timestretching and resampling

hey guyz, I'm just mad curious about a thing about 2kxl. Since I got me a 950 I've been working alot with it and I find it incredibly comfy. They have been produced 12 years apart one from the other so the 2kxl should be a lot faster and "technologically advanced". So why is it that in the 950 you just push a button and you have instantly them both with a cool sound (gritty but cool nonetheless) while the 2kxl takes AGES to do both and sound reaaally bad? Shouldnt technology make steps forward in 12 years, instead of backwards?

Hmm all I can guess is that on the xl the audio is rendered/resampled in the machine, or reconstructed in the RAM for higher quality and less aliasing. while on the s950 it might be less destructive? More like pitching up and down on the xl, applied to the sample instead of reconstructing it? Best guess though, I know I've used both kinds of time stretching (instant vs. rendered).

im sure they just got ahead of themselves thinking they were writing some awesome timestretching code, when the fact is the shitty timestretching algorithm in the 950 is actually what is iconic today.

tapedeck wrote:im sure they just got ahead of themselves thinking they were writing some awesome timestretching code, when the fact is the shitty timestretching algorithm in the 950 is actually what is iconic today.

I always got the feeling too... I figured the machine wasn't designed from the ground up for that but they tossed it in at the end. I am not sure if the xl has the same cpu as the 2000classic... but those two features weren't on the classic.. so .. who knows.

Cool that they have it, some people use it a lot... I personally wouldn't want to have to plan ahead and time stretch something the night before just so I can get a beat done.

as bad as the timestretching is - the resampling is the real killer.

it could be a useful feature, but in one experiment i did it took 20 minutes to resample a 4 second sample.

tapedeck wrote:as bad as the timestretching is - the resampling is the real killer.

it could be a useful feature, but in one experiment i did it took 20 minutes to resample a 4 second sample.

yeah same here... the only time I tried using resample was the time I decided to try it out to see what it did to the sound.... it took so long I forgot what the original sounded like.

tapedeck wrote:as bad as the timestretching is - the resampling is the real killer.

it could be a useful feature, but in one experiment i did it took 20 minutes to resample a 4 second sample.

Damn I knew it took forever but I thought you would be good with one shots. That's insane. I don't understand how the mpc works, I know it's really slow isn't it like 100mhz? Debauch you posted the speeds recently I think? Yet on something like windows 95 machine that's 133mhz and 32mb of ram could resample a wav almost instantly on save... Did they just cram in a bunch of crap into the alorgithms, instead of just dropping bits? Why is it so damn slow! Also it's way fast in terms of buffering, faster than a computer double the speed could trigger WAVs live like the mpc. Is it all just widows, cause everything has to be routed through the entire OS?I notice when I'm booting in ms dos and running programs you can play faster, but still not as fast as an mpc 10 years older... Someone educate me

m56p87 wrote:Did they just cram in a bunch of crap into the alorgithms, instead of just dropping bits? Why is it so damn slow! Also it's way fast in terms of buffering, faster than a computer double the speed could trigger WAVs live like the mpc. Is it all just widows, cause everything has to be routed through the entire OS?I notice when I'm booting in ms dos and running programs you can play faster, but still not as fast as an mpc 10 years older... Someone educate me

you don't just 'drop bits' to resample - it involves some interpolation algorithms here and there to make it sound 'good' - which is probably what they thought they were doing.

i wasn't able to find a very consistent number - except that the 'quality' setting made a big difference.

as for the old windows thing - yea likely the software not being optimised to run realtime. however, if you use something like a soundfont, the sample playback at least feels instant.

before i had a real sampler i would generate loops / hits and then had to build a soundfont to play them back as a sampler.

I usually am not really into emulation or having to use an out of the box product to emulate what I can otherwise do in the box if I have the actual hardware (this especially applies to video games)

but someone posted in a few threads the website for a program that runs on windows specifically to time stretch audio files the way it was done on the "iconic" (lol) akai s1000.. I believe that was what it did anyhow... I didn't try it. It supposedly is accurate and gives the metalic sound you heard on amen breaks in old dnb.

Now, I think it may actually be worth doing that in the pc and sampling the result to the mpc... simply for time saving reasons.

^its called akaizer, and I believe the URL is something like akaizer.blogspot.com great little free app that renders way fast. I've always wanted to try using it to time stretch down and back up and see if you get any cool results. I think it's available for Mac and pc now

haha i was resisting the urge to post a link to it - cause it is an awesome program:
http://akaizer.blogspot.com/

cross-platform (windows mac linux)

I have not played with the S950, but are you comparing the same thing?

Playing back a sample at a different pitch is different to actually modifying the sound file.

Is the S950 actually modifying the sound file, or just playing it back at a different sample rate to change the pitch, and therefore the sample is shorter in time at higher pitches and longer in time at lower pitches?

Isn't the MPC modifying the pitch in the actual sound file, keeping the same length of the sample no matter what the destination pitch, and isn't that different to what the S950 is doing?

Wouldn't the MPC 16 levels or a chromatic assignment be the equivalent of what the S950 is doing, rather than a destructive file based operation such as the MPC timestretch?

Also, the MPC2000/XL CPU is approximately equivalent to a 286 PC in terms of processing power, so I wouldn't expect much from it.

the s950 has a timestretch as well.

it is based on a simple cyclical crossfade-overlap technique as far as i can tell.

http://www.youtube.com/watch?v=O3-rr_oGkgc
http://www.youtube.com/watch?v=8C82ocjwOeA <-this one jams

That one with the song is awesome.