emusonacid!

Hi there...

I just wanted to try using the Emu as a synth.

When I load up samples and try to loop them, the sample start is always offset by 2 samples and 4 if I turn the encoder. Same happens to the End parameter.
Whats wrong?! How do you loop your single cycle waveforms?

cheers

The same issue occurred to me when I tried to create a single-cycle square wave for PWM.
The sample-points in the display are also not very accurate.
You cannot rely on the visual information when editing samples. There is also some kind of offset.

You have to use an external sample processing software and import the sample afterwards.

I don't know why people always use the C-note for sampling & looping.
The best note from a mathematical point of view is an 'A' with a base-tune of 441Hz when sampling at 44.1Khz.
When sampling at 48Khz you have to detune the A-note so that it has a tuning of 480Hz.
When looping you have an integer value for the loop-length.
At a last step you have to correct the sampling rate to 44000Hz with your sampling processing software. This is possible in Sound Forge.

sounds fairly complex tbh. So the EMU samplers are not capable of looping from the very beginning to the very end of a sample?
I then should add some silence to the beginning and end of a sample, which obviously isn't a good work around as it would change the beginning of it. Damn...

edit: Found a workaround from the Emu sampler users group on facebook:

Petróci Gábor: "Or copy the sample and paste after it, and then you can create loop"

I think with 128 mb of ram this should be a no brainer :) I will try it right now.

Quote: "Or copy the sample and paste after it, and then you can create loop"

As I said: "you cannot rely on the visual information"
If you want to have a sample-accurate editing you have to use an external sample editor.
The inaccuracy is about 4 to 10 samples when using the graphic display of the EMU for choosing the right settings.

nah, that's not true. I just tried it and did the exact same thing on the computer, it sounds the same. There is no offset, it just won't let you go to the very beginning or the very end.

Well, your truth is not my experience.
When you say "it sounds the same" it does not mean that this is exactly the same.
As far as I remember did I cross-check the result of the edited sample by sampling it via a Digital I/O into Sound Forge and counting the samples.

It's OK if you are satisfied with your result.

Yeh the sample loop start/end point is always a few ticks short of the actual sample length, so if the sample you're working on is precisely one cycle long then your loop will be at the wrong pitch and contain digital glitchiness because the loop start/end points transitions are not aligned.
TIPS
Use longer samples.
Paste it onto the beginning of a very short silent sample.
Copy/paste it onto its own end point.

Rob you are dead right about the 441 = A sample being best and not C but bad habits die hard and I do it all the time,it's generally OK with longer loops but it can create a few headaches when making super short loops as the tuning goes out, but you can use the onboard sample rate converter to rectify this.

Sorry mosrob no offence meant.

I compared the results to the file on the computer and the held pitch and had a constant phase offset between each other. So I guess I get sample accurate looping?

ok looping of single cycle samples for emu samplers

now emu rules you cant loop on the first sample or last sample you cant do fractional loops

this has been the same since the EII and has been carried on for backwards compatibility

now a loop need to have a minimum of 3samples from sample start to loop start

now a loop needs to have several samples after loop end for its interpolation engine

sf2 also uses this same type of extra material after loop end and its also needed for crossfade looping as you cant perform a crossfade loop with no data after loop end

when truncating emu samplers always leave extra samples after the loopend point

and is why loops created with ensoniqs or akai can click with emu samplers as emu on conversion always allows the loop to be extra samples after loop end aand on 1 sample loop increments unlike ensoniq and akai allow fraction of 1 sample

if you want to use single cycle loops with emu you need to create a single cycle with 3 loops of a single cycle then find the perfect loop with the middle cycle when you have a perfect loop you can then truncate this and emu will leave extra samples after loop end for its interpolation engine

to stop aliasing its better to down sample for transpose than uptranspose so because samples aliase at the upper notes if you sample root and then use your root as top key it wont aliase because your not
uptransposing if you put it through a analyzer you can see that after 3 or 4 semi tones up you get overtones where the frequencies above NyQuil start to fold back as lower tones

but if you just downtranspose it does not happen as you actuall slowing down the samples

jhulk
Nicely put.
A down sampled copy of a sound works really well when used on the higher range of a preset while the original sample is keymapped onto the mid and lower keys
It is good when you need extra notes higher up the keyboard without digital artifacts.

