So I have a 100W M90 MH ballast for powering a 100W H38 MV lamp, but they have different specs so the MV lamp is overdriven. I have compensated for this slightly by putting ceramic resistors in series with the primary but they get real hot! So I started experimenting with capacitors in series with hte lamp on the secondary (CWA style). Well many capacitor tries later I discovered that this is either impossible or I am just not doing it right.

Starting with .22uF in series, I get less than a mA through the lamp. Similar with 1uF, but 2.2uF yielded a few mA. Then trying my next larger value (3.8uF), it drew around .5A with a horrible horrible buzzing from the ballast. It was drawing reasonably current on the primary as well.

So if the primary and secondary current is so tame, why is the noise it's making sound so wrong? I tries it again without any capacitor and it lit up just fine (overdriven as usual but quiet).

Is there any way I can do this without winding my own choke or using resistors? I can wind my own choke and it would probably work but I would prefer to do anything but that first.

Well, it is known for very very long time that that using capacitive ballast on a line frequency is bad. Just in some specific circumstances it is tolerable.

Use an oscilloscope with a high voltage divider and a current probe or shunt to investigate. You are likely getting high current waveform distortions and may be some part of the transformer core is saturating.

Yeah I know, maybe it could be better to overdrive the lamp a little than it would be to drive it with pulsy capacitor currents. At this point I may just deal with it

I am not sure about American CWA practice, but in the world of fluorescents, leading ballasts roughly have the capacitor impedance about twice of the choke impedance. That way, choke inductance cancels about 1/2 of the capacitor impedance, and you end up with a leading ballast with about the same impedance as the choke intended for lagging circuit. Important: this is true only for the 1'st harmonic.

So you likely can try to calculate a capacitor having 2X of the nominal impedance of your M90 ballast, and then decrease it just slightly to adjust to the proper current. This is of course somewhat simplified by not accounting for the choke resistance.