AD633 - trim or not?

[Grant Richter]

> But then I noticed:
> When I removed the ground from an INPUT, I had to re-trim.
> So I guess that when a chip input pin has series
> resistance, the bias current (or whatever it's called) out of the
> pin develops a voltage offset. So then you need to re-trim
> depending on the impedance of what's attached to the INPUT.
> I do remember that when I was using it as a ring-mod I
> would have to re-trim depending on what I connected to it.
> (A Fisher-Price Sing-Along must be AC coupled)
> 

The input stage has to be biased by a current from somewhere. This current
will develop a voltage across a resistor.

> Note 1: This is at the breadboard input, so the chip sees 100k
> to ground. Didn't know that it mattered at the time. When INPUT
> is grounded the chip sees 100K to ground. When open,
> it sees 200k to ground.

Bear in mind the real signals present. They may not be DC trimming related.
If there is any noise or hum on the input when not grounded, that will
appear as carrier feedthrough, but the signals are real and the circuit is
operating properly. Good ring mods should have 60db carrier suppression, but
only if the modulation input can be quieter than 60db itself. And -60db is
very hard to achieve on an unshielded breadboard.

Also, any DC component in the modulation waveform will cause a constant
carrier feedthrough (like a pulse wave with other than 50% duty cycle). This
is the purpose of AC coupling, so the DC waveform components are blocked. A
waveform has a DC component anytime the area above zero does not exactly
match the area below zero. It only takes a tiny amount of DC from any source
to produce objectionable carrier feedthrough. Consider a gate circuit on the
output to cutoff remaining carrier artifacts.