you get me wrong with wavetables that i create and generate with mathlab

i have a sample for every octave from c1 to c7

then unlike wavetable osc that then use speed and interpolation as they uptranspose them which causes aliasing as your doubling the frequency of the first wavetable

as if there are any partials just under nyquill it will when doubled one octave higher by an octave will fold over

and why they use bandlimiting techniques but the problem with that is lots of computation and through the process the upper keys drop harmonics and you can tell that

with samplers they use oversampling chips x64 delta sigma and brickwall filters that stop anything above 18to20khz on older samplers they were set to around 12khz-16khz

now because i make the wavetables or any sample i will sample root for each multisample then i set the keymaps so root is top key

on my single cycle loops with half cycle for pwm i set them like this

root c1 top key c1 low key c-1
rootc2 top key c2 low key c#1
root c3 top key c3 low key c#2

and on and on until i get a full keymap from c-1 to c7 this will give you an 88 note keyboard range if center is c3 or c4

now because the my top key is played at root frequency of the sample as long as the spectral content is 0-20khz at its played at 44.1khz which gives me a frequency content for nyquill of 22.05 khz

then there will be no aliasing now because im not uptransposing so its not going to double i wont get any aliasing

but it will only downsample the root when down transposing as it reads it slower as long as its not downtransposed to much the transient of the wavetable will still punch

but if you down transpose to much the sample slows down and the transient is lost

just listen to a piano sample transient when transposed 2 to 3 octaves below root or a drumloop and you can here that the transients loose there punch

this is how to make sure that the wavetables never alias

here is a test file for you its the virusmod wavetable set from the z3ta+ synth giving you the 60 waveforms of the wavetables of the virus c

this is one of the wavetables set for looping a single cycle and an extra half a cycle for pwm with the eultra samplers

this is just a keymap of samples as i explained no synth params

i saved it as a b4e floppy bank its 23k per multisample so can get a lot on one floppy disk

should be able to fit all 60 virus wavetables on one 1.44mb disk

heres the first one https://www.dropbox.com/s/ff80amk4enl2v ... 0files.zip

oh wow, that looks nice thanks!
I have no trouble getting stuff into the machine, I have a cf card reader :)

Yes, this looks good.
I would love them all please

how many i have thousands as i create them for many synths

in various sample rates ill give you a hint of what i have available to render these files

all 2000 waves from the ppg2.3
all the prophetvs plus loads of my personal single cycle 8bit waves
all the k1/k2/k3/k4/spectra10
all the uwavext
all the esq1
all the jd800 single cycle waveforms theirs about 40 of them
all the single cycles from the wavestation /t1/x5dr/dwgs 64 waves so thats the full dwgs from the korgdss1 dw8000 and 6000
lots of analog waveforms from lots of synths

lots of string synthesizer waveforms and lots of russian string synth samples

loads and loads of generated transwaves 128 x single cycles in one wavetable multisample files

on the x2 and x3 emulator x you can do transwave synthesis never tried it on the ultra but i can do the full tables so that if any one wants to try they can but doing it on the x2/x3 it was not as fast as on the ensoniqs so dont know if the ultras processor will be fast enough to impliment it as the ensoniq uses a custom asic voicechip for it

casiocz waveforms but also as morphing tables

http://www.youtube.com/watch?v=A43RzNlEXuA here a video of a user using my pcm in a korg t1 1990 synth this set is just 1mb in size

https://soundcloud.com/jhulk

here are some doodles showing synthesis and pcm examples and multi velocity layered instruments for the asr10

and some dsp stuff im working on