Typical dpi of dot matrix printers available in the 1980s
- 60, 72, 80, 90, 120, 240 and 360 dpi
- 72 dpi for 8/9 pin and
- 120 dpi for 12/24 pin printers.
- Multiples thereof by printing a line twice.
The Long Read:
However, it's difficult to determine exactly where the individual pixels of the monochrome image lie...
Which of course is intended, after all, they should give letters, not dot-clouds.
This alone can't account for some of the problems I've had trying to restore an original raster of the image... the width and height of features don't appear to be multiples of some common factor.
They rarely are as the horizontal resolution is not defined by pin distance but timing. Of course, 'square' resolutions were offered to simplify bitmap graphics. Usually 72 dpi.
Now, my understanding of the mechanism that a dot-matrix printer uses for printing is that there will be a single vertical row of pins,
True for some, but others, like the excellent 24 pin NEC P6/7 used multiple columns (two in case of the NEC). Each moved a bit, so they could print in high resolution and part overlapping without running twice over the same line.
and then a stepper motor moves that array horizontally across the paper. Unless great care was taken in the engineering of the mechanism, there's no reason to expect that the distance between the pins vertically would equal the shortest distance the array can be moved horizontally.
No. The printer does not move a minimal distance, stops and then prints, but moves continuous, only stopping when there is no more (buffered) data. To generate different DPI the head is simply moved at different speeds. Thus horizontal resolution is mostly a software thing (*1).
For example the quite successful (since cheap and good) STAR LC-10 of the mid 1980s used a 9 pin head with a vertical pitch of 72 DPI, but offered horizontal resolutions of 60, 72, 80, 90, 120 and 240 DPI. The increased resolution was achieved by moving the head at different speed (*2), so for 120 dpi the head simply moved half as fast than at 60, and again half of that for 240.
The NEC P6 offered similar resolutions up to 360 dpi - so technically on par or even better than laser printers at the time.
There's also no reason to expect that the horizontal movement would be a minimum of the width of a pin... overlaps might be possible (speculation on my part, I don't know if this was ever offered, or what the point would be).
While the very early ones (70s) simply outputted a 7x5 matrix, higher resolution did become standard during the early 80s. The mentioned LC-10, using a 9-pin head, offered various character sets between 9x9 with next to no overlap - great for fast listings - all the way to 18x23, usually called NLQ or Near Letter Quality. Here the head only printed in one direction and used overlapping not only horizontally due slow movement, but as well vertically by advancing for only half a pin height (1/144 inch).
So, what shapes would these pins be? Were they square or round?
Most often round or close to round, as this gives the best edges. Especially early on. Having a round pin gives better movement and less need for fine details manufacturing the head. A round hole is the most simple one to drill.
Would they ever be wider than tall (or taller than wide)?
Yes, there were rectangularish (horizontal wider) ones as well, as this ends up giving a better picture to thin vertical lines, while at the same time blending in well on overlapping dots.
Did the mechanism make overlapping of the individual pixels a possible, and in what circumstances did that occur?
As said, horizontal overlapping was easily done by reducing print head speed. So this could be done even early on and with quite cheap mechanics.
Vertical overlapping needed a good paper transport, or a print head with overlapping columns of pins. Either way they came rather late and were more expensive due the need for more exact mechanics.
What number of pins were available commercially?
Anything from 7 pin with the original Centronics 101 of 1970 to 36 pin ones in the late 1980s. Sizes I remember are 7, 8, 9, 12, 18, 24, 36.
Most common during the early 1980s were 8/9 pin, some with double printing for effective 16/18 line charsets, like the quite successful Epson MX-80/100 (*3) and the later Star LC-10 - some in use up today. During the mid to late 80s 24 pin printers like the NEC P6/P7 became quite popular around the world as replacement (or cheaper alternative) to type head printers.
While high count pin printers were successful in Europe and the US, their true home was Japan, as high resolution is mandatory to print their finely detailed glyphs. Thus printers with up to 36 pins were created and sold in good numbers, but rarely seen outside Asia.
Long Story Short:
The question sounds as if you're working on a look-alike emulation for some old printer. To close in you may need at least a manual of the printer used with hardware specifications and modes offered. Using this and round dots (with some blurring) will get you close.
For more detail you may need to acquire a printer (or at least the print head) to check the pin shape, orientation and exact measures. This will get you close to a perfect cleanroom emulation.
Next would be to make it lousy again. After all, these pins did not really fire in sync. Due basic component variance (there are coils and analogue amplifiers) but as well rust and component aging, fire time will vary, giving the very common sight (back then) of having a single line moved a bit to the left or right.
*1 - Embedded that is, as only very few printers allowed to fine regulate the head movement via host commands.
*2 - To be correct, this is not entirely true, as the 72 and 80 dpi resolution was done at the same speed as 90 dpi, but with slower firing.
*3 - I wouldn't be surprised if the MX-80 is the most OEM'ed device ever. Even IBM rebadged it as 5152 for the IBM-PC (5150).