|
|||||||
![]() |
|
|
Thread Tools |
|
|
#1 |
|
DriverHeaven Newbie
Join Date: Apr 2004
Posts: 4
Rep Power: 0 ![]() |
aliasing at rates other than 48000
i was wondering if anyone had noticed when recording audio from a wave output
that is playing at any rate other than 48000, to any rate other than 48000, aliasing occurs because of the resampling filter not being perfect. i havent tested this completely or tried with creative's driver, i have an sblive value, newest version of kx. i'm not sure if the problem is constant or only when recording in this specific case (non 48000hz wave loopback) is the resampling filter in software? it isnt a big deal, but i just thought i'd point it out. |
|
|
|
|
|
#2 |
|
Apple Fanboy?
|
i think resampling is done after the DSP runs at 48khz… it is resampled before the DSP aswell so i would record is 48khz… les errors that way
__________________
Chris - The Aussie Super Mod
Hardwareheaven Rules - Sig Request Thread How you can help HardwareHeaven by using Digg! Hardwareheaven Super-Moderator |
|
|
|
|
|
#3 |
|
DriverHeaven Addict
Join Date: May 2003
Location: Philadelphia, PA
Posts: 380
Rep Power: 0 ![]() |
I also think alot of the aliasing is being done by whatever program is playing back the sound file. On way of seeing this is by using timidity that has an option to turn off aliasing of samples it play's. I think foobar player may have output plugins that has options to turn on or off the aliasing. Can't remember though, but check if it does.
|
|
|
|
|
|
|
|
DriverHeaven Newbie
Join Date: Apr 2004
Posts: 4
Rep Power: 0 ![]() |
i discovered the problem when working on some dsp code for minBLEPS oscillators,
which are used to generate perfect bandlimted impulses for waveforms such as pulses or ramps. i spent several hours scratching my head until taking a nap, where the idea came to me in a dream, the aliasing wasnt occuring in my program. i set up a wave logger, and behold, the direct output of the program was alias free. i then changed the sample rate output to 48000hz, and the recording rate to 48000hz, and no aliasing was generated. the problem is most definitly exactly what i've discribed it to be :P this would mean any program which outputs at any rate other than 48000hz is aliasing, meaning even winamp, when playing normal mp3 files. the aliasing comes out more as a kind of distortion, actually. the problem still occurs, even when sampling at 48000, with an absolute direct connection from the wave source to the record input in the dsp routing application. the drivers, or soundcard, seem to be generating distortion of the sound and creating frequencies that simply do not exist in the wave source! the problem goes away when using a playback rate of 48k. the problem is present at ANY recording rate if the playback is not 48k there is not any problem with the downsampling filter, it records absolute alias-free at any rate, from the soundcard input (with limitations on the analog filter, depending upon card model) and from the wave output on the card, assuming the rate is 48k the problem is either with the driver or the soundcard generating some form of distortion when upsampling wave data at rates other than 48k infact it may even be something to do with directx, i havent tried using asio or winmm anything like that to playback alias free sounds and test it. like i said in my original post, i have not completely tested this bug. it is a moderately serious bug though, as like i said, anything played back at rates other than 48k is aliasing. a simple upsampling filter like cubic spline would be better than the current proformance... of course i'd think the driver (if that is the source of the problem) could use sinc, or minBLEPS perfect resampling instead. i was just posting to notify people of the problem so i could see if it occurs for others, it may just be my version of the card. i actually have no idea.... but as it isnt noticable to most people (the distortion at 44100 is above 18k) like i said in the original post, it isnt an absolutely serious problem, i'd stick to implementation of driver features before smacking bugs like this. |
|
|
|
|
|
|
|
DriverHeaven Newbie
Join Date: Apr 2004
Posts: 4
Rep Power: 0 ![]() |
further analysis
i've found the problem does not occur when using winmm for wave playback.
the problem is with directx. it may have nothing to do with kx. should i shoot someone at microsoft? i suppose its all fair. considering how i code. tee-hee. |
|
|
|
|
|
#6 |
|
Apple Fanboy?
