Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Should I post a new sample reg alt-preset normal ? (Read 9587 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Should I post a new sample reg alt-preset normal ?

I have found a new sample where Dibrom's alt-preset normal (all revisions) produces very audible artifacts and distortion compared both to alt-preset high and r3mix settings and I was ready to post it for us to see what's the problem with it. In light of the recent developements (Dibrom's departure), I'm wondering what to do, as, at the moment, I don't know if Dibrom has released the source for alt-preset normal and if anyone else will take his place in the developement of the preset. In other words, is there any point in posting this sample or not?

Cheers.
Zbutsam

Should I post a new sample reg alt-preset normal ?

Reply #1 –
I think you should post the sample anyway...

you never know what might happen... once we have the source code some people might try to fix it... and they might succeed...

Should I post a new sample reg alt-preset normal ?

Reply #2 –
Yeah, zbutsam. Could you post the sample.
Juha Laaksonheimo

Should I post a new sample reg alt-preset normal ?

Reply #3 –
http://www.geocities.com/sad_baffoon/love.zip
(Queensryche - I don't believe in love, LPAC 663 KB)

With alt-preset normal, you get very audible background distortion of the guitar and loss of high frequencies. Doing a spectral analysis in Cooledit Pro, the alt-preset normal mp3 lowpasses to about 11 KHz, everything above that is distortion. alt-preset high and r3mix go up to 16 KHz and show no apparent problems.

I hope this is helpful

Should I post a new sample reg alt-preset normal ?

Reply #4 –
Thanks. Have you done any blind testings?

I hear very bad backround distortion also in the original and r3mix especially during 2-4seconds. In my opinion --alt-preset normal is not worse, but since I'm having headache right now I'm not very eager to do ABX, but will do later.
The guitar has no high freqs at all. The "missing" high freqs are VERY low energy high frequency noise which should be totally inaudible.

I asked also Beatles' brief opinion and he actually said that r3mix might have a bit more distortion than --alt-preset normal. Also he confirmed that there's no audible high freq loss with --alt-preset normal. But mostly he was terrified about the quality of original...

Anyway, can somebody else except zbutsam hear very audible artifacts (missing from original of course) with --alt-preset normal and this sample?
Juha Laaksonheimo

Should I post a new sample reg alt-preset normal ?

Reply #5 –
Thanks for testing guys 
I've uploaded the "alt-preset normal" mp3 at http://www.geocities.com/sad_baffoon/love_rev_6.zip
Perhaps I'm doing something wrong here, cause the difference is so big to my ears I don't need to ABX it. I hear lots of "air" mssing (like an 11 KHz brickwall) and serious deformation of the sound like somethink Xing at 128 would produce.
Please can someone else also test it cause I'm going crazy with this sample

Should I post a new sample reg alt-preset normal ?

Reply #6 –
The mp3 you provided seems correct. I don't believe in this sample the "airness" could come from over 11kHz very low energy noise.

But you can try abx for example --alt-preset insane and --alt-preset insane --lowpass 11.2 (Even better way would be to lowpass the original with cooledit and compare orig wavs)

The thing is, in my opinion the original sounds also like 128kbps Xing-encoded.  One possible reason might be that you are expecting the original to sound good (while it's not), and only hear the problems with mp3 when listening more carefully? I'm not sure what you are using for mp3 playback, or is any EQ in use. The most secure way would be to decode the MP3 to wav and do some blind tests.

Could anybody else say something (anything) about this?
Juha Laaksonheimo

Should I post a new sample reg alt-preset normal ?

Reply #7 –
I hear no difference from the original using either --alt-preset normal or --r3mix in Dibrom's rev6 compile.

I can see the cutoff at 11 kHz in spectral view of the --alt-preset normal encode.  --r3mix encodes up to about 13 kHz.  But I don't hear any of this.  Also, I don't hear any added distortion in the guitars, although the original is pretty bad.

Since I had nothing to latch on to, I didn't ABX.

ff123

Should I post a new sample reg alt-preset normal ?

Reply #8 –
First of all, regarding the quality of the sample, it was taken from a CD and, of course, the rest of the song doesn't sound as distorted as the short sample I provided.
Second, I decoded the alt-preset normal through Winamp's Disk Writer (without any EQ of course) and in a blind listening test in ABX I scored 20/20 against the original WAV.
Seeing how this sample and the artifacts I heard started giving me a headache, I browsed in various mp3 sites trying to identify the nature of the problem. Finally I found it:
Ringing artifacts in Lame from ff123's site
(http://ff123.net/ringing.html).
The artifacts I was hearing were the same as the one I heard in the unt.mp3 sample from that site, the ringing in the drum fills in seconds 8 and 20 of that sample.
Realizing how this ringing can be reduced by --nspsytune I did the followng test: I encoded the love.wav sample at 128 Kbits both with and without --nspsytune using these settings:

a) -h --nspsytune --athtype 2 --lowpass 16 --ns-bass -8
b) -h --athtype 2 --lowpass 16 --ns-bass -8

Whereas the (a) mp3 is pretty much OK quality wise, the (b) one showcases the exact same artifacts as alt-preset normal, only VERY VERY pronounced. I hope that at 128 you will be able to hear the problems I was talking about.

