I got into an argument that 192kbps mp3 is transaparent and there is no audible difference between 192kbps and 320kbps.
The other person says with the appropriate equipment and samples/songs the difference is easily noticaable.
So:
- how transparent is 192kbps compared to 320kbps
- if difference can be heard, then on what type of music in general and what equipment can the difference be heard on?

In other words, if there is an audible difference, how hard/easy is it to be heard?

It's all very subjective. Get this person to listen to some samples of the recent 128k listening test, which showed that for many (most?) people 128 VBR is transparent (or very close) nowadays, even with very good equipment. There are some rare problem samples which may benefit from 320k as compared to VBR~192k, but unless you have very good training in spotting compression artefacts (this has only little correlation with being an "audiophile" with expensive equipment), I doubt many people will notice an improvement. Get him to perform a blind listening test to prove his claim - as he said it's easy it shouldn't be a problem for him

VBR or CBR? Which encoder, and which version? Your question is too general to give a definitive answer.

The APS switch in the recent LAME builds, which is VBR at a target bit rate of 190kbps, is transparent for the vast (and I mean vast) majority of music, unless you are actively looking for (and have extensive experience in looking for) encoder artifacts. Better equipment does not necessarily mean artifacts are more apparent, in fact it's possible the opposite can occur.

Sorry, I was in such a hurry I forgot. I'm talking about LAME @ any bitrate but MOSTLY CBR. Version? Well, can't it be answered in general? If not, then versions 3.96, 3.90.3 and 3.97 b2 for example. Althoug for normal people they are all very close and it doesn't matter that much which version is concerned. Or am I wrong?

Even CBR 192kbps encoded with a recent version of LAME should be transparent for most material, although unless there's a reason you absolutely have to use CBR there's no point doing so.

Question is what kind of music could the possible difference be heard on and how expensive should the equipment bo to hear the difference?

For CBR 192, try some heavy metal. The large amount of HF noise could probably use up lots of bits and make artefacts more likely to occur. Also music with lots of stereo separation, because Lame then cannot exploit bitrate savings due to joint stereo and has to encode the two channels separately, which uses up lots of bits. Or music with sharp attacks - short blocks use up more bits to encode and may give 320k an advantage.

OK, even if there IS a difference how big would be? Would it be CLEARLY noticeable or slightly?

You are the only one able to answer to this question. As said before, transparency or the lack thereof is a matter of subjective perception. If when ABXing samples yourself you find a difference, then you'll be able to quantify it, but noone can do it for you, for once again it's a purely subjective thing.

Invite the one who's making the claims to perform an ABX test and see if he can truly ear a difference.

For Lame --preset cbr 192, I'd say extremely slightly in most cases for most people; If you are talking about older Lame versions with CBR 192 and simple stereo, it may be easier to spot. It all depends on the listener, it's impossible to make big generalizations

I have had problems in cbr 192 simple stereo particularly with heavy metal music. In some cases phasing effects in guitars and cymbals are added. In other cases distortion is added.

In THIS sample; In ABX tests, slightly high frequency distortion is added.

LAME at 128 VBR is more than transparent for me

A problem with ABX testing is that often the equipment used is not specified. An mp3 will be much more "transparent" (can't tell from the original well-recorded/mastered CD) through an ipod with stock buds, for instance, than through a high-end stereo system.

This makes the blanket statement "LAME xxx 128kps is transparent" based on the ABX tests highly suspect. Was the equipment used indicated? If so I missed it.

If you *really* read the sentence, he is completely right. The simple fact of the existence of "killer" samples is enough to demonstrate it.

sTisTi has mostly answered the rest of your questions.



tman, your sentence has two meanings, one that can be accepted here (with exceptions), and the other that cannot.

Equipment *can* make a difference when spotting differences. Yet, doing an ABX (ideally) in a discotheque can be more difficult than with an ipod. The in-ear buds can let you perceive better the small differences, because there isn't anything else disturbing.

Finally.. ABX means you *CAN* spot a difference. Failing to ABX proves nothing, except if it is the general case.
So for all usefull interpretations, 128kbps VBR LAME is near transparent. That wont prevent someone to spot a difference with 320kbps with a problem sample.

LAME was tuned using shitty laptop speakers. And I don't think equipment makes that much of a difference : I only heard the diff in the 128 kbps test once or twice in the case of LAME, and I have 800$ of listening eq...

Remember the preset is calibrated to make use of Joint Stereo. I think by specifying simple stereo youve broken the preset by forcing lame to use an encoding mode which is usualy less efficient. Quite likely the distortion you are hearing is the result of pychoacoustic discernments designed for 192kbs joint stereo not fitting into 192kbs simple stereo. Idealy the preset would adjust itself to cope, but it could be just far too much work, to have to test and tune, not only individual presets but also their dynamic adjustments to users favourite extra switches.

If you really want to use simple stereo, maybe use a lower bitrate (default joint stereo) preset to set up the internal thresholds and then raise the allowed bitrate and stereo mode with switches.

