Best way to split ore into foundries?

[orcs]

I now have access to tier 5 belts, and I am designing a megafactory.

Currently my design involves bringing raw ore to my megafactory for smelting.

I have watched many YouTube videos and people who build mega factories and I see so few people using splitters and mergers efficiently, so I am just wondering if maybe there is something I am misunderstanding.

If I want a bank of iron smelters operating as efficiently as possible (while keeping the overclocking low), then I believe I should be splitting the 720/min ore on my Tier 5 belts, into 3 lines, which will result in 240 / line. Then split that again into 3, 80's, and then into 2, 40's (smelters will be overclocked to produce 40/min). All in total, that 1 line of 720 will feed 18 smelters, all at 100% capacity.

However, I have seen many people who will just have a bank of smelters to either side, and will then just run a belt down the middle. However, this does not make sense to me since at every split, you are splitting the line load into thirds (which will include any overflow from prior split). This means that eventually, at the end of your line, you will have smelters starving for supply. When I work this out in a spreadsheet, I see that at the 7th row of smelters, they are now not getting enough ingots to produce 40 each. So this means that 12 smelters are at full capacity, and 2 are at about 66% capacity.

Does it not make more sense to split product off evenly, rather than just running it down a single feeder line?

[Suzaku]

Copying this from another thread a long time ago:

Manifolds work by splitting the available resources in half (or thirds, if you're using all sides of the splitter) over and over as the resources travel down a single line. This will eventually fill the first machine in the row, which then leaves the remaining resources to fill the next machine, and the next, and so on. Once a machine has been filled, it can then reliably draw at its normal rate of consumption. This process repeats the entire way down the line until, assuming you did the math right, all machines are running at max efficiency.

As a simple example:
You're producing 120 iron ore/min from a miner.
A smelter smelts iron ore at a rate of 30 ore/min. This means you need 4 smelters.
You place a row of 4 smelters. In front of each one is a splitter.
As the ore hits the first splitter, the resources are split; 60 ore/min to the smelter, 60 ore/min to the rest of the line.
The second smelter gets 30 ore/min from its spltter. The 3rd gets 15/min. The 4th gets 7.5/min.
As time goes by, the first smelter, being supplied with too much ore, fills up, and can now only obtain ore as quickly as it goes through it, meaning the original 60 ore/min it was getting now drops back down to the 30 ore/min that it can handle.
This process repeats to the second smelter, and the third, and finally the fourth smelter receives what remains, which is the precise 30 ore/min that it needs to function at full efficiency.

This method allows you to quickly set up simple, compact, flexible, expandable resource distribution in almost any situation. This can then be used in a similar fashion with mergers, as you then merge all of those ingots back onto a single belt to be moved to the next step in the process.

[Andromeda_Warrior]

Yea that seems about right. Mind if I help?

[Andromeda_Warrior]

I have great knowledge on this kind of stuff, but it would not hurt to use it here and there.

[DrNewcenstein]

The issue you eventually run into is time + distance.

The farthest machine from the source will always run low, while the ones closest will always back up, unless you place the machines equidistant from each splitter output, each one facing directly outward from the splitter's outputs (not in a row, but in a cross pattern).

[Suzaku]

Originally posted by DrNewcenstein:

The issue you eventually run into is time + distance.

The farthest machine from the source will always run low, while the ones closest will always back up, unless you place the machines equidistant from each splitter output, each one facing directly outward from the splitter's outputs (not in a row, but in a cross pattern).

Distance is a non-factor. All that does is determine how long it will take to eventually become 100% efficient, which is usually no more than just a couple minutes.

[tritscwa]

Perfect splitting's advantage is in start up time. Since you're properly splitting all resources as required, all the machines should reach maximum production immediately. However, this is a very short lived advantage as once the manifold split gets saturated, there is no effective difference between the two setups. So then it just comes down to ease of setup, and manifolds are generally just easier to setup. So in general, outside of early or trivial setups, most people tend to move to manifolds as their production lines become more and more complex as they, long term, have equal production and are far less of a headache to setup. Another advantage to manifolds is they are easily expanded. You may not need full production yet, and thus can just lay out the first few machines and some empty space for the rest, and it is easy later to extend the manifold to the full production you want. Perfect splits are generally far harder to do this.

[orcs]

I've worked out the math on this now, and while I thought the far more efficient system would be be even distribution, instead of the manifold system, it now appears they are the same. The manifold system is easier to set up though.

