and I just ripped about 400 CDs in FLAC at the "5" setting.....onto the SSD of my Eversolo A8
So now I want to A/B the rip with the original CD, because the rips sound fine to me but haven't compared them back to back.
Why do Wav and Flac Files Sound Different?
This article is from 2016, so outfits like JRiver may have developed workarounds for the metadata/sound quality issues sussed out below. Inquiring minds want to know.
Why Do WAV And FLAC Files Sound Different?
"Based on these results, we attempted to pinpoint which section of the metadata might be responsible. Since the cover art file associated with the metadata is the largest contributor to the metadata header size, we began by examining the effect of deleting cover art prior to the WAV-to-FLAC-to-WAV conversion protocol. This proved fortuitous, as our first suspicion proved correct."
53 responses Add your response
The more interesting point of this article seems to be that both Flac and Wav are sonically adulterated by metadata during playback to the extent that fidelity is somewhat harmed. A significant portion of my digital library contains flac files downloaded from Allflac.com and converted in JRiver to Wav. Now that Allflac accepts only bitcoin from overseas customers I just use Quobuz - downloading their Wav files into JRiver and then burning to CD's for my transport. A useful conclusion for my situation would seem be that I not include files such as "cover art" in the download since it appears in the article's graphs to pollute the sonic data. |
I think that in order to get a productive discussion, could be good to have some mark of reference. The article goes back to 2016 and many things have changed since then. Can you tell us about your experiencia? Do you feel some difference when playing different formats: WAV, FLAC, ALAC, indeed mp3? Have you tried different "apps" to play your files: JRiver, Audirvana, Roon, ....? Do you feel any difference playing the same file with metadata and totally stripped from metadata?
|
Wav take up much more space, the key more so is the quality of the recording it is a fine line ,I just use Flac files ,but do have a few files with Flac and wav files ,the differences is so small it is not worth taking up all that extra space IMO. Qobuz uses Flac files ,and hirez masters in Flac ,and saves tons of space and sounds fine. |
From my experience with Tidal, FLAC files overall sound much better than WAV and comparable to MQA (although a much larger selection of FLAC since MQA clearly very selectively started with top tier original recording). MP3s are very subpar/compressed as always and not worthy of a place in this conversation other than a footnote. |
Another nope.....they sound the same. Don't believe the hype. |
When I ripped my CD collection (maybe 400 +-) to a hard drive I ripped in both FLAC and WAV. To my complete surprise I could hear a small difference in sound on playback. The FLAC files sounded a little more laid back. Could be how the streamer, a Node 2I, or the DAC, a Bel Canto, process the data, I don't know. I preferred the WAV files. Memory is so cheap these days, 4TB for around $100, file size is not an issue for me. Jim S. |
And of course the listeners were unaware of file type and metadata status for each trial, correct? And the tests were randomized to prevent observer bias? And the results were corroborated across multiple listeners, using multiple trials? And all listeners were unaware of the nature of the changes? And dummy A/B trials with identical files were conducted randomly during resting to create a baseline? No, not all that was done? Not any of that was done? Because absent that minimal level of basic care in the testing protocols, this 'study' has zero credibility. It's just poor experimental design. Worse, the authors propose no causality attributable to metadata, no 'how' to account for any statistically meaningful differences. |
Here is a somewhat more nuanced take on the issue. The author is responding to the article that started this thread. I should stress here that I am not playing my files off of my computer - merely using the computer to store these files and then burn CD's to them which are then played on a Jay's Audio CDT3 Mk3. I am using a 2 year old Dell. What more interests me is the interaction of the computer buffer and perhaps more specifically the buffer involved in CD burning with the quality of the resulting burned CD. By the way, whether an improvement or not, the "cover art" file is now deleted by me before copying the downloaded Quobuz .wav files to JRiver. I miss the visual tag of cover art in the JRiver front page, but the text caption is sufficient for locating files.
