I thought to route bus between ICs legs so they could run the entire height of the board. DIP legs are 2.54mm apart (0.1”) so I needed holes 1.27mm apart - one set for bus and second set for ICs. Evenly distributed on 10x15 cm board was over 8000 holes. Also the holes need to be at least 0.8mm in diameter to fit in pin headers. This gave very little room for soldering pad but I squized it in… PCB manufacturer complainde for the number of holes and trace spacing :( and wanted to charge extra.

So I did the 2nd attempt - I removed every 2nd hole on the buses and every 2nd hole on chip traces and shifted them vertically 1.27mm. This got accepted. Manufacturer kindly suggested gold plating (soooo expensive) but agreed to go with HASL noting the quality may be worse than usual. But the boards look great and the cost of making 10 is not different than buying 10 proto PCBs of this size. I made them 1mm thick for a test - they are rigid enough.

The design: Buses are numbered from 1 to 58 - I marked few of them in green.

There are 3 rows of ICs A, B and C which have short traces reaching PCB legs - marked in magenta. To connect ICs to buses and power you only need mostly horizontal connections - cyan.

The same board can be used as a backplane and as a card. You can connect boards in one after another in a flat arrangement

or perpendicular (as a backplane and cards)

And an example of how ICs could be placed on the board:

References:

• The project: GitHub

This is my first Amiga but I know there are well known solutions to that. But they suffer from being more expensive than Amiga itself (my Amiga was 65€) or requiring custom plugs like 23 pin video connector. (or both). My 1st idea was to just build A520HD video adapter. But since I was building it from scratch I thought I’ll make an internal version. Then I though I wanted SVideo more - my 4/3 monitor does not have component input. Common solution seems to use Sony CXA1145 chip from A600 but it is unavailable (I think) and it requires an external delay line. I went with more modern AD724 instead which is much simpler. Designed and made a board:

I don’t have the schematics from that time - never thought to publish it… Anyway - it didn’t obviously work. I naively assumed that DENISE pin 18 called “burst” on the schematics will output PAL 4.433MHz color burst. So it doesn’t. :) Turnes out the same people who made Atari 800 worked on Amiga and did it the same way. So burst is “base clock” / 8: 28.37512MHz / 8 = 3.54689MHz which is 4/5 of the required frequency. Hmmm Atari… Let’s borrow colorburst from Atari then:

And it kind-of worked:

Hardly ideal…I need to move PAL clock to my board and colors are off… The colors issue was somebody mixed op-amp legs… and most of distortion is gone once we have own clock.:

Not ideal but promising. The board looks bad now though:

The issues:

• New board needed - this one served well as a prototype but it is not reliable at all.
• PAL clock is not synchronized to pixel clock - there was visible moving artefacts on the pixels edges.
• Still a lot of noise

I decided to re-do the schematics, add A600 /4*5 clock circuit (almost identical as in Atari) and order board from PCB manufacturer.

Sooooo 2years have passed. I did other things :) But finally:

With some more fixes - I made a mistake and swapped R and B AD724 inputs and also replacing input cables with coaxial grounded at input - the image quality is great. So here it is at GitHub. Fixes are not there yet but maybe sometime ;) an eagle eye can spot external PSU replaced with 2 internal boards and 12V input next to new video output.

The image finally looks great over s-video:

And even beter over component although TV makes it darker and little over-saturated. This does not happen on a monitor with component input.

References:

• The project: GitHub
• A520HD project: A520HD
• AD724 datasheet: AD724.pdf; source
• Amiga 500 Board Rev. 6: A500_R6.pdf; source
• Amiga 600 Board Rev. 1.5: A600_R1.5.pdf; source
• Atari 800XL: 800xl_01.gif; source