[gregorio]

I tried one of these tests online the other day and got 2 out of 6 right. Between uncompressed wav, to 320 mp3 and 128 aac.

Yep, that’s entirely in line with the expected result that everyone else gets.

This was not conclusive to me as I believe I would have to test the same file that I am used to hearing compressed down to the differing amounts to form a conclusion.

It doesn’t make any difference if you are used to hearing it. What can make a difference is training to identify the difference but even then you still won’t be able to hear the difference between wav and 320 MP3, only with 128 AAC and only with certain types/pieces of music.

I don't believe my transducer is limiting.

The transducers are easily the weakest part of the reproduction chain, with way more noise/distortion than even relatively cheap DACs, amps and cables. In this case however, the limiting factor is human hearing. This is completely deliberate of course, lossy codecs are designed to achieve their compression using “psychoacoustic models”, they remove the audio data that the human ear cannot hear.

Maybe this has the opposite effect and the sh!ttier the can the worse it sounds?

A sh!ttier can is always sh!ttier, it just tries to convert an analogue signal into a sound wave, it doesn’t know if that analogue signal has come from a wav or an MP3 and plays the wav less sh!tty. The only possibility would be if: The sh!itty cans produce an audible amount of inter-modulation distortion specifically in response to freqs >16kHz and you have a wav with a significant amount of >16kHz content and you compare it with a 128kbps MP3 (which has all the >16kHz content removed). In which case, you’d get audible IMD with the wav but not with the MP3. Apart from this potential but unlikely scenario, you’ll be more likely to hear the difference between wav/320 MP3 and 128 MP3 with better cans, as the audible differences may not be audible above the noise/distortion of the sh!ttier cans.

G

 

[bigshot]

I wish my ears heard as well as my cans sound!

 

[71 dB]

I tried one of these tests online the other day and got 2 out of 6 right. Between uncompressed wav, to 320 mp3 and 128 aac.

If you don't hear any differences, the probability of getting a correct answer by luck is 50 % or 0.5.
Since every sample can be answered wrong or right, the amount of possible rows of answers is 2 * 2 * 2 * 2 * 2 * 2 = 64.
There is only one way to get all of the right: The probability to do so by luck is 1/64 = 1.5625 %
The same goes for getting nothing right: The probability is also 1/64 = 1.5625 %
If you figure out how many ways you can get 2 correct answers you get 15 way (probability = 23.4375 %)
Intuitively getting 3 correct and 3 incorrect out of 6 randomly is the most probably outcome of all and the math(s)
does agree: there are 20 ways to do so giving us the probability of 20/64 = 31.25 %.
The probability of getting 2, 3 or 4 correct answers is (15+20+15)/60 = 50/60 = 83.33...%

In general you use "n choose k" formula n! / (k! * (n-k)!) to calculate the number of ways to get k correct answers out of n tries and the formula:

Pk = (0.5^n * n!) / (k! * (n-k)!)

gives the probability of getting k correctly out of n. As a graph it looks like this:

If you can hear differences, the probability of getting correct answers is larger than 50 % and the results disagree with the estimated probabilites above. For example, if the differences are so easy to hear you get always everything correctly, the probability of getting 6 right answers becomes 100 %. If a small difference can be heard with considerable effort, you probably get 70 % correct and the graph above becomes this:

So, most of the time you should get 4 or 5 correct out of 6. If the difference is a bit easier to spot and the probability of getting a correct answer is 90 %, the graph looks like this:

The bigger the difference, the more the graph is "skewed" to the right approach the situation were the probability of 6 correct answers out of 6 is 100 % which make perfect sense. The generalized formula to calculate probabilities is

Pk = (p^k * (1-p)^(n-k) * n!) / (k! * (n-k)!),

where p is the probability of correct answers (e.g. 0.8).

 

[Blackwoof]

For 160kbps VBR - all codecs

LAME 3.100 at V3 : Pretty much transparent 99% of the time, Only some tracks need V0(48KHz) cause of pre echo.

QAAC at Q82 : Would be 100% if I didn't need to use FHG AAC to cover stuff Apple AAC chokes on?.

Opus 1.3 at 150kbps : It pretty much 160kbps CVBR but It seems bloat hard on pure synth based ambient reaching 195kbps when others are 48 ~ 115kbps.

Musepack at Q4.9 : Reaches 100% despite not being MDCT based no idea why Streaming & Broadcasting ignored this codec?.

