Hi,
I' m new in the forum, I am happy to enter in this forum (very nice people) I hope to find many friends and tech info ... and why not... to give some contribute in the future with my background about software programming.

(sorry my english is bad.. i know)

so... this is my question:

I had traied to make my mp3 with this settings:

--altpreset extreme --verbose ...

but the hight frequnces is cutted to mutch for me

so ... i had traied with this settings:

--altpreset extreme -k --verbose ...

better... but sometime the hight freq. is metallic, so i try this
settings with compromise:

--altpreset extreme -lowpass 20 --verbose ...

Of course the best performance is "altpreset", and i try this setting only for experiments, in other way can use insane settings... is clear i don't want to use this settings for headphones ... but to listen my music with a good hi-fi system.


I'want the hight freq. "more brillant" without an 'insane' setting... can Help me.

I am sorry for my english... and if this question is already maked in the past (but i am new in the forum).

Thanks

I use lame 3.90.3

--alt-preset extreme has its lowpass cutoff between 19383 Hz - 19916 Hz. I highly doubt that you can hear such high frequencies with your stereo system, unless you paid more than 300$ for your speakers. And even then, your ears might not make out the difference.

I suggest you enable your equaliser settings to increase treble, as postprocessing, instead of trying to change the sound in the encoding.

Good luck,
T.

Try an ABX test. I highly doubt you can really really hear the lowpass, but because you are paranoid about it, you *think* you hear it.

I think it would be nice for you if you tested your lowpass threshold. I did this in the past and it really helped clear up some things that were in my mind. Do the following:

1. access ff123's samples page.
2. scroll down until you find mustang.flac. This is the original music sample. Right below you find links to lowpassed clips of the same samples.
3. start with the 12KHz lowpass one (mustang_12kHz.flac)
4. use an ABX program and test if you can hear a difference between the lowpassed one and the original one (ff123's ABC/HR comparison tool, foobar2000 with the ABX comparator, or Java ABC/HR comparator by schnofler).
5. if you can consistenly hear a difference move up the lowpass (13kHz, 14kHz etc. etc.) until you can't.

You'll be amazed with the results, I assure you.

hope this helps.

1. Search the forums before posting. This has been asked just a few days ago.
2. Don't make such claims without providing ABX-logs and samples
3. Dont troll
4. If 1-3 does not seem acceptable to you, then dont post on ha.org

I'm very sorry if my question sounds like a troll... i had make a research, but my question is for other target (my experiment?)... i think?
In the future I make more attention.
Sorry

So... I thank you for the answers, with a very accurate ABX test I have found the right way to encode my mp3.
Thank you.

Did you try to find your lowpass threshold? If that's the case could you please share your results? meaning lowpass threshold and ABX logs?

I have called 2 my friends (piano and guitar palyers... i am voice), in my studio I have Infinity system play back for player, and speaker home made (renge response 25 Hz ~ 23000 Hz).
I have downloaded this files
http://ff123.net/mustang/mustang_17kHz.flac
http://ff123.net/mustang/mustang_18kHz.flac
http://ff123.net/mustang/mustang_19kHz.flac

I have maked 3 track in mp3 ' --altpreset insane -k', played in random mode, my girl-friend log the answers for quality (A=low B=medium C=hight)
result:


http://ff123.net/mustang/mustang_17kHz.flac = 10 A , 4 C, 1C
http://ff123.net/mustang/mustang_18kHz.flac = 7 A, 3 B, 5 C
http://ff123.net/mustang/mustang_19kHz.flac = 0 A, 5 B, 15 C

But the sounds play good in any situations, only an impercetible natural sounds in 19Khz, more hight lowpass is really very difficolt to say the better.

I don't think making MP3 tracks first is a good idea, because the sound will be altered especially in the higher frequency area. When doing this tests you should use the original lossless files.

Furthermore, it is hard to say anything from your results, I think. I'm not a statistics expert but I don't know if it is possible to say whether or not you really could distinguish the different samples from your "ratings".

ABX tests are better, from the results you can say something like "I could hear the difference between sample A and B with a probability of at least 95%". See this thread for more info.

Well, maybe i overreacted a bit. The lowpass-discussion had been solved years ago and we didn't get repeated topics about it for a long time. However, in the recent time, some people seem to have found a hobby in registering new accounts simply for starting trolling-attempts - and just a few days ago, another lowpass topic(the first one since a long time) was started - also by someone who just registered... so maybe i counted one and one together a bit too fast. Sorry.

