The WAV had more air and flow, and most if not all heard it and commented on it. This comparison was playing files by both upnp /DLNA server (Unitiserve) and USB into the NDS.

The sound with FLAC didn'tsound broken, but did sound uninvolving, and with the Titans and 500  series amplification was quite marked. 

We also heard some mediocre as well as outstanding hidef... It shows a good CD rip / spin can stil hold it's own.

I think with the Naim architecture the secret is to transcode or convert to WAV if you are critically listening or want max musical enjoyment.( assuming you can hear these different cues between WAV and FLAC and I accept we not all do). I would hate Naim making the sound the same by compromising WAV. I suspect if with NDS they don't sound the same it will require a major architectural change from Naim to make a difference, and after all the NDS is a refinement on the NDX not a complete redesign.

Guy, my collection has FLAC as well as WAV, however I simply real time transcode to WAV before sending to the Naim and there is no problem. BTW I find FLAC metadata options sometimes limiting, I love the, albeit optionally supported, ID3 extension to WAV that allows bespoke and custom meta data tags with some players and servers. Having said that standard EXIF tagging (using dedicated stanardised INFO tags) with WAV, that all modern wav software *should* support is quite effective for artist, album, track, sequence, date, but admittedly is a subset compared to FLAC  ID3 and even more so WAV ID3.

Simon

This may very well be true, I do not know.

I am (clearly) not qualified to discuss the technicalities, but I confess that the difference of opinions here as to whether there was or could be a difference in the sound of different lossless formats made me apathetic about doing my own homework thoroughly.  Now that I have begun to seriously compare AIFF to ALAC in my system (SBT/nDAC/555PS,, etc.), I certainly like the sound of AIFF better.  Whether or not the SBT is not playing ALAC properly, and does better with AIFF I cannot answer.  I guess when I get around to doing similar tests on my Mac Mini, if the delta is the same, it might mean others have similar difficulties with the respective formats, or maybe they are using similar or shared processing technologies, distinct for each format.

If Naim decided it was easier to overcome the tagging issues uniquely painful with WAV than make the other lossless formats sound as good, it was either the lowest-cost to the best overall user experience, or a cunning way of delivering a great-sounding solution that defies easy comparison to, and imitation by the competition, on some levels.  Maybe we should just trust Mother-Naim.

Interesting business.

Nick

The question was "WAV vs FLAC. Should naim aim for no difference?"

Are you arguing that Naim should aim for a difference? That it is desirable? 

Maybe you have better information than I do.  I am inclined to assume Naim did their best with all of the formats.  I suppose, without actually knowing what is ultimately possible, my assumption is really just speculation.

Nick

Naims "best" with FLAC was originally absolutely exist non-existant. They only used WAV. Now they do offer FLAC but haven't yet achieved the same standard. Shouldn't you hope they do? Shouldnt you hope that their best gets better?

And noone really answered the point I raised concerning an NDS plus nDAC. Does this combination also make FLAC less good than WAV? If it does, then the nDAC must be incredibly sensitive to processing/power supply issues in the NDS which is surely disappointing. If not then such differences aren't inevitable.

Yes - i understand that, and having designed within mixed signal environments, i appreciate that the task facing the Naim engineers with a box like the NDS isn't trivial and compromises must be made somewhere - perhaps if cost is no object then things would be different. As Simon has said if you hear the difference then transcode at the server.

I do find it amusing though that owners will quite happily accept the need for careful cable dressing, correct placement of components etc then complain that the the NDS doesn't sound as good when playing files which Naim say are compromised compared to their prefered Wav.format

I seem to have stirred up a hornets nest...

I was trying to come at this issue from a slightly different angle and explore 2 things. One is whether it is possible to design a system capable of reproducing FLAC and WAV with no compromise on quality and no discernable difference. The other assumes the first point is true, and asks whether Naim do themselves a disservice by highlighting the limitations of today's hardware and encourage critisism from others who claim to have this problem nailed?

On the first point, I think it is widely accepted that any perceived difference in FLAC is due to the additional processing required and the resulting incremental noise, either transferred by electromagnetic radiation in the chassis, or through power supply fluctuations. As Naim's core philosophy focuses on big stiff power supplies, star earthing, and isolation I would have thought they are better placed than anyone to have a stong defense against the effects of processor noise on the delicate analog circuits that follow?

Is it suggested that other manufacturers who do not highlight this issue are deliberately sub-optimising WAV playback to level the playing field, or are there competing software or hardware architectures that maximise the potential of both? If so, what other compromises are being made?

Other than an interesting debate, I'm also looking to get an NDS soon so am interested in getting as much insight into optimising its performance. At present all my music in stored in FLAC format and I've invesing a lot of time getting the tags correct as I want them. At present I stream FLAC to my Transporter which decodes the FLAC and drives SPDIF to the Naim DAC / 555 PS DR. With this set up I have never been able to detect any differece between FLAC and WAV, so have been happy to stick to FLAC.

