Atom Colour Board Thread

[trixster]

https://shmups.system11.org/viewtopic.php?t=52172

I gave this a quick whirl last year; the pi seemed to work ok and i could access the menu options from the pi's composite out, but the gbs did not seem to like having input through the RGBHV. I didn't have time to try it further.

[trixster]

I’ve got one of Prime’s Colour Boards Mk5.10c in my Atom and I’m attempting to connect it to my GBS8200. Sadly I’m not getting a picture. I’m using a DIN to 15HD (VGA) cable which works great for connecting either a beeb or Master to the 8200.

The cable is wired for standard Beeb on the DIN side:
1(R)
2(G)
3(B)
4(Sync)
5(GND)

and on the VGA side it’s:
1, 2 and 3 for R, G and B,
6, 7 and 8 are ground
13 is Sync.

The R, G, B, and sync all have 200 ohm resistors in the line, wired at the 15HD end.

I suspect the resistors are the problem. Can anyone advise what the problem might be?

Thanks!

[hoglet]

I suspect the resistors are the problem. Can anyone advise what the problem might be?

You've tried cycling through all the different video modes I guess?

The only thing I recall is that too high a SYNC level causes the GBS-8200 to stop recognising the video signal. So it might be worth trying a larger resistor in line with the sync signal. Looking at the schematic I posted in this post, I ended up with a 430R resistor inline with sync:
viewtopic.php?f=44&t=5994&p=56123#p56123

Dave

[trixster]

Yeah, I’ve tried cycling through the 4 different resolutions and have tried altering the geometry variable (clamp, size, posn etc). I’ll have a look at the cable to see about swapping out the sync line’s resistor.

[danielj]

Isn't prime's board at proper vga levels? Not ttl? You'd want no resistors in that case!

[hoglet]

Yeah, I’ve tried cycling through the 4 different resolutions and have tried altering the geometry variable (clamp, size, posn etc).

Just to be clear, I wasn't talking about the different resolutions.

I was checking you had cycled through the different input sources (YPRPB, RGBS, RGBHV) - it needs to be set to RGBS.

Anyway, I've just dug out my GBS 8200 and connected it to Phill's Mk 3.5LS colour board and it works fine with or without resistors in the sync signal.

So I'm not sure what to suggest.

Does your GBS 8200 work with a Beeb?

Dave

[trixster]

Yes, I’ve tried cycling through RGBS, RGBHV and YPRPB.

The GBS8200 works perfectly with my Beeb and Master with the cable I’m using.

[hoglet]

Do you have another monitor or a SCART TV/Cable you could test the Atom Colour board with?

Dave

[trixster]

A standard Beeb scart cable from RetroComputerShack on eBay is on the way, should be here tomorrow!

I did try it with a dell 2001fp which accepts 15khz signals but it stated mode unsupported which I suspect is because it’s expecting RGBHV over vga and not what a beeb/Master/atom outputs?

The dell supports composite in - is it possible to get composite from the atom / colour board?

[trixster]

Scart cable arrived, Atom works a treat!

[hoglet]

Scart cable arrived, Atom works a treat!

To be honest, that will probably look better than the GBS 8200.

Dave

[trixster]

Indeed, but the tv with scart (and the ossc) is in the lounge, and all the computers (and gbs8200) are up in the loft with monitors that have no scart inputs. Sadly as handsome as the Atom is it’s not going to get wife-approval to live downstairs so I need to find an acceptable solution where the Atom can live upstairs.

I’m assuming your RGB2HDMI will not work with the colour card?

[hoglet]

Indeed, but the tv with scart (and the ossc) is in the lounge, and all the computers (and gbs8200) are up in the loft with monitors that have no scart inputs. Sadly as handsome as the Atom is it’s not going to get wife-approval to live downstairs so I need to find an acceptable solution where the Atom can live upstairs.

Why not look for a cheap old LCD TV, like the Sony Bravia KDL 15G2000?

There are pretty common on eBay:
https://www.ebay.co.uk/sch/i.html?_from ... acat=11071

I’m assuming your RGB2HDMI will not work with the colour card?

No, it would likely look dreadful, if you got a picture at all.

It might be possible to add a Atom mode to the software. But there is a hardware issue as well: the Atom actually supports two levels of green, so it's not simply digital/TTL level RGB like Beeb is.

Dave

[John Ferguson]

For my Atom I bought a second hand TV and I found it had manual tuning to receive the output from the modulator. My colour board also displays on this.

