I had first convinced myself that ceramic caps didn't work when I restored my very first electrostatic set about 15 years ago. It was a Motorola VT-71, and maybe the results would actually be different on that set (maybe I'll try it sometime; I still have the set). But on this Sentinel 400TV, the results are as shown below.
The pictures show the raster obtained with a test pattern so that scan nonlinearity can be seen fairly easily. I had to readjust the vertical linearity and vertical height controls quite a bit between the various capacitor choices, but the bottom line is that I was able to get a pretty decent raster scan with any of them within the adjustment range of the controls.
1. Raster with the array of four .005 uF 3 kV gold disk caps (equivalent to .005 6 kV) on the vertical plates. This required significant correction of vertical linearity, but looks OK once adjusted. These are in fact the exact same capacitors I had installed in my VT-71 15 years ago that I thought showed bad behavior in that set!
2. Raster with the single .005 uF 4 kV red disk caps on the vertical plates. This required much less correction of vertical linearity.
3. Raster with ASC .0047 uF 6 kV tubular caps on the vertical plates. This was the original case for which the vertical linearity had been adjusted.
4. Raster with large .015 uF 6.5 kV blue laser power supply caps on the vertical plates. This required a little vertical linearity correction in the opposite direction, but not much.
Note that they actually all look pretty good! The only difference is some modest improvement in the linearity at the very top and bottom of the raster when the tubular caps (especially the large value .015 uF) are used.
I also tried to see if there was a difference in "bunching" behavior at high contrast with the various caps. Bottom line is that this set, for whatever reason, does not show bunching until the contrast and brightness are high enough to show other problems with the picture (a little blooming, focus suffering), and there was NO notable difference in this behavior between the various cap choices above.
So then I got to wondering...
Was the problem actually with the horizontal deflection when using ceramic caps? So I tried it. For these experiments, I left the big blue laser caps on the vertical and swapped out the ASC .001 uF 6 kV tubulars that were on the horizontal plates with the following:
5. Raster with two .001 uF 3 kV gold ceramic caps (equivalent to .0005 uF 6 kV) on the horizontal plates - note this is less capacitance than originally there. Overall raster width just a little decreased from what I had with the tubulars, but no other obvious problem.
6. Raster with the array of four .005 uF 3 kV gold ceramic caps (equivalent to .005 uF 6 kV) on the horizontal plates. Same width as I had with tubulars, and no obvious problems.
So there.
What does it all mean? Do different sets and different CRTs behave differently with different kinds of coupling caps?
In any case, we certainly cannot conclude after this experiment that "ceramic caps always cause problems when used as deflection coupling capacitors in electrostatic deflection sets."
On the other hand, I'm not about to start recommending people to use ceramic caps for this purpose, or to start using them routinely myself in future restorations. Will probably do more experiments on other sets at some point.
Perhaps the key point is that the ceramic caps needed significant correction of vertical linearity, which on the Sentinel set is a provided adjustment. I believe quite a few other electrostatic sets don't have a linearity adjustment, so perhaps without the capability to adjust linearity, the ceramics result in poor vertical linearity. Maybe that's the story.
Sorry for the confusion, everyone!