I now decided to let you know more about the details of the design. I share now the principle of operation along with the schematic drawings.
So this design still uses the good old AD835, but in an unusual application. The whole thing is centralized around the fact that the AD835 has an output function of W=X*Y+Z (the AD834 don't have the "+Z" input).
The video input is directly tied to the input Y, the contrast signal is fed to input X, at input X there is an analog switch, that can toggle between the contrast voltage (normal operation) and the ground (blanking the input). Then there is the Z input where again there is an analog switch, that is toggled between the OSD/testpattern signal source and the ground.
-During normal video, the contrast voltage is fed to X and Z is grounded. so the OSD/testpattern is bypassed. The analog switch might have a crosstalk, but even then it is conducted down to the ground through its Rds(on), so in fact a quite good rejection is achieved this way. The output therefore is: W=Y*X+0
-When OSD/test pattern mode is on the X input is grounded (aka. muted video input) but the Z (OSD) signal is let through ie. W=Y*0+Z.
-During blanking the Marquee switches back to OSD source and injects blacker than black level through the Z input (OSD)
The AD835 when properly used has about -60dB rejection to the input signal when it is multiplied with 0, on the other hand the dynamic range of each color is only 48dB (as it has 8bit resolution)
However in this way the OSD/test pattern is not affected by the contrast setting, and makes the built in stair step pattern meaningless, but I don't think this would be a real problem.
So all needed feature was done only by playing around the AD835, but nothing was put in the video path, only the AD835, but you can't omit that, since some sort of contrast controlling is needed anyway...
The AD835 is slew rate limited by the first hand, that means high output voltage cost a lot of bandwidth, on the original 02p VIM therefore the output was not even properly terminated for the 75Ohm transmission line, still about 25% signal loss occuring on the coax cable through the VNB. By putting this small board right at the VNB you gain this 25% signal level (that translates to bandwidth overhead), and the input level requirement is still the same.
With this circuit I measured around 1,5ns rise/fall times at 0.7V output (1 pixel at 1080p 72Hz is 5ns wide) which is still plenty fast.
There is a separate low noise +/-5V power supply for each board that also feed the adjacent VNB's +/-5V line too.
The gamma correction is also "interesting" for the blue. With these multipliers used in CRT applications it is often overlooked that all inputs are actually high bandwidth capable, So unlike to the VIM the gamma is not added to the output signal as a serial element.
The input signal is monitored (before contrast control) and is fed to a high speed difference amplifier whether the input is higher or lower than the threshold and this simply modulates the contrast voltage, even if it is only one pixel wide. Again, the AD835's all input is high speed capable.
This gamma circuit is a so called ECL inverter with emitter degeneration. It is tuned such way that if the input voltage is larger that the threshold level it produces a (linearly) proportional output voltage, depending how much the difference is between the input and the threshold. This voltage modulates the actual contrast voltage and yield a square-like mathematical function (see simulations) - the gamma curve.
This thing is basically working, but there are problems still operating in the Marquee: The Marquee is a damn huge noise generator. Putting the VNB-DB on a VNB in a bare stock machine will make a lot of running "sparks" appear on the screen. A noise that is picked up by the daughterboard (interestingly in the NEC XG I had no such problem with a similar setup). Better grounding / shielding does not help. The noise is coming from two sources as it turned out, once it comes from the HDM's scan regulator supply (later ULTRA HDMs came with proper filtering for that rev. 8 or newer I think). The other source is the LVPS,
especially the standby section (that generates the filament voltage as well). I could lower the LVPS noise with adding LC filtering, but the noise is still not completely gone, and I am currently working on the ultimate solution here. The main reason why I revealed these things is to help others (who are technical minded enough) to mate this design with other top projectors like Barco 909, or Sony G90 -as I don't have any of these machines. A marquee VNB paired with my daughterboard result in a complete video block, only the interfacing have to be solved for other machines (and the power requirements of course).
Is it now OK to build a board for personal use (not for resale), slthough I won't give away PCB designs, so either you have to invest your time in the PCB design, or buy a set of blank PCB from me at 30USD. I'll still undertake making complete board / modsets, once the power supply issue is fully sorted out.
