We know that modulator feedback is possible through the mod matrix. E.g. FM'ing an LFO with itself is useful for 'pulse width' (literally for the square wave) control of the waveform, but the feedback also alters the frequency and the linearity. Extra fun is to be had by cross modulating the two LFOs, giving unusual shapes and complex equilibria. It can be hard to control the frequency or to achieve a specific shape though.

As it happens, LFO 1 also has a phase control, and it turns out it's useful for more than setting the starting phase. It can be used to control the periodic symmetry of the waveform without altering its frequency, using feedback. If you set LFO 1 Phase as a mod target in one of the user assignable columns, and then use LFO 1 as a modulator for it, you get to control the periodic symmetry without changing the frequency and with optional control of linearity.

For a sine or triangle you can vary the waveform between a rising and falling ramp. A mod amount of -70 will give a rising ramp, 0 gives the original waveform, and +70 gives a falling ramp - all this with linearity retained (with the Phase knob at 0). For the square you get almost a PW control, but it also introduces a 0 value for part of the period, which means that the square takes on 3 levels through its period. The up and down ramps can be made to have chaos on the ramp and even 'stutter' at extreme settings, and the S&H and random waves are affected as well. The Phase knob itself can alter the results by introducing a nonlinearity, and perhaps even some (phase feedback) clipping or wrap-around (not sure what happens in some cases).

Some of this can be quite subtle, some less so - and in all cases the frequency stays the same. Of course there's also the opportunity to combine self FM and PM, even with cross modulation, but that yields results which I'm not going to attempt to describe...

Here's an image of waveforms for various settings for LFO 1 phase feedback (I apply the LFO modulation to the VCA in order to make usable scope shots in the form of amplitude contours):

1st column:

tri: mod = 0, phase = 0 (straight up triangle)

tri: mod = -55, phase = 0 (half way to an up ramp)

tri: mod = -55, phase = fully CW (nonlinearity introduced)

tri: mod = -55, phase = fully CCW

2nd column:

sin: mod = -80, phase = fully CCW

sin: mod = -55, phase = fully CCW

saw: mod = -65, phase = 0

sqr: mod = -55, phase = 0 (note the three levels(!))