|
WAV files are so old that there was never a schema developed for them to hold metadata internally. FLAC is merely a "container" for the WAV files having more robust places in the design (schema) to hold metadata, things like track name, number, and yes, cover art. Even lowly mp3 files have by design, places to put that data. If your streamer is working right, it will use the metadata but there should not be any sound artifacts from that metadata being present because they are not INSIDE the WAV file that is decoded. Free LOSSLESS Audio Codec is the nature of the beast. The same can be said for Apple's ALAC. I've encoded wav files as level 8 FLAC (the best compression that can be done and still be lossless when decoded) and level 0 (no mathematical compression at all) and can hear no difference in my headphones. But then I'm 65. Still, by design, these things should have no sonic difference at all. Even whatever chips were in a lowly 1998 PC could handle the decoding without issue, let alone modern, much faster chips in phones, and well-designed streamers. |
A little while ago I downloaded a CD quality CD's worth of music (Sarah Janosz) from Quobuz. Once on my hard drive I deleted the "folder" (image of Janosz) from the list of tracks, uploaded the files to JRiver and was interested to see upon opening the file in JRiver that the "folder" image was still active and came up as the usual jpeg in JRiver where it acted as the visual header for the group of files. It would appear to a layman such as myself that this file is permanently embedded in the download even if "deleted." Any ideas on how to really and truly remove it before burning a CD of the tracks? |
@bolong You can have the image "embedded" in each track as FLAC allows it, in fact you can put a different image in each track of a "disc". If you try any tagger (I use Tag Editor) you will see that and you can also delete the image for each track if you want. When you use a program like JRiver, Audirvana, Roon, ..... to play the files they can work in two ways: choosing the "folder / disc" image over the track image or the contrary. If one the images is missing, they will choose the one that exists. And if there is no one, no image will appear. Regarding your question: if you burn your CD (physical disc) as a real CD the tracks are transcoded to WAV and the images are not included in the process. |
@aldnorab You are opening another big and old discussion. On one side, when hard disk space was not such a a commodity as it is today, people tried to reduce the size of the music files as much as possible. That implies the use of algorithms to "fold" them. But also implies the use of algorithms to "unfold" them just in the fly while or just before being played. This implies the use of processing capacity of the computer ant that means noise. Again, in the "old" times, this operation could demand a slightly significative "effort" for the computer. Today you can say that this "noise" is absolutely negligible. It will depend on your computer (streamer, server or whatever) If you are curious, play the same file / track as FLAC level 0 and level 8 and trust your ears: I think you can use any transcoder to do ver both FLAC files. And my advice: once you have tested this in YOUR equipment and with YOUR ears, forget it: there will always be people saying that they hear the difference in their equipment and there will always be people saying that with the computer of such equipment is "scientifically" imposible to feel any difference. |
@aldnorab I do not know and it is not important regarding quality of the audio file. A "song" can be transcoded as many time as you want and no "musical" information is gonna be missed (remember we are in digital world and information means 1s and 0s) This is assuming that you transcode among lossless formats: ALAC, FLAC, AIFF, WAV and any other: so streaming company can receive a WAV file from the majors and transcode it to FLAC as it is the most "commercial" one The only thing that could be lost is metadata (like image, composer, group, orchestra, year, ....) Regarding your question, I guess that they supply FLAC because is universal: can be played in almost any system and has a very good capacity for metadata. Anyway the streaming company can receive a WAV file from the majors and transcode it to FLAC as it is the most "commercial" one. But honestly, I do not know. |
@bolong I do not know for every case. What I have seen when I have bought music from Qobuz is that they provide an image that you can see in the FOLDER where the music is and when I have retagged every SONG FILE in order to add information important for me, sometimes there was the same image in each song file. The same applies in the case of music that I have bought from other online shops (Presto, Highresaudio, etc.) For each song the cover art and all metadata is "separated" from "music" in each MUSIC FILE: for understanding this, it helped me very much to realize that all the information (music, song name, duration, composer, singer, orchestra, group, ...) is a very long row of 1s and 0s but similar to a very long train, the 1s and 0s for music are in one big wagon, and metadata goes in one or two or three small wagons. Following this idea, WAV only admits one metadata wagon whilst FLAC, ALAC, AIFF admit all the wagons. Metadata wagons can be full, empty or something in between Please, this is not "scientific" at all, but this idea helped me very much. I hope it can be useful for you and not confusing. |
