0.13 Rainbow Belts and other madness

[BlakeMW]

I've been having a blast with the new belt circuit connections.



By far my favorite creation is what I call the rainbow belt, it's a simple, efficient and reasonably robust way of packing 3 or more different item types onto a single belt


Rainbow Belt: Labs



How it works is as follows.

• The belt forms a loop
• A circuit wire is connected to the belt at the end of the loop right before the splitter.
• That connected belt is set to always allow flow, and "read contents, hold", its function is to detect items which are recirculating.
• The belts entering the splitters are wired to the detector belt and given conditions such as "Alien Science Pack = 0", meaning they only allow items to enter if those items are not recirculating, meaning they are being consumed.

This system only allows items to enter the loop if they are actually being consumed and this makes the system highly resistant to jamming up. It works so efficiently you can basically get the full useful throughput of the belt, which for labs is about 60-70 labs on a basic belt. Furthermore the setup is absurdly cheap, being just a few splitters and a few circuit wires, that it is actually much cheaper than a second belt and all the extra long-handed inserters would be.
The system handles all sorts of contingencies really well, like if one of the resource types dries up then the belt will generally not saturate, it will just end up with a smattering of the other types recirculating, leaving plenty of room for the missing type to enter when it becomes available again. It also deals well with uneven consumption, as whichever resource is consumed most will be allowed to enter the most.

Rainbow Belt: Science Pack 3



In this screenshot the output isn't being used much and it has backed up. The rainbow belt is wonderfully mixed up. Under clogged conditions the occasional spurt of resources will be allowed in, which is why there is banding (the banding is moderated by the action of the splitter). But when the resources are being consumed quickly all the resources will be allowed to enter simultaneously and will be mixed homogeneously - this is why a rainbow belt setup generally permits the full throughput of the belt to be used, if everything is being used it is perfectly mixed.

Rainbow Belt: Red Circuits (1 lane only)



One use of the rainbow belt is economizing on belts, another use is when you literally only have room for one belt due to beacons (note: I am religiously opposed to underground belt braiding).
In the case of advanced circuits resources are used in a 2:1:1 ratio, in such a case one lane can be dedicated - in this case the copper cable is allowed on unconditionally and the circuit belt logic is used to moderate plastic and electronic circuit inflow.

As I noted before, the rainbow belt setup is really quite robust. It handles things like being starved well, without saturating or jamming. (Note: In the above setup I used a filter inserter for each resource type, there's no particular reason, I was just trying something)

Count Perfect Mixing
A rainbow belt is a thing of terrible beauty, artistic, slightly chaotic, possibly evil. It probably drives some pedants loopy just looking at it (I on the other hand, love watching my rainbow belts).

But happily, I also have a count perfect system, once again with advanced circuits which use the handy 2:1:1 ratio:



This one uses an arithmetic combinator as a signal repeater to remember the count of items which have passed through the connected belts.

• Both connected belts are set to read belt contents in pulse mode.
• The plastic belt is set to "Plastic < Electronic Circuits"
• The electronic circuits belt is set to "Electronic Circuits < Plastic"
• The constant combinator sends a "Electronic Circuits = -1" signal to one of the belts to break the deadlock.
• Since either one belt or the other is always operating there is always a full belt of content entering the splitter.

As in the previous case the copper wire is allowed to enter unconditionally. Once again a loop is required, even though the items are count perfect it is possible that the inserters will for example take a bunch of plastic, leaving enough electronic circuits in a row to completely block an assembler. The loop prevents that happening.

[Kelderek]

Wow, very creative, I'm impressed! It looks like a real space saver and even if space isn't a concern it ends up being simpler and more elegant. I can't wait to give it a try.

[MeduSalem]

This thing is epic to say at least.

I was looking forward to trying something like that but you beat me to it, BlakeMW.

(note: I am religiously opposed to underground belt braiding).

We should found a sect, because I am strongly opposed to the braiding cheat as well.

[Guu]

I also recently did something similar https://youtu.be/WNcb8Om6gpE

[Qon]

Love the look of it, even though it is probably not much better than what we have used so far. Can you show me a production unit where this is preferable due to reasons other than aestethics? q:

[pieppiep]

I think this is really nice with a mod like bob's modules where it can take up to 14 ingredients to make a product and the creation of the product takes enough time.

