Atom Colour Board Thread

[1024MAK]

When I had a wander around our local Currys a few weeks ago, I checked a lot of the current range of TV's for sale, the ones I looked at all either had a SCART socket, or a RGB socket (that a SCART adaptor can connect to).

So analogue is not dead yet

Mark

[hoglet]

I've always had an irrational obsession with getting things perfectly "pixel mapped", so that when you have a regular images (e.g. alternating black and white lines) it looks very uniform. I don't think this is technically possible with SCART on most displays, which is why I'm so fixated on VGA.

For normal non-obsessed folk, SCART will work just fine. After all, this is how many BBCs are connected. Once Phill finishes his Mark IV board I think the results will be pretty good.

Dave

[hoglet]

Morning all,

So far we have had
- Episode 1 : Acorn Colour Board
- Episode 2 : Phill's Colour Board
- Episode 3 : Hoglet Mk2/Mk3 Colour Board (which TC is playing with now)
- Episode 4 : External VGA Adapter based on Papilio FPGA Board

and now, the (hopefully) final instalment, the GODIL VGA adapter.

The GODIL is a tiny FPGA module designed as a DIL replacement, and in this case it's replacing the 6847 with a "Super" 6847 that contains 8K RAM, a 6847 core (from Alan's Atom FPGA) and a line doubler to output VGA directly.

Here's a couple of pictures of the GODIL, with a VGA cable wired in to the test connector via a few resistors:

Here's the empty Atom, ready for installation. The 6847, a 74LS138 and the 12 2114 Video RAMs have been removed:

Here's the GODIL installed into the 6847 socket:

There are two flying leads:
- a Write signal for the internal 8K of Video RAM
- a Reset signal to try to retain the original behaviour of staying in text mode when break is pressed. The problem that the 8255 outputs (controlling the graphics mode) go high impedance during Reset. On a real atom, these float low. Where as on GODIL the are pulled high by 1.5K pullups, flicking it into Graphic Mode 4 when break is pressed, which is just wrong! To retain the original behaviour I needed a reset signal, which isn't present on the 6847 socket. Anyway, I digress....

Here's a couple of screen shots of the final results:

The Xilinx VHDL Code is on GitHub, if anyone is interested.
https://github.com/hoglet67/AtomGodilVideo

Best picture I've ever seen on the old girl!

There might even be room for an AtomSID in the FPGA, but it would have to be addressed at 9Bxx.

Dave

[hoglet]

Sorry, I hit submit prematurely.. A few more pictures have been added to the previous post....

Dave

[1024MAK]

Nice



Mark

[danielj]

There's no cap doffing icon so it's simply

d.

[jonb]

Ooh la la Monsieur!

Je pence c'est un jeu du "Pimpez mon Atom", n'est-ce-pas?

[oss003]

Hi Dave,

nice job ........ again ......

This looks very good, no shading blocks and sharp edges ...... very nice!
Is the other colourset with orange?

Did you also test it at 1,79 mHz clockspeed or can we simply use 2 mHz now?

Greetings
Kees

[AlanD]

Hello Dave

Nice job

AlanD

[hoglet]

Hi Kees,

I tested it at with the Atom running at 1.79 MHz and it worked fine, so I would expect it to be OK at 2MHz as well.

Here's a couple more pictures showing the two colour sets:

The only thing I'm not 100% happy with is the cable arrangement - I don't like the cable being fixed. I'd prefer to have some kind of a socket that i could detach the cable from, but then I don't want to butcher the Atom case in any way, as mine is pristine.

The other possibility is outputting the VGA from pins on PL4 that are driven by the 6847. In some ways that would be neater, but it might mean sacrificing orange. There would then be a short cable that would connect the 10 pin IDC to 15 pin VGA socket. A normal VGA cable could be attached to this.

This only possible problem with using PL4 is there are really only 4 easily usable free pins driven from the 6847 (OA, OB, CB and HS). We need one more for VGA, which would mean a mod to free up pin 9, or putting a VGA VSync out in place of the Atom FS. Any idea how sensitive software on the Atom is to the duration of FS (on the 6847 it is ~2ms and on VGA it is much shorter, 128us I think).