The only way you’ll ever convince me there’s a lick of difference between WAV and FLAC is with the jitter measurements. It is _possible_ that the decompression has so much of an effect it overworks limited or pooor buffers and jitter control mechanisms, but IMHO, it’s the 21st century and any streamer at all in this day and age should perform equally well with either. I should point out that any of this should be VERY streamer dependent. Unique CPU's, algorithms and real-time software is probably running on each of them, so the idea that this is something shared among all of them is not believable. Also, take Roon, which essentially reads the files in the server and then seends them to each endpoint separately. The data stream sent by Roon has been stripped of it's original format. What happens then? |
The gist of this thread had to to with the possibility that any metadata was a potential pollutant of the musical information. When I used to download albums from allflac.com removing the "cover" file from the track list before loading it into JRiver would nix the cover art showing up in the thumbnail header for the albums which seems to indicate that this data was no longer involved in the track files. Not the case with Quobuz. None of this may matter sonically. Or it might in some small. way. |
OP: The metadata and audio are separate. They share a file, but it’s not as if the album cover art is being sent per musical frame. It is in a different location, along with all the credits, so once the music starts to play, it is completely out of the picture. At best, these should could cause picoseconds of delay in starting to play the music, but zero when playing has begun. HDCD and MQA however are interwoven into the music data itself. It does take a little more CPU power to decompress FLAC or ALAC vs. WAV, which is going to take more network time, but if your player sounds different, it's messed up. |
Maybe I’m reading something into your last message, but can you explain what HDCP has to do with FLAC v. WAV? I’m not trying to hassle you, this is an honest question posed to an obviously knowledgable colleague. FWIW, this is a topic long of interest to me. I wrote some lengthy articles about digital copy-protection the topic for mainstream tech publications back when the first round of DVD CP standards emerged, and at the time, HDCP’s only application for audio had to do with DVD-Audio media (and then hirez SACD DSD-encoded stereo). I understand that today HDCP is incorporated into HDMI/DVI/etc.-transported signals, but that would not distinguished between FLAC & uncompressed PCM content. Btw, +1 re: your comments about metadata being processed independently of digital-audio content. Jeez. Yes, nothing’s impossible, but Sagan’s Law ("extraordinary claims require extraordinary proof") applies here. D |
@stilljd - The FLAC compression levels are aspirational. Think of them more as setting how much time to spend attempting to compress a song, more than how much they will compress a song. That is, as you increase FLAC compression it spends more and more time to compress the data, but may not actually be able to do so. @cundare2 No offense is taken. HDCD was a way of encoding a variety of features into a 44.1 kHz/16 bit signal, most famously it was a way of encoding 24 bit data into 16 bits. This compression did in fact affect the original music, in the sense that the original 44.1 kHz/16 signal was no longer bit identical as it now had information encoded about dynamic range. HDCD was complex and encoded more features than just this. MQA is also not a bit-perfect conversion. It attempts to encode in a low resolution signal (44.1/48.1, etc) high resolution content (96 kHz/24, or higher). In the case of HDCD or MQA you are left with 44.1 kHz/16 bit data which is no longer bit-perfect of the original. On the other hand, ALAC, FLAC and WAV formats though result in exactly the same bit-perfect signal resulting from decoding their files. Conversion between them to each other should result in exactly the same set of 44.1 kHz/16 bit data streams (if that’s what you started with). How they handled metadata might be different but in all cases the metadata is NOT encoded into the music stream. |
this is ludacris wav and flac sound identical. flac is completely lossless it uses an on the fly decompression wav does not contain metadata while flac does there are absolutely 0 advantages with wav. we have been dealing with computer audio servers and streamers fo over 10 years
Dave and Troy audio Intellect NJ Streaming specialists
|
Um, beg to differ. You are correct in that the existence of album art or metadata should make no difference to a properly behaving streamer. Perhaps @audiotroy you are conflating album art with metadata? I just converted a file from FLAC to WAV and it had artist, album, year of release, etc. but the album art did not make it over. I found this interesting article on album art however, and it seems this is a rather new capability for WAV files. Important to note that some metadata is required for WAV and FLAC as it describes the data in the audio data container that explains how the stream of bytes should be interpreted. Floating point vs. integer, number of channels, sample rate, etc. is all necessary for WAV files and part of the file format definition you may read more about here. There is a second, optional part of metadata associated with human legible, descriptive tags. Not actually needed to play a file, but certainly needed to organize a music library. |
It may help to think of an audio file as having 3 parts:
Each of these chunks of information are entirely separate in the file, so as @audiotroy points out, the existence or absence of album art should make zero difference to the DAC. The byte stream presented to the DAC should be identical regardless of optional metadata or indeed file format (FLAC, WAV, ALAC, etc.). The one possible/plausible area I can see FLAC vs. WAV causing a difference is with poor streamers that don’t implement decompression (FLAC or ALAC) with adequate buffering and that somehow the time taken to decompress each chunk is affecting the rate at which samples are presented to the DAC. IMHO, this is possible, but certainly signs of poor design, not a feature of the file format. If there IS a difference in a file because it has or does not have album art that would be a significant functional error/bug. To end with a metaphor, imagine a vinyl record. The metadata is the text on the label. The cover art is on the cardboard sleeve, and the byte stream is the music in the groves. If you can hear a difference in an album because of the sleeve art you really have surpassed the boundaries of human capabilities. |