[BlakeMW]

Love the look of it, even though it is probably not much better than what we have used so far. Can you show me a production unit where this is preferable due to reasons other than aestethics? q:

I consider it a practical way to handle science packs. As I noted, you can run about 60-70 labs off a single basic belt - that's a lot of labs, more than most players ever build (unless you're speed running completionist or something), using a rainbow belt is cheaper and more compact than a pair of belts with long handed inserters. It's also nice being able to just stick with a single belt, I'll certainly be using this setup in future games since the way I see it, it is every way superior.

I also consider my red circuit setups practical - now clearly underground belt braiding is superior especially in terms of raw throughput, but I consider that a grievous exploit and won't do it. The other alternative is logistic bots. But something worth considering here is I started out those setups as prod1/speed1 and progressively upgraded them. At the start I didn't have the logistic bot upgrades so this was a quick and easy way to start using prod/speed. Being able to use two rows of beacons is nice, especially as you upgrade to prod2/3.

It would also be a practical way to do science pack 3 with prod/speed, as that recipe is slow enough that a single belt can easily handle the requirements. I don't normally prodify Science Pack 3, but it is one of recipes where prod modules are quite justified. Also you can reuse the belt to carry the output from the assemblers (for this it helps to exploit the fact that underground belts give precedence to items being inserted - but even without underground belts it'll work if you have several belts set to detect recirculation, ensuring that enough gaps are left)

Finally it'll also be a practical basis as a kind of "logistic belt" for things like engine, electric engine, robot frame manufacturing - though I consider using bots for that a no-brainer, unless playing a logistic network embargo. Something about the rainbow belt is it can support multiple entry points and multiple measuring points, you can also link inserters to the condition for directly moving things to/from the belt.

[MeduSalem]

The other alternative is logistic bots.

Not to forget that Logistic Bots are much, much, much more costly with 0.13 now. The energy consumption skyrocketed and I am really considering to go back to belts due to how inefficient bots have become for mass delivery.

[DOSorDIE]

Nice. Can we have a savefile from it?

[ske]

You should also have a look at this thread: viewtopic.php?f=8&t=19609

Crosslinked your post there as you are perfectly on topic for the other thread.

[Qon]

The other alternative is logistic bots.

Not to forget that Logistic Bots are much, much, much more costly with 0.13 now. The energy consumption skyrocketed and I am really considering to go back to belts due to how inefficient bots have become for mass delivery.

Not really my results. I have a pure bot megabase (and pipes ofc) WIP that used about 100MW for roboports in 0.12 and 150MW in 0.13. And it uses about 1.2GW total (not counting smelting and mining). The robot energy charge increase seems fairly insignificant when looking at those numbers. I haven't really done any real measurements though. Maybe the factory wasn't running at 100% at that moment? Haven't played a lot of 0.13 yet.

I had a lot of roboports though so the lowered transmission energy cost mitigates the energy cost of robots a bit. But it's a fairly small share of the overall energy used so wasn't ever a big contributor really.

Consider making your robot factory more compact. If it's more compact robots travel shorter distances and thus use less energy. And use more roboports. The transmission energy required is very low, especially now. More roboports means robots don't have to do long detours or wait in a queue to charge, which decreases the amount of energy you need to charge the robots which is the factor you want to minimize.

[Qon]

I consider it a practical way to handle science packs. As I noted, you can run about 60-70 labs off a single basic belt - that's a lot of labs, more than most players ever build (unless you're speed running completionist or something), using a rainbow belt is cheaper and more compact than a pair of belts with long handed inserters. It's also nice being able to just stick with a single belt, I'll certainly be using this setup in future games since the way I see it, it is every way superior.

But you have to loop it back. So do you really gain any space used? 2 belts or one looping back, think the loopback is even bigger because you have to let it travel a bit further than your lab line to get it around and back. Or you could loop it back the same path that your second belt would take. What you gain is just a few long handed inserters. That reason alone isn't worth it for me. I mean I like it, and maybe would use it even if it cost me 200 long inserters extra. But only for the fun of it. A loopback at the same space your second belt would otherwise occupy still eliminates the need for long handed inserters though, even if it takes up the exact same space.

I also consider my red circuit setups practical - now clearly underground belt braiding is superior especially in terms of raw throughput, but I consider that a grievous exploit and won't do it. The other alternative is logistic bots. But something worth considering here is I started out those setups as prod1/speed1 and progressively upgraded them. At the start I didn't have the logistic bot upgrades so this was a quick and easy way to start using prod/speed. Being able to use two rows of beacons is nice, especially as you upgrade to prod2/3.

