Honktown wrote: ↑Fri Oct 18, 2019 4:59 am
foamy wrote: ↑Fri Oct 18, 2019 4:19 am
Why not just read the input side of things, and add exactly as many fuel cells as there are reactors, every time you activate the inserters? Loop the belt back in to the feed via a priority filter splitter so's any excess is preferentially consumed and you can send the used cells on their way from there. Basically a sushi belt with just one item per lane. Use a priority splitter to feed a patch of yellow belt so that the cells are guaranteed to try and fill each piece of buffer in turn.
That should put your total belt buffer for the reactor proper at just 4 cells per reactor, which is probably acceptable. You'd want to do some kind of manual start routine to feed it with to that level initially, but that's not tremendously hard.
I thought of something which is even better and takes no manual setting: insert a lead item onto the belt and take it off at the end. As the lead item passes over the front of the inserters, insert a fuel cell onto the belt. It will be exactly timed so every space in front of an inserter has a fuel cell at the same time (one inserter swing from when the lead item was detectable on the belt). This synchronizes the inserters, so they can either be all inserting at once, or all picking up and disabled at once, with no extra fuel cells on the belt.
Edit: I was thinking about spacing vs inserter timing.
As long as the reactor inserter arms are more than a tile apart, a basic inserter is fast enough for cells. The cells will all be placed ~1.125 tiles after the lead item (1.2 second round-trip swing time, 1.875 tiles/second, one cell per swing), which makes them out-of-sync, but they're all delayed the same amount, so they're still together. To keep them within-tile takes a blue inserter (80% of the way down a single yellow belt tile round-trip), but it's not necessary unless the inserter arms are one-tile close.
Edit 2 corrections corrections corrections. You need to use a fast inserter and yellow belts.
While that'd be cool I don't trust tick-precise belt/insert synchronization. Stuff on belts can do weird things when it hits splitters and inserters have all sorts of weird things to trip up on. Sooner have local buffers and pull from that, and guarantee they're full via circuit network, although TBH I just monitor the buffer on the feed belts. As long as the belt is fully loaded on entry the reactor layout beyond that guarantees the reactors will all have fuel, and if there's a cycle where one doesn't because of a fuel shortage, it'll be corrected on the next one.
Not that a fuel shortage is
likely. 200 reactor-seconds with full neighbour bonuses is 32GJ. A single L3 assembler, no modules, can give you 15,000 reactor-seconds per minute, or, in other words, it can power
250 reactors. Assuming no modules anywhere in the chain, that needs ~110 uranium mines and ~38 centrifuges to drive things. If you need 40 GW of power, I think that's not really a major deal.
If the whole setup excepting the mines is prod&beaconed, that drops to 60 mines and 5 centrifuges (and uses less than a fifth of the capacity of the assembler :v). Speed-moduling the mines brings that down to around 25. And that's before mining productivity comes in!
Nuclear fuel is
really cheap once you have Kovarex going, all told. There's arguments to be made for not bothering to instrument a nuclear plant
at all and just continuously feed it fuel because who cares. It's
inelegant, but if your factory has a known average draw you can just build to that and let accumulators handle the smoothing.