WAV vs FLAC revisited

[mep]

I'm curious as to what the majority of people find provides them with the higest fidelity for digital playback (if there is such a thing). Is it Flac, WAV, or something else? I thought Flac was the way to go based on RR using it for their HRx releases and I have converted all of my RB CDs to Flac and I download Flac files from HDtracks.

If I could start a digital food fight that would distract digital people from my other thread that would be OK.

 

[Bruce B]

The choice of FLAC has nothing to do with audible choice and everything to do with size (storage space and download bandwidth) and tagging.

 

[Ron Party]

If I could start a digital food fight that would distract digital people from my other thread that would be OK.

Ha! Good one.

Mark, stick with flac. Your ears and your pocketbook will be very well served.

 

[fas42]

The trouble is that you cop too much flak from others if you use Flac ...

Frank

 

[mep]

Ha! Good one.

Mark, stick with flac. Your ears and your pocketbook will be very well served.

In all seriousness, I was/am interested in what the digiphiles think is the best software and software format to play back digital files with. I started off with Mediamonkey and settled on Foobar. I have always downloaded Flac files and ripped my CDs to Flac files. I know some people like JRivers software and some like WAV files. And yet I read other discussions where people claim none of it can make a difference because you are dealing with digital files. I don’t know what the ‘truth’ is or if any of it really matters in the ultimate sound quality. I have three different SACD players and the format never excited me. I can play back digital files up to 24/192, but I can’t play back the higher rez DSD files that some claim are the bee’s knees.

 

[Ron Party]

Again, I'd be surprised if you could reliably distinguish between JRiver and Foobar. To me the decision is one of user friendliness and this is where JRiver shines. I know you have faith in Mike Lavigne's take on matters audio; he has already posted about his experience in playing DSD files with JRiver. JRiver has an small entry fee whereas Foobar is free.

I still use Media Monkey myself, not for playback, but for tagging. I really like its user interface for this purpose.

 

[bblue]

Well I'll be ......

It's been a few years since I last compared the playback of flac and wav files (flacs made from the same wav, or wav made from the same flac. And up until now I firmly believed there could be no difference in playback between them.

I was wrong.

They aren't huge differences but they are readily identifiable. I normally use Foobar2000 (latest beta). What started me thinking about this again was I noticed when playing music in Foobar, that a change of clock source made quite a difference in the sound. More than I would have expected.

I normally use a Delta 1010lt and its internal clock for ordinary playback via the ASIO audio drivers. Its S/PDIF output goes directly to my Crane Song Avocet monitor controller. Just for the heck of it I redirected the output (still ASIO) to my Profire 2626 digital interface which has its own clock (supposedly pretty good according to the guys at Black Lion Audio) and interfaces to the PC over Firewire. The audio quality improved significantly -- better detail, cleaner soundstage, etc. Then I replaced the Profire clock with an external clock generated by an Apogee Big Ben word clock. There was a subtle but easily identifiable improvement again. At this point it seemed like any hash/noise/artifacts I had been used to in the audio just disappeared. So that became my new reference.

Leaving everything that way I took several flac files (88.2k and 192k) from HD Tracks and converted them to wav's. This is music I've heard a lot and am quite familiar with (Ray Brown Live at Loa, Oscar Peterson Night Train and We Get Requests, Modern Jazz Quartet Last Concert and Blues on Back, and a few others a bit more raucous). I took one song at a time, listened to the flac a few times, then the wav. Surprise! It really was different. On the wav's overtones cleaner and more distinct, more space between instruments, deeper and more detailed bass, all the usual stuff like that. It was subtle the first few times I compared a track and I wasn't sure of the results. But the more I listened to more and more tracks comparatively, it was absolutely different. How can that be?

I reasoned that it must have something to do with how Foobar is decoding the flacs with its internal plugin. All files were on the same HD and were well buffered by now, so it wasn't disk access. Then I tried Jriver player (which I've never used before). The difference between its sound of flac and wav was less obvious but still there.

