IanB wrote: ↑Tue Nov 07, 2023 5:58 pm
Looking at those scope traces, there is a colour burst present so is the video output in colour?
A colour subcarrier superimposed on the video will mess up RGBtoHDMI's video capture causing dot crawl.
Yes and no — the Presenter has a standby screen which has a rather nasty purple background and some white text. It almost seems like they chose those colours to look awful with composite colour fringing!
When connected to a calculator, however, it seems to only output an image in black and white (and all the calculators it works with only have a black and white display). So, aside from the colour burst itself, I don't think there is any colour in the image (whether there's still something there that would upset the RGBtoHDMI is another aspect) but the whites and blacks look fairly clean (at least in the first field — see below), albeit noisy and it looks like some sort of smoothing might be going on:
When I set the Y Hi level on the RGBtoHDMI correctly, it's easily able to reliably distinguish things cleanly on the first field.
IanB wrote: ↑Tue Nov 07, 2023 5:58 pm
Can you post a closeup photo or video of some characters when displayed via the monitor's normal composite input
The other possibility is that it is using half pixel character rounding like the BBC's mode 7 which would require sampling at double horizontal resolution with double clock frequency and double line length.
I was wondering if there was some sort of effect going on but I assume it would have to happen in the Presenter: the 240x128 image is what's on the calculator screen and the sharp version of the images I sent before; if the other field is some sort of rounding, I think it would have to be inferred by the Presenter, using some of smoothing effect based on the image, rather than knowledge of what's actually being displayed, in terms of fonts. I have tried doubling everything, as you suggest, but that doesn't seem to give me anything better.
The image looks fairly clear using the composite input on my Dell LCD monitor and doesn't display any of the effects in the 'dodgy' field you can see in the RGBtoHDMI sampled output. The same goes for my old green screen CRT (which my phone seems to think is cyan).
Another thing that suggests to me the Presenter is doing something odd is the scope trace from near the bottom of the screen: on the source image, there is a single pixel high black line from the left to the right edges of the display, a blank white line and a few exclamation marks. This is clear in the first field of the signal (and there's a blank white line in the signal before this):
The same line in the second field in the frame looks weird, though:
... the black line is two scanlines high and the voltage for it seems to be halfway between the white and black levels in the first field, there's also no blank white line between the black line and the start of the exclamation marks, and the exclamation marks voltage level differs on each line.
It looks like the first field is a perfect image from the calculator, then there's some sort of anti-aliasing going on in the second field to smooth things out: zooming in on the LCD photo, it seems like this might be happening, as there looks to be a grey line either side of the black line, and possibly some grey above the 6 and 3, rather than a clean white line, but it might be the monitor doing something aliasing to the LCD pixels, although the monitor does have a 1:1 mode (photo below) that just displays a small image in the middle of the screen and I think I can clearly see the grey bits here (assuming it's not filtering things itself):
If this is the case, is there any way to grab just the first field of the image and discard the second? Or do you think there's something else wrong?