When the NDS arrives I will have removed the SPDIF compromise and moved to a later generation suspended chassis DAC with all of Naim's latest tweaks. However, the streamer and DAC are in the same box and subject to the same electromagnetic environment (I recall there is some extra internal screening in the NDS?) Initially I will still be using WiFi as I can't easily drop a cable into my listening room without incurring the wrath of Mrs B.... a new room is planned for next year which I will be able to fully optimise for listening. You may argue that the WiFi compromise has a far more profound effect than WAV vs. FLAC

I plan to use Asset to stream so I guess the obvious answer is to transcode to WAV on the server, and stop worrying.

Anyway, let the debate continue...

It was highlighted above that Linn extols FLAC and it maybe that their 'test' ears found no audible difference to WAV on the SPECIFIC setup WHEN they did it. As we know streaming setups are many and varied and the whole concept is evolving rapidly.

I personally decided that, as mentioned, Disk space is really not a constraint and since the potential audio result was my main determinant, rather than tags and even album art, I have always ripped to WAV. The fact that I stumbled into using dB Poweramp and Asset means, it seems, that I can have tags and Album art etc with no issues with WAV.

That said I have had plenty of opportunity to compare FLAC with WAV on my Linn KDS/1 and personally hear 'something better' with WAV rips than FLAC. In some cases I have intentionally bought the FLAC of a CD I have already ripped at home to WAV and even have the LP version as a 3rd test disk. I just prefer WAV and thats that. It stacks up better against the LP replay, for example, to my ears. 

However there is the obvious other factor here and that is the attraction of a less resource hungry audio library. Linn for example has a growing relationship with multiple labels and regardless of other issues debated, HD tracks also has a large library. They are both, as are other providers I am sure, fully supporting FLAC, I believe because it is an easier and less data intensive format to create fully tagged, store and deliver to customers.

How much that motivates their opinions about FLAC is not in the public domain but I would bet it is high on their lists of determinants.

As has also been pointed out FLAC can be transcoded BEFORE delivery by the likes of Asset. I do this as a matter of course, again because I can, but it doesn't seem to come out sounding quite as enjoyable as direct ripped WAV.

If Naim, and for that matter, other serious HiFi suppliers want to use a 'yardstick' against which to optimise, as a first priority, then IMO that should be WAV. Certainly they should aim for the other formats to sound as good as they intrinsically can but for me WAV should be the reference.

Oh my, this is getting rather confusing.

A while back, using my own setup (see profile), I asked Mrs. Hook to listen to FLAC versus WAV versus FLAC transcoded on-the-fly to WAV using Asset UPnP server.   In other blind tests, she's been very good at hearing subtle differences (24/96 versus 24/192, Tibia versus Powerline, NDX S/PDIF versus RME 9632 S/PIF, and so on), but I am afraid in this case she could hear no difference amongst the three options.

I rip using FLAC level 0 (minimal compression), so perhaps that contributes.  Were it an option at the time I started ripping, I may have chosen dbpoweramp's "uncompressed FLAC" option.

In any case, I've left Asset's transcode option turned on because occasionally I listen to music using an old AVR which can act as a UPnP client, but can not understand FLAC.

@Simon - in the past, using your NDX->DAC/555PS, I recall you saying that you could not hear a difference between FLAC and WAV, so either the NDS is different, or the rest of the Naim demo setup you heard is more resolving than your Cyrus setup?  As you said, the NDS is a refinement on the NDX, so perhaps those refinements really do bring WAV versus FLAC differences to the fore.  

But WAV versus FLAC transcoded on-the-fly to WAV?  That one I can't get my head around...

@Geoff - I think yours may be the first post I've read where someone says they can hear a difference between WAV and FLAC transcode on-the-fly to WAV. Which UPnP server are you using?  Obviously, the difference here can longer be attributed to the streamer's level of CPU processing, so the only explanation I can think of is that the UPnP server is somehow not uncompressing the FLAC file properly (which is surprising, given how proven a process converting FLAC to WAV is).  I am stumped, and would appreciate hearing any theories anyone has to explain this particular difference.

My goal is try and understand if there is any benefit to re-ripping my CD collection.  I have over 2000 CD's, so the effort would be substantial.  Can't imagine re-doing this using a PC, and would almost have to resort to a more automated solution (UnitiServe or a 3rd-party all-in-one ripper/tagger).  I thought I had a pretty resolving system, but I am well aware that others, like Geoff's, are even more so.  And even though I cannot hear these differences today, my thought is that at some point my system will be upgraded to the point where presumably I can...and perhaps as soon as the NDS replacing the NDX->DAC.   Right now, however, I am more than a little confused by these reports, and it would be understand if it down to the streamer, the UPnP server, or the files themselves.  I know that previous efforts by Simon and others showed there were no differences between the PCM in WAV files created by a UnitiServe and other ripping methods.  It would be great to see someone figure out how to capture the output of Asset (or any UPnP server) converting FLAC to WAV on-the fly, and see if those bits match up exactly to streaming the original WAV file!