Then I installed an M2Tech HiFace USB to S/PDIF converter and ran its output to another input on my monitor controller, and set both Foobar2000 and JRiver to use it instead. The HiFace was another (subtle, but very real) improvement over either clock mode of the Profire. With it, the differences between flac and wavs on both players were the same sonically, but more obvious overall.

Then I tried Foobar with both WASAPI and Kernel Streaming drivers. KS seemed to improve it but WASAPI didn't seem to make much difference. There don't seem to be WASAPI or KS drivers for the JRIVER, at least not that I have found.

So now, in order of clarity and detail:

Foobar2000 WAV HiFace KS
JRiver WAV HiFace Direct Sound
JRiver FLAC HiFace Direct Sound
Foobar2000 FLAC HiFace KS

I still have other players to try (Real, VLC and Audiogate) but now I'm not thinking it's directly player related. The computer is my DAW which is running W7-64 bit, 3.2Ghz Core i7 processor with 12GB of Triple channel 1600 RAM. It's loafing along at a couple of percent utilization during all this (but higher utilization doesn't seem to change anything).

My Avocet controller drives a Pass Labs X250.5 amp which drives B&W Nautilus N801 studio monitors. Speaker cable and IC between Avocet and Pass are Litz construction from BPT, very high bandwidth. Speaker cable is equivalent 7.5 AWG, IC is balanced. Everything in the studio is fed with balanced power from Equi=Tech Son of Q conditioners. I think I've posted this info somewhere else in the Forums but don't recall where.

I'm still waiting on a Halide Bridge for additional testing.

I'm a non-believer that this can be happening -- really can't believe that it is, and haven't yet rationalized what this difference might be caused by. Anyone with other suggestions to try while I'm in a testing mood?

--Bill

 

[Vincent Kars]

There don't seem to be WASAPI or KS drivers for the JRIVER, at least not that I have found.

WASAPI is intergral part of Win7
It is supported by JRiver from V16 on.
http://thewelltemperedcomputer.com/SW/Players/MC14/MC14_audio.htm

BTW: interesting experiments including the different clocking schemes

 

[Phelonious Ponk]

I'm a non-believer that this can be happening -- really can't believe that it is, and haven't yet rationalized what this difference might be caused by. Anyone with other suggestions to try while I'm in a testing mood?

--Bill

Sure it can be happening if, somehow, in those changes, you're changing audible levels of jitter and/or noise. Or not. Really want to test? Set it up so you can switch between variations rapidly and compare them blind, starting with what you now think is the biggest gap. Easier said than done, I know, but you really don't know much until you can do that. Audio memory is far too short. Then again, so is life. You might just want to hear what you hear and listen to some music instead.

Tim

 

[mojave]

There don't seem to be WASAPI or KS drivers for the JRIVER, at least not that I have found.

Go to Tools > Options > Audio and you can pick from ASIO, Direct Sound, WASAPI, WASAPI-Event Style, Kernel Streaming, or Wave Out.

Here are the differences between WASAPI and WASAPI-Event Style from JRiver's Wiki:

WASAPI

This output mode pushes data from Media Center to the sound device. It works with nearly all hardware.

WASAPI - Event Style

The output mode lets a sound device pull data from Media Center. This method is not supported by all hardware, but is recommended when supported.

WASAPI - Event Style has several advantages:

It lets the audio subsystem pull data (when events are set) instead of pushing data to the system. This allows lower latency buffer sizes, and removes an unreliable Microsoft layer.
It creates, uses, and destroys all WASAPI interfaces from a single thread.
The hardware (or WASAPI interface) never sees any pause or flush calls. Instead, on pause or flush, silence is delivered in the pull loop. This removes the need for hacks for cards that circle their buffers on pause, flush, etc. (ATI HDMI, etc.).
It allows for a more direct data path to the driver / hardware.
The main 'pull loop' uses a lock-free circle buffer (a system that J. River built for ASIO), so that fullfilling a pull request is as fast as possible.

 

[bblue]

WASAPI is intergral part of Win7
It is supported by JRiver from V16 on.
http://thewelltemperedcomputer.com/SW/Players/MC14/MC14_audio.htm

BTW: interesting experiments including the different clocking schemes

Thanks. The JRiver comment was bad wording on my part. I was using the JR Jukebox, not the Media Center and there was no support for KS or WASAPI in it. Just finished installing MC 17 and everything is there. Still need to re-do the comparisons with this but the ears aren't tuned into it tonight...

