While this just looks so lovely for science packs has anyone considered using this for anything else? Like make a iron/copper sushi belt for green circuits as a very simple case. The red circuits and blue circuits straight from plates for advanced use.
Re: Beautiful Sushi Belts
Posted: Mon Sep 19, 2022 1:45 am
by BicycleEater
I've built a version which divides the belts using splitter-logic, so that they naturally give 1/3 or 1/4 as relevant. This means it doesn't suffer from items missing.
It can have issues if the sushi output is disconnected, but with about 3 looparounds installed, I haven't been able to create any broken cases.
It does have some circuit network (1 wire), just to empty the loops out a bit if they get overfull, which can happen when the sushi belt is cut. It should be unnecessary (the priority splitters should do the trick, but can often filter out the wrong item, so this works way better).
Its very similar to many of the other sushi designs seen here, but with many layers of over-fill/underfill protection, and a tiny bit higher throughput by running one side of the belt a little denser than otherwise. Specifically loopback splitter-ratio sushi belts can end up with their loopback on the splitters getting full, resulting in the belts hopping a little, which takes a very long time to recover - I'm not even sure it always can, and resulting in very slightly higher ratios of that item in the belt. By adding the priority splitters at the end, this excess causes the belt to overflow a little. The problem is it doesn't always overflow with the right item, so this triggers a section of the sushi to be dumped into the overflow section.
This in turn goes into the input with priority, and so goes back into the system. The next problem is that this overflow often has a lot of the same item, so I added bypasses for the filter splitters, and a loop-around of the overflow. This loop also captures any items which should not be in the sushi. This loop should be very low utilisation, so even with blue belts, a yellow looparound should suffice, and could be shared across multiple sushis.
There is one assumption in the Sushi belt that isn't always true. Actually mostly not true in fact.
The designs here try to put equal numbers of every science packs on the belt. But the consumption isn't equal numbers. Not every research needs every science pack. And sometimes they need more of one than the other.
To make the belts perfect this should be added to the design. Since this forum is about circuit-free solutions this would have to be something you turn on/off manually. E.g. connect or disconnect an extra red science pack belt for the 2x red researches.
Since this forum is about circuit-free solutions this would have to be something you turn on/off manually.
I'm not so sure about that.
Using a feedback measuring system, and clever belt jamming techniques, you could conceivably vary the levels of potions until the number received was proportional tot he number sent. You could do this with each pack holding at least say 1/32th of the belt, and then giving extra 1/32ths to the sciences which have fewer returns.
I'm imagining some combination of other items circulating, etc. such that a small rate of returns is acceptable, but more will cause the belts carrying that item to be 'poluted' by e.g. iron, which is filtered out at the last minute, resulting in lower throughput.
The presence of these items could be used to cause a belt of the polutants to overflow, resulting in polutants being put on the belt.
I'll put some effort into designing such a system.
One problem is ensuring that the total amount adds to 1 - i.e. the belts don't overflow (or underflow) the sushi capacity.
To do this I'd guess you'd have to have a single stream of polutants, and divide it up amongst the sciences in proportion to their return rate. One issue would be avoiding oscilation, so I'd guess you'd have a timer, which was the same length as the sushi-return, resulting in the return rate being measured relative to the original send rate.
Re: Beautiful Sushi Belts
Posted: Mon Sep 19, 2022 8:52 pm
by BicycleEater
Well, I've made it work - at least kinda. It oscilates massively, overfills its output belt, etc. But it is a belt only self adjusting sushi-belt system.
Note that it only uses yellow belts, with some reds for length purposes.
I have a huge number of improvements I'd want to make, but this is kinda version 1 at least.
Anyway, while pretty much everything is possible (did you try to build a computer out of belts and splitters?) if you have complex needs belt-only is hardly for you. Too much hassle for too little reward usually.
While this just looks so lovely for science packs has anyone considered using this for anything else? Like make a iron/copper sushi belt for green circuits as a very simple case. The red circuits and blue circuits straight from plates for advanced use.
Yeah I've been working on that.
From what I found out it works very well if you only feed a few assembly machines (like one or two).
The most common issue to fight with is the fact that the pseudo-random way the inserters will pick items up might lead to a succession of more than 4 items (the exact number being 4 plus the internal buffer for this particular recipe) which will stop the last machine, thus turning its neighbor into now being the last machine which spreads this clogging uphill.
One of the solution I found was to feed the end of the belt back into itself, some sort of weird mix between a loopless sushi and a looped sushi:
And it works flawlessly for many recipes if not all of them.
While this just looks so lovely for science packs has anyone considered using this for anything else? Like make a iron/copper sushi belt for green circuits as a very simple case. The red circuits and blue circuits straight from plates for advanced use.
Yeah I've been working on that.
From what I found out it works very well if you only feed a few assembly machines (like one or two).
The most common issue to fight with is the fact that the pseudo-random way the inserters will pick items up might lead to a succession of more than 4 items (the exact number being 4 plus the internal buffer for this particular recipe) which will stop the last machine, thus turning its neighbor into now being the last machine which spreads this clogging uphill.
One of the solution I found was to feed the end of the belt back into itself, some sort of weird mix between a loopless sushi and a looped sushi:
And it works flawlessly for many recipes if not all of them.
That's because you are counting and timing it wrong.
In your picture you have 1 pipe and 1 gear alternating. But thanks to inserter bonuses the inserter will pick up more than one item. Alternating items means the inserter will wait longer to get a second gear so that is inefficient.