It would also be a practical way to do science pack 3 with prod/speed, as that recipe is slow enough that a single belt can easily handle the requirements. I don't normally prodify Science Pack 3, but it is one of recipes where prod modules are quite justified. Also you can reuse the belt to carry the output from the assemblers (for this it helps to exploit the fact that underground belts give precedence to items being inserted - but even without underground belts it'll work if you have several belts set to detect recirculation, ensuring that enough gaps are left)

Maybe for beaconised setups with many ingredients that are also slow then. Not many things require more than 4 in vanilla though. I think many cool belt solutions are looking for a complicated recipe with many ingredients to solve though.
A loopback takes the same space as 2 belts but doesn't have the throughput of 2 belts. You could probably do output on the input belt on both types though. Worth examining.

Finally it'll also be a practical basis as a kind of "logistic belt" for things like engine, electric engine, robot frame manufacturing - though I consider using bots for that a no-brainer, unless playing a logistic network embargo. Something about the rainbow belt is it can support multiple entry points and multiple measuring points, you can also link inserters to the condition for directly moving things to/from the belt.

Yeah that would could be a real application for this. Because ...

You should also have a look at this thread: viewtopic.php?f=8&t=19609

Crosslinked your post there as you are perfectly on topic for the other thread.

.... doesn't give you a compressed input belt. Is your solution quicker to wire up and build than that though?

[BlakeMW]

I have a pure bot megabase (and pipes ofc) WIP that used about 100MW for roboports in 0.12 and 150MW in 0.13. And it uses about 1.2GW total (not counting smelting and mining). The robot energy charge increase seems fairly insignificant when looking at those numbers.

One thing about bots, before you get the upgrades they are *much* more energy intensive. Basically they consume energy for every second they move so the slower they are and the smaller their stack bonus the more energy they consume to move a certain number of items. In principle a logistic bot with no upgrades uses 4 (stack size) * 3.4 (speed bonus) = 13.6x as much energy per item delivered as a fully upgraded bot.

But you have to loop it back. So do you really gain any space used? 2 belts or one looping back, think the loopback is even bigger because you have to let it travel a bit further than your lab line to get it around and back. Or you could loop it back the same path that your second belt would take. What you gain is just a few long handed inserters. That reason alone isn't worth it for me. I mean I like it, and maybe would use it even if it cost me 200 long inserters extra. But only for the fun of it. A loopback at the same space your second belt would otherwise occupy still eliminates the need for long handed inserters though, even if it takes up the exact same space.

A loop takes almost no space at all for labs - you run the belt down and build 16 labs on both sides, then run it back up and build 16 labs on both sides, 64 labs in total, and only about a dozen extra belt tiles to close the loop. I mean heck, in the past I've usually looped the belt most the way back up anyway just because I want my labs to be more of a square block than a long thin rectangle.

A loop also takes almost no space at all when you use underground belts to pass under the inserters - usually that space is wasted as normally you only need 1 output inserter and maybe 1 power pole which leaves plenty of room for an underground belt. In my red circuits build I managed to fit the return loop under both output inserters and substations. That space would just be wasted otherwise. I often use that dead space to route resources through (I <3 spaghetti) but not so often that I can't use it for loopbacks (btw the 0.13 underground belt speed laying is neat for this).

.... doesn't give you a compressed input belt. Is your solution quicker to wire up and build than that though?

I would say it is embarrassingly easy to setup and it allows a fully compressed input belt.

However considerably greater care is required when there are multiple points of entry. The system largely leverages the power of splitters to evenly merge the input belts, when you have more than one point of entry especially when those entry points do not use a splitter (such as inserting directly onto the belt) there is a greater scope for problems.

[Qon]

I have a pure bot megabase

One thing about bots, before you get the upgrades they are *much* more energy intensive.

Well, my megabase bots are fairly upgraded :]

A loop takes almost no space at all for labs - you run the belt down and build 16 labs on both sides, then run it back up and build 16 labs on both sides, 64 labs in total, and only about a dozen extra belt tiles to close the loop.

A loop also takes almost no space at all when you use underground belts to pass under the inserters - usually that space is wasted as normally you only need 1 output inserter and maybe 1 power pole which leaves plenty of room for an underground belt.

