How to divide by 5 with conveyor belts

[buchanant]

This problem has been daunting me for a year. I finally figured it out. First split by 2. Then split each of the 2 by 3. So that you have 1 2 3 4 5 6 splitter nodes. Splitter nodes 3 and 4 divide by 2 and combine 2 of the splits into a 3.5 combiner node and the other 3.5 split feeds back into the initial 2 splits. This evens out all the resources to 5 exits. Enjoy!

To simplify this down, this can be accomplished with 5 splitters and 3 mergers.

https://steamcommunity.com/profiles/76561198004665858/screenshots/

[Stress Tensor]

Originally posted by hurenogeko:

here is my design laid out more better visibility
https://steamcommunity.com/sharedfiles/filedetails/?id=2920109935

Yep - looks good. Not sure how I was mis-seeing the other one. Thanks.

[Die Hand Gottes]

I build more of a band usually also a sushi band
https://steamcommunity.com/sharedfiles/filedetails/?id=2877842726

[Huren Ogeko]

Originally posted by Die Hand Gottes:

I build more of a band usually also a sushi band
https://steamcommunity.com/sharedfiles/filedetails/?id=2877842726

that usually called a manifold. its the easiest way but some people want to split evenly which is why we get posts like this. I prefer manifold setups 99% of the time but every now and then there is something nice about splitting things up evenly no matter how much space it needs.

[Die Hand Gottes]

With me, it is also evenly distributed, you just have to calculate the product line exactly, I also make sure that I rather underclock and save power.
this splitter balance I have never liked and never used.

[DaBa]

Originally posted by hurenogeko:

Originally posted by Die Hand Gottes:

I build more of a band usually also a sushi band
https://steamcommunity.com/sharedfiles/filedetails/?id=2877842726

that usually called a manifold. its the easiest way but some people want to split evenly which is why we get posts like this. I prefer manifold setups 99% of the time but every now and then there is something nice about splitting things up evenly no matter how much space it needs.

Which I still do not understand, there is no practical reason to perfectly split the belts rather than using the manifold. Both accomplish the same objective, the only difference is that manifold takes significantly less space and is easier to scale up at a cost of needing a bit of time to sort itself out initially. And even that last part can be avoided if you can provide the initial buffer upfront.

So... why? Do people just not understand how those things work, or is it purely for aesthetical reasons? Or maybe just a leftover habit from Factorio where balancers actually help in a lot of situations.

[Die Hand Gottes]

yes no idea maybe too lazy to calculate it perfectly for that the game has an ingame calculator and also use an external one.
you just have to calculate the product line from the beginning to the end product exactly but is better than this splitter Gedöns what is not even so flexible I calculate partly to 4 digits behind the decimal point and that works great.

[Stress Tensor]

Originally posted by DaBa:

Originally posted by hurenogeko:

that usually called a manifold. its the easiest way but some people want to split evenly which is why we get posts like this. I prefer manifold setups 99% of the time but every now and then there is something nice about splitting things up evenly no matter how much space it needs.

Which I still do not understand, there is no practical reason to perfectly split the belts rather than using the manifold. Both accomplish the same objective, the only difference is that manifold takes significantly less space and is easier to scale up at a cost of needing a bit of time to sort itself out initially. And even that last part can be avoided if you can provide the initial buffer upfront.

So... why? Do people just not understand how those things work, or is it purely for aesthetical reasons? Or maybe just a leftover habit from Factorio where balancers actually help in a lot of situations.

In general I agree with this sentiment, and I also use manifolds almost exclusively. But I do see two reasonable use cases for load balancing, both of which are related to nuclear power:

The first (lessor one) is where the output numbers of one stage map pretty squarely with the inputs of the next stage (either even, or 2-to-1, 3-to-2, or something fairly simple). I say "lessor" because only in the case of even output to input numbers will the complexity/space be better with load balanced than manifold, but in some other *simple* mappings the increase in complexity and space usage may be acceptable if fast ramp-up is desired. With the _very_ slow rate of production of nuclear fuel rods, with a manifold you will be waiting a fairly long time for saturation of the first nuclear power plants before you reach stable power generation of all power plants, much worse than with coal or regular fuel power.

