Audition Cuts Frequencies Above 16 kHz While Exporting MP3 Files

[Fatih5C10]

Hello everyone, I use Adobe Audition 2022 Build 22.0.0.96 and as I made the problem clear in the title when I export a file as MP3 with 320 kbps quality via Audition, frequencies between 16 kHz and 20 kHz are cut off.  this problem is independent of the source file. for example, when I export a FLAC file as MP3 this happens. When I edit a full frequencies range MP3 file (20 Hz - 20 kHz) and than export it this problem also happens too. The funny thing is that when I do the export process with Audacity which is a free and open source software, this problem does not happen. I tried the same experiment with other paid softwares such as Acoustica 7, iZotope Ozone 9, iZotope RX 9 Audio Editor, and as I mentioned, frequencies above 16 kHz are not cut like Audition in any of them. I think this problem is caused by Fraunhofer MP3 Encoder but I can't do anything to change it. Therefore I will try to find the solution by writing my problem here. I also want this problem to be seen by the authorities. I will also share the screenshots containing the problems I mentioned with this post.

[SteveG_AudioMasters_]

Yes, almost certainly the Fraunhofer encoder, but a) you can't change it and b) 'the authorities' (we call them the developers) won't see this, as we're a U2U forum. If you wish to make a feature request that they will see, you need to post it here. 

[Fatih5C10]

First of all, I am relieved that you answered in a way that proved my guess. Secondly, I'll reopen the thread where you directed me, thanks for that.

[PaulVin]

Changing the mp3-export "Format Settings" from "Constant" to "Variable" seems to have mostly solved the 16k/20k cutoff issue... but it also produces a much more "digitized" looking spectral image... I have no idea if this is wise, a bad idea, or simply of no consequence either way.

[SteveG_AudioMasters_]

Setting it to variable is a very bad idea if you want to retain any degree of accuracy of timings - it's all pure guesswork on the part of the encoder and is notorious for being wrong! The other thing, of course, is that now Audition doesn't use the Fraunhofer encoder any more, but uses LAME 3.100 (there are now no licensing issues) and this behaves rather better in the upper ranges.

[PaulVin]

Got it - won't be doing the VBR, then - thank you!

Puzzleingly, I exported mp3s on my two computers, both running the latest version of Audition, with identical export settings (as far as I can tell). The result from M2 Mac Mini still displays the 20k cutoff, while the M1 Macbook Pro file does not.

[SteveG_AudioMasters_]

You have to be careful when comparing MP3 files made in Audition. If you do a 'save as' and then think that what you have displayed on the screen (which incidentally says that it's the MP3) is actually that, then you'd be mistaken - that's still the original file! To view the actual MP3 file it has to be decoded - which means that it has to be opened again. Only then will you see any differences...

[PaulVin]

Got it - thanks again. On re-testing, the 20k cutoff is still present on both computers. Is this something you'd have expected to go away with the LAME 3.100 encoder... or is it's presentation simply less egregious than it used to be? In the final analysis, is this just something to live with when creating MP3s in Audition?

[SteveG_AudioMasters_]

When you consider everything - the way that MP3s are encoded, the range of human hearing in adults, whether there's an R in the month, etc - it's quite simple; it really doesn't matter. MP3s are encoded using quite an arcane process involving critical band analysis and selective coding losses (aka masking) and are pulling the wool over your eyes about what you think you can hear anyway. Adult humans very rarely have significant hearing above 16kHz, and it's also pretty difficult to capture anything above that anyway, because it all gets very directional, and the majority of mics have to be in exactly the right place to capture anything that's there - which your average pair of speakers and headphones won't reproduce anyway...

The general consensus for quite a while now is that LAME sounds 'better' than Fraunhofer. It's quite a reasonable position; the rules for MP3 playback insist that all players have to be able to handle any encoding scheme and that nothing about playback can be altered in the player. So players are pretty much a level playing field. But there's quite a bit that can be manipulated in the encoder, and that's where any differences manifest themselves. But they all rely on critical band encoding, and it's that top band - the one centred around 16kHz - that is handled differently. It's all a matter of where the masking is applied, which is why you will usually see the spot immediately above 16kHz bouncing around a lot. They get away with it here precisely because humans can't hear these frequencies very well - so that's where a significant coding gain is made. You can reduce the amount of this biouncing by increasing the bit rate (I never use less than 192k), which means that not so much of the band has to have its bit depth reduced in the masking process (that's how they do it).

So the whole thing is pretty moot. You can't hear it - they don't fully code it. And the answer to your question is undoubtably 'yes'...

[SteveG_AudioMasters_]

I should perhaps mention that the reason that I know far more than is good for me (or you) about this is because the whole MP3 coding thing formed a significant part of one of my acoustics special studies, which was all about whether you could make meaningful acoustic measurements from MP3 files, and what, if any, the constraints were.

[PaulVin]

Wow, thank you, Steve. I really appreciate a good deep dive... there's a big difference, IMHO, between, "there's not a monster under your bed - don't worry about it", and, "that rumbling is the floorboards resonating with the furnace that keeps our house warm - so take comfort, it's a byproduct of our house keeping us safe and comfortable."

By which I mean to say, I want to make sure I'm providing a quality product - and now I have an informed response to my own concerns, and any questions I may get. So again, thank you so much for taking the time and effort to explain all that.

Your last comment brings to mind a question I've seen thrown around: the usefulness of YouTube-based audio comparisons, such as mic shootouts... (sorry for picking at a whole new ball of string!)