Factorio Forums

I love fractals (I have a tattoo of the dragon curve) and would love to use these, but are they non-throughput-limited (i.e. clos networks)?

papercrane wrote:I love fractals

Same!

papercrane wrote: and would love to use these

You must have a big belt base then. Really big. Show me

papercrane wrote:but are they non-throughput-limited (i.e. clos networks)?

Yes, they all allow full belt throughput. Had no knowledge of telecom "clos networks" before though so can't answer about if that applies.
The 512 belt balancer has the same throughput as a 512 belt wide bus.

2 questions,
1) are the blueprints updated to 0.15?
2) I found the blueprint for 16 here, is the blueprint for 64 around here somewhere? Couldn't see it (probably just missed it)

Bomaz wrote:2 questions,
1) are the blueprints updated to 0.15?
2) I found the blueprint for 16 here, is the blueprint for 64 around here somewhere? Couldn't see it (probably just missed it)

(1) No, but they still work fine.
(2) 64 is in there. Look again

Looks amazing When I was about 15 yo I rendered my first fractals in VB6.0, though not understanding it fully. Since that time I've just watched some "zoom-in" videos on Youtube.

Qon wrote: (2) 64 is in there. Look again

Ok, I saw it in the imgur library but unlike all the other balancers there the blueprint string is missing for 64.

Amegatron wrote:Looks amazing When I was about 15 yo I rendered my first fractals in VB6.0, though not understanding it fully. Since that time I've just watched some "zoom-in" videos on Youtube.

Thanks!

I can watch fractals forever. Fully understanding them kind of simultanously requires very little effort and infinite effort.

Bomaz wrote:

Qon wrote: (2) 64 is in there. Look again

Ok, I saw it in the imgur library but unlike all the other balancers there the blueprint string is missing for 64.

Well the blueprints for the big ones were too big to post here, so I just uploaded the whole save file which includes them all as blueprints and all of them already built. That way you can just load it up and take what blueprints you need. Also, I'm lazy q:

Qon wrote:

papercrane wrote: and would love to use these

You must have a big belt base then. Really big. Show me

Well, it's actually a large rail base but I find myself in need of some larger balancers to deal with ore in particular. You can download a save of my base from this bug report:
viewtopic.php?f=7&t=48762&p=282409#p282409

Qon wrote:

papercrane wrote:but are they non-throughput-limited (i.e. clos networks)?

Yes, they all allow full belt throughput. Had no knowledge of telecom "clos networks" before though so can't answer about if that applies.
The 512 belt balancer has the same throughput as a 512 belt wide bus.

Well....after playing around quite a bit with building my own balancers and going through the posts on throughput issues with large balancers I have an example of the problem with your 8-belt balancer: Note that 4 full belts are going in the top-left and only 4 half-belts come out through the outer four belts on the bottom. The issue isn't that it can't handle 8 full belts going in and out, it's that in the case of some belts being backed up or empty you won't get full throughput of what is there.

papercrane wrote: Well....after playing around quite a bit with building my own balancers and going through the posts on throughput issues with large balancers I have an example of the problem with your 8-belt balancer:

Screen Shot 2017-05-27 at 8.03.21 PM.png

Note that 4 full belts are going in the top-left and only 4 half-belts come out through the outer four belts on the bottom. The issue isn't that it can't handle 8 full belts going in and out, it's that in the case of some belts being backed up or empty you won't get full throughput of what is there.

Well if you have 4 outputs and 4 inputs you use a 4 belt balancer, right?
If you absolutely need a 4 - 8 balancer where some outputs are blocked dynamically then you can just put two 8 belt balancers in a row for desired effect. Not sure yet if you can do it in a more efficient way. Have you seen any other solutions?

Qon wrote:

papercrane wrote: Well....after playing around quite a bit with building my own balancers and going through the posts on throughput issues with large balancers I have an example of the problem with your 8-belt balancer:

Screen Shot 2017-05-27 at 8.03.21 PM.png

Note that 4 full belts are going in the top-left and only 4 half-belts come out through the outer four belts on the bottom. The issue isn't that it can't handle 8 full belts going in and out, it's that in the case of some belts being backed up or empty you won't get full throughput of what is there.

Well if you have 4 outputs and 4 inputs you use a 4 belt balancer, right?
If you absolutely need a 4 - 8 balancer where some outputs are blocked dynamically then you can just put two 8 belt balancers in a row for desired effect. Not sure yet if you can do it in a more efficient way. Have you seen any other