Next you know the inserter will pick up N gears, then swing before coming back and picking up M pipes. So you need to leave a gap between the gears for the first assembler and the pipes for the first assembler. You can fill it with gears for the second/third/fourth assembler. It's all deterministic (assuming you have no backlock or power shortage throwing off the timing) so you can figure out the exact item pattern to put on the belt so every inserter will have the right items in front of it the moment it wants to pick them up. But that be tideous to figure out exactly.
The easiest solution I can think of preventing a deadlock would be to first send gears for all assemblers. Then send pipes for all assemblers and last send steel for all assemblers so they produce engines. This works as long as you can send all the required steel faster than the first assembler will produce an engine. So the length of the assembly line is limited by the recipe time (and really short for e.g. electronic circuits). It's easy to figure out the perfect pattern for the inserters and a single item.
Side note: Is it still a sushi belt if you send 200 gears followed by 200 pipes followed by 200 steel and then repeat?
Re: Beautiful Sushi Belts
Posted: Mon Mar 20, 2023 2:59 pm
by FuryoftheStars
(Disclaimer: I don't use sushi belts)
In the image above about the engines, considering the gear:steel:pipe ratio is 1:1:2, would it work/make sense to just have one lane of the belt be all pipe, and the other lane be mixed gear:steel?
While this just looks so lovely for science packs has anyone considered using this for anything else? Like make a iron/copper sushi belt for green circuits as a very simple case. The red circuits and blue circuits straight from plates for advanced use.
Yeah I've been working on that.
My question was about doing it with multiple recipes though. So you start with a sushi of plates (or even ores) and I guess gaps and pass it by assemblers with different recipes. One takes iron plates and outputs gears, the next pipes, until finally you produce engines. At the end you remove the engines from the sushi belt for consumption and loop back anything else.
Doing this for multiple recipes and the output going back on the belt add an extra level to the complexity.
Re: Beautiful Sushi Belts
Posted: Fri Apr 12, 2024 8:07 am
by xaetral
The thread is nearly 5 years old and reached 100k views, looks like the perfect time for me to share my latest tech.
I edited the first post, feel free to check it out
Re: Beautiful Sushi Belts
Posted: Fri Apr 12, 2024 9:51 am
by Tertius
It's still looking beautiful, however I ceased using sushi belts for science production. On the green table and as long as every science pack is fully supplied, everything runs nice and fine.
However, if for some reason one science pack isn't produced any more and the labs stop research and you don't notice this in time, the belt loop will slowly fill with every science pack still available until it's completely filled. If you now fix the science pack production, there is no space on the belt to insert the new packs. Theoretically, it works because splitters alternate between both inputs, but actually the splitters stall and output no items, because the amount of returning items from the belt loop have no gaps, and you're trying to add additional items. This happens, because the sushi belt is a loop. You have to remove a significant part of the items from the belt loop until the stalling doesn't happen any more. It also cannot fix itself, because space on the belt can only appear if some packs are consumed, but there are no packs begin consumed by the labs, because one pack is missing for the current research.
So sushi belts need buffers such as buffer chests for the returning items. The belt loop with its gaps as buffer isn't sufficient.
However, if for some reason one science pack isn't produced any more and the labs stop research and you don't notice this in time, the belt loop will slowly fill with every science pack still available until it's completely filled.
That's why I made the new design I just talked about, it now features a rate limiter on the output so that no science pack fills more that 1/4 of a side belt (and 1/3 of the other one).
Hence why this update is great.
Re: Beautiful Sushi Belts
Posted: Fri Apr 12, 2024 1:25 pm
by FuryoftheStars
2.0 may make this easier to accomplish, too, with the full belt reader.
2.0 may make this easier to accomplish, too, with the full belt reader.
What's easier than circuitless sushi? XD
Have you tested it under the concept of one of the science packs running out completely and then making sure it can re-insert even if the current belt completely empties out of it and then fills up with everything else?
(As an unrelated side note, and not something that has to be done (simply because it's a personal pet-peeve of mine), but on each of those splitters with one output side disconnected, I'd set an output filter (my preference is the decon planner) to prevent those few, poor science packs that end up being in a place where they could never, ever be used. )
Have you tested it under the concept of one of the science packs running out completely and then making sure it can re-insert even if the current belt completely empties out of it and then fills up with everything else?
Of course, that's exactly the nature of the improvement
Re: Beautiful Sushi Belts
Posted: Fri Apr 12, 2024 3:07 pm
by Tertius
I'm sorry, but the added setup will also stall on the outer lane, if just research is stopped one or two times for a moment so the blue/purple packs accumulate in the splitters of the returning belt. It's worse than other solutions due to the sideloading, because the mixing of the lower 4 packs don't have any rate limiting, so the blue pack has first prio, purple next, yellow next, white last. This will create sections on the belt without any pack, so labs will not pull items from these, and if these sections return, they clog and block the splitter section.
To unstall, you have to remove a significant amount of packs from the belt manually.
Re: Beautiful Sushi Belts
Posted: Sat Apr 13, 2024 10:00 am
by xaetral
Just because the research is stopped doesn't mean the belt is
I've added the blueprint to the post, feel free to try it out to see the limits by yourself.
As long as the science packs arrive on the correct side of the correct belt and the sushi belt itself is a continuous piece of express belt, it should never ever break, whatever input or output pattern you have.
This is a generic problem with every sushi belt - not just your beautiful mixer implementation. The handling of the items returning from the loop is not trivial.