“Double blind” make it sound hard to set up. Foobar2000 (on WIndows) or ABXTester.app (on Macos) make this trivial to set up: load a pair of files (one ALAC and one WAV, say) and the App will administer the test. No need to invoke the help of a friend, set up a mess of cables, …
I will await a printout of the Foobar2000 test report from anyone who actually thinks they can distinguish different lossless file formats.
Double blind tests are flawed: https://www.audioholics.com/how-to-shop/revealing-flaws-in-the-loudspeaker-demo-double-blind-test
Sighted tests are more flawed.
Exactly! The only way to go is measurement and analysis!
ABXTester doesn’t open AIFF, ALAC or FLAC files, useless. If there is a similar Mac app that does, let me know, I’m happy to do the tests.
But we can perceive even beyond what is measurable. From Rob Watts, designer of Chord DACs:
the ear/brain is very sensitive to this issue - noise floor modulation makes things sound bright, hard and aggressive. But actually we can hear levels well below the measuring limits of the best test gear; I have had digital FFT’s with noise floor of -200 dB modulating to -190dB - something you would never be able to measure as it would be swamped by analogue noise - but removing the noise floor modulation results in the SQ becoming warmer and smoother.
with some measurements as well there too.
It doesn’t (natively) support FLAC. But it certainly does support AIFF, WAV and ALAC:
N.B.: Apple uses the same (.m4a) file extension for both AAC and ALAC files. You’ll just have to believe me that the file in the above screenshot is an ALAC file.
Apple uses the same (
.m4a) file extension for both AAC and ALAC files. You’ll just have to believe me that the file in the above screenshot is an ALAC file.
I was not able to open either an aiff or m4a in the app, wouldn’t recognize them and they were greyed out. I will keep trying.
Hah! That’s amusing.
From Wikipedia:
There is also a compressed variant of AIFF known as AIFF-C or AIFC , with various defined compression codecs.
…
The file extension for the standard AIFF format is .aiff or .aif . For the compressed variants it is supposed to be .aifc , but .aiff or .aif are accepted as well by audio applications supporting the format.
The “uncompressed” audio file you are trying to test is actually a compressed audio file (perhaps one even using the FLAC compression algorithm).
Try loading an actual uncompressed AIFF file, and you should be fine.
What? DACs are not “optimized” for a particular file format/container. They either are capable of receiving PCM or DSD or both. Software decodes PCM from ALAC, FLAC, etc. and sends it to the DAC. They don’t send WAV format…
It doesn’t even open an ALAC .m4a file either though. And, the AIFF file is not a compressed one, that app is just saying it is.
Does the App open anything at all? Perhaps it is pooched.
I have no idea what’s going on with your system. The App works fine for me on WAV, AIFF, ALAC (and MP3 and AAC as well).
FWIW, I’m running Macos 10.14.6. ABXTester is version 0.9, downloaded from the Mac App Store and (not that it should be relevant) iTunes is 12.9.5.5.
Edit: Perhaps the issue is that the App does not have the permissions needed to open files in your Music Folder. Open the Privacy Tab in the “Security & Privacy” in “System Preferences”.
I didn’t find that I needed to, but perhaps you need to grant ABXTester.app Full Disk Access (or, at least access to your Music folder).
Could be the app is not compatible with Catalina. I also just updated to the 10.15.5 update yesterday.
WAV = FLAC = ALAC = AIFF in terms of resulting PCM…they are all bit-identical. No side is (or should be debating that). This is a non-issue for Roon endpoints which sees PCM. Use whatever format (FLAC preferable for metadata) on Roon Core.
With devices, where the audio device is doing the decoding itself, you can always compare for yourself. Whether you hear any difference or not, ALWAYS ALWAYS IMO rip or download your master library in FLAC. For metadata, FLAC, ALAC, and AIFF are better than WAV. Between ALAC, FLAC, and AIFF in terms of metadata, I’ve come across on some devices where certain fields such as “album name” or “album artist” were incorrectly rendered or missing with AIFF and ALAC. I’ve found metadata universally to be better compatible and less troublesome with FLAC. Opt for the highest compression to save space for master library or with Roon Core and Roon endpoint schemes.
If you think WAV sounds better, simply derive a secondary library by batch transcoding the master FLAC library. Always have a master FLAC library for Roon and/or backups. In those systems, where the decoding is happening in the endpoint, use the WAV library.
As far as evidence, I’ve read comments by Linn engineers and in Naim’s own manual that they have measured differences in noise floor between FLAC and WAV decoding. Each has a unique pattern of noise floor. With FLAC, it’s more broadband and random, whereas with WAV it’s periodic and with well defined peaks (what’s more noticeable or harmful?)…Linn recognizes this and makes adjustments to render this inaudible by utilizing extensive filtering. Whereas, Naim simply recommends the use of WAV for best SQ.
Nonetheless, mechanisms and evidence do exist. There are different patterns and levels in noise caused by the decoding of these two formats. Linn even had measurement graphs of this. The more interesting thing is at what levels do these differences become inaudible. I’ve read Rob Watts speculate that it may be much lower than common accepted values such as near -180 db based on his internal testing.
When I use the BDP-1 (which is over 10 years old and limited in CPU power…so who knows how recent CPUs in recent streamers compare?) as a Roon endpoint, you can use FLAC or WAV on the Core and the sound is the same. However, when I use the BDP-1’s internal MPD decoding, then the differences show up.
Not only are there differences between FLAC and WAV, but there are also differences whether they are sourced from a NAS (ethernet) or an attached USB drive.
