Analouge vs VA vs Sample Based
2006-05-17 by Alien
The following is a reprint of a write I recently made to an associate
of mine with minor edits. There's a section of the letter that deals
with TB-303 emulation via the JV synth engine and as the principles
described can be applied to a rather large variety of sampler/sample
based synths (in cluding the P2X), I thought others may find the
technique covered relevant as well. I will note that the given
technique assumes the user is starting from a fairly blank base and
not with parameter values thrown all over the place; further, the
foundation of the technique can be taken farther to squeeze out other
"analouge" type sounds from sample based synths.
Regards,
Al
Synth BS
In regard to your analouge comments, I've always been a big lover of
analouge (probably largely due to me getting my start in sythesis with
analouge pieces), and in many ways, I do think it carries some
qualities that Rompler/sampler based affairs and VA's can't
touch.....but in many regards, I think those qualities are misunderstood.
For one, most analouge synths were based on subtractive synthesis; but
to that same end, most Rompler/sample based synths were/are also
subtractive synthesizers. In essence, in the analouge/digital switch,
the main synthesis method employed NEVER got touched. Oscillator/Sound
sources and filters (voltage controled to digital) were all that got
affected. In this regard, I think it would make sense for one to want
to expand on the amount of waveshapes available for an oscillator to
employ, and in this regard, the progression to a sample based approach
was a logical and sensical move as it allowed for the general
waveshapes generated by VCOs to still be employed while also allowing
for an entire entourage of waveforms difficult to achive via VCO to
also be brought into the fray. In this regard, the "Oscillator" of a
typical rompler synth BLOWS away VCOs/DCOs in the area of overall
flexibility. I truly can understand why you have a primarily sample
based basis in this regard.
If you are getting into a lot of waveshaping, I don't feel that VA's
quite match the quality of TRUE oscillator sync at this juncture.
Also, the sound of analouge filters and plain digital (or physical
modelling analouge/VA) still sound different; I would not say either
sounds BETTER or WORST than the other, simply different.
To this end, I actually ran an experiment recently, testing out the
qualities of sound between different Oscillator/Sound Source types;
filters etc. To begin the experiment, I made a simple acid line on the
777 and gave note to it's timbraic qualities. I then did
individual/seperate routings of a sqaure wave and saw wave from my 777
into the SuperNova II and utilized the 18db filter on the SNII (had
been using the same type of filter on the 777). I set up a filter
envelope on the SNII to act like the filter on the 777, ran the same
sequence information into the SNII, and then proceeded to attempt to
get the same sound out of the SNII with it utilizing a 777 oscillator.
Aside from the SNII's filter having a wider range, issues still came
up and ultimately it was not possible to replicate the 777 sound
despite filter poles and filters EG setups that mirrored each other. I
then reversed the scheme, utilizing SNII oscillators through the 777
filter. I ended up having to utilize 2 SNII pulses to replicate the
777 square oscillator sound but after that, no noticable difference
could be made out (the SNII saw took some fiddling, but two
oscillators did not need to be used to replicate a 777 saw). In
essence, the major difference between the two proved to be the filter.
(I will further note that it was not possible to get the 777
oscillators to replicate the sound of SNII oscillators AT ALL).
I then replicated the experiment once more, this time utilizing the
777 and the XL-7. There proved to be saw and square waveforms within
the XL that matched the sound of the 777 oscillators pretty much dead
on, and from here on, once again the notable difference came in the
form of the sound of filters.
I actually plan on taking this experiment a bit further, but swapping
out the 777 with the MFB Synth II (which uses a 24db filter), I'm
fairly confident that the sound of filters will prove to be the
difference once more.
There was another aspect of my experimentation that I went through
which was simply replicating patches on different synths. To begin, I
developed a patch on the SNII and simply made sure to ignore aspects
of it's architecture that differed from Roland's JV synth engine and
then sought out to replicate it on the 505 (which has a JV base).
Timbres of STRIKINGLY similar quality were able to be developed in
this regard. So similar that ultimately, unless one had each synth
side by side, I highly doubt one would have been able to differentiate
between which synth had actually produced the timbre (espically within
a mix of things).
I'll note that in the area of patch replication, I had carried out a
similar experiment years ago with Native Instruments Generator
(precursor to Reaktor) and my Juno 106 where Generator proved to
capable of matching the 106 pretty much dead on. I later would put
Reaktor to a MS20 test where Reaktor would prove capable of not exact,
but pretty darn close to perfect emulations.
I further plan to extend this aspect of my experimentation some time
in the future.
In any regard, if my current hypothesis proves correct, analouge would
seem to have a big edge simply in regard to waveshaping methods,
oscillator sync and continously variable wave shapes, in particular.
