Twin Shifter mod Prototype build results

[mrbombermillzy]

This thread will document whatever information can be gleaned (by me) about the working of the twin shifter mod, as remade by Sporniket.

This initial test post will be literally just testing the board actually displays an image, so dont get too excited.

So to start with, here's the machine we are using:

P1040208.JPG (60.01 KiB) Viewed 3348 times

P1040210.JPG (133.73 KiB) Viewed 3348 times

Its a Mega 2 converted to 4Mb. There's an IMP chipset in there including the CO70713-002 shifter IC which is what I will be initially testing:

P1040211.JPG (150.94 KiB) Viewed 3348 times

P1040212.JPG (170.55 KiB) Viewed 3348 times

Ok then, so identical shifter, along with a duplicate copy inserted into the board. A DIP socket was needed as a riser to actually bed the unit down, as the other nearby ICs were too high for a single socket.

Also, not sure how well this board is going to fit any of the ST/F/M boards and their revisions, but with this C100167-001 Rev.B Mega board, the board has enough space around (bearing the above height issue in mind) with literally no space to spare on the PSU side. It's a tight fit, but it fits!

P1040216.JPG (152.73 KiB) Viewed 3348 times

So while I was just checking the measurements and whether the board will actually fit in place, I decided to take a sneaky attempt to boot the machine without jumpering the three 3.6Kohm resistors to see how that affected the picture:

P1040215.JPG (117.81 KiB) Viewed 3348 times

Well, it looks identical to the original machines output (in monochrome anyway). I also have not needed to grab an extra power line from the motherboard either. So far, it seems fine.

I am currently finishing a new test to display the newly theorised possible flawed colour errors in certain bits: https://www.exxosforum.co.uk/forum/viewt ... 100#p80986 so will try to find the colour leads and a suitable RGB monitor at some point in the week.

That's all for now.

[sporniket]

it's good news that it works in monochrome, as only the main shifter drive it (and it has a separate circuit than the RGB pins).

Now waiting to see how it looks in color low and medium

[mrbombermillzy]

@sporniket I know. Couldn't resist though. Was more a quick preliminary fitment check.

However, I didn't realise the RGB signals are separate to the mono display on the ST machines. (Lucky I'm just doing the software )

I fitted it and literally don't have the right lead<>monitor connections for a colour ST to continue.

Generally, I'm quite well kitted out with CRTs, but the ones I have which are 15khz (Aydin Ranger, C=1084) I don't have the right leads for, so its either wiring up something, or heaving a 21" CRT TV on the desk and using the SCART.

Ran out of time for either option, so more later.

[sporniket]

Ok. I hope you won't have to carry the 21" crt tv

[mrbombermillzy]

Apologies for the delay. I have had various real life issues crop up recently, so I've tried to quickly get some data on this board out there for those of you waiting for some news.


Ok, so here I present the initial findings from studying the colours via tests on the twin shifter board described by Barry Orlando and brought to life by Sporniket.


General Software notes:
======================

* The original test software design goal (as discussed by Sporniket and myself) was to produce a line of 14 x 16 pixel colour segments, (colour incrementing from a palette value of 1>) which are changed on a
per line/raster basis. This would then show 2800 colours per screen and therefore the full 65535 colours in less than 24 screens.
However, after testing a more rudimentary version of the code, I found that colour visibility viewing may be a bigger problem than anticipated (see below), so have sacrificed a reduced screen amount for better visibility,
as I believe it's the more important issue at this stage. Hence, there are currently only 14 x 16 pixel colour columns PER SCREEN (i.e. 14 colours per screen). I hope this is ok for now.

* I have also thought it prudent to code an identically working set of the colour test program; one version for the regular 9bpp ST palette and the other being the twin ST shifter one. This should hopefully help diagnose/confirm
any faults/issues, helping deduce that the hardware is working as required. Other than the 9bpp version being (ironically) far more elaborate to code, the two are identical, other than the colour value increment routines.

* Regarding STe: The twin shifter palette bit significance/positioning has been designed, it seems, with STe compatibility in mind:

GRGBRRRRGGGGBBBB
R_analog = -1--2543
G_analog = 0-1-----2543
B_analog = ---1--------2543

Therefore, the twin shifter version of the software should show identical colour position results to a regular STe, up to the 4095 colour limit anyway, so can be used to confirm correct colour values up to this point.

* All testing/pictures produced on a Mega 4 ST with twin shifter mod and a C=1084 CRT, along with Hatari v2.3.1 Windows64 version emulating standard Mega ST and STe machines.



Testing the colour differences:
==============================

9bpp vs 12bpp vs 16bpp:
======================

* As the regular 9bpp shifter will be visually more difficult to show that it cannot handle a higher bitplane colour amount, we have to look at the more striking methods for confirmation of palette depth differences:

* On the 16bpp prg, the regular ST shifter will produce a black (0,0,0) or a previously displayed (repeated) colour whenever the 4th and/or 8th bit of the palette colour value are set (as I have not coded any bit transform routines to avoid null bits like on the 9bpp version). The Twin shifter will however, show a valid colour for this position, as (like the STe) all the lower 12 bits are used for RGB colour information.

