The attached patch "should" produce a bandlimited square wave at 523.2 Hz (using the sawtooth/delay/subtract trick). But the oscilloscope array shows a smaller-than-expected pulse width.
I think my math is right: delay time in ms = 1000 / freq * pulse width. But the delay time is slightly off from where it should be.
I'm pretty sure, if I do the same with SuperCollider's Saw and DelayC units, the pulse width renders as expected. So I think there is some inaccuracy with delread4~'s delay time. Or something else?