NEC XG neckboard performance problem - SOLVED at last

That is a good idea to use a spice simulation to determine resistors. Are there special libraries for these opamps?

You will need more than feedback resistors to tame that chip. I have tried them on the Vim for the marquee and they oscillate a lot. you'll have to try a lot of pico farad cap values across the feed back res. and some other things.

I stopped working with them but might play this winter. I have like 20 of them. I think Jarmo played with them too and got them to work somewhat in my Marquee thread.

Look for the chips with the highest Flat GBW product. That spec is key above all others.

Nashou

Today I inspected and played with my (Marquee 9000) 2005-03p VIM. It clearly seems the PCB layout is not ideal for the task. It was unstable even with all original parts in it, but still the circuit could be measured.
This board had a peaking capacitor in the HFA's feedback that was tuned to 140MHz, after removing that and with the increasing of the gain, -regardless of the slight low freq. oscillation- I must say that it worked quite good even with the HFA1100 in it. I measured 2ns rise time at 3Vpp output which equals to about 230MHz (with the correction of my measuring rig). Also this was with the MMBT3906 transistors in it. With the BFT95s it became more unstable making it utterly useless. So the MMBT3906s went back and replaced the HFA1100 with CLC449, the LF oscillation remained about the same, and the rise time went down to about 1.8ns at 3Vpp (again with the compensation of my measuring rig) this euqlas to approximately 280MHz.
These are very good results to start with, because this bandwidth would directly reach the cascode stage of the VNB.

EDIT: the measured rise time contains the pulse generator's own limited rise time (about 1ns), which was not taken into account, so the circuit's actual bandwidth is higher than the indicated 280MHz.

I didn't make screenshots, and the boards are packed away now, but there was nothing special about it, it was just a nasty oscillation.

On the other hand on friday I've discussed this project with some other engineers at work, who generally know nothing about CRT, but they encouraged me to not stick to this original Electrohome circuit, rather tke advantage of the symmetrical output of the AD834, since the output stage is also symmetrical.
Initially I was skeptical but now I also see a great potential in this to and now explain what are the pros and cons.

PROS:
-Using symmetrical output would require two output opamps but only working at half gain, meaning the bandwidth (slew rate) effectively doubles in that stage.
-The noise introduced by the opamps would be less than in single output case.

CONS:
-Maintaining symmetrical output costs ($$$) one more opamp for each circuit.
-New inserting point would be needed for the bright/bias+clamping circuit.

That is indeed true. So I made a small excell worksheet "simulating" the effect of cascaded stages with individual bandwidth limit.
As it can be seen you have to input what is the frequency of the signal in which you are interested, then there comes the bandwidth limitation is the DAC (let it be a VGA card or Moome card or HDFury), and there is 3 more custom stages, this mainly reflects the construction of the this XG project: stage 1 would be the "GAIN CTL" board, Stage 2 is the contrast controll daughter board and stage 3 reflects to the endstage (VNB). Of course exact numbers are not know but everyone can play with different numbers to see how is the performance related to different bandwidth distributions.

This calculator assumes simple 1st order filters at each stage at the given frequencies, resulting in a 4th order roll off in the end. The amplitude is directly calculated from the transfer functions instead of rise time theorem (which I commonly use).

Hint: if you want to bypass one stage from the calculation just enter some crazy big number in, if you want to add more stage, feel free, you can easily find out how the end result is calculated.

So it seems I finshed the PCB design too. Actually I designed two different PCBs, one with the already mentioned BFT95 and THS3201, and the other is more faithfull to the original design, with CLC494 and MMBT3906. I'll order 10-10 PCB from each (because they're cheap) and build one of each type and will play with them to see how they behave.