This can be seen immediately on the first screen (albeit with difficulty) as the 8th colour bar being black on the regular shifter:

MST1.JPG (28.57 KiB) Viewed 3182 times

Whereas on the twin shifter, the image is a (hard to see) dark blue:

TS1.JPG (76.53 KiB) Viewed 3182 times

STe version matches the twin shifter:

STE1.JPG (30.29 KiB) Viewed 3182 times

Other colours which show the difference are colour 128/bit 8 and colour 136/bits 8 and 4 (2nd and 10th colours from left on page 10):

MST2.JPG (35.83 KiB) Viewed 3182 times

And the twin shifter version:

TS2.JPG (68.88 KiB) Viewed 3182 times

STe version again, matching the twin shifter:

STE2.JPG (35.41 KiB) Viewed 3182 times

* Unfortunately, the lack of colours on the regular 9bpp shifter won't manifest differences in an easy to see way (did I say the regular shifter doesn't easily let on that it cannot display >512 colours?) and will only really be seen by a SLIGHTLY DIFFERENT set of colours on the twin shifter ST, than that displayed on the regular ST, when the count goes over the 511 colour value. (The regular ST can still increase the colour value and set unused bits higher in the full 16 bit colour field, it's just that they will be null and other repeated colours further down will still be producing repeated colours over the 512 limit.

Anyway, the colour difference can be shown (if you have good eyes!) starting on page 147; colours 2045-2058 (4th colour in) onwards, once the 512 colour limit of the regular ST have produced all the unique colours. (Any further colours on regular shifter are just repeats):

Regular shifter:

MST3v2.JPG (34.29 KiB) Viewed 3182 times

Twin shifter: (Note: Zoomed in and digitally enhanced to show red content of the colour bar (which is evident here):

TS3v2.JPG (36.82 KiB) Viewed 3182 times

STe version matching twin shifter one:

STE3v2.JPG (35.08 KiB) Viewed 3182 times

I know it's very hard to see the differences (hence one of the reasons I decided to leave the more visible single, larger columns of colour in the program for the time being), so I have also included a colour analysis of the two (MegaST and STe) screen shots above, showing that the twin shifter and STe colours actually start to include some element of red, whereas the similar looking MegaST version does not:

Regular Shifter:

MST3.JPG (51.38 KiB) Viewed 3182 times

STe:

STE3.JPG (76.48 KiB) Viewed 3182 times

As I am using an emulator vs a CRT, visually distinguishing between the slight variance in colour at 12bpp on a TFT screen and 16bpp on an old CRT is next to impossible, while colour picking from two sources which are not in any way calibrated with each other is a more error prone task too. So perhaps some digital image enhancement to boost the red levels on the twin shifter picture and prove there is red after the 4th colour on page 147:

TS3v3.jpg (82.04 KiB) Viewed 3083 times

Continuing higher up the colour scale proves inconclusive between the poor colour matching of the emulated Ste and the CRT display of the twin shifter, but what the hell::

Page 1000:

MegaST:

MST4.JPG (35.23 KiB) Viewed 3182 times

MST4v2.JPG (22.55 KiB) Viewed 3182 times

Twin Shifter:

SEE NEXT POST FOR THIS IMAGE!!!

STe:

STE4.JPG (27.32 KiB) Viewed 3182 times

STE4v2.JPG (26.48 KiB) Viewed 3182 times

Page 4680:

MegaST:

MST5.JPG (35.71 KiB) Viewed 3182 times

MST5v2.JPG (22.1 KiB) Viewed 3182 times

Twin Shifter:

SEE NEXT POST FOR THIS IMAGE!!

STe:

STE5.JPG (35.59 KiB) Viewed 3182 times

STE5v2.JPG (23.44 KiB) Viewed 3119 times

ST 9bpp vs 9bpp on twin shifter:
===============================

So far I have dealt with ways to produce/prove the existence of extra colours on the twin shifter. However, we must also look at any hardware errors that may manifest in accurately producing those colours. This is where I pause for feedback from the HW gurus. I confess, I have yet to bypass the RGB resistors to the shifter on the mainboard (which is the next task and for me, personally in my situation, the hardest thing by far ), but from what I can see apart from a slightly murkier than normal black level, is some 'chroma noise' in the solid colours, (also on the green desktop); some colours even exhibiting different colour bars/bit values in the colour blocks.

If you can think of any HW adjustments that need doing, perhaps I can kill two birds with one stone and get them done when the resistor bridges are getting done.

Anyway, not a really deep write up, but it's a start! A better comparison may be picking better contrasting colours when digitally enhanced or better still, running the MegaST without the board and with the board on a TFT display. But that's a thing for next time.

Any comments or noticed errors in my writeup are welcome.

[mrbombermillzy]

Here are the 2 missing twin shifter images:

TS4.JPG (72.8 KiB) Viewed 3180 times

TS5.JPG (68.72 KiB) Viewed 3180 times

Sorry, but I seem to have gone over the single post image limit; hence the 2 missing images above.

[sporniket]

Nice, the first set with the "just above dark" stripe on STe/Twin shifter, "still black" on the ST is the most visible, to me.

I confess, I have yet to bypass the RGB resistors to the shifter on the mainboard (which is the next task and for me, personally in my situation, the hardest thing by far ), but from what I can see apart from a slightly murkier than normal black level, is some 'chroma noise' in the solid colours, (also on the green desktop); some colours even exhibiting different colour bars/bit values in the colour blocks.

yes, you should get a less dampened picture once you bypass the 3 R,G,B resistors of the mainboard. I guess that

EDIT: also, it confirms that the address registers are in the 0xxxx range.

[derkom]

Also, I don't know why SOME images are shown at the bottom as attachments?

You didn't place those inline anywhere in your post, so they remain as attachments at the end of the post. Normally I would try to find where someone would have liked to inline such images, but your post is so full of images that I wouldn't know where to start.

[sporniket]

@mrbombermillzy

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

TEST.zip

(402 Bytes) Downloaded 115 times

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

test-reference-ste.png (24.99 KiB) Viewed 3139 times

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

test-reference-stf.png (26.33 KiB) Viewed 3139 times

[mrbombermillzy]

@derkom Thanks! Was in a bit of a rush and forgot the last 2 example colour value analysis photos. (Now added).