Access Virus classical osc model

[dr660]

Dear KVR people,

I am interested in reverse engineering of Access Virus classical oscillator. It is just like a sport for me.

I discovered that actually it is not alias-free at all. Although the saw and square waveforms are not ideal, so it helps to reduce aliasing a little bit. I attach the sin-saw-sqr morphing (automated by MIDI) file. Don't blame me for 96 kHz.
http://efshare.com/?s=EHWYB5

There are two different morphing stages. One is sin to saw, and another is saw to sqr (or saw to triangle).
Observing the way the waveform changes, I thought that it is closer to PD (Phase Distortion) synthesis with interpolation between points during morphing. I include my PD table for sin-saw stage. The sin of course produced with a linear phase and a Virus saw is somewhat on the picture. The sharp rise at the upper part is supposed to model the sine-like lower edge of the saw.

So far I got the result pictured on the screen I also attach. It is good but a bit crispy anyway.

My question is could you suggest any ideas of how to improve the model, or maybe point me to some other approach. Please, understand me correctly, I do not want to sample it or WT it. The goal is to find the algorithm.

Thank you!

[dr660]

Also there is another link to audio file with morphing in case the first will not work
http://www.filedropper.com/virusc