There are many users that do not hear any differences between formats and/or USB/NAS with their BDPs and various DACs in their systems. I totally understand, believe, and respect that.
Between the remaining users that DO hear some differences, what I find very interesting is that we usually can hear similar differences and can describe the same thing that we are both hearing in our systems involving BDPs, HOWEVER, we end up with different interpretations of whether which dichotomy sounds better and/or is a more accurate rendition.
Similar discussions involve the Jitterbug. Lots of agreement between what happens to the sound with or without Jitterbug in objective attributes, but difference in interpretation of whether it sounds more real life, or more accurate, or more enjoyable.
If someone wants to suggest that that it’s all in our heads, you absolutely have 100% right to do so…and who knows, it really may even be the truth 
. However, if you for a second stick with the gut feeling of what we are hearing, I’m more interested in understanding how so many people can roughly hear the same sonic changes, yet come up with different interpretations. I wonder how much age (FR audiograms), internal ideologies of what better sound is and other biases?, reference to previous and current systems and thus experience factor in towards whether you like A or B.
FWIW, between my BDP-1 and various gear, I tend to group FLAC, USB, and no Jitterbug as together and WAV, NAS, and Jitterbug on the other side. Obviously, you can mix and match these for number of combinations, but the dichotomy stands.
The FLAC, USB, and no Jitterbug sound signature is one that is more energetic, more upfront, narrower and together. In quick blind testing, one can easily prefer this sound as it’s more noticeable and appears to be more detailed. However, in both my rigs, I tend to get more fatigued of this as the listening session goes on.
The WAV, NAS, and Jitterbug group gives a sonic signature that initially appears more diffuse and doesn’t jump out at you. In fact it can sound less detailed and flat as first. However, as you listen more, rather than the sound jumping out at you, you can notice more deeper into the sound. It sounds more relaxed. In long term listening, it doesn’t bother me as much and I stop noticing the sonics and just kind of zone out. This is especially important when watching movies.
At this point, I’d love to know and see these measurements to explain these difference in SQ and which group yields the more accurate sound to see whether I prefer the more accurate or less accurate sound. I’m always up for some ear training.
(I hear similar trends in balanced interconnects as well: Grimm TPR, Mogami 2549, Mogami 3173, Ghost cable. I prefer the more boring cables that give the depth and don’t appear overly detailed at first. Well shielded and twisted with good symmetry and geometry. Keep the noise floor as low as possible.)
FLAC requires decompression, which results in (fairly slight) CPU activity. If your computer is directly connected to your DAC then varying levels of CPU activity can be audible. The Roon devs tell a story of how this was demonstrated to them at Meridian, where they heard a difference in audio playback when a mouse pointer was moved onscreen.
This is why Roon recommends that Core and Output be separate devices. If you use a network connection to a Roon Ready or Roon Bridge device then any processing noise is avoided.
So the first question I have in mind when reading accounts of folks who hear differences between lossless formats is system topography. Some people prefer a direct connection to a computer and can spend thousands of dollars optimising it. I prefer to use a separate Output and avoid the Augean stables of computer audio optimisation.
The CPU on my server runs full speed all the time upsampling to DSD 256 with EC modulators in HQ Player and convolving a room treatment WAV. But this has zero influence on the music, which is sent over the network to a microRendu.
The above 2 posts are exactly right and say things better than I could. One simply cannot and should not make a blanket statement that there are no differences, there are many variables involved.
Much as I admire (and spend on) Chord gear, I wish Rob Watts would be careful with what he says.
200 dB — no, that’s far beyond human hearing, the audible range is about 120dB. At 0 dB SPL, the ear drum moves less than a molecule, so to sense 80 dB below that level, it would need to move one 1/10,000 of a molecule.
I have even heard Rob speak of sensitivity to 300 dB changes. Let me tell you a story from my college days. A party, a bunch of engineering students sitting around drinking beer, shooting the ■■■■■ the conversation is flagging, kinda gloomy — and somebody says “300 dB” and everybody bursts out laughing, we all have another beer and the party continues. What’s so funny about 300 dB? There aren’t 300 dB. If you can measure earth to the precision of 1 kg, that’s only 240 dB. If you can measure a typical electrical current to the precision of individual electrons, that’s about 320 dB but Heisenberg’s uncertainty principle forbids that.
There is one result in science where measurement agrees with theory to ten significant digits, a single scientific measurement and it is considered a marvel, Nobel price stuff, and ten significant digits is 200 dB. (This is the electron’s anomalous magnetic dipole moment, “which is just like the regular magnetic dipole moment only more anomalous.”)
All other measurements of the fundamental constants of physics are known to 8 significant digits at best, which is 160 dB. So please stop this talk. It is nonsense. Not physically possible.
And did you get your permissions-problem fixed, so that you can run ABXTester.app, now?
Wow! I did not know that. That is just gob-smacking. I shall have to view everything he says with extreme skepticism going forward.
To add some some content, let me make a small correction here. There are two measurements in science with 10 significant figure (200 dB) accuracy. This is necessary to make contact between theory and experiment.
The theoretical prediction for the anomalous magnetic moment of the electron is expressed as a power series in the “fine structure constant,” α. Tom Kinoѕhita spent his entire adult life computing the first 5 terms in that power series. But to compare Tom’s prediction with experiment, you need to plug in a number for α. “Everybody” has probably heard that α is approximately 1/137 ~ 0.0729… But that’s way too inaccurate. You need an experiment which can measure α to the required precision. That’s the integer Quantum Hall Effect.
Aside from those two measurements, there’s nothing in science that comes remotely close.
Except for the hearing of audiophiles …