@m2team00 It seems to be a relatively new feature to have album art in WAV files and not sure all converters support it. See more here. |
I find it an interesting question. Lossless should be lossless, i.e., bit perfect, whether a WAV file or FLAC. This brings me to another age-old debate on here. If FLAC is bit-perfect, and it supposedly is, if one rips a vinyl album to FLAC using a good quality analog rig to FLAC, shouldn't the FLAC file have the same "analog" sound as if one were playing the album on a turntable? I have done that experiment and on my system (all digital), the vinyl inevitably has some "crackles and pops" from the album that is absent from the same music ripped from CD or streamed. That plus the convenience of digital makes me not miss my turntable! |
@moto_man We should separate two issues. Lossless, and recording quality. WAV, FLAC, and ALAC should all be lossless formats, but I don’t do quality assurance on every algorithm out there which is converting from one to another, or used in CD ripping. It is technically possible that someone has a FLAC compression or CD ripping code that is not lossless, but if so it’s a bug. The other part of this is the recording and playback quality which is affected by things such as the ADC, recording settings, digital filters on your DAC, etc. In an ideal world, of course, the playback is identical to the recording. If you are missing events there may be something going on. There are a number of music analyzers out there that might be worth experimenting with to give you an answer.
|
Anyone who agrees with me that since WAV files are what nothing else is allowed to differ from, and are automatically reference listening compared to anything else, is already my idol, even before they agree that WAV sounds like it can’t be wrong another noisy way like FLAC does. I have to listen to heavy metal WAV’s only, because of what a slacker the correct untampered with copies make me. Streaming FLAC Beethoven is still why you’re dead, though. Oh, and from an SSD clearly beats reads with ecc from an optical disk. I’m stuck with USB for my I2S signal for the time being only, but the read sounds more confidently clear than my 80’s cd player ever gave me. After that, I can’t believe my DAC designer smoked my first Yamaha cd player with each separate balanced side for only that price... |
@erik_squires, I think my point is slightly different. When one is ripping from a vinyl album, I would think that the analog signal from the cartridge would be captured and converted "lossless" to FLAC. That should mean that the FLAC so captured, when played should have the exact sound as if the signal was played from the turntable. Therefore, if you are playing vinyl because of its "analog" sound, shouldn't that analog sound be captured bit for bit on a FLAC and then sound "analog?" I guess that the mere act of converting an analog signal to digital and then back to analog through a DAC might possibly change the sound so that the FLAC is not "identical" to what is coming out of the cartridge or phono preamp, but . . . |
@moto_man, interesting question. I did some analog recording from vinyl to a CD using a Pioneer PDR 04. The original vinyl recording does sound different compared to the CD recording ('better', to my ears). I could imagine that some 'parts' of the overall sound played by a TT including all the peripheral equipment used (including the pre amp and amp) is not transferred to the CD. This experience is supporting your guessing, it seems. |
“I know that this revelation of Mobile Fidelity cutting from transfers made to digital is sort-of anathema to some of our purest fans, and I understand that strictly from what you might read on a piece of paper. But the truth is that us cutting from digital transfers that we ourselves made is not an example of us losing our way, it’s actually an example of us adhering to it adhering to the purest aesthetic and perfecting things beyond anything in the past and the past goes back now 45 years…. Preserving the tape.”
— Rob LoVerde, Mastering Engineer, Mobile Fidelity Sound Labs |
WAV has always had the ability to have metadata it just wasn’t well supported as an official standard. This changed over time. Qobuz downloads now embed the cover art and metadata into each track for WAV or FLAC. You can strip all metadata from an entire folder of music files using dbpoweramp tag editor. Just highlight all files and open (right click …) in tag editor, then delete all rows of metadata and cover art. You don’t have to do it for each track. It would be interesting to see if the resulting cd burnt with ‘clean’ WAV files sound any different from one using the tagged WAV files. |
Thanks for the suggestions. I am assuming that stripping the folder of all metadata means the CD cannot be used in a manner such that individual songs can be called up by my transport, i.e, the whole file becomes a monolith that can only be played from start to finish of the album without the ability to call up individual tracks. |
Thanks for the suggestions. I am assuming that stripping the folder of all metadata means the CD cannot be used in a manner such that individual songs can be called up by my transport, i.e, the whole file becomes a monolith that can only be played from start to finish of the album without the ability to call up individual tracks. |
Each track should still be an individual file, even without the metadata. |
