View Single Post
Old 2nd October 2020 | Show parent
Gear Addict
creativespiral's Avatar
Originally Posted by cbmd ➡️
It has the same oscillator parts as the original. Natural oscillator drift is still there. You were speaking of the Vintage knob which is not a slop parameter.

We analyzed a bunch of different Prophet REV2s and REV3s and found that they all had a unique sound. We identified several key parameters which affect this and mapped them to the Vintage knob in varying amounts to simulate the older instruments. Again, it's not a randomized slop parameter like that on the Prophet 6 and OB-6. It's quite musical across its entire range and really gives the experience of using a vintage Prophet 5.
I think I've got some more specific info on this Vintage Knob. Pym and I discussed Voice Modeling last year on the Sequential Rev2 forum. I wrote up a bunch of info on the topic of per-voice variance measured in classic synths. Back in May 2019, Chris/Pym basically previewed that he would be working on this exact type of implementation - A single knob macro that swings through a table of offsets for various parameters.

You can read in the Seq forum for more details:

Slop vs. Stable Voice Modeling Tables
Slop and Osc Drift Circuits are basically complex LFOs that swing back and forth the frequency of each oscillator independently... there's always this drifting motion to frequencies. The difference is that Per-Voice Modeling uses stable offsets per voice, and per component value, rather than having everything in motion. There are a few advantages to modeling with tables of stable offsets:

1. There's less chance of weird dissonance where oscillator slop LFOs are in divergent phases and create extreme un-musical phasing.

2. For pads/strings and other patches with longer release times, you're not getting a bunch of extra "artificial motion" on top of the "natural phasing motion" of just having oscillators with stable offsets.

3. With voice modeling tables, the synth performance is more repeatable from session to session.

So, if this implementation has been done how I suspect, it will be a big improvement over previous "slop" implementation... and the single knob will make it easy to swing a large table of offset values to osc frequencies (and other parameters- ie: env timing, vcf cutoff precision, etc)

@ Pym , feel free to correct me if I'm mistaken, but based on our previous conversations, I'm guessing this is how this knob behaves on the new P5/P10 Rev4.

If anyone else is interesting in reading more on the topic of Voice Modeling, I wrote up an article here:

Finally, CONGRATS to @ Pym and @ cbmd and team! This synth is gonna make a lot of people happy!