And another thing; I hope I haven't insulted anyone by persisting with this sample. It's just difficult for me to accept that noone is hearing some quality degradation which to my ears is very obvious

Should I post a new sample reg alt-preset normal ?

Reply #9 –
Quote
Second, I decoded the alt-preset normal through Winamp's Disk Writer (without any EQ of course) and in a blind listening test in ABX I scored 20/20 against the original WAV.


Well, this convinces me you hear something.  And if it's ringing above about 12 kHz, I won't be able to hear it.  Apparently, not too many people can hear (much less are bothered by) ringing at higher frequencies.  But you appear to be one of those unlucky few.

Quote
Whereas the (a) mp3 is pretty much OK quality wise, the (b) one showcases the exact same artifacts as alt-preset normal, only VERY VERY pronounced. I hope that at 128 you will be able to hear the problems I was talking about.


If it dips down low enough in frequency, I should be able to hear it.

Quote
And another thing; I hope I haven't insulted anyone by persisting with this sample. It's just difficult for me to accept that noone is hearing some quality degradation which to my ears is very obvious


That's why ABX tests are useful.  Listeners often disagree with each other -- you find it amazing that something you hear so clearly can't be heard by others.  I find it amazing that you do!  ABX allows us to trust each other's opinions.

ff123

Should I post a new sample reg alt-preset normal ?

Reply #10 –
This sample revealed something I didn't know could be so extreme! With --alt-preset normal, listening with the HD580 it sounds OK, but with the HD250 linear II the artifacts are all over the sample. They coincide with the spikey stuff above 11 kHz. Sounds irritating because of its randomness.

The HD250, apart from being closed, has a 'thinner' sound and I knew I used to hear more HF artifacts with it, but I'm amazed I heard none at all with the HD580. So here could be one more reason why there's such different perception with HF artifacts: try some different phones!

The alt-preset normal has a very 'tight' ATH, eager to cut away the higher sfb's. Another example castanets.wav, notably between the guitar chord and the first click I hear a disturbing vibrating hiss.

Like the 'gpsycho sound', you hear it or you don't, there's no point arguing much.

Hans

Should I post a new sample reg alt-preset normal ?

Reply #11 –
Ok, I'll try again, although in my opinion the original is one big audible distortion... But if you hear something more and can abx then it's there. The spectal looked pretty solid upto 11kHz, though.

Quote
a) -h --nspsytune --athtype 2 --lowpass 16 --ns-bass -8 
b) -h --athtype 2 --lowpass 16 --ns-bass -8

Ok, will try. Btw. the section b --ns-bass -8 does nothing. It only works with nspsytune.

Quote
And another thing; I hope I haven't insulted anyone by persisting with this sample. It's just difficult for me to accept that noone is hearing some quality degradation which to my ears is very obvious.
Of course you haven't insulted anyone. You are obviously sensitive for ringing. I'm obviously not that sensitive for ringing, or then I can't very well discern (or care about) the added distortion since the original is horrible enough..

I would be interested whether the ringing is gone if you lowpass at 11kHz, or does it happen lower? Those high freqs just look SO low energy that I find it hard to believe anyone could hear those behind the original distortion.
Juha Laaksonheimo

Should I post a new sample reg alt-preset normal ?

Reply #12 –
Lowpassing the mp3 at 11 kHz gets rid of it all.
While it's silent above 11 kHz, the guitar only sounds a little dryer, cleaner than the original. It's this intermittent content above 11 kHz that irritates.

What phones do you use BTW?

Should I post a new sample reg alt-preset normal ?

Reply #13 –
How high volume you guys use. Yeah I can hear the difference clearly between alt-normal and r3mix if I highpass at 10kHz, still it amazes me that somebody could be bothered with it because the original is absolutely horrible and the distortion in it takes my full attention. The distortion in the original is bothering me too much. Oh well..
Juha Laaksonheimo

Should I post a new sample reg alt-preset normal ?

Reply #14 –
Quote
Originally posted by JohnV
How high volume you guys use. Yeah I can hear the difference clearly between alt-normal and r3mix if I highpass at 10kHz, still it amazes me that somebody could be bothered with it because the original is absolutely horrible and the distortion in it takes my full attention. The distortion in the original is bothering me too much. Oh well..

Hehehehe...I confess I feel the same way.....

Should I post a new sample reg alt-preset normal ?