Sorry to ramble.  Any thoughts and/or advice is much appreciated.  Maybe I should I just stick to vinyl?

Hook

Dave - i think we can only really speculate and use our ears - the implementation of the streaming section is different between the Linn and Naim players and so the effects on other parts of the player when performing more processor intensive operations will be different. Linn have gone and measured various parameters and have stated there is no measurable difference whereas Naim have stated a musical preference for uncompressed data citing an increase in the power supply noise floor as the issue. I'm sure Naim could design the NDS to be format agnostic, but it's a complex design and in the end there is a budget to work to. At least by deciding on Wav to be the prefered format, they can characterise the streamer board performance when running under these conditions and engineer the rest of the design, both mechanically and electrically to suit.

Geoff  +1

James

The difference between Linn and Naim is not a difference between measurement and musical preference.

Naim have stated that for their units their is an increase in the power supply noise floor. This is a measurable phenomenon (otherwise how would would they know it exists) though with musical consequences. Linn with a radically different approach to power supply design can measure no increase in noise in the power rails that feed the clock and DAC in a DS, though they can indeed measure differences in noise earlier in the system. So it would seem that Linn have found ways to keep the noise away from the clock and the DAC but Naim haven't.

If you stream FLAC to an NDS feeding an nDAC does the increase in power supply noise floor in the NDS affect the nDAC? Even with all the power supplies/upgrades and rebuffering?

Linn presented measurements stating that measureable artefacts should be low enough not to make a difference. Naim have said there is a difference (which as you state is measured) and have stated a preference for a particular format.

It's not really a radical difference in power supply design - both use multiple rails, with multiple regulation - the main difference is linear vs switch mode.

How do you come up with the assumption that Linn have solved the issue and Naim haven't ?. Users on the Linn forum find differences (and i found with my Akurate DS that i preffered AIFF to Apple Lossles - the same i find with my NDX). - have you done the comparison yourself ?

And yet Geoff (whose ears I trust) still hears a difference between FLAC, WAV and WAV transcoded to FLAC on-the-fly.   And between WAV rips and FLAC rips converted to WAV (and then streamed).

So, broken transcoding?  Broken format conversion?  Or maybe, just because Linn cannot measure the noise, doesn't mean there is some other side effect of the additional processing?  

No answers here...just grasping at theories in attempt to understand how these differences can be audible.

Hook

If you check the link to a similar discussion I posted earlier you'll see that Linn can measure different types of noise for WAV and FLAC earlier in their system, but by the time the power supply gets to the DAC and clock there are no such effects.

Im happy to accept that there are  differences between formats in Naims current products - though still everyone here has bodyswerved the question I raised concerning the NDS/nDAC combo - but I wouldn't be happy to accept that these differences are inevitable, as it implies that engineering has reached its limit.

So, to answer the o/p's question, imo Naim should definitely aim for no difference.

I don't think you are going to find anyone planning for that combo as the nDAC will only downgrade the NDS.

maybe in the future the market and customer will decide wav/flac .  the majority of direct usable files are flac, seems there is a difference in conversion to my ears/system a very very small one.  in ones systems other changes will give more rewards.  guess this generates upgrading and selling more equipment

Beg your pardon, I meant the NDX.

Well it is kind of you to be trusting Hook. Just don't make a major ripping decision based on little old me alone.  

As a comment I do believe my current setup is pretty well optimised and uses an HQ audio chain. ( 552/500/KDS1/Kharma Ceramique 2.3s on two discrete dedicated 16mm spurs using wired gigabit connections for both the KDS and the NAS running Asset UPnP ).

What I haven't done for some time is RIP to FLAC myself. When I first setup a streaming system a few boxes and years ago I did a few tests of the first CDs I ripped to compare FLAC & WAV and felt then that FLAC's took on a slight metallic edge when the volume ramped up that didn't set in with WAV. As a result I decided up front to rip to WAV.

All FLAC's I have now and am comparing in this discussion are downloaded versions. This means I have no idea and have not investigated the FLAC compression settings used by the likes of Linn on their music sites. 

The audio subtleties of the original CD master and whether the actual CD is a reissue etc. are a significant element to consider I feel. To be precise I am not saying every FLAC file sounds  as clearly different from the WAV equivalent. If it is a slightly crap recording the small differences I pick up on with the very best recordings start to drop away. 

Interestingly I have a few DVD-Audio discs with 192 and 96/24 files on them which can be ripped with DVD Audio extractor at the FULL resolution and I have been able to compare these LPCM hi res rips against downloaded hi res FLAC versions. Again I prefer the WAV files to FLAC, but this is only from a very few comparisons mind you.

As to picking up on differences between straight WAV delivery vs Asset transcoded WAVs of FLACs that maybe me being preprogrammed to prefer direct WAV rather than a real difference. I just don't know anymore....and anyway it is not like the FLACs sound awful. They are just as enjoyable really.

Geoff

.