The second is what I mentioned previously in this thread where someone recommended it again for nuclear items but for a different reason. As we all know, in manifolds the first machines in the line will accumulate in their input buffers until they are full before the later machines in the line will run at 100%. With radioactive items, you may not want them to hold this excess input which would unnecessarily increase your radiation exposure. Load balancing, if done right, can eliminate this unnecessary input buffering. Having said that, I did not implement load balancing for my nuclear setup because I did not find (yet) the radiation to be that big a deal.

[Die Hand Gottes]

With the nuclear power I make also simply a distributor have with it no problem is worse overclocked reactors with water to supply there one must operate much expenditure that the stable run however not because of the fuel rods.
https://steamcommunity.com/sharedfiles/filedetails/?id=2919803117

[Mister Fabulous]

Load balancing vs. manifold is certainly personal preference. I like to load balance just the assemblers for things like modular frames in the early game because I make 10/min. That's gonna take a while if I don't preload the assemblers.

To reiterate the nuclear example, late game if you're going big and build something like 200 reactors, those are going to take hours to ramp up and it's going to produce a large radioactive area because of the concentration of rods in so many of the reactors and on the belts if you care about that as well.


I'd like to comment on kluns post. It's the same as the non-bottlenecked version in the wiki reference I posted. If you go with the version that uses only 1 merger and 3 splitters, the belt between the merger and first splitter needs to be able to handle 120% of the incoming rate. So if you're doing something with a single mk 5 belt you want to split 5 ways, you can't have more than 650 coming in otherwise it will start to back up. The version that uses 2 mergers and 4 splitters doesn't have a restriction on the incoming rate.

Both setups work on the principal of splitting one belt looped back 6 ways over and over and adding that to the other 5. 1/6 + 1/36 + 1/216 + ... Adding 1/(6^n) from 1 to infinity = 1/5. Yay math! I didn't figure it out, just wrote a line in Java that was close enough:

IntStream.rangeClosed(1, 100).asDoubleStream().reduce(0.0, (a, b) -> a + (1.0 / Math.pow(6, b)));

EDIT: Thanks buchanant. Took a little bit to wrap my head around it. Each side from the first splitter looks like the splitting result 20%, 20%, 10% of the original input and you're combining the two 10% belts from each side. I couldn't see it at first. It's a little more bulky than the others, but yup it works exactly as advertised.

[Huren Ogeko]

Originally posted by DaBa:

Originally posted by hurenogeko:

that usually called a manifold. its the easiest way but some people want to split evenly which is why we get posts like this. I prefer manifold setups 99% of the time but every now and then there is something nice about splitting things up evenly no matter how much space it needs.

Which I still do not understand, there is no practical reason to perfectly split the belts rather than using the manifold. Both accomplish the same objective, the only difference is that manifold takes significantly less space and is easier to scale up at a cost of needing a bit of time to sort itself out initially. And even that last part can be avoided if you can provide the initial buffer upfront.

So... why? Do people just not understand how those things work, or is it purely for aesthetical reasons? Or maybe just a leftover habit from Factorio where balancers actually help in a lot of situations.

I rarely do it now bur in the past I used to prefer balancers mostly as an ecstatic or just because I wanted a challenges with some of the crazy ratios you can find in the game.

These days I am nearly 99% manifolds I prefer to do direct connections where I can. say I have 15 constructors feeding 5 assemblers. I will pair them up 3 constructors manifold into a single assembler rather then have all 15 on a single manifold feeding 5 assemblers on yet another manifold.

[Wolfgang]

I am already so glad I bookmarked this thread. I will certainly use one of these layouts in my factory.

[The Big Brzezinski]

If at all possible, I would use conveyor rate splitting. To split 150 into five belts of 30, say, I'd use a smart splitter to saturate a 60 belt and overflow the remaining 90 to a 120 belt. Then I just need to split the 60 in two and the 90 in three.

[kLuns]

I appreciate the input and it seems we are a little divived on this subject.

Originally posted by buchanant:

kLuns, that looks like it will have unequal distribution. Half goes left and half goes right. Left puts out 1/6th per outlet and recycles 1/6th to split between left and right, resulting in 5/24th per outlet. Right puts out 1/6th per outlet + the recycle from left, which put out 7/36th per outlet. It doesnt add up to me.

I'm not sure if that's what happens.
let's figure this one out:

All outputs get 1/6.
5/6 Gets used by the connected machines
One feedback splits 1/6 to 2* 1/12 into both mergers
Both splitters split those 1/12 into 3 equal parts of 1/36

This 6/36 (all outputs total) gets divived into 30/36 into machines and 6/36 into redividing.

