Swallowing my pride and asking for help :-/

[TotcProductions]

Ok folks, here's the deals...

I'm helping out on a project...my assigment is to take a set of pre-recorded samples, from various synths, of a few waveforms and make single cycle waves out of them.

So, for example, I've got 1 sample here of a pulse wave from an ARP synthesizer...it's been sampled at 96khz, 16 bit...My job is to isolate 1 cycle from the waveform, as close to zero crossings as is possible, and then, make that single cycle exactly 1,024 samples long (so if the cycle is, lets say 2,197 samples long, I need to resample the thing to make it 1,024 samples long). The process I've been using is as follows (in sound forge 8 by the way):

1. I load the ARP sample I was provided with.

2. I zoom in as far as SF will let me, and I select the beginning and end samples that are closest to zero crossings, and should provide a smooth loop.

3. I copy that selection to a new wave file (also 96khz, 16 bit).

4. I multiply the samplerate by 1,024 samples (ie: 96,000 * 1,024).

5. I take the product from the above equation and divide it by the number of samples that are actually in the single-cycle wave at the moment. (so, continuing with the example wave, this would be 98304000 / 2,197. The second of those two numbers is the actual sample length of the single-cycle waveform I've selected).

6. The answer to that equation is 44,744.651blahblahblah. So, I round it to 44,744, and then resample the waveform to a sample rate of 44,744hz.

7. The above operation has now made my nice little waveform exactly 1,024 samples. I then go in to the resample dialog, set the thing to "set sample rate only", and switch it to 44.1khz. Now I've got a 44.1khz, 16 bit single cycle waveform that is exactly 1,024 samples in length.

8. this is where my f**king problem comes in...I copy the wavecycle, paste it in front of the first cycle, and they don't make a smooth waveform! There are gaps and shit in between where I pasted them together.

Natrually, I figured I hadn't selected beginning and end samples that would allow me to make a smooth transition from one cycle to the next, so I open the original file, and make a new selection, being careful to make sure they match up properly, run the process and boom, same shit, different gaps. WTF! No matter which samples I choose, which cycle, etc..etc..etc...once I'm finished, I paste the cycle to make a continuous waveform and they don't match up. There are steps in them.

So, please consider all of the above, if it makes any goddamned sense, and help me out here. Is what I'm trying to do even possible??? (by the way, the original samples I was provided with were sampled at 43hz). Am I just being stupid and missing some obvious detail? Are my eyes really that f**king bad that I'm unable to select proper start and end points? ( ). What's going on?

Any and all help is greatly appreciated,

Thanks kvr

[ChimeraMan]

"6. The answer to that equation is 44,744.651blahblahblah. So, I round it to 44,744, and then resample the waveform to a sample rate of 44,744hz."

Just a thought (coz i don't really know shit all about this) but;

Have you tried rounding UP to 44745? I noticed you're decimals were higher than .500. Wouldn't you normally have to round up in this instance. Plus my guess would be you'll be more likely to get pops than gaps if it didn't work out perfect. Just a thought.

[Benedict]

I'd love to offer some constructive help but I can't count past Boom, Cha, Boom, Cha without having to start again...

Best of luck with that Totc. I hate all that maths stuff, does my head in - and I'm starting to learn C++

Maybe that the original samples were an odd rate.

I'd suggest that if you wanted the samples to represent that ARP then you may well need more than one cycle to allow for 'analog variation' but I'm sure you have considered that.

[TotcProductions]

"6. The answer to that equation is 44,744.651blahblahblah. So, I round it to 44,744, and then resample the waveform to a sample rate of 44,744hz."

Just a thought (coz i don't really know shit all about this) but;

Have you tried rounding UP to 44745? I noticed you're decimals were higher than .500. Wouldn't you normally have to round up in this instance. Plus my guess would be you'll be more likely to get pops than gaps if it didn't work out perfect. Just a thought.

Tried that...I've noticed that when I round up, if the sample-length is higher than 1,024 (like 2,197) then I end up with cycle that's 1,025 samples in length. So I round down, as it results in a cycle of exactly 1,024 samples every time. However, if the sample-length is less than 1,024 (for example: 1,003), then I need to round up...otherwise I end up with a cycle that's 1,023 samples in length.

confusing

@Benedict: Boom, cha, boom, cha is about where I am at the moment.

Peace

[bobsled]

Sorry...best I can do is bump this to the top o' the pile...

good luck and all that

[Teksonik]

All I can say is..........Magic! You know Eye of Newt,Wing of Bat, that sort of thing. Ever since I tried to loop my first sample in an Ensoniq Mirage with a two digit display in Hexadecimal I've been of the opinion that sample manipulation is the work of Wizards,Sorcerers, and Shamans.

[Ildon]

All I can say is..........Magic! You know Eye of Newt,Wing of Bat, that sort of thing. Ever since I tried to loop my first sample in an Ensoniq Mirage with a two digit display in Hexadecimal I've been of the opinion that sample manipulation is the work of Wizards,Sorcerers, and Shamans.

Let's hope Shamann shows up then.

[TotcProductions]

All I can say is..........Magic! You know Eye of Newt,Wing of Bat, that sort of thing. Ever since I tried to loop my first sample in an Ensoniq Mirage with a two digit display in Hexadecimal I've been of the opinion that sample manipulation is the work of Wizards,Sorcerers, and Shamans.

It very well may be in this case. Usually I'm good at it...but the need to keep everything at 1,024 samples, and using material sampled at 43hz (F0 I think), it's making it really tough. I'm not even sure it's possible. Gonna keep trying anyway though.

peace

[funkadil]

I think it is possible. Check out the wavetable of surge. I remember the developer on the forums there explaining how to do it in soundforge. He had a similar method. I gotta go though so sorry! no link!

[Mutant]

Maybe you should try first resample the whole wave and then find and cut out best single cycle ?

[duncanparsons]

window it to ensure proper zero crossings.

do a tiny fade in on the original sample (more room in 2k+samples than 1k..), and a tiny fade out to -INF at the end. Make the fade in/out about 5-10 samples long, so that when scaled they are only 3-5 samples long. When looped the fade out/in will only be 6-10 samples which shouldn't be a problem (<1% deviation form original waveform).

I could be wrong, but it's worth a go

HTH
DSP

[Paulie Phonick]

Sorta what Mutant said - cut the wave to start at a zero crossing and have a bunch of cycles to choose from, resample then cut the best individual cycle. Fades can help as well, as duncanparsons said, but cross-blending the beginning and the end could produce better results.

[RTaylor]

Audacity?

[the_nihilist]

Audacity has a keyboard shortcut that automatically picks zero crossings.

EDIT: Tuning, trimming, and looping samples also makes me want to kill myself. ESPECIALLY tuning single cycle waveforms. You edit a single cycle sine sample at 440 hz, but as soon as you put in one interesting sounding peak the whole thing goes completely out of tune. GRRR. Trying to do any more than a dozen at one time? I'd beat myself to death with the keyboard.

[TotcProductions]

Thanx guys. I'll give it a shot and see what happens.