"We took the drums into loads of different rooms in the studio, whacked them and recorded them onto DAT". comes the explanation. "We sampled the best sounds into Mike Johnson's Greengate and then we recorded the drums that were on tape into the Master Tracks sequencing program on a Mac using a Syco PSP. It was the first time I've managed to get one of those to work. I've tried triggering it off tape before but never with any success. I always got a delay, but once it's recorded onto a sequencer track and you know what that delay is, you can move everything back so it all lines up. We didn't quantise it that much, we just kept the feel and replaced it with our wazzo drum sounds."

Occasional drum patterns are also worked out on an Apple Macintosh with the aid of Intelligent Music's Upbeat software. Although primarily designed for drum programming, Morris also uses it for writing sequences - especially basslines.

"I love the fact that it's so interactive", he enthuses. "It's like a massive TR808 really, because everything you put into it gets automatically quantised, and there's no way to turn it off. I generally save sequences from it as a MIDI File and load it into Master Tracks for humanisation. The Mac has become increasingly important to us as a songwriting tool. Not quite 50% of this album was written through jamming for ages and picking up on bits we wanted to work on. The rest started out as ideas that someone had got which we then all worked on with the Mac. The great thing about a program like Upbeat is that it's dead easy for anybody to knock out an ace drum riff. It's equally as easy to come up with a wazzing bassline, whereas I couldn't sit down in a practice room and come out with an ace keyboard riff because I'm not really a very good keyboard player. I'm not even that good a drummer I suppose, now you come to mention it."

Also on the higher notes you're suddenly left with just a sine wave. That's not how a modelled filter would behave.

My guess is that it's just a set of saw wavetables for Reso = 0 and Reso = 127.
Reso values in between are just crossfaded.

Square is generated by adding saw + delayed inverted saw. That would also explain, why the saw is one octave up and there is saw "pwm" (which is just 2 delayed saws).

And it will only pick waveforms in the wavetable, that won't alias too much (unless you use the pitchbender, it obviously doesn't have the processing power to take that into account).

Not if you want to recreate a traditional filter, but the D50 is different anyway.
I've attached a file. First half are all 30 Reso values straight out of the D50, second half is me crossfading between reso 0 and reso 30. Sounds pretty much the same to me.

And the saw PWM thing? With my theory, it would make sense, since the difference between saw and square would be just flipping the polarity of the delayed saw.

Have you ever looked inside the ROM?

Really appreciate your input and your blog btw