Re: [Mellotronists] sampled (was "tron hate")
2005-04-28 by ferrograph@aol.com
<< I've always thought that the result of an analog summing buss is much more 'natural' sounding than digital. My guess is that the upper harmonics are more fairly represented in the analog sum. >> ah... time to chip in, I think, given the number of samplers & romplers I've used in the same room or on the same stage as my old 400 & it's nine frames. emu have had a few goes at capturing the mellotron, most recently just before they stopped making anything useful, in the proteus "vintage keys" rom. I think it's 32Mb total, but there're a zillion synths, e-pianos & clavs on there aswell. obviously, it sucks, but not quite as badly as their earlier efforts. I've never used their cd, or the pinder, though I do have the hilarious propellor island disc somewhere. I also have a few alesis boxes- quadrasynths.... they had a "vintage" rom card too. 8Mb. must've seemed like a lot at one point, back in the late 80s perhaps. anyway, what with this parlous state of affairs, the practical problems & general unwillingness to schlepp the precious 400 itself, & a bad experience with a hard-drive right before a gig, I decided to go down the flash-ram route & sample my own frames for live-use. first, the alesis. actually, no, forget that. this sort of reminiscing is contra-indicated by modern psycho-therapy. suffice to say that even one 'tron sound can't be fairly represented by the 8Mb card. I gave it my best shot. the emu, on the other hand.... the proteus series can accommodate up to 4 32Mb flash ram sticks, but you need one of their top-of-the-range samplers to make them with. I have done this, with the e6400 ultra sitting right next to 1098 & just a guitar lead between them. I have the sounds from 6 frames on one stick, this by dint of sampling every second or third note & then downsampling to reduce the required memory. obviously, you can downsample some sounds more than others, so I'd a-b each sample to see what I could get away with. I doctored one patch to have clarinet & bass clarinet combined, with the split almost inaudible. my church organ splits into the same tapes running at half-speed. & so on. no looping, so some of them run out at less than 8 seconds. once the basic patches are assembled in the sampler, you blow them onto the ram stick & put this into a proteus box. this is where the fun starts. the proteus line are very complex little romplers &, with lots of programming, are capable of great things. sadly, most users just dialled up presets & never dug deeper. I created a basic patch for mellotron sounds, & have found that the same patch makes other samples sound a bit "tronny" too. it uses one layer, no chorussing, no filtering (except for my "phaedra" tron-through-lfo'd-vcf patch) & a gated vca. once this generic patch is built, I just load a different set of samples into it, rename it & move onto the next. so the emu just plays the sample exactly as it was recorded? no. it sounds a bit lifeless like that, as has been noted, & especially so when you play a chord. hmm.... how to fix this? well, without hooking up my oscilloscope, I can confirm that playing a chord on the real thing is very different than playing the same notes back all at once from a sampler or rompler. it's louder. maybe the sampler has some logic that reduces the level depending how many voices are active- this'd make sense, since these things are expected to deliver 64 or 128 synth channels through a single stereo output; that's a hell of a dynamic range. each additional note raises the level by up to 3dB on the tron..... so paradoxically, we have to ride the level more with the samples than with the real thing. some sounds are just plain shrill if you play chords; the pre-amp in my 400 is certainly being overdriven. figures, though- oboes are like that in real life too, & they're not supposed to play chords anyway. single notes seem to sound the same, more-or-less, from the sampler or the real thing. but as soon as you play two keys on the tron, all sorts of new variables come into play... vibrations up & down the capstan from uneven pinch rollers? extra load on the motor? I spent a few hours investigating this, then set to work on the proteus. the patch I've ended up with uses small (really small) amounts of pink noise & other random control sources to vary pitch & amplitude. I also modulate the start-point of the sample to simulate incomplete rewinds using velocity control. aftertouch (channel pressure) is used to lower the pitch by a tiny amount- you really have to lean on an sms equipped 400 to make it sound like a cmc 400, but it still happens, so it's in the patch. someone actually noticed this at a gig a few weeks back. I have even used a "key-random" mod source to switch between two different samples of the same note... I have some duplicate sounds amongst my frames, & wanted both versions to be represented. where I've used the same sample for two notes, one of them will have it's timbre & other attributes altered by the proteus. the whole patch also takes advantage of the user-tuning tables available in the proteus; you can make microtonal adjustments across the keyboard so the tron patch is perfectly out-of-tune. or not. I've been at this for 14 years now, with as many samplers/romplers, & this is as close as I've managed to get. we still use the real thing in the studio, but I can live with the sampled versions I've made in there or on stage. the same techniques will work to some degree on other boxes; I went with the proteus because I could get multiples of 32Mb into a 1U box with all this programming capability & not have to worry about carrying a hard-drive around. duncan/r.m.i./400 nr1098