John

[trixster]

I removed the 200ohm resistor from the HD15 cable on the sync line but still don’t get a picture. I’ve also tried altering the r g and b resistor pots on the gbs8200 behind the vga input but again I get nothing. The master and Beeb still happily return a picture on the gbs. Not sure what next to try!

[trixster]

Right, I have an rgb to HD15 cable now with no resistors on the r g b and sync lines... reckon this is safe to test with the colour board and gbs8200?

I’m just a little paranoid that something might get fried on either the Atom, the colour board or the 8200! Or am I being silly?

[1024MAK]

The variable adjustments (preset resistors) for the colours on the GBS8200 board are in series with the red, green and blue input signals anyway. So turn them to midway.

Mark

[trixster]

Okeydokey, I’ll give that a try

[edit] alas, still no picture. But crucially no dead atom

[hoglet]

Hi all,

I've made a prototype of yet another colour board for the Atom:

In its current form it sits externally, connecting directly to The Atom's PL4 connector with a short cable. It needs nothing special inside the Atom. The video output is HDMI from the Pi Zero, at pretty much any resolution you like up to 1920x1080.

If you look at the prototype board there are three sections, flowing left to right:
1. Some analog circuitry to digitise the low-level Y, PA and PB signals from PL4
2. A XC9572XL CPLD module, because that's what I had to hand
3. A Pi Zero to generate the HDMI output

The analog circuitry is mostly borrowed from my earlier colour board and can be seen in the schematic here:

rgb-to-hdmi.pdf

(67.92 KiB) Downloaded 90 times

The new addition is a improved clamp circuit (a back porch clamp, bottom left corner) to restore the black level of luminance signal (Y). Without this the black level varies depending on what's on the screen, making it hard to convert to digital levels.

The CPLD has several functions:
- it takes in the (now-digital) colour (AH, AL, BH, BL) and luminance (Y) signals from the analog section
- it performs some very simple noise filtering
- it samples them at exactly the right time (using an 8x clock, so 8 possible sample positions)
- it collects four pixels worth of data (a quad)
- it passes this quad to the Pi for further processing in software
- it generates a valid CSYNC signal from the 6847's HS_N and VS_N
- it generates a valid CLAMP signal that's active during the back porch for the clamp circuit

Fitting all of the above into a single CPLD has been quite a challenge. Currently there are just 3 spare registers in the CPLD:

Code: Select all

Function    Mcells      FB Inps     Pterms      IO
Block       Used/Tot    Used/Tot    Used/Tot    Used/Tot
FB1          18/18*      28/54       53/90       8/ 9
FB2          15/18       25/54       28/90       4/ 9
FB3          18/18*      30/54       69/90       8/ 9
FB4          18/18*      28/54       39/90       7/ 7*
             -----       -----       -----      -----
             69/72      111/216     189/360     27/34

The Pi (in software) does all the scaling/resizing and outputs a nice clean HDMI signal. This is running the same RGBtoHDMI software we're been developing in the other monster thread.

All the current design files are in github:
https://github.com/hoglet67/RGBtoHDMI/t ... ad_atom/v1
https://github.com/hoglet67/RGBtoHDMI/t ... /vhdl_atom

Some observations on how this is working in practice...

1. It's the best quality output I have seen from an Atom to date
2. Once the Y threshold is correctly adjusted with the small trim-pot, the black level is good for all the different possible screen modes
3. It doesn't seem to need continual re-adjustment
4. Auto-calibration works well on RGB-to-HDMI as long as a fairly complex test screen is used (e.g. CLEAR 0 full of random data)
4. The noise margin is still not as good as I would like, but most of the time the noise filtering works well
5. It seems to work slightly better with Philips LM319s than TI LM319s (Phill also observed differences between manufacturers)
6. At the moment it's limited to 8 colours, so yellow (CSS=0) and orange (CSS=1) are the same - this might be fixable, I'm not sure yet

At the moment this is a one-off prototype, the main goal being to ensure the RGBtoHDMI software works with machines other than the Beeb. It certainly be possible to shrink this work using SMT parts, and possibly fit it inside of the Atom. I'm not sure how the HDMI cable would exit in this configuration.

I'll certainly bring this along to the ABUG meeting in January, if anyone is interested.

Dave

[trixster]

[roland]

Great job Dave

That could be the second Pi connected to my Atom. Imho this is an easy and affordable colour board solution for many Atoms.

[John Ferguson]

It would be nice if this was available for owners to buy.

John

[oss003]

Wow .... looks great ... nice job Dave ...

Greetings
Kees

[hoglet]

A few more developments to report...