Vorbis at Q5 : 85% at best gave up ABX'ing since it so bad at handling pure noise when AAC/MP3 have no issues?. Starting to wonder if the dev's are lying as I'm not the only one who noticed that both 3rd party & main encoders just cover issues by using high bitrate(350kbps). No idea why Spotify picked this codec when LAME MP3 at V3 & V0 would've been good enough?.

 

[Sterling2]

These days Apple Music seems to deliver AAC and ALAC files which I cannot distinguish from SACD, so for now I am quite satisfied with Apple Music from an M1 iMAC to Parasound 2.1 Preamplifier via usb.

 

[Blackwoof]

These days Apple Music seems to deliver AAC and ALAC files which I cannot distinguish from SACD, so for now I am quite satisfied with Apple Music from an M1 iMAC to Parasound 2.1 Preamplifier via usb.

Yeah I use AAC as It has near MP3 level support unlike Musepack, Since I can switch to a DAP without being locked out if my phone failed.

 

[gerelmx1986]

Ive adopyed on the following philosophy.

At home with a DMP-Z1 lossless
Pn the go with a WM1A lossless
On the go with an NW-A25 wallman lossz AAC VBR 240kbps

 

[Roland P]

My music while running: 128kbps mp3 (so I can put as much music as possible on my 4gb player/headphones)

 

[71 dB]

My music while running: 128kbps mp3 (so I can put as much music as possible on my 4gb player/headphones)

I wouldn't recommend 128 kbps mp3s in high fidelity home listening situations (at least 192 kbps or better imo), but for running music masked partly by environmental sounds (traffic noises, wind, birds chirping, etc.) it is totally ok. Admitting you are fine with such a "low" bitrate in situations were higher bitrates aren't really needed isn't a sign of wooden ears, but a sign your brain isn't wooden.

 

[Svperstar]

Been in this hobby since 2003 and seen so many people claim they can tell the difference between 320 k mp3 and lossless, one even claimed he guesses "9/10 correct on blind tests" but plenty of people say stuff on the internet doesn't mean its true IRL....

My music while running: 128kbps mp3 (so I can put as much music as possible on my 4gb player/headphones)

I used to have a 128 MEG mp3 player for the gym in 2003 lol. Then I got an iPod. Then smartphones got big enough where I didn't need an iPod.

Over the years I have had to use all kinds of strategies to cut down the music library I store on my PC vs mobile. I used to use APE to store my CDs because it was smaller then FLAC, the problem with APE is it uses waaaay more CPU then FLAC to decode. More CPU = more battery drain.

I used to convert my library from APE to 256k vbr -> cellphone.

I converted all of my APE into FLAC a few years ago and also finally cellphones got big enough storage I was finally able to just drag and drop my music library onto my phone.

 

[Ryokan]

128kbps sucks even I can hear that.

 

[TheGiantHogweed]

On my FiiO X3 that I've used for nearly 10 years on public transport, I use 320kbps mp3s for almost everything. Even on Spotify, I tried comparing the difference between 160 and 320 and the difference was noticeable but only minor.

What I concider more importent is the production of the music you are listening to. If it is pretty badly mastered, I think you would really struggle to tell appart 160 and 320. Given the amount of space 320kbps MP3s save over CD quality flacs, I would pretty much always go with MP3s. I only go for flacs if the recording is good enough to make it worthwhile.

 

[bigshot]

I don't hear any difference with well recorded music either. Lossy is definitely able to achieve transparency.

 

[gerelmx1986]

Im usinf db poweramp music conv rter to conv rt ftom lossless to AAC 224k vbr, whivh is the max VBR that the included FDK AAC encoder allowa

133 albums on 16GB walkman internal syorage, pretty decent ampint. Anyways i belive i wont tell thedifference on suvh a tiny DAP the size of a 2nd generation iPod nano , wont give thequality sound that a home rig gives and used on public transport where noise of all kinds os present... 224 vbr is more than enough

 

[Blackwoof]

I wouldn't recommend 128 kbps mp3s in high fidelity home listening situations (at least 192 kbps or better imo), but for running music masked partly by environmental sounds (traffic noises, wind, birds chirping, etc.) it is totally ok. Admitting you are fine with such a "low" bitrate in situations were higher bitrates aren't really needed isn't a sign of wooden ears, but a sign your brain isn't wooden.

Lossy audio these days are now transparent at 96kbps VBR with Opus 1.4, xHE-AAC, AAC-LC(Apple or FDK).