Max-moduled stone brick smeltery
Posted: Sun Nov 26, 2017 11:01 pm
I've been working on a series of max-moduled mechanical smelting designs for a while now. Many of them are kind-of stuck due the limitations of 0.15 splitters but I've found that if I slightly overproduce and join lines with "a bit more" than is needed then I can achieve full compression quickly and reliably. This is an example of that mechanic at work. If we replace the productivity modules in the smelters with speed modules (or provide precisely 8/5 blue belts of stone), we would in theory get fully compressed output with no buffering at full tilt with this design, but, alas, it will tend to sputter and cough for astronomically long periods before settling down into a fully compressed rhythm (if it does so all). It's hard to be sure if this is due to splitter misbehavior or an intrinsic issue with my design but I strongly suspect the former.
Some aspects of the implementation might seem kind-of wierd but there are reasons for most of the design choices I made. For now I won't elaborate too much, for fear of tldr problems. I'd be happy to consider any ideas or answer any questions.
One thing worth mentioning is the decision to feed the bottom row separately from the top two rows, which share the same belt (but not lanes) for input and output. If we tried to put the ore for all three smelters on one belt-lane, we create a maths problem. If we want each column of smelters to produce 1/3 of a belt of output this will require 2/3 of a belt of input. Obviously (but not obviously enough that I failed to discover this the hard way) only one belt-lane cannot provide this. Productivity modules almost-but-not-quite rescue us from this problem -- each column would require 8/15 belts of input, just 1/30 more than one lane can hold. But the result is less than satisfying. We could solve the problem by sharing input and output lanes; but even if we get rid of or replace the productivity modules, inevitably, a bit of raw stone slips through. I did manage to produce some designs which solved this using belt circuit conditions, but that kind of violated the no-circuits constraint I was trying to work within, and resulted in a big ugly mess. The three splitters and reliance on backpressure seemed like a lesser concession. Earlier designs used a second layer of 1->3 splitters and routed ore to each smelter individually which worked fine but seemed like overkill; this seems to work just as well and is presumably much more UPS-efficient.
- a max-moduled stone smelter
- MaxModuledStoneSmelter.png (2.99 MiB) Viewed 9920 times
Blueprint
I have included an input-lane balancer, as otherwise this thing consumes ore in a profoundly lopsided pattern; but this won't matter for many applications; in the picture above the first five rows of tiles (and resulting stray underground) could be removed to omit this feature and preserve a tiny smidgen of UPS.Some aspects of the implementation might seem kind-of wierd but there are reasons for most of the design choices I made. For now I won't elaborate too much, for fear of tldr problems. I'd be happy to consider any ideas or answer any questions.
One thing worth mentioning is the decision to feed the bottom row separately from the top two rows, which share the same belt (but not lanes) for input and output. If we tried to put the ore for all three smelters on one belt-lane, we create a maths problem. If we want each column of smelters to produce 1/3 of a belt of output this will require 2/3 of a belt of input. Obviously (but not obviously enough that I failed to discover this the hard way) only one belt-lane cannot provide this. Productivity modules almost-but-not-quite rescue us from this problem -- each column would require 8/15 belts of input, just 1/30 more than one lane can hold. But the result is less than satisfying. We could solve the problem by sharing input and output lanes; but even if we get rid of or replace the productivity modules, inevitably, a bit of raw stone slips through. I did manage to produce some designs which solved this using belt circuit conditions, but that kind of violated the no-circuits constraint I was trying to work within, and resulted in a big ugly mess. The three splitters and reliance on backpressure seemed like a lesser concession. Earlier designs used a second layer of 1->3 splitters and routed ore to each smelter individually which worked fine but seemed like overkill; this seems to work just as well and is presumably much more UPS-efficient.
[edit: this seems to depend on quasi-random start conditions, presumably i.e., unsmelted bits of ore sitting in the smelters, etc. With further testing, I've seen times as low as 22 and as high as 27]
I tested with the rightmost column favored and it achieves backpressure in about five minutes, but has trouble compressing the output, possibly due to splitter bugs or complex polyrhythms at the final join. Almost an hour in it still could not sustain compression. Probably one could attempt to fight this by screwing around with the inserter stack sizes but my conclusion is, as I expected, the input-side balancer is apparently doing something useful in v1.03 above.