With a few tweaks to the CPLD in the new colour board, I am able to discriminate the following distinct colours:
- black
- red
- green
- yellow
- blue
- magenta
- cyan
- white
- dark green background (text)
- dark orange background (text)
- bright orange (text)
- normal orange (semi graphics)

That's a total of 12 colours, including three shades of orange.

Here'a a sample:

I think it's a matter of personal preference whether you like the non-black text background. But the 6847 does indeed generate it, so depending on what colour decoding hardware you have connected, it is possible to see it.

With the RGBtoHDMI thingy, there is a palette allowing each of these 12 distinct colours to be redefined, so you get to choose anyway. So if you want it to look like a classic Acorn colour card, which shows orange as red, than that's completely possible.

Dave

[HenrikErlandsson]

From memory, I seem to recall a color mod for the Atom in an ancient club magazine. I don't think it involved a board though? (Or is it the "Classic color card"?) Different thing maybe for TV? And this works on RGB monitors, which is what I have planned next.

Is this project still active? I read a post on here that it was canceled or something.

[Prime]

Right I've patched the mk5.1 board to take input from PL4 pin 9, and changed the CPLD so it only takes input from one comparitor.
I've also changed the resistors so that the adjust produces sensible output and I get text fine

I've also updated the CPLD code to generate the back portch clamp signal (for 17 cycles of VCLK after the rising edge of HS).

But what I don't currently have is the 74LVC1G66 About the only analog switch I have like that is a 4053....might try that

Cheers.

Phill.

[Prime]

I can't get this to work

Replicating the Input to the Y comparitor C1, R1, R8, R9, RV1, C7. Also replicated the clamp circuit R6, R7, C4, I don't have any of the LVC gates but I did have a 4016, which according to it's datasheet is good down to 3V so I used that.

Anyway I can't get any meaningful output from the Comparitor with these values, adjusting the pot goes straight from a black screen to a green screen with no inbetween no matter how slowly I adjust. If I remove C1, and connect the Y input straight to the comparitor I get what I would expect the comparitor turns on and off for each pixel (and I get the signon message as normal).

It could of course be that my comparitor is fubar

This is a trace of the video just sitting at the signon screen.

Traces from my Atom.....

The Yellow traces is the level out of pin 9 of PL4, the blue trace is the other side of the 100nf Cap.

Question is the clamp signal meant to be active (high, therefore switching the 74LVC1G66 / 4016 on), during the back porch?

Cheers.

Phill.

[hoglet]

Phill,

Question is the clamp signal meant to be active (high, therefore switching the 74LVC1G66 / 4016 on), during the back porch?

Yes, the clamp signal should be active high during the back porch, and active for as long as possible. (i.e. it should go high just after that start of the back porch, and stay high until just before the end).

It looks like for some reason it's clamping the black level to 0v, rather than 1v, so something is broken.

A couple of things to check...

1. Measure the signal across C4; it should be ~1V DC.

2. Capture the clamp signal and the comparator input at the same time.

3. Check that C4 is 10uF and C1 is 0.1uF. It's important that C4 is much larger than C1 (100x larger).

4. Check there are no other components pulling down the comparator input to ground.

Dave

[hoglet]

Phill,

I've just reviewed the datasheets.

The 74LVC1G66 has an on-resistance of 6.5 ohms (VCC = 3.3V)

The CD4016B has an on-resistance of 800 ohms (when run off 10V, so more when run off 3.3V)

Possibly that's just too large, i.e. there's not enough time during the 4us back porch to replenish the charge on C1.

Dave

[Prime]

Phill,

Question is the clamp signal meant to be active (high, therefore switching the 74LVC1G66 / 4016 on), during the back porch?

Yes, the clamp signal should be active high during the back porch, and active for as long as possible. (i.e. it should go high just after that start of the back porch, and stay high until just before the end).

Right cheers.

It looks like for some reason it's clamping the black level to 0v, rather than 1v, so something is broken.

A couple of things to check...

1. Measure the signal across C4; it should be ~1V DC.

0.98V close enough I'd say

2. Capture the clamp signal and the comparator input at the same time.
3. Check that C4 is 10uF and C1 is 0.1uF. It's important that C4 is much larger than C1 (100x larger).

Yep used the values on your schematic, so that should be OK.

4. Check there are no other components pulling down the comparator input to ground.

There shouldn't be I'll try unplugging pin 4 of the other comps (that I used for colour and mono graphics), to see if that helps.

On the issue of the on resistance of the 4016, I'm hopefully going to go down to RS in Nuneaton tomorrow and get some of the 74LVC1G66s, I only used the 4016 because it was what I had available

Cheers.

Phill.