Like --preset cbr 128 -m s -b192 -B192 -etc

-but always when you try tweaking the presets - ymmv

For most music in general, 192 is probably transparent to most people. But (IMO) your friend still wins the argument because of the "with the correct samples" bit. Namely, I can ABX 10/10 256 kbps lame from an original wav file with the correct sample. (It was that castinets sample if anyone is curious). As for how easy it is, to me I have to focus to hear it but it is clearly there. Again, however, spotting the difference between 256 and something else on a killer sample is very different from doing it in actual music.

The fact that joint stereo is a lot more efficient than simple stereo is obvious (at least for Lame) and this has been discussed even the death here; and it was not my point. You severely misunderstood my post.

In addition, it was not my intention "tweaks" the presets as you erroneously are suggesting.

My point is that lamentably, with all the good information about joint stereo, a lot of people still believe that old and obsolete settings as CBR 192 simple stereo are a good solution for transparency; which is not true, at least for heavy metal music; that's all. My intention was with my sample to inform this to that people.

Sorry I did misunderstand, why simple stereo cropped up confused me.
Isnt the thread about comparing default 192 to 320?

edit: cleanup

I'm new here and am not trying instigate, but I think you missed or ignored my point.

Regarding the first part: I thought my point was obvious - that the better the equipment (and listening environment, obviously), the more likely you are to hear a difference. A high quality (revealing) home stereo will allow you to hear any differences a lot easier than POS portable player stock ear/headphones. (Why would you try to ABX in a discoteque?) If you want to use headphones - custom-fit IEMs will also allow you to hear what cheap buds can't (for the same 128kps sample, obviously). So, person A may claim something is transparent, where person B, with better equipment/(environment) would not. That is why it is important to mention what equipment was used for the test. [To clarify even further, if the test was done on highly revealing equipment and no difference detected, it's pretty safe to say that the tested sample is "transparent", for all practical purposes. If it was done on an ipod with stock buds, then it may be fine for that use but not for anyone using better headphones, or, especially, piping through a stereo system.]

For the second part, I don't see how your 128kbps VBR LAME being transparent claim was justified/supported.

[Note - I haven't done an ABX test so I don't have an opinion on whether any encoding is transparent or not, but I think it is important that the test itself is "transparent" (sample info, equipment, etc.) so the info can be useful to others.]

Equipment doesn't really seem to matter unless its broken. A cheap pair of headphones and onboard sound (that is not resampling to 48kHz) is generally enough. Maybe theres some extreme case it where it would make a difference, but I doubt that really matters for practical testing.



Yes thats evident! Try doing some ABX tests. You'll be surprised what encoder artifacts really sound like I think. At least I was. Its not like the normal distortion you expect to hear from bad audio equipment.

To add to the above thought, I think headphones do matter to some extent. I noticed that my closed headphones tended to work a little better since they blocked out some more ambient noise then my old open cans. But that was mostly just because they were helping to compensate for the problems with my listening environment.

I just don't think an expensive stereo helps. Actually I can't imagine doing codec testing on a stereo period. Headphones seem a lot easier for whatever reason, at least to me.

Thanks. yes, I didn't realize that any differences between encoders would be so obvious. so there is never a difference in the sound quality - the only difference is "artifacts" added?

Sound quality generally is the quality of the audio output from various electronic devices.
en.wikipedia.org/wiki/Sound_quality

when you saturate 5W laptop 'speakers', it might cause you to perceive 'artefacts' where the speakers are to blame, but generally the only difference between 'bad' and 'good' speakers is in the actual frequency range, and mentioned maximum output levels.. So while it might be harder to perceive high-freq artefacts in cheap-ass speakers than in good quality ones with tweeters, the artefact is in the data, yes, since it's produced by the encoder

well there are a lot of differences among speakers besides freq response and power handling (despite what Consumer Reports would have you believe), but that is an entirely separate issue that I don't think we should get into here.

Sound Quality - what I meant was that, based on a previous response above, it was implied that the only potential difference between original (BTW, are people comparing the mp3 to the original CD, or to the WAV file only?) and compressed is added artifacts, so I was trying to clarify whether db level, dynamics, detail, distortion, stereo separation, image, etc., etc. are not affected at all by any encoding, and the only potential change is adding "arti(/e)facts", which were implied to be easily discernable on even low resolution (/cheap) equipment. Your comment:

"So while it might be harder to perceive high-freq artefacts in cheap-ass speakers than in good quality ones with tweeters,..."

supports my initial statement that the equipment DOES matter when ABXing, so there appears to be some disagreement among the "experts" (surprise, surpise ) (of course, speakers that distort while playing the sample would not help the test reliability).

It is not the same to listen to something through two monitors attached to a digital equalizer than to listen to it with a mobile phone's speaker, using the built in "equalizer preset".

Of course equipment does matter, but not to that degree you want to imply. Many of the people ho did that ABX test used headphones, and most than not, of good brands.
In other words, what is NOT true is that equipment in the thousand $s is needed to perform a valid ABX.

You're right aswell assuming that there are other differences than just artifacts. In fact, Vorbis tries to make differences in favour of artifacts ( an artifact is something disturbing, a difference is not, if it isn't being compared to the original).