Dave

[danielj]

Dave, if you're taking most signals from PL4, (I assume we're talking small VGA dongle out the back) why not just use a slightly wider cable and take the extra lines back to the fpga? It's not as neat as attaching to a single point but it would appear the same from outside (you can't disconnect something from PL4 without opening the case anyway) and you'd get all the signals?

d.

[Multiwizard]

Hi Dave,

perhaps you can use this idea for your output, without making new holes in you Atom housing, I did this for my joystick inputs.

viewtopic.php?f=3&t=7250


Greetings, Wim...

[hoglet]

Thansk for all the connection related suggestions, I'll keep pondering...

Dave

[roland]

Hi Dave,

First of all, a wonderful job that you have done and .

This only possible problem with using PL4 is there are really only 4 easily usable free pins driven from the 6847 (OA, OB, CB and HS). We need one more for VGA, which would mean a mod to free up pin 9, or putting a VGA VSync out in place of the Atom FS. Any idea how sensitive software on the Atom is to the duration of FS (on the 6847 it is ~2ms and on VGA it is much shorter, 128us I think).

I prefer modding pin 9 as this is easy, remove C2 and R6. Solder a piece of wire between the negative pole of C2 and the "right" side of the resistor and the mod is ready without cutting traces. So it's quite reversible.

FS is often used in programs because of its constant timing. It's a 60Hz signal (which makes the duration ~16 ms) and many games use this signal for delaying. Also some sound statements use it. I suggest to keep this signal as is.

Regards,
Roland

[hoglet]

Hi Dave,
FS is often used in programs because of its constant timing. It's a 60Hz signal (which makes the duration ~16 ms) and many games use this signal for delaying. Also some sound statements use it. I suggest to keep this signal as is.

If I re-purposed FS as VGA VSync, it would still have a period of 60Hz. The only difference is that the width of the low pulse ~128us vs ~3ms. I think the WAIT routine at FE66 would work just the same. I might give it a try.

Dave

[roland]

Ok, I didn't notice you are only varying the duration of the pulse width. In that case I don't expect many problems with software, as long as they do not call #FE6B.

I built once an I/O card for my atom-in-pc with a simulated 60hz from a NE555. The duty cycle was 50%. I never noticed any problems with that setup.

[hoglet]

On ebay:
http://www.ebay.com/itm/Vintage-Acorn-A ... 1713196396

This expensive Atom has a colour board I have not seen before:

Can anyone identify what this is?

The MC1372 is a colour modulator designed to work with the MC6847:

The MC1372P is available on eBay, so I'm going to try and make one of these I think for my original Atom.

Dave

[PhilYoung]

On ebay:
http://www.ebay.com/itm/Vintage-Acorn-A ... 1713196396

This expensive Atom has a colour board I have not seen before:

SVideoAdapter.PNG

Can anyone identify what this is?

The MC1372 is a colour modulator designed to work with the MC6847:

SVideoAdapterSchematic.PNG

The MC1372P is available on eBay, so I'm going to try and make one of these I think for my original Atom.

Dave

I've just had a quick scan of that data sheet, and it implies that the intended RF output is on Ch3 or Ch4, so between 60MHz and 70MHz. So you'd need a VHF NTSC set. Adjustable via the tank circuit on pins 13/14, but I doubt that it'll work at UHF (CH21 starts at 470MHz). Also remember that NTSC as broadcast normally uses +ve modulation, rather than -ve modulation as was used for PAL on UHF in the UK.

So sourcing a TV might be a problem unless you have access to a suitable one.

If you make one up you could check if it's basically working using a spectrum analyser on zero RF bandwidth and maximum VBW to decode and get some sort of picture though. the video. Not an actual picture of course.
Cheers,

Phil Young

[hoglet]

Hi Phil,

I think it's possible to use the MC1372 to generate baseband composite video. In the data sheet Test Circuit 2 does this.

There are also a family of composite video boards for the CoCo2 that use the MC1372. See this page for more details:
https://sites.google.com/site/thezippst ... 2-projects
There are some screen shots and the quality looks quite good.

