Apple][VGA

Development news




11.5.09 First PCB running (at least parts of it)

After having a LOT of trouble with the PCB company I finally got a bunch of PCBs - but no CPLDs and other stuff yet. At least I started populating a RamWorks III daughterboard. The -5V switchmode power supply has issues (haven't found that yet - the circuit seems to be correct, but no oscillation and thus no output) Because I won't be able to do color anyway (missing parts) I left it for now and installed the CPLD, the 74HC86 clock circuit and the RAM and prepared for the smoke test...
No smoke, but correct sync and after hooking the B/W signal to my old IBM 12" there is a correct B/W picture without noise. The digital part is error-free! I didn't take a picture because I think everyone has seen a catalog and a half-populated PCB.
The RamWorks III doughterboard is a little smaller than the original RGB board and occupies the space left to the RW3 - if you have a european //e you'll lose slot 2 - I wasn't aware of that before - where shall I put my SSC then? There is also a "force monochrome" jumper.

19.1.09 More successful tests and PCB design

I had to go to school again - very bad. But there is no school at weekends - for luck.
After finding out that all the color issues were due to a blown CPLD, I first shouted a couple of words that I'm not going to repeat...
Then I chose to use another CPLD. I still had a 10ns CPLD on hand - but it was too old, the Xilinx ISE 10.1 Software refuses to program it. Xilinx told me that I have to use the 2001 version, which only runs on Windows - which my PC is not running.
So I got some stuff to set up a PIII-450 system and installed the forth worst operating system on earth - Windows 2000 - on it. The 2001 version had even more bugs than my version but after two hours of messing with it I had the CPLD programmed - with the old design, of course (the new design required a higher chip count). After resoldering everything it worked instantly. So let's untwist the cabling and set up for a photo. I got zapped while reconnecting it, and - guess it - the CPLD was blown. Some research uncovered that the //c's power supply (a vanilla 12V/4A PSU) was not earthed despite having an earth connector. I just fed 220V AC straight into pin 84 of the CPLD! No wonder it failed! I'm also quite sure that the first CPLD failed for the same reason.
While knowing that it works, I designed a PCB today. It is 84x96mm (3.3x3.8'') in size and also includes a switching power supply because the linear regulator produces a lot of heat (too much for a small device I think). The supply has been tested and runs like a charm.
This is the prototype:

And this is what the final PCB will look like:


11.1.09 Tests of a prototype of the //c version - continued.

Today I fixed an issue with the color circuitry (lack of colors 0-7). Because it seemed to be a race condition (which was possible there), I redid the design of that part. It now equals the old discrete design. After changing some wires (why does the Xilinx ISE program always mix up pins that are completely unrelated to the change?!) I fired it up and saw the correct signals on my scope. I told Henry that I'll send him a screenshot in color this evening.
But things went differently. Started with the NTSC TV Emulation circuitry, and fired it up to see whether it was working - black screen. After swapping two wires (and noting that in the schematic) I had a picture. I then adjusted the sharpness and took a photo:
The noise issues have disappeared - the bright spots on-screen are dust, not noise ;-) .

I then wanted to finish the color stuff. Before wiring the DAC I did some more tests - and bit 3 of the color number is replaced by bit 2! The signal is in the wrong place - and I don't know why. I really have no clue - I double - maybe triple - checked the equations: they are correct.
I really haven't thought that the simplest part could turn in such a nightmare - I'll give up for today.
Some photos of the grown test board:

10.1.09 Tests of a prototype of the //c version are running.

This new design is CPLD-based and will only need 4 ICs (the original design needed 29) and should not have some compatibility and reliability issues the original design had. It did have some other issues, though. The high pin count wires of per area caused the board to become noisy. I fixed most of the noise issues and a bus fight today. Pictures of the first tests:
This image is taken with a better monitor - the old IBM 12'' is fairly blurry. You can see some spots - this are the remaining noise issues - They are due to poor ground wires and will not be present in the PCB version (I hope ;-) ) The image is only in red, because there is no color circuitry running at the moment. I simply fed the monochrome video signal to the red channel via 220Ω Resistor (It's the red channel, because it is pin 1 ;-) ).