A Bad Casting Day

Yesterday I made a try at casting a font of 12-point Caslon Old Style using composition mats and the computer interface for the caster.The results were less than stellar. A whole bunch of things went wrong:

1. The matrices and/or mould are worn so type casts with bits of flash all around the face and a bit down the side corners, making it stack into the type channel unevenly.
2. The H airpin became stuck up by about the third line of type, which meant that about half the letters cast the matrix from column H instead of the desired matrix. So I got plenty of em-dashes, lowercase y, apostrophes, etc. They were, however, casting from the correct row in the matcase and so also casting to the correct width. This seems to be due to the valve in my interface becoming stuck partly open, but I’m not sure why. I tried the interface (without running the caster) today and it worked fine.
3. Lines with many narrower casts ended up longer overall than lines of mostly em-sized casts. I suspect this is due to having a worn Normal Wedge. These tend to wear more in the center, where the size for intermediate-width type is set, causing these to cast wider than they should. In mixed running text this would not be much of a problem, but when you are casting a font and can get a line of m’s and M’s followed by a line of i’s it is very difficult to keep a stable line length.
4. About 2/3 of the way through the job I started to get type cast with bent tops, and eventually type with broken tops, or split open, even some sticking together from still-molten metal coming out of the broken type. This would happen when I got a long run of wide letters. This increases the average metal flow and so increases the cooling load, heating up the mould and causing type to be ejected before it is completely hard.
5. All that sticking and broken type caused the entire column of type to get skewed as partial lines pushed onto the galley.
6. There still seem to be some issues with the software and communication between the caster and computer. Although the job ran to completion, when I was fiddling around afterwards (looking into that stuck H airpin) the computer lost the ability to see the caster signal for a new ribbon code.

So I have to try a few solutions to these problems:

1. I may tear down and rebuild the mould, in a separate workshop free of cat hair. I will also verify that the bridge carrying frame height is set properly.
2. This problem magically went away the next day, so I don’t know if I should open up the air valve to check for dirt blocking the seal.
3. As long as I don’t mind a quad being exactly 12 points wide, I can adjust the casting width to make all the casts a tiny bit narrower, which will have more of a shortening effect on the lines that contain many narrow letters and so make the line lengths match up better. In the long run I have to get some normal wedges that are not so worn.
4. I hope that I can do some combination of reducing the metal temperature, slowing the caster, and increasing the mould cooling water to ensure the large type comes out solid without causing problems like nozzle freezing when a long run of smaller type is cast. I should also check how much the water in my closed-loop cooling system has warmed up through the job; if the water going into the mould is hot it will not cool the mould as effectively. This might explain why it was only a problem later in the job, though that could also be due to there being more long runs of larger letters closer to the beginning of the alphabet.
5. This will fix itself once everything else is fixed
6. I have reviewed the software again, looking for things like poor I/O error recovery and concurrency issues, made a few changes, and will try out the new version next time.

When looking into the incompletely-hardened type problem I also found that my metal thermometer seems to be broken. Its lowest reading should be about 350°F, and the metal should be around 650°F but I could not see any mercury (assuming that is what is used in these thermometers) column at all.