Dave

[PhilYoung]

Hi Phil,

I think it's possible to use the MC1372 to generate baseband composite video. In the data sheet Test Circuit 2 does this.

There are also a family of composite video boards for the CoCo2 that use the MC1372. See this page for more details:
https://sites.google.com/site/thezippst ... 2-projects
There are some screen shots and the quality looks quite good.

Dave

Oh right, I should have read it more thoroughly before posting. Plus checking ebay they are only a couple of quid so I might get a couple myself, I'd assumed they would be @@@@@RARE@@@@@@.

Cheers,

Phil

[hoglet]

Here's a mostly discrete transistor approach to getting RGB out of a 6847:
http://alice.system-cfg.com/files/A4K_schema.png

It uses 13 transistors and some kind of sync combiner chip.

This is Alice:

Dave

[oss003]

Hi guys,

I did found this on the internet. It was a game design for a lightgun with a 6847.

I had the idea to build it with a PAL but until now, it's still an idea .....

Greetings
Kees

[hoglet]

Hi Kees,

If you go right back to the beginning of this thread, that is pretty much the design I started with:
http://www.stardot.org.uk/forums/viewto ... 994#p56123

I seem to have mislaid this board, as I can't find it anywhere

It worked quite well, but the problem was there is very little margin for error when setting the comparator thresholds, and the optimal values seem to vary depending on the screen mode.

Dave

[hoglet]

Here's a link to another design using the MC1372P:
https://archive.org/stream/color-comput ... 9/mode/2up

Dave

[hoglet]

I've done a bit a experimentation with the 1372 and a 6847 on a breadboard, and the results are a bit rubbish. The colours are correct but the picture is very soft and fuzzy.

I think there are several things contributing to this:
- there is loads of stray capacitance on a breadboard
- I don't have the 2N4401/4403 transistors, and am using 2N3904 and BC557 instead.
- my TV (an LG 22MN43) thinks the signal is 480i and is trying to de-interlace it, adding to the overall blur.

I'm trying to decide whether I can be bothered to make this on a small PCB.

I think that long term solution that outputs RGB (like Phill's board or the GODIL) will always be better.

Dave

[1024MAK]

... the picture is very soft and fuzzy.

What, on a composite output, never

If produced on a PCB and if you used transistors designed for baseband video amplification, it would likely improve the picture, but would still not get close to RGB. If you could get S-video out (separate luminance and colour signals) that would give a better picture than composite.

Keep in mind these chips were intended for use with domestic CRT TVs, normally with 21 inch or smaller size screens.

Mark

[hoglet]

Here's a colour board I haven't seen before:

It's in an eBay Atom:
http://www.ebay.co.uk/itm/121756495316
Thanks Wim for spotting this.

It looks like a more complex, DIY version of this:
http://stardot.org.uk/mirrors/www.bbcdo ... er_cct.png

But it has 6 chips rather than 3.

Ring any bells?

Dave

[oss003]

Could it be a combination of the PAL board with the colourboard Delft: www.acornatom.nl/atom_nieuws/1984/nr2/19842086.htm

Greetings
Kees

[trixster]

Dave, in the GBS post on page one of the thread you mention tweaking the GBS to get it to output 50hz rather than 60hz. Is this a simple thing to do or did you have to hack the firmware?

I'd love to have my gbs8200 output 50hz as default as I use it with a Beeb, a Master and an Arc all of which suffer noticable 'judder' due to the gbs's 60hz output.

[hoglet]

Dave, in the GBS post on page one of the thread you mention tweaking the GBS to get it to output 50hz rather than 60hz. Is this a simple thing to do or did you have to hack the firmware?

Gosh that was a long time ago!

As I recall, it was more of an experiment. Rather than trying to hack the firmware, I just hooked up a Pi to the SPI bus (between the microcontroller and the 5725 video chip) and just directly re-programmed the registers in the video chip. This is a bit naughty, as the SPI bus is meant to have just one master, but it worked for an experiment.

Clearly the right way to do this is to figure out how to hack the firmware. I didn't get that far...

Dave