|
i think it is due to the internal processings of the DSP locked at 48khz… using 44100 or another sample rate will cause resampling both before and after the DSP processing… so i would not recommend using any other sample rate than 48khz
__________________
Chris - The Aussie Super Mod
Hardwareheaven Rules - Sig Request Thread How you can help HardwareHeaven by using Digg! Hardwareheaven Super-Moderator |
|
|
|
|
|
|
|
DriverHeaven Newbie
Join Date: Apr 2004
Posts: 4
Rep Power: 0 ![]() |
directsound buffering doesnt work well at 48k with my setup, i dont know why
but anyway it doesnt matter if you should use 48k, i usually do, its a problem that other rates are not being correctly resampled by directsound, i'm not sure if its kx's fault or directx's fault, but somebody has to fix it :P |
|
|
|
|
|
#8 |
|
Apple Fanboy?
|
yes for some reason with the DSP, it resamples even if it is already at 48khz, so you will never get bit-for-bit (this is a hardware issue), so i doubt there is a fix available for it, as for the DS resampling… that is done post-DSP, not sure whether at a software or hardware level, my guess is software though - mayb an issue that needs to be looked into
__________________
Chris - The Aussie Super Mod
Hardwareheaven Rules - Sig Request Thread How you can help HardwareHeaven by using Digg! Hardwareheaven Super-Moderator |
|
|
|
|
|
#9 |
|
DriverHeaven Newbie
Join Date: Apr 2004
Posts: 4
Rep Power: 0 ![]() |
I don't speak english very vell. Really, I don't read english very well.
I have a SB Live! Value, and I have this problem with CubaseSX1.06. I can listen and record my VSTi very well, but when I listen the record, the sound of the VSTi are different. It's revolutionated. I'm going crazy!!!!!!!! If you have the solution, please tell me or send me to stbn7777@hotmail.com Excusme for my english!!!
|
|
|
|
|
|
#10 |
|
Apple Fanboy?
|
alright um… does it sound like there is an echo with the VSTi record? if so go to the Record Windows, and look for the "wave" slider and mute that
__________________
Chris - The Aussie Super Mod
Hardwareheaven Rules - Sig Request Thread How you can help HardwareHeaven by using Digg! Hardwareheaven Super-Moderator |
|
|
|
|
|
#11 |
|
DriverHeaven Newbie
Join Date: Apr 2004
Posts: 4
Rep Power: 0 ![]() |
the record sound it's no like a echo, it's like a phaser. I try with your answer.
Plesase, excusme my english. |
|
|
|
|
|
#12 |
|
Apple Fanboy?
|
that sounds like something that happens when i used ASIO…
are you Recording or are you bouncing a track?
__________________
Chris - The Aussie Super Mod
Hardwareheaven Rules - Sig Request Thread How you can help HardwareHeaven by using Digg! Hardwareheaven Super-Moderator |
|
|
|
|
|
#13 |
|
DriverHeaven Newbie
Join Date: Apr 2004
Posts: 4
Rep Power: 0 ![]() |
I don't understand you. Bouncing a track? What's the meaning of bouncing? I record a VSTi (for example, IM-7 of cubaseSX) and I listen ok, but after the recording, I listen the "new MIDI track" and listen a diferent sound. This new sound it's like a phaser. Tell me your email and I send you a sample (100 kb).
|
|
|
|
|
|
#14 |
|
DriverHeaven Newbie
Join Date: Apr 2004
Posts: 4
Rep Power: 0 ![]() |
excuse me. Not sample. I send you a example.
|
|
|
|
|
|
#15 |
|
Apple Fanboy?
|
alright, i think your problems is that you are playing the recorded sound AND the midi sound you recorded at the same time
__________________
Chris - The Aussie Super Mod
Hardwareheaven Rules - Sig Request Thread How you can help HardwareHeaven by using Digg! Hardwareheaven Super-Moderator |
|
|
|
![]() |
| Thread Tools | |
|
|