Ah, true. Not if you are doing it between alternating rows of beacons though.

[js1]

I have done something similar with 0.12, here: viewtopic.php?f=8&t=25107

But I agree that 0.13 makes some things like this easier, you don't need the counting chests anymore (but also some more difficult in fact, for example in newer revision of the above base I used inserters to put stuff on belts instead splitters, but stack upgrade bonuses will break that setup in 0.13).

Anyway, I don't think this is very practical except if you want to make your base really compact. Which is what I like to do.

[MeduSalem]

The other alternative is logistic bots.

Not to forget that Logistic Bots are much, much, much more costly with 0.13 now. The energy consumption skyrocketed and I am really considering to go back to belts due to how inefficient bots have become for mass delivery.

Not really my results. I have a pure bot megabase (and pipes ofc) WIP that used about 100MW for roboports in 0.12 and 150MW in 0.13. And it uses about 1.2GW total (not counting smelting and mining). The robot energy charge increase seems fairly insignificant when looking at those numbers. I haven't really done any real measurements though. Maybe the factory wasn't running at 100% at that moment? Haven't played a lot of 0.13 yet.

I had a lot of roboports though so the lowered transmission energy cost mitigates the energy cost of robots a bit. But it's a fairly small share of the overall energy used so wasn't ever a big contributor really.

Consider making your robot factory more compact. If it's more compact robots travel shorter distances and thus use less energy. And use more roboports. The transmission energy required is very low, especially now. More roboports means robots don't have to do long detours or wait in a queue to charge, which decreases the amount of energy you need to charge the robots which is the factor you want to minimize.

Well to be honest on the map where I noticed the effect I have driven a hard energy bargain. I've almost been to the limit of my power plant in 0.12 already and the bots were almost the only thing that still drew energy (factory was almost idle otherwise)... so if you say that they require 1.5 times the energy than before in average... then it's no wonder that my power plant became insufficient. I should have doubled the plant a long time ago.

And yeah... I am a robot mining sinner. So sometimes 5000 bots flying all over the place because I am too lazy to build a new train station.

[ske]

but stack upgrade bonuses will break that setup in 0.13

I really like the stack inserters, they are a fine addition to the game but the stack bonus for normal inserters is just... off... it's ugly and should be removed. If you need throughput, you can take a stack inserter and if you don't you take a normal inserter.

[British_Petroleum]

This is really nice. I'm a big fan of using belt loops.

[piriform]

Sorta, kinda related (actually pretty much the opposite )

blackorwhite.gif (240.6 KiB) Viewed 21360 times

The idea is to steer the materials to a desired location. in other words, a distributor.

[BlakeMW]

Processing units (with prod4) uses and outputs resources in a ratio of 20:2:1.4, so ideally you have a belt practically full of electronic circuits, and scatter in advanced circuits in a 10:1 ratio. Also since the electronic circuits are being removed at a phenomenal rate there is plenty of room on the input belt to accommodate the output.

So here is my latest feat of belt madness, count perfect belt mixing with asymmetrical input ratio and belt reuse for output. It feeds in green and red circuits in a 10:1 ratio.


Notes:

• The two circuit-connected belts are set to "Read Belt Contents" (in "pulse" mode) and their enabled condition is set as shown.
• The wires are shown, essentially red is input, green is output. The decider combinator at the bottom is set as a counter, the input is wired to the output. The reason to use two different wires is to prevent the "A" signal being accumulated in the counter.
• The express belt forms a loop. In this case there's nothing you can do with the dead space between the pipe and beacons so the loop has no real cost.
• This setup supports about 5 Assembler 3 with prod3 + speed3 beacons, equal in output to 30 Assembler 3 without modules. It's actually a very compact way to feed so much resources into processing unit manafacturing.
• The first assembler outputs directly to the out belt - this is to break the "can't pick up, can't put down" deadlock that will otherwise occur if the processing units output (temporarily) gets backed up. With this special first output inserter it seems to be reasonably deadlock resistant.
• To make stronger guarantees against deadlock you can read the number of electronic circuits recirculating and use that as a condition to stop one of the input belts.
• The system is quite robust. Feel free to vacuum items off the belt or "spike" the belt with extra advanced circuits, the assembler input buffers are more than large enough to accommodate minor inconsistencies in the number of items. I've never had to reset the counter.