Reply #15 –
was the sample removed?  I went to the link posted near the beginning of the thread to try and hear for myself and got an error message from geocities.

Should I post a new sample reg alt-preset normal ?

Reply #16 –
No it's not removed. Try right-click the link and "save target as".
Juha Laaksonheimo

Should I post a new sample reg alt-preset normal ?

Reply #17 –
This is quite interesting.
Could you please try
a) -h --athtype 2 --lowpass 16
b) -h --athtype 2 --lowpass 16 --athlower 10

I know b) will probably be lower quality than --nspsytune, but I'd like to know if b reduces problems over a.

Should I post a new sample reg alt-preset normal ?

Reply #18 –
Unfortunately b (-h --athtype 2 --lowpass 16 --athlower 10) to my ears sounds just as bad as a. On the other hand, messing around with the options, I found that

-h --athtype 3 --lowpass 16 

reduces ringing considerably compared to

-h --athtype 2 --lowpass 16

The problem with this sample is not only ringing as an annoying artifact, but the fact that ringing destroys all mid-higher frequencies muffling the sound as if it was sampled at 22050 Khz 8 bits. I say that because in the original sample, in spite of the distortion of the guitar I can hear enough "airness" and clarity whereas the mp3 without --nspsytune sounds like on old tape which has corroded by constant use over the years

Should I post a new sample reg alt-preset normal ?

Reply #19 –
I noticed that the sample was relatively low volume.  Perhaps this fact makes things worse.  I'll try this later tonight, but maybe something like:

--alt-preset normal --scale 2.0

makes things better?  It's just an experiment because that's obviously not something for general use.

ff123

BTW, what kind of headphones or speakers do you use, zbutsam?

Should I post a new sample reg alt-preset normal ?

Reply #20 –
Looks like with --scale 2.0, lame encodes more of the higher frequencies (now up to 13 kHz), although there's still quite a bit of the "speckling" above 13 kHz. which causes the ringing artifacts, according to Dibrom.

The athadjust probably needs tweaking to make this sample sound better to zbutsam's ears.  However, getting low-volume stuff to sound good comes at the expense of higher overall bitrates.

That mythical adaptive lowpass filter is starting to sound better and better.  Maybe it could at least take care of the ringing.  For example,  I bet

--alt-preset normal --scale 2.0 --lowpass 13

sounds pretty good on this particular sample.

ff123

Should I post a new sample reg alt-preset normal ?

Reply #21 –
Using --scale 2.0, the artifacts are greatly reduced. I can still hear them faintly but for the most part they are masked behind the doubling of the volume and I have to pay attention to be able to discern them. I doubt I'd be able to score very well in an ABX test.

My PC speakers are a pair of mid-price Yamaha YST20s (25 Watt RMS). Quite good for every-day listening at medium volumes but nothing special and not intended for artifact-watching 

My headphones are the famous (-> takes a look at the brand)
FIRSTLINE H1096 DIGITAL SUPER BASS MONITOR (!!!!) of the El Cheapo variety. I was out to buy the Grado's SR60 model but the guy who imports them here in Greece asked for a retail price of about $180. I got so pissed off that I walked into the first electronics shop on my way and bought the first pair of headphones that looked faintly respectable. Anyway, they have exceeded my expectations so far as they have a more linear frequency reponce and a less boomy bass than my previous Technics. Also they have a higher upper range and I can now prowdly hear much more MP3 artifacts than I did before   

Should I post a new sample reg alt-preset normal ?

Reply #22 –
Quote
For example, I bet --alt-preset normal --scale 2.0 --lowpass 13 sounds pretty good on this particular sample


Truly, the addition of the lowpass switch, eliminates and the last few traces of ringing. The mp3 is now totally transparent to my ears even with careful inspection.

Should I post a new sample reg alt-preset normal ?

Reply #23 –
Quote
My headphones are the famous (-> takes a look at the brand) 
FIRSTLINE H1096 DIGITAL SUPER BASS MONITOR (!!!!) of the El Cheapo variety. I was out to buy the Grado's SR60 model but the guy who imports them here in Greece asked for a retail price of about 0.


I wonder if you would have the same experience as Hans if you were to use another pair of headphones, like perhaps the Sony MDR V6 (aka MDR 7506)?  The Sony's are used professionally because of their "flat" response, and perhaps your Firstlines have some peaking above 11 kHz.

ff123

Should I post a new sample reg alt-preset normal ?

Reply #24 –
I'm afraid I won't have access to higher quality headphones for the near future as I'm on a very tight budget for the time being and I can't think of somone to borrow them from   
For what it's worth, I tested the untweaked -alt preset normal mp3 with two pairs of earbuds I usually use on the road (one Philips and an old noname one). If I turn the volume loud enough, the ringing is still there, although less distinctable due to general sound degradation. The fact that I know for what kind of artifact to look for may have helped as well.