Help - Search - Members - Calendar
Full Version: problem with lowpass filter
Hydrogenaudio Forums > Lossy Audio Compression > MP3 > MP3 - Tech
schytluisje
hey, i use lame 3.93.1 with --alt-preset cbr 256 but i am consedering to change it to --alt preset cbr 256 --lowpass 22.5 , normally the 256 kb/s give me really good sound, but cause i don't want to change to a 320 bit rate, i want to change my lowpass filter, i saw regular audio cd's cutt off @ 23 khz, so, if i change the lowpass to 22.5 i would have an mp3 equal to cd ( i know i can't hear that frequency but still), but does these change of lowpass filter infectes my sound quality? if this is already posted on this forum, please forgive me

greetz koen
criZZb
First of all: the Nyquist's theory says, that this cut-off frequency is 22050 Hz rolleyes.gif

Preserving the whole spectrum @ 256 kbit/s will decrease the quality.

Have you considered --alt-preset insane preset instead?
schytluisje
do i get the whole frequency if i use --alt-preset insane?
problem is the file size is bigger then...
Peter
Do you prefer high frequencies you definitely can't hear over possibly hearable artifacts in lower frequencies ?
ScorLibran
QUOTE(criZZb @ Jul 19 2003, 03:30 PM)
Preserving the whole spectrum @ 256 kbit/s will decrease the quality.

Have you considered --alt-preset insane preset instead?

If I may add to criZZb's point about preservation of the whole spectrum negatively affecting quality...

The reason it's generally not preferable to use a lowpass set that high is that the encoder would then use bits to encode frequencies which you would never hear, where those same bits (since with cbr you have a set number of bits/second) could have been better used to improve sound quality in the audible spectrum. In other words, if your personal audible limit is 17kHz, then a lowpass much over that (perhaps >18kHz) would be wasting bits to define details of inaudible sounds at the expense of the quality of sounds well within your hearing range.

--alt-preset insane, as criZZb points out, may be worth considering (although it is 320kbps) since it's not so much more than 256kbps in terms of filesizes, and the default lowpass of that preset is already ~20500Hz. I use --alt-preset insane myself, and have never felt the need to manually change the lowpass setting at all.

Also, have you considered using --alt-preset standard or --alt-preset extreme? If you don't already know about them, they are highly-tuned VBR modes that provide very desirable results for many people. In fact, --alt-preset standard averages 180kbps to 220kbps and is transparent to many people according to what I've read here. For comparison, --alt-preset extreme averages about 220kbps to 260kbps. The lowpass filters for both seem to be sufficiently high for most people, with both residing a little below 20kHz.

The FAQs and the forum search function can provide any further information you want to know about the LAME --alt-presets.
DigitalDictator
Why use CBR? Just use "-preset standard" or "-preset extreme". Higher quality and lower/comparable file size. (unless you for some reason unknown to me HAVE to encode with CBR)
schytluisje
thanks for the reply's guy's, now i really know something more about mp3, so instead of increasing the lowpass, it would be better to decrease the lowpass?

greetz
sld
erm, it would be better to just use the --alt-presets.
They have been tuned to give people the best possible lowpasses at the various settings in the presets (whether --alt-preset standard, --alt-preset extreme, etc).
Of course, if you know (after testing with ABX, eg) that you cannot hear up to the lowpasses in the alt-presets, you can always set a lower one to save file sizes.
schytluisje
ok, i now still use --alt-preset cbr 256 , standard command line (personal prefer), but the lowpass is around 20064 hz i guess, how come noises over 20 000hz are still noticable, the lowpass cut's the music > 20 000 hz ? or not, here is a pic, so you will see what i mean, http://users.skynet.be/fa093969/screenshot.jpg,
next question :my computer is a pentium 4 ; 1.7 ghz; 256 ddr ram ;don't know wich soundcar ; 2 years old;
my other computer is a pentium 1; 200mhz, 64 ram, creative labs sound blaster awe 64, yet 6-7 years old, wich computer have theoritical the best performace or sound output to play mp3 or doesn't that matter?

greetz koen
schytluisje
sorry if my english isn't good, my native language is dutsh
PoisonDan
Why are you so obsessed with the lowpass ? Do you have any sample where you think the lowpass is audibly too high or too low ? Is there any particular reason why you don't trust the lowpass setting of the --alt-preset values ?

Oh, and using graphs is almost useless for judging the quality of psychoacoustic codecs.

About your computers: how can we judge which one offers the highest playback quality if you do not know which soundcard is present in your P4 ? The soundcard is of course the most important factor for overall sound quality.

I do seem to remember, however, that the sound quality of the AWE64 is not very good.
schytluisje
reason why i'm obsessed with the lowpass: last week i did tried to convert my mp3's from my harddisk (--alt-preset cbr 256 - m s) to smaller mp3's for my portable mp3 player cause i noticed how smaller the mp3 files how longer my mp3 player could play with my rechargeble batteries ;-)
first some info about my mp3 player, i bought it about 6 months ago for around 70 € or 65-75 $, don't know the exact rate €-$, and now i know it's crap, doesn't play mp3's good + high electricit consumption (=> play time around 2.5 h (mp3 with 256 bitrate) with GP 1300 mah rechargeble batteries) and that too my opinion too short. (some mp3 player from the same type go to 10-12 hours ;-))
So i tried to convert my archived mp3 files to --alt-preset 112 ( with play's very good, nearly no difference to hear with an 256 kb/s file). When playing with the command line, i did come to a command line --alt-preset 106 -h --nspsytune --athtype 2 --lowpass 20 --ns-bass -8 --scale 0.93, i saw that how higher the lowpass the better the sound on my portable mp3 player, here below some sounds i recorded with my computer so you can hear how bad my mp3 play's, mp3 player line out -- cord -- line in computer ( analog ;-)), i can hear difference between the 106 and 256 but not when i'm @ work and with the 106 my mp3 player plays longer :-)). Back to the lowpass : don't have the files anymore wich i tested the sound of the lowpass but i do remember when the lowpass was @ 20 i had the best quality with my mp3 player sound quality - file size, you have to know the earphones that where with the mp3 players can go up to 25 000 hz if i still can remenber me, correct me if it's impossible ;-)
gonna post the link soon because i'm uploading it now to my ftp, so you can download it @ high speed (350 -400 kb/s)

