4 replies to this topic

[5-11under]

Hi there,

I've been doing a lot of work over the past year, trying to create a good component (YPbPr) output from a ColecoVision. The picture I'm getting is quite clear, but some of the colors are muted, especially the yellows.

I'll describe the schematic I'm currently using... The regular ColecoVision VDP outputs have been completely disconnected from the regular ColecoVision circuitry. All three outputs are dealt with the same, but separately, of course: I've got the output (each of pins 35, 36, 38 of the TMS9928A) connected to one end of a 500 ohm potentiometer. The other end of the potentiometer is connected to ground. The tap of the potentiometer goes to the input of a buffer amplifier (LMH6734 in this case, either running from a +5V supply, or a +/-5V supply) with gain of +2 (although I think it works fine with a gain of +1, too, with different potentiometer settings). The output of the buffer amplifier goes to the TV/monitor, usually directly (adding 75 ohm resistors doesn't seem to make any effective difference in the colors).

I have control over each of the lines, Y, B-Y, R-Y, via the three potentiometers. Adjusting the Y potentiometer changes the luminance, and adjusting the B-Y and R-Y potentiometers changes the B and R levels, respectively (and the -Y too, I guess). Viewed with an oscilloscope, the signals are not clipping. If I turn up the B potentiometer, I can get the yellows to become more saturated. However, at that point, the background becomes too bluish (instead of black or near black), and the blues become too saturated.

I've also tried some level of DC restore, using diodes, with no effective difference.

I've tried this with two ColecoVisions, with the same results. Replacing the TMS9928A with a TMS9918A gives good colors and saturation levels, similar to what I get with Ben Heck's composite mod. Unfortunately, with the TMS9918A, you get weird rainbow effects (the picture doesn't look too clear, either, but maybe that part could be fixed).

According to the datasheet, the TMS9928A and TMS9918A do output slightly different colors. However, the results I'm getting are vastly different from each other. I'm hoping there's a simple way to get the component output to closer match the composite output.

I've fed the YPbPr output to a YPbPr to VGA/XGA converter, with the same results. At that point, I've also reduced the B output and increased the R and G outputs, using the same potentiometer and buffer circuitry that was used for the YPbPr. That can help, but as the yellows get more saturated, the blues become less saturated. This also makes the circuitry much more complicated.

If any of you have done a component modification to your ColecoVision, I'm interested in knowing if you're getting muted yellows or not. If you are getting muted yellows, do you care? If you're not getting muted yellows, could you tell me/us how you've done this? If I find the solution, I'll share it in this thread. At some point, I'll also take some pictures of the oscilloscope vector results (R-Y vs. B-Y), as well as some YPbPr monitor results versus composite results.

Any other help or suggestions are also appreciated.
Thanks for your time,
5-11under

As a bonus, whoever helps the most in solving this, I'll reward with a free copy of my next ColecoVision project, which I'm hoping will be pretty special.

Edited by 5-11under, Sat Sep 26, 2009 1:47 PM.

[intvnut]

5-11under, on Sat Sep 26, 2009 1:46 PM, said:

Hi there,

I've been doing a lot of work over the past year, trying to create a good component (YPbPr) output from a ColecoVision. The picture I'm getting is quite clear, but some of the colors are muted, especially the yellows.

[...]

As a bonus, whoever helps the most in solving this, I'll reward with a free copy of my next ColecoVision project, which I'm hoping will be pretty special.

I'm not sure how helpful this is, but here's how TI told people to do it back in the day:

http://spatula-city....to_Monitors.pdf

Includes schematics. It even mentions the correct scaling to apply to the R-Y and B-Y signals, and that the components in the circuit were selected to ensure that various colors (including yellow) are fully saturated.

Enjoy!

[5-11under]

Hi intvnut,

I've built part of the circuitry before, with no success. One thing, though, is that I don't want to convert to composite or RGB. Composite won't look as good, and I have no way to display RGB at home without further conversion. However, reading it again, I do have another idea I'd like to try out. Details to follow... .

Thanks,
5-11under

[5-11under]

Okay, here goes...

First of all, I feel much relief at the moment, having found the solution that has bugged me on and off for much too long. As is often the case, the solution isn't as complex as the problem seems to be.

As you may know, the ColecoVision uses a TMS9928A Video Display Processor (VDP). The native output consists of 3 signals: Y, B-Y, and R-Y. Y is the luminance, B-Y is the blue color difference (relative to the luminance), and R-Y is the red color difference. Most people who want to output to a YPbPr monitor/TV start by taking the signals from the VDP, adding a load resistor to ground (keep in mind that the ColecoVision already has some loading, if that part of the circuitry is left connected), and connecting the signals to the Y, Pb, Pr inputs of a monitor, either directly, or through a buffer amplifier with unity gain (a buffer may help, because the monitor is likely loaded at 75 ohms to ground - I'm not sure of the implications of loading the VDP with 75 ohms or less). If you do this modification, you will find that, amongst other things, the display is oversaturated with blue. To remedy this, you can replace the load resistors with (500 ohm) potentiometers, tapping the wiper connections to the monitor or buffer amplifiers. This allows control over the Y, B-Y, and R-Y levels. Unfortunately, after adjusting the controls to optimal levels, the colors will still not match what was intended. The colors will be likely muted, especially the yellows.

Recall that the blue level was much too high when connected directly to the monitor. In order to get a half-decent display from the method described in the previous paragraph, the B-Y signal needed to be greatly reduced. However, when doing so, not only is the B reduced, but the -Y is also reduced. By adding more -Y (or subtracting more Y) to the scaled down B-Y signal, the proper signals can be nearly attained. The same procedure could be done to the R-Y line, although the improvements would be less striking (note: I haven't tried the R-Y correction yet).

In order to perform the mathematics on the B-Y signal, the B-Y and Y signals need to be linked in some fashion. I don't think this can be done with simply a resistor or other passive circuitry between the two lines, because the Y line would be adversely affected. I used a buffer amplifier to keep the Y signal path clean and strong, before tapping it to subtract from the B-Y signal. I used an LMH6734 for the buffer amplifiers, and an LM318 op amp for the subtraction function (I'm sure other "high speed" buffers and op-amps could work fine). I still need to optimize the schematic - I think there's some room for decreasing the parts count I'm currently using. I'll post more details, as well as pictures, at some point, but that's enough for today.

Thanks to intvnut, who wins the as-yet-unknown prize. Not so much for the specific solution, but in leading me back to the application report, where the solution was found.

Thanks and enjoy,
5-11under

[intvnut]

5-11under, on Sat Oct 3, 2009 9:20 PM, said:

Thanks to intvnut, who wins the as-yet-unknown prize. Not so much for the specific solution, but in leading me back to the application report, where the solution was found.

I'm glad you got decent output! That's the biggest prize of all, IMHO. :-)

I'm glad I could be of help, however indirectly.