BigDumbDinosaur suggests "learn how to fly a Piper Cub before you climb into the cockpit of a 747." BigEd notes "It is faster to make a four-inch mirror then a six-inch mirror than to make a six-inch mirror."
I have taken this more literally than most by limiting computer board construction to a four inch disc (or metric equivalent thereof).
This may seem limiting but I have found 78cm^2 (12 square inches) to be a very convenient size for design and manufacture. It is approximately 1200 perfboard holes and it is suitably large for a combination of 8 bit DIP processor, ROM, RAM and I/O without increasing risk of failure too significantly. In particular, by removing the square corners, considerably bad design is required to make any signal longer than 100mm. Indeed, I've found that parallel buses will typically be 80mm or less.
The rounded shape is suitably portable for mobile applications, such as a music player, without precluding desktop applications. TUIT100 is suitably sized for cheap 100mm*100mm Asian manufacture - and there is a small amount of space in the corners for unrelated circuits at no additional cost beyond shipping weight.
TUIT100 has symmetric fixing holes in 60mm square. This allows a board to be mounted with four orientations. I considered 70mm fixing holes. However, 60mm fixing holes is compatible with the two way symmetry of 160mm*100mm EuroLozenge format. Both are compatible with smaller mezzanine cards, 128mm/192mm/256mm desktop enclosures and a forthcoming range of miniature racking. Mezzanine cards are designed with a straight or curved inner edge. This allows maximum safety, maximum flexibility and minimum waste when manufacturing in volume. The main board was designed mostly on a 5mm grid. The exception was the 1.5mm radius for M3 fixing holes. Mezzanine cards were mostly designed on 1mm grid with Pythagoras 3:4:5 ratios, although to get focus of the 50mm card correct, 0.25mm was required.
I found the curved edges and fixing holes to be a minor inconvenience. Although the effective area for DIP placement is less than 10 square inches and shaped in a very fat cross, minor components are naturally relegated to four curved regions. It is possible to move fixing holes into a diamond. This provides a small increase of contiguous area for DIP placement. However, minor components, such as power connector, may be relegated to eight regions with less area and more tenuous wiring. This is not likely to be an advantage.
TUIT100 started as a parody of PC/104 format. However, advice from experts, such as GARTHWILSON, is to avoid a parallel bus card system. This is specifically to increase project success. So, TUIT100 is a card system without a card bus. If you require off-board signals then a serial bus, such as 65SIB, is recommended. For compactness, ease of operation and reliability, several incompatible tweaks are also recommended. Most significantly, I'm an idiot. Therefore, I made 7SER electrically safe when connected in reverse.
To maximize board area, I devised several KiCAD schematic symbols and footprints. However, using them will complicate a design and reduce chance of project success. Obviously, I started with Dr Jefyll's LE74HC574K245N latch with read functionality. For my own purposes, I devised stacked latches for bank switching. I also devised 24 decoded outputs with LE74HC238M3N and similar. The schematic symbols looks particularly mad. Unfortunately, I had insufficient space for three 8*2 IDC headers.
RAM is easily stacked although power consumption should be calculated if CMOS RAM is operated at speed. It is most common to decode RAM chip select signals. It is also possible to selectively decode read/write signals. In this case, one conceptual row of RAM receives chip select and one conceptual column receives read or write signal but only one chip receives both. In the general case, two sets of RAM stack footprints may be desirable. One set offers distinct chip selects and common read/write signals. The other set offers common chip select and distinct read/write signals. This is obviously more work for minimal gain. However, height, decode fan-out or other property may favor the obscure case. There is also the useful case of byte write/word read in a video system. This allows character or palette value to be written sequentially from 8 bit bus while allowing wider reads when rastering. This requires separate data lines and write signals but allows all other pins of a RAM stack to be shared.
While working on
XR0284 stub memory module, I accidentally discovered a very helpful technique to maximize board area and maximize project success. It is possible to make DIP pads which are enlarged on the inside only. These are easier to solder but do not increase the perimeter of a DIP component. In particular, where socketed 0.6 inch wide DIP is placed over unsocketed 0.3 inch wide DIP, both may have pads which are enlarged on the inside only. The disadvantage to this technique is that a board must be rotated to solder both sides. However, this is only required by people who are inexperienced at soldering. Experts may ignore the convenience feature and solder without rotation.
Finally, daft backronyms for TUIT100 are highly encouraged.