now i think i told you the whole storry, sorry if i did or typed anything wrong, i'm still an amateur @ ripping mp3's

greetz koen
DickD
QUOTE(schytluisje @ Jul 21 2003, 11:25 AM)
ok, i now still use --alt-preset cbr 256 , standard command line (personal prefer), but the lowpass is around 20064 hz i guess, how  come noises over 20 000hz are still noticable, the lowpass cut's the music > 20 000 hz ? or not, here is a pic, so you will see what i mean, http://users.skynet.be/fa093969/screenshot.jpg

Koen,

Like me you were probably taught that the human ear can hear from 20 Hz to 20 kHz.

That is only an engineering rule of thumb, and it's not completely accurate.

When engineering music playback equipment, you could aim to be flat to 20 kHz and you won't usually be disappointed, because that includes engineering margin of safety.

Psychoacoustic encoders, like LAME have been tuned more precisely to what people hear. The ear gradually gets less sensitive as frequency goes up, so more distortion is allowable. Very very few people can tell music lowpassed at 19 kHz from the original. Many can't even tell 16, 17 or 18 kHz lowpass from the original.

Many of us thought we could, but only because we wanted to believe it. In a randomised test, like ABX, we find the truth.

In your graphs, there's some sort of content at >20 kHz, but the vertical scale is logarithmic, so it's MUCH smaller than the signal at < 20 kHz. It may be that it's actually very low level noise in the smallest bit of the signal (quantization noise) or it's caused by decoding clipping, which is common in MP3s made from modern CDs but usually inaudible.
schytluisje
i know you're right, but i have listened closly and concentrated to my mp3 player and maked my own decision that my mp3 player plays better sound when the lowpass is @ 20 instead of 16, it could be just my imagination, don't know but it sounds better and i have read somewhere in this forum ( i have read a lot as guest before i joined this good forum) that high frequency improve some hearable frequency, correct or not??

just for the record here are the links so you can jugde my bad mp3 player

[URL=http://users.skynet.be/fa093969/--alt-preset 106 (mp3 player).wav]--alt preset 106 playback with mp3 player[/URL]
[URL=http://users.skynet.be/fa093969/--alt-preset cbr 256 (mp3 player).wav]--alt preset cbr 256 -m s playback by mp3 player[/URL]
[URL=http://users.skynet.be/fa093969/Bruce Springsteen -01- The Ghost Of Tom Joad.mp3]--alt preset cbr 256 -m s (original form computer) ( during mp3 laws not complete)[/URL]


kind regards koen
sld
"ABX: The only sure way to make a sonic decision!" ©

Well, we know that you are trying hard to prove to us that your listening experiences may be valid, but this forum absolutely demands ABX as a method of distinguishing between two different files.
I don't know much about frequency issues so I do not dare to comment on them. What I can only do is to insist that you do ABX smile.gif.
2Bdecided
At low bitrates (and anything below 128kbps is certainly a low bitrate), the mp3 encoder will probably dump most of the high frequency content, even if you over-ride the default (and sensible) low pass value, and replace it with something higher. Any extra high frequencies that are stored due to your use of a higher lowpass are robbing (or taking) bits that would be much better used to store lower frequencies more accurately. At 128kbps (and even more at 112kbps) there are not enough bits available to store everything below 15kHz properly - by asking the encoder to store everything up to 22kHz, you are asking the impossible.


However, your ears and brain may be more sensitive to a simple high frequency cut off than to nasty coding artefacts. What's more, on this sample, there is considerable HF information in the harmonica (which means that many people would prefer 20kHz lowpass against, say, 14kHZ lowpass), AND the recording itself is almost mono, which makes it quite easy to encode.

Try other music. If your ears still tell you (in a blind test - i.e. when you don't know which one you're listening to) that the higher low pass sounds brighter, ANd you prefer this AND it doesn't introduce any other audible problems for you, then use it! It will be adding more artefacts than the default low pass, but if the dull sound is real (not imagined) and you find this more annoying, then use your high lowpass and be happy. btw, 23?!? is overkill - 18-20kHz is already too much at 112kbps IMO!!!

I'll bet the lowpass in --alt-preset 128 is already high enough for you - trying listening blind.

Cheers,
David.
Mike Giacomelli
Download Winabx and see for yourself if you can hear >16khz.

Anyway you're transcoding from 256 and using a 20khz lowpass? Seems pretty pointless since you probably lost 90% of the high frequency stuff the first time you encoded. You're just wasting bit on garbage most likely. But run the ABX test and see for yourself whats happening.
KikeG
QUOTE(schytluisje @ Jul 21 2003, 03:07 PM)
... i have read somewhere in this forum ( i have read a lot as guest before i joined this good forum) that high frequency improve some hearable frequency, correct or not??

Not. Not if the high frequencies you are talking about are not "hearable".
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.
Invision Power Board © 2001-2008 Invision Power Services, Inc.