To add a bit more to the topic of vertical linearity using different types of coupling capacitors, I did a few more experiments.
In this case, instead of readjusting the vertical linearity control to optimize the picture with each type of capacitor, I set the linearity control for best linearity with the .005 uF 6000 V ASC tubular caps which are generally thought to work well in these types of sets. Then, to simulate how other types of capacitors would behave in sets that don't have a linearity adjustment (e.g., Motorola VT-71 and many others), I substituted different capacitors and took pictures to make the linearity changes apparent.
Here are the five types of capacitors and their corresponding images:
1. ASC .005 uF @ 6 kV tubular cap (linearity adjusted for optimal picture with this type of capacitor):
2. Network of four gold ceramic caps with Z5U dielectric, .005 uF @ 3kV (series parallel configuration, with net capacitance of .005 uF @ 6 kV):
3. Red ceramic with Y5S dielectric, .005 uF @ 4 kV:
4. Two of the above red ceramic Y5S caps in parallel, for .01 uF @ 4 kV:
5. Blue .015 uF @ 6.5 kV large tubular capacitor from laser power supply:
Comments:
The really bad actor is the gold Z5U capacitor, with bad compression near the bottom of the image. Note that the red Y5S capacitor is much better, with just a little compression at the bottom, even though both are ceramic disk capacitors.
Looking at the specs for the two types of dielectrics, we have the following:
Z5U: +10 to +85 C, with capacitance variation of -56% to +22% over this range
Y5S: -30 to +85 C, with capacitance variation of -22% to +22% over this range
Now of course we aren't varying the temperature here, but it is also known that ceramic capacitors have ferroelectric (piezoelectric) dielectrics, whose capacitance changes with the applied voltage. That's the effect that's causing the problem here. If we assume that the capacitance change with voltage bears some resemblance to the capacitance change with temperature (which may or may not really be true), maybe the Y5S rating is a better choice here than Z5U. To really check, one would have to look at some other examples of these types of capacitors and compare.
It has been mentioned above that using a higher capacitance can provide better linearity when using ceramic capacitors. Note that when two of the red ceramic caps in parallel are used, for a .01 uF capacitance, the linearity problem actually goes the other way (compressed at top), suggesting that some capacitance between .005 and .01 uF would actually give the same performance as the original .005 uF tubular cap used here.
I don't have enough of the gold ceramic caps to see what would happen with adding extra caps for more capacitance with that particular type.
Finally, when an even larger capacitance is used (this time with another type of tubular cap), the linearity goes even farther in the other direction (compressed at top).
This all goes to show that the .005 uF tubular cap gives a somewhat nonlinear scan, but the set has some compensation in the circuitry to linearize things. This makes sense, since these caps were expensive and they could save money with a not-too-huge cap and a bit of compensation. Putting in too large a value today requires some modification of the linearity correction (either adjustment of linearity control, or modification of compensation circuitry if no control is provided).
So the bottom line is that ASC tubular 6 kV caps seem to do a good job of emulating the original tubular caps of the same value (based on experience in many different types of sets), but ceramic caps can indeed be used if some measures are taken to adjust the linearity. One simple approach is to increase the capacitance by the right amount.
There are some hints here that Y5S dielectric is better than Z5U, but that needs to be checked further before we can really recommend it with any confidence.