Outside of this, filter sound difference would be the divide (on that
note, just about all filters tend to have small degrees of sound
difference IMO though). Further, in regard to waveshaping, in the area
of oscillator sync, it's also worthy of note that the amount of
analouge synths out there with OSC sync capabilites is not really all
that high (the amount of analouge's OR VA's for that matter with
continously variable wave selection is another area of low selection).
The Supernova series arguably becomes unique in it's OSC sync
implementation due to it's Virutal slave implementation, which really
allows it to dive into areas actual analouge could only hope to
achieve.....one considers the amount of polyphony VAs offer WITH
oscillator sync, the fact that a small line of VAs have a virtual sync
feature, and well, it can be argued that despite the sound quality of
the sync feature, there is something VAs have to offer in sync areas
that analouge does not.....in this regard, I'd fall just short of
saying actual analouge based sync is BETTER and say again, it's simply
different.
Ultimately, the analouge/VA/Sample Based differences I think are much
smaller than many people make them out to be and ultimately reside
more in people not truly understanding synthesis theories or having a
lack of synth programming knowledge than in a giant difference in
sound. I'm not sure if you've carried out experiments similar to the
ones I've recently dove into, but if you havent', I'd highly suggest
carrying out at least a patch replication test utilizing your romplers
against VA softsynths you may have.
Finally, something Romplers/Samplers have over analouge's and VA's
lies in the fact their architectures typically allow for a synth stack
to be created immediately. That is to say, allowing 4 (or MORE,
sometimes MUCH MORE) waveforms, each with their own given synth
architecture, to be utilized in one patch. Thus, even when limited to
waveforms common to analouge synths, a patch created on a
Rompler/Sampler could necessitate multiple synths, a high amount of
voice eat up, or a fair amount of timbral eat up, to accomplish on an
actual analouge or VA synths.
I am by no means out to knock analouge synths, as I love them,
currently own my own share of them (with even more I've owned in the
past), and don't ever forsee myself ridding myself of them (the 777 is
my favorite synth bar none). BUT, I equally love Romplers/sample based
modules and my sampler use is pretty frequent. I've found both to have
their own unique brand of strengths and weaknesses.....and in general,
actually give a lot more credit to sample based approaches in regard
to overall flexibility than to analouge.
For example, one may be shooting for the classic "acid" sound. Someone
may say the JV engine (or 505 to be specific) is horrible for acid,
but I couldn't disagree more. The acid sound simply necessitates a
basic saw or square wave, ONLY zero or positive EG effects on the
filter, that the EGs have zero placed settings on the attack, sustain,
and release (decay can be varied), and finally, some possible playing
around with portamento values, after all of this, the option is left
open for whether or note polyphonic voices are allowed (unlike on the
303), if detuned oscillators/waveforms will be implemented (unlike on
the 303), and ultimately, figuring out ways to compensate for the JVs
lack of a 7pole/18db filter.....the filter problem often being able to
be compensated for simply through keeping EG filter placement a fair
deal above 0 as standard and modifying the sustain level to be kept
somewhere above 0 as a standard rather than at 0. The 505/JV now
becomes an acid device with all the same 303 controls (minus accent)
at hand AND MORE (just need to sequence in 303 stylings from here on).....
concerning the shortcoming in the accent area, I'll note that even
here there are ways for approximation though. If Velocity controll of
the amp is nulled (as is the case with the 303) and velocity levels of
resonance are given reasonable values, though it would still not be a
total emulation of a 303 accent, it would in fact allow for a partial
emulation of the 303 accent....so long as all "unaccented" values were
of the same velocity, "accented" values were of all the same velocity,
and that velocity levels of "unaccented" notes was a fair amount below
planned "accented" values, any time an accent velocity value was
placed a resonance spike would occur and thus emulate some of the 303
accent properties.
There is a lot of good to analouge, VCOs and VCFs in particular (I'm
not a particularly big fan of DCOs outside of their stability). There
are definately some annoyances that come along with it as well (though
most newer VCOs tend to be fairly stable, be prepared to dig out your
voltage gun every now and then for some of the older ones). BUT, the
amount of sound quality difference between analouge and sample based
synths generally falls far short of what many make it out to be (IMO).
My 505 is a JV synth base (predcessor to your Fantoms XV base), and of
course the CS is a P2500....if you'd like, I'd be willing to make some
quick recordings of the JV and or Pseries being put to test against
Analouge and/or VA for you in regard to patch replication just to sort
of demonstrate how slight the differences can be at times. If you
think you might find such usefull, let me know and I'll make some up.
Regards,
Al