Example #1: Even Distribution. 720 splits 3 ways is 240 each. Then split this 3 ways is 80 each. Then split 2 ways is 40 each. This requires 18 smelters, overclocked to make 40/minute, for a total of 720/min.

Example #2: Dual Manifold. (For comparison sake, smelters are overclocked to 40)
1. 80 smelted. 640 passed down. Smelters 100%+ efficient.
2. 80 smelted. 560 passed down. Smelters 100%+ efficient.
3. 80 smelted. 480 passed down. Smelters 100%+ efficient.
4. 80 smelted. 400 passed down. Smelters 100%+ efficient.
5. 80 smelted. 320 passed down. Smelters 100%+ efficient.
6. 80 smelted. 240 passed down. Smelters 100%+ efficient.
7. 80 smelted. 160 passed down. Smelters 100%+ efficient.
8. 80 smelted. 80 passed down. Smelters 100%+ efficient.
9. 80 smelted. 0 passed down. Smelters exactly 100% efficient.

Dual Manifold also requires 18 smelters, overclocked to make 40/minute, for a total of 720/min.

[Suzaku]

I would've just went with 24 smelters at normal speeds, but yep, all of the math checks out.

[Maehlice]

Looks like you've got your answer about the manifold, but I'll add this:

Remember that the larger your manifold and the slower your initial PPM, the longer it takes to reach saturation and begin running at 100%

Bulk items like ores and ingots take just a few minutes to reach saturation, but slower items may take hours. In those cases, I highly recommend hand-feeding the machines' inputs as you go along programming them.

Similarly, I also recommend feeding the manifold and powering the machines as you build them versus waiting until it's all done. That way, construction and saturation coincide.

Finally, building the entire "trunk" of the manifold with the highest belt speed (same as the input) and using the slowest possible belt for each takeoff also speeds up saturation time.

[John Hadley]

The designs you are seeing are designed for maximum ease of layout and neatness, not for most efficient splitting behavior. Such designs are easy to copy and combine blocks of to make massive factories without having to manually lay out every machine. They are ideal for blueprint usage. This game doesn't have blueprints but if you use mods I think you can copy rectangular sections of factories with the area mod.

If you use parallel splitting solutions then you have to modify the splitter design depending upon how exactly how many machines you are going to use and the exact split that each machine needs. If you extend the line to add more input then you have to redesign the splitting solution. In a serial splitting solution you just assume that the input hoppers of everything will be saturated and in that case they will just take as much they can use and as long as you didn't put in more consumers than you can provide resources everything will be produced at its maximum rate. In practice you have to wait a few minutes for the input hoppers of such a factory to fill up before it will have the ideal split. Until then the factory will produce unbalanced outputs. This is the price you pay for the simplicity of the design and layout requirement.

You can decide whether you want to make a custom parallel splitting solution for each factory that is more difficult to duplicate and expand later or you want to easily expand your factory by using a splitting method with no thought required, but you have to have a bit of wasted materials sitting around in hoppers before the machines run the way you are intending.

[luntacarsus]

I learned the hard way do NOT use a manifold for uranium fuel rods. They each take two and a half minutes to construct and each reactor can potentially hold a stack of 50 - it will take longer than the lifetime of your save to get more than a few reactors to full efficiency.

[Obz3hL33t]

The Wiki has a pretty good rundown of load balancing versus manifolds.

When I first started (am still on my first game) all I had were Mk1 belts, so it was obligatory load balancing for most things.

If you like running belts along the wall and hooking them with Splitters you're already using a manifold, but you can load-balance from there. If nothing else is hooking that belt it's not really a "manifold" yet anyway, just some extra belt.

If uranium fuel rods are radioactive that's an even better reason not to line the walls with them ...

[Vectorspace]

Originally posted by luntacarsus:

I learned the hard way do NOT use a manifold for uranium fuel rods. They each take two and a half minutes to construct and each reactor can potentially hold a stack of 50 - it will take longer than the lifetime of your save to get more than a few reactors to full efficiency.

I've been working on my first nuclear plant, with 24 reactors. I am using a manifold, but it isn't hooked up yet. I am stockpiling fuel rods until I have 26 stacks (one per reactor plus 2 stacks for filling the manifold) so when I do hook it in it will saturate immediately.
Since it depends on multiple trains for supplies, I'm not going to activate it until after signalling lands so my nuclear fuel trains don't crash and stall my fuel production

[Mojo]

https://www.youtube.com/watch?v=gscXx9YqDEE

[Veeshan]

Now the question of whether to use max speed belts down the manifold or scale down as ppm reduces... probably no effective difference but I've been doing both just to see.