Originally posted by The Big Brzezinski:

If at all possible, I would use conveyor rate splitting. To split 150 into five belts of 30, say, I'd use a smart splitter to saturate a 60 belt and overflow the remaining 90 to a 120 belt. Then I just need to split the 60 in two and the 90 in three.

This is a good method and doesn't require much space.

[DaBa]

Originally posted by Stress Tensor:

Originally posted by DaBa:

Which I still do not understand, there is no practical reason to perfectly split the belts rather than using the manifold. Both accomplish the same objective, the only difference is that manifold takes significantly less space and is easier to scale up at a cost of needing a bit of time to sort itself out initially. And even that last part can be avoided if you can provide the initial buffer upfront.

So... why? Do people just not understand how those things work, or is it purely for aesthetical reasons? Or maybe just a leftover habit from Factorio where balancers actually help in a lot of situations.

In general I agree with this sentiment, and I also use manifolds almost exclusively. But I do see two reasonable use cases for load balancing, both of which are related to nuclear power:

The first (lessor one) is where the output numbers of one stage map pretty squarely with the inputs of the next stage (either even, or 2-to-1, 3-to-2, or something fairly simple). I say "lessor" because only in the case of even output to input numbers will the complexity/space be better with load balanced than manifold, but in some other *simple* mappings the increase in complexity and space usage may be acceptable if fast ramp-up is desired. With the _very_ slow rate of production of nuclear fuel rods, with a manifold you will be waiting a fairly long time for saturation of the first nuclear power plants before you reach stable power generation of all power plants, much worse than with coal or regular fuel power.

The second is what I mentioned previously in this thread where someone recommended it again for nuclear items but for a different reason. As we all know, in manifolds the first machines in the line will accumulate in their input buffers until they are full before the later machines in the line will run at 100%. With radioactive items, you may not want them to hold this excess input which would unnecessarily increase your radiation exposure. Load balancing, if done right, can eliminate this unnecessary input buffering. Having said that, I did not implement load balancing for my nuclear setup because I did not find (yet) the radiation to be that big a deal.

Oh yeah, I have totally forgotten about that one, haven't played the game in a while so maybe that's why, or that I have done nuclear power only a single time so far. It is a good example!

[Mister Fabulous]

The main argument for load balancing in my opinion is ramp up time for a factory to run at 100%. How long do you want to wait? For items that are produced at a low rate and need many buildings if you're not overclocking and also take certain items at a low rate as well like plutonium rods, super computers, regular computers (crystal computers in this example), it's going to take a long time for the input buffers to fill even with small stack sizes. Yeah it'll get there, but sometimes you just gotta have it !!!NOW!!! and balancers in all their forms get you to 100% right away.

Originally posted by The Big Brzezinski:

If at all possible, I would use conveyor rate splitting. To split 150 into five belts of 30, say, I'd use a smart splitter to saturate a 60 belt and overflow the remaining 90 to a 120 belt. Then I just need to split the 60 in two and the 90 in three.

This one is a specific solution and needs 2 splitters and one smart splitter instead of 3 splitters and a merger assuming the bottlenecked version is applicable. I'm with you and just nitpicking because the 1:5 setups (and all prime splitter arrays) are general solutions that work in all cases at the cost of space. There are many other combinations where using belt rates to do the work can provide a simpler setup.

Balancers:
Also, you don't need a smart splitter. You can build a short length of a faster belt connecting to a slower belt at a splitter so you don't run into problems with a splitter deciding not to put an item on the slower belt you want to saturate because an items got to the splitter 1/2 a second sooner than expected and in that moment there wasn't room on the slower belt yet.

Here's an example of splitting 2 incoming belts of 600 to 3 belts of 360 and one belt of 120: https://imgur.com/QA60doE. If you try it out, the numbers in the containers at the end will be very slightly off due to the small bit of excess that is accumulated on the short fast belt segments.

Here's another, little harder to see belt types from the distance, going from 3 @ 240 to 4 @ 180: https://imgur.com/Dvestkt

Now those only work because the incoming belts have the same rates and the numbers work out nicely. The latter in my opinion is a nicer solution in that instance as opposed to the 3-4 balancer shown here (not my drawings): https://imgur.com/a/TnomMFk. Which is a general solution. Which is me making the same argument you did where a specific solution is better/smaller/cleaner than the general one. Where was I going with this?