Phase mod with PWM

[Martin Czech]

:::Now, PWM could be done with two 50% duty cycle oscillators, slighly
:::detuned and logical ANDed. If the two waves are in phase  you get 50%
:::duty cycle, out of phase means 0%. A kind of nonlinear beating (of course
:::this would be stupid, this is only an example).
:::
:::In order to modulate you have to phase shift one of the
:::waves. Unfortunately frequency is change of phase, or :
:::
:::F=d/dt P
:::
:::So , changing the phase implies changing frequency at least of one of
:::the components, and this is audible.
:::
:::So , from this reasoning I'd say: No, you can't get one without the other.
:::

Uuaaaah, sometimes the "deliver" knob should be locked...

F=d/dt P is true. One could exploit the repetitive nature of the square
waves and shift one of them always forward: P(t)=C*t+w*t.

w being omega, the normal frequency,
and C beeing your phase shift constant

Frequency will then be constant: F=C+w, this is detuning, but constant.

Your complaint was about a phase modulation like P(t)=C*sin(w1*t)+w*t

and this will give the annoying F=w1*C*cos(w1*t)+w,

this means frequency up and down, the faster (w1), the worse.

So, in other words, you should test, if the PWM done with two detuned
square waves (logical AND) is free from the unpleasant effects.  If this
is the case (which math can not show allone, perception is perception, and
not math, and I took a big step turning the logical AND into something
linear, which is of course not quite correct) one should change the
usual up-and-down modulation scheme into a cyclical up-only, or down-only.

I think one has to look at the scope to see the difference between both
methods, then the circuit should be clear. I'm too lazy now to draw
these waves...


m.c.