--Bill

 

[microstrip]

You might just want to hear what you hear and listen to some music instead.
Tim

Brilliant! But why not:

You might just want not to hear what you hear and listen to some music instead.

 

[bblue]

Sure it can be happening if, somehow, in those changes, you're changing audible levels of jitter and/or noise. Or not. Really want to test? Set it up so you can switch between variations rapidly and compare them blind, starting with what you now think is the biggest gap. Easier said than done, I know, but you really don't know much until you can do that. Audio memory is far too short. Then again, so is life. You might just want to hear what you hear and listen to some music instead.

You'd think, right?

The problem is that once I've heard a better rendition of a particular set of music, re-listening to it without the improvement is no longer enjoyable. Yeah, I've always been like that, and I can recall the 'visual' representation of my listening experience over several days. So if after a couple of listens to a new configuration I don't get the same impression as originally, either my ears are 'off' for the night, or something else isn't right. The extremely subtle differences can be difficult to deal with, though, that's for sure.

In doing these comparisons I am not able to do an immediate A/B switch, but about 1-2 seconds apart, which is more than enough once you have zeroed in on a particular sound or characteristic that sets the sources being compared, apart.

The other thing to be careful with is the interpretation of which is actually the most accurate. Yes, there may be differences, but which is right (as opposed to preferred)? Usually it's the one with the deepest soundstage and clarity between instruments, but it's possible to be fooled on that if you don't repeat the testing over and over with different sources. Each source has to provide the same characteristic difference, or you are somehow misinterpreting the results.

--Bill

 

[bblue]

WASAPI is intergral part of Win7
It is supported by JRiver from V16 on.
http://thewelltemperedcomputer.com/SW/Players/MC14/MC14_audio.htm

Ok, I got rid of JR Jukebox and got Media Center 17 installed now using Kernel Streaming and the USB M2Tech HiFace device for output, same as with Foobar.

The difference between FLAC and WAV files is still quite noticeable on both programs when playing from local disk, even with JRMC set to play from buffer not from disk. The differences in either format between JRMC and Foobar is very close. Sometimes I think there's a difference and other times I'm not sure. If there was a winner I think it would be JRMC in my present configuration. Foobar has been my reference standard for *years*.

JRMC is pretty nice software, but takes up a whole lot more memory than Foobar does because of maintaining its online database. It's like over 300M (in my case) versus Foobar at around 35M. Plus there are additional processes and services at play with JRMC, but in a system with sufficient RAM, it shouldn't be a big deal.

Back to WAV and FLAC differences, the soundstage on WAV files from source to source is consistently a little deeper, a little wider and much clearer (space between and detail of instruments) compared to the FLAC file version, which sound slightly closed in and restricted at both ends of the spectrum in comparison.

I still need to try the same comparisons using a remote fileserver (NAS) for both filetypes. Mine's really busy right now doing a synchronization so it's not a good time. Maybe tomorrow.

--Bill

 

[Vincent Kars]

Probably the Media Network (DLNA, server)
If you don't use it you can disable it

 

[bblue]

Probably the Media Network (DLNA, server)
If you don't use it you can disable it

Do you mean for the high memory usage? Their DLNA server takes only about 5MB of memory. It's the media player itself that really consumes when it is doing tag, picture or library processing in the background. I've seen it as high as 350MB and as low as 8MB. Doesn't seem to bother anything, I was just pointing this out.

I've finished another round of WAV vs FLAC comparisons and have these comments:

1. The sound of WAV files is always superior on Foobar and Media Center (and other players).
2. Load on the (my) computer does not affect it
3. It doesn't matter what drive it's on, or whether it's buffered in memory
4. It's the same if all source files are on a remote server connected via ethernet. (i.e. there is no difference between a wav played from a NAS to one played on the local system.)