- Lyx

I'm glad you tried but your test protocol does not make it valid. It has to be a double blind test, so it is unbiased. The way you performed does not let you draw any conclusions. The whole idea of an ABX test is that you don't know which sample among the two being tested is playing while evaluating.
Also, you don't have to convert the samples to MP3. Use the original ones.
You don't have to perform it with fancy equipment. Just use your computer speakers.

I have a similar requirement of higher lowpass filter. I shall elaborate. When I rip a CD to my HDD I use vorbis at -q 7. Now, ABXing suggests that this is pointless because -q 5 is already transparent to me. But I choose -q 7 anyway because it isn't just me that will be listening to the compressed music. Plus, I'm a bit of a perfectionist and even though I cannot tell the difference I still prefer the high frequencies to be left there as much as possible. Using a spectral analyser I can see a big difference (lowpass frequency) between -q 5 and -q 7. This difference is what justifies the extra bandwidth.

Basically, I don't rely soley on ABX testing to determine better compression ratios.

But without ABX you have no backup that raising the lowpass would improve quality - which in turn means that you are proposing something without knowing what you're talking about. You're just saying "i *asume* this would be better", but you dont know. Search the forums and you will find more evidence that your proposal does not improve quality, but more probably *degrades* it.

Basically, your call for "improving the warm-fuzzy-feeling" will not be taken serious on hydrogenaudio.org.

What's the point if you can't hear the difference? I don't understand.

If you want everything to be there and don't care about the space (it seems like you don't care since you use -q7 over -q5 even being unable to tell the difference) why don't you use lossless? It is the only way to be sure you don't miss anything, but for casual/portable listening is way too much IMO of course.

Moving the lowpass up in perceptual codecs is not that simple. It would bloat the bitrate and quality for mid/low frequencies would probably go down, but I'll let someone more savvy elaborate on that.



edit: added quote

I doubt -q 7 quality is degraded compared to -q 5.
Are you assuming that I'm deliberately increasing the lowpass filter? I'm not. -q 7 happens to use a higher frequency for lowpass than -q 5.

But also -q7 bitrates are larger than -q5. Of course it will be better. This is not related to the lowpass only.

Because I'm not the only one that will be listening to the compressed songs. When -q 5 was my standard there would occasionally be complaints. The small increase in bitrate was all that was needed. Lossy is still too much, since resulting files are over four times bigger than the ogg files. I didn't claim to feel or hear the difference, infact I said otherwise.

Obviously. But it's only 64kbps extra, which doesn't create storage problems.

And what makes you think that with -q7 there won't be complaints? You're not making any sense....

Sorry guys, should've read the posts more thoroughly...

The thread starter wanted to increase the lowpass filter without changing from APE to API... Obviously a bad idea. My intention was to say I have a tendancy to prefer higher frequencies to be preserved even if I can't hear them.

Using a higher vorbis bitrate I found, with unorthodox methods what I wanted. Much to my, and my other listeners' content.

Are you aware of the fact that bandwidth is not unlimited and that lossy encoding means that some data is lost?!?

Uh, yes. Because it's a lossy codec, I know. But with a higher bitrate, (god forbid the semantics) _less_ data is lost...

If what you're talking about is not raising the lowpass ->at a given average bitrate<- but simply about increasing the bitrate...... so that you're just encoding at higher quality than necessary.... then what does this have to do with this topic?

What you're saying is just "i like to encode at higher quality than necessary". But this topic was about raising the lowpass alone without adding more bandwidth.

^ I know, sorry, as I wrote just a few minutes ago about not reading it thoroughly, I realised what I done

I've tried this test. There is a problem with the samples. The lowpass filter is a brickwall one, and the ringing is obviously audible at 12 and 13 kHz for me. I can ABX them easily, but if the lowpass filter had been softer, there would not have been any audible ringing, and maybe I could not have ABXed the samples relying only on treble loss.

I agree. The 14khz sample has the ringing too.

Excuse my ignorance, but is this the kind of filter generally used in lossy codecs or is it different?
If this is the kind of filter that creates the worst artifacts then I assume one can be safe of one's lowpass threshold using this method, I mean in the worst case one's lowpass threshold would be lower than the one found using samples lowpassed with this so called brickwall filter. But, as I've seen before one should never assume...

As far as I know, all major lossy encoders use smooth filters instead of brickwall filters.

I've uploaded some new samples, using cosine filters instead of brickwall ones. See this thread.