[miniman82]

My opinion at this point is that the flyback is wound in such a way that it is somehow current limited. In other words once you get to a certain point, perhaps the core begins to saturate. Can't do anything about that, because messing with the core (or its gaps) changes the magnetic properties of the entire transformer in possibly undesireable ways. Lets remember the whole transformer is a precision engineered part, especially with regard to frequency. Mess up one thing in any of the windings, and it goes haywire.

Or maybe the primary simply has too much resistance, since the amount of current supplied is influenced by the resistance of the primary. More primary resistance means less current when HO tube conducts, which means a weaker magnetic field and corresponding flyback pulse when the stage suddenly cuts off. Obvisouly, there is nothing short of rewinding the flyback that can cure it if that's the problem either. Folsom is the only one I know of that's tried (re?)winding one, and the results were less than satisfactory.

The secondary might not have enough turns in it, but that seems unlikely since it does in fact produce the required voltage at low brightness. That tells me the problem has more to do with the supply side of things, since the problem doesn't manifest till more demand is placed on it. It's not that there's a lack of HV per se, 21-22kv would be just fine for a 21" CRT. The problem is lack of adequite current delivery, which is why HV sags when brightness approaches normal viewing intensity- the beam current required by the picture tube outstrips the flybacks ability to produce it, so the picture blooms and lacks brightness. The situation is analagous to power grid sagging during the hot summer months, when everyone is running their AC units. Too much demand, and line voltage sags. The fact that the horizontal scan decreases in width is another clue, pointing to a lack in magnetic flux to power the scan coils. Again, I can only think of too much primary resistance being the cause.

If I'm right, about the only thing that's likely to make things better is an increase in B+ voltage. Since you can't get a load with less resistance, you can always try driving a larger waveform into it in an attempt to extract more power on the other side. This would likely require vast changes to the LV power supply, including a switch to solid state rectifiers and juggling the dropping resistors to all circuits besides horizontal scanning. Careful attention needs to be paid to cathode current in this case, since a small change in input voltage can push current way overboard in a hurry.

A good test would be to put the set on a variac, with the brightness set so that the picture just starts to bloom. At this point, you should be able to read a slight drop in HV. Try increasing the AC input voltage to say 125 from 117, and see if HV tries to come up at all. If it does, there may be something to investigate.