5. On the Korg Audiogate player the differences between the two formats is considerably less. However the overall sound of that player is slightly different from the others. Sometimes it sounds cleaner, other times it sounds like either the HF is rolled off a little, or has some sort of damping applied. It's a very nice listen, but there can be subtleties or details in the source (FLAC or WAV) that are slightly obscured compared to Media Center playing WAV's.

Unfortunately, the Korg Audiogate requires you to manually set the playback bit rate to that of the source. If you don't it will just quietly re-sample it to whatever is currently set. It's pretty good re-sampling, though, and is fairly easy to miss. Also, using ASIO or default device is the only way it will allow you to use all the sample rates of the sound card. If you choose WASAPI and (e.g.) the M2Tech HiFace USB adaptor it will only allow 44.1k even though the HiFace is fully capable of up to 192k.

A THEORY.

I'm pretty sure that the FLAC problem on most players is that they are using the same Real-Time decoder engine, and there are flaws in that engine. I've seen folks comparing bit by bit output of flac.exe to those of WAV files and finding no difference. Of course not. That's not the real-time playback engine.

It seems as if the processing time for real-time decoding of FLAC inside a player changes from bit to bit and is somehow affecting the resulting timing (though I don't know how) or dropping bits if they can't be delivered in time for the next cycle at the current clock bit rate. If this is the case, the errors and resulting degradation should be much greater at higher bit rates than 44k, but FLACs seem to be quite audible even at 44k. Whatever it is, it's got to be related to the real time decoder plug-in, since we know that the significant bits of a WAV file are completely reconstituted from a FLAC file by flac.exe (which is not a real-time engine).

Any other thoughts?

--Bill

 

[Vincent Kars]

To test this theory, record the SPDIF out of both. Load them in a editor, time align and do a null test.

http://www.audioasylum.com/cgi/vt.mpl?f=pcaudio&m=67015

 

[bblue]

To test this theory, record the SPDIF out of both. Load them in a editor, time align and do a null test.

http://www.audioasylum.com/cgi/vt.mpl?f=pcaudio&m=67015

If the problem is caused by timing as I first considered, I don't believe this test would show anything, because the problem wouldn't resolve until it hits the DAC. But it sure would show bit drop if any. I'll give that a try.

Thanks.

--Bill

 

[sasully]

You'd think, right?

The problem is that once I've heard a better rendition of a particular set of music, re-listening to it without the improvement is no longer enjoyable. Yeah, I've always been like that, and I can recall the 'visual' representation of my listening experience over several days. So if after a couple of listens to a new configuration I don't get the same impression as originally, either my ears are 'off' for the night, or something else isn't right. The extremely subtle differences can be difficult to deal with, though, that's for sure.

In doing these comparisons I am not able to do an immediate A/B switch, but about 1-2 seconds apart, which is more than enough once you have zeroed in on a particular sound or characteristic that sets the sources being compared, apart.

The other thing to be careful with is the interpretation of which is actually the most accurate. Yes, there may be differences, but which is right (as opposed to preferred)? Usually it's the one with the deepest soundstage and clarity between instruments, but it's possible to be fooled on that if you don't repeat the testing over and over with different sources. Each source has to provide the same characteristic difference, or you are somehow misinterpreting the results.

--Bill

None of this makes it a well-controlled comparison, which is essential to rule out all sorts of psychological confounders in perception of difference and preference. What you've written up there boils down to 'I trust my ears'.

 

[fas42]

Bill, I've never gone into the playback from computer resources, it's always been straightforward CD usage, but I'm now doing some low level experiments and getting interesting results ... see my posts on the Redbook thread.

My shtick in the audio game for a long time now has been that it is all the subtle, less obvious, things that do the real damage to subjective sound quality. And one of these elements is cross interference: one piece of the electronic circuitry that's in the arrangement that produces the end sound negatively affects another. Everyone pays lip service to this, but it's hardly ever treated or dealt with seriously by the manufacturers of the gear.

So from my point of view FLAC will always suffer, or at best be equivalent, compared to WAV. Simply because more work is being done in electronic circuitry, somewhere, to get the sound from the source to the speakers. If a cruddy MP3 can be dramatically improved in quality simply by resampling to hi-res then that gives the game away in that regard ...

Frank