Twin Shifter mod Prototype build results

[mrbombermillzy]

@mrbombermillzy

Do you have a simple PI1 viewer, to display this test image ? TEST.zip

on the ste/twin shifter, you should see 16 colors, like this (forget the UI of deluxe paint :D ) :


test-reference-ste.png


on a regular ST, you should see only 8 colors :


test-reference-stf.png

I've displayed your test image in Neochrome Master:

P1040237.JPG

However, I'm not sure how the program will decide to use the STe (or higher) palette on a Mega ST? ;)

The blue hues in my post above (which you mention) reveal at least an STe palette match, if that's what you are looking for? (Also look at TS3v2.jpg for a digitally boosted red to prove STe palette availability in the output.) I have coded the program to output up to 16bpp, so this is a better way of showing enhanced palette colours. (Just need to unify the test displays and ditch the CRT). :roll:

[mrbombermillzy]

I've updated some errors/typos on the long post from earlier. Hopefully it's all correct! :D

[sporniket]

I've displayed your test image in Neochrome Master:

However, I'm not sure how the program will decide to use the STe (or higher) palette on a Mega ST? ;)

I can distinguish the intermediate values of grey instead of 8 solid rectangles :cheer: , so this test is conclusive too, and it seems that when loading the file, Neochrome blindly copies the palette data into the registers.

[Tbs123456]

So is the STe shifter just basically 2 ST shifters in a single package?

[sporniket]

Not exactly, as the Ste Shifter can process natively the RGB-444 color rendering, it outputs natively the 4 bits of each channel. And the STe shifter is also tasked with dma sound.

[mrbombermillzy]

After a bit of thought, Ive found a way of showing some better results with the setup at hand.

If I actually set the top four bits of the palette word (the 4 bits that are not mapped to the STe palette) in various combinations, then perhaps we can see the 16bpp palette colours more clearly.

So here we are:

TSTOP4v1.JPG

TSTOP4v2.JPG

The STe version, just to confirm, shows no colours in this palette range:

STETOP4.JPG

So, despite the top 4 bits being LSB values, (bottom 2 LSBs for G) we can see the colour bars with no image processing of the photos.

Unfortunately, we are only seeing 3 colour bars (along with what I call some chroma noise columns) where there should be 7 :(

The program should be displaying 7 colour bars, with a black bar between each one.

The colours SHOULD be (from left to right):


-------GRGB
4096, 0001 0000 0000 0000 BLUE [Shown]
8192, 0010 0000 0000 0000 GREEN [Missing]
16384,0100 0000 0000 0000 RED [Missing]
32768,1000 0000 0000 0000 GREEN (LSB) [Looks more like red with moving green interference lines]
12288,0011 0000 0000 0000 GREEN/BLUE [More like Green/Yellow?]
28672,0111 0000 0000 0000 RGB (GREY) [Missing]
40960,1010 0000 0000 0000 GREEN/GREEN [Missing]

Any thoughts @sporniket? :/

[mrbombermillzy]

Here is the photo with chronic saturation to try to pick up any hint of the other colours:

TSTOP4v3.jpg

[sporniket]

Good experiment @mrbombermillzy :)

The first thought that I have on the result, is that since the blue bar is showing, I would have expected to see the GREY bar. In that respect, it would be interesting to try the missing combinations.

To sum up

• BLUE, GREEN, RED (resp 0001, 0010, 0100)
• GREEN(LSB) (1000)
• GREEN + GREEN(LSB) = STRONG GREEN (1010)
• BLUE + GREEN = CYAN (0011)
• BLUE + GREEN + GREEN(LSB) = GREEN LEAF (1011) (not tested yet)
• BLUE + RED = PURPLE (0101) (not tested yet)
• BLUE + GREEN(LSB) + RED = MAUVE (1101) (not tested yet)
• GREEN + RED = YELLOW (0110) (not tested yet)
• GREEN + GREEN(LSB) + RED = LEMON YELLOW (1110) (not tested yet)
• BLUE + GREEN(LSB) = SKY BLUE (1001) (not tested yet)
• GREEN(LSB) + RED = ORANGE(1100) (not tested yet)
• BLUE + GREEN + RED = GREY (0111)
• BLUE + GREEN(LSB) + RED = PINK (1101)
• BLUE + GREEN + GREEN(LSB) + RED = LIGHT GREEN (1111)

About the noise and all, as long as you do not install the jumpers on the R/G/B resistors of the main board, the signal is weakened. And since this test is about the least significant bits, it translate as very weak signal to start with, so I am not surprised that the noise starts to be visible. (reminder : the least significant bit on the extension board goes through a 121K resistors, the second least significant bit goes through a 60K resistors ; then it goes through the 1K resistors on the main board, filtered and amplified, impedance matched and then sent to the monitor )

That and maybe some signal enhancing/filtering on the side of the CRT, and maybe that explains some bar not showing.

So I would wait for the resistors on the main board being bypassed before trying to infer more from those result (by the way, the bypass of the resistors is reversible, once you wish to go back to stock.)

[mrbombermillzy]

Been sat on this for nearly a day now...have just been waiting for it to get dark enough! :)

@sporniket: As requested, here is a display of ALL the top nybble colour values.

I've basically started from colour value 0 for the background (and last 3 16px bars), and the rest from left to right are just adding #4096 every time to the colour value (so an incremental count up of the high four bits in the palette word):

HICONST.JPG

Hard to see low significance colours on a CRT, so digitally colour boosting the image is our friend:

HICONS1.jpg

And lastly, what those colour values should theoretically be:

GRGB
0000 - Black (background)
0001 - Dark Blue
0010 - Dark Green (slightly lighter than LSB dark green)
0011 - Dark G+B (Greener shade of Cyan)
0100 - Dark Red
0101 - Dark R+B (Magenta/Purple)
0110 - Dark R+G (Greener Yellow)
0111 - Dark R+G+B (slightly greener tone of grey)
1000 - Dark Green (LSB)
1001 - Dark B+Green (Cyan)
1010 - Lighter Green
1011 - Dark B+ Lighter Green (slightly Greener Cyan)
1100 - Dark Green (LSB)+Dark R (yellow)
1101 - Dark Green (LSB)+Dark R+Dark B (Grey)
1110 - Lighter Green+Dark R (Greener shade of yellow)
1111 - Lighter Green+Dark R+Dark B (slightly greener tone of grey)

Not too bad a show really. Although some of the colours are a bit suspect: (Greener yellow 0110, Dark Green (LSB) 1000, but this could be due to CRT display/camera errors in colour reproduction).

Next job is definitely getting the RGB resistors bridged. Should be able to get that done this week.

[sporniket]

That's great @mrbombermillzy :)

Finally all the colors are rendered. True, the "dark green 0010" seems darker than "dark green (lsb) 1000", however "Dark R+G (Greener Yellow) 0110" seems ligther than "Dark Green (LSB)+Dark R (yellow) 1100" as expected.

We will see what happens when you bypass the resistors, I expect that you will have a less tough time looking at those faint shades ; and that with better signal-noise-ratio, the CRT will give a better rendering of the colors.

In the mean time, I will craft a special PI1 picture to assess the effect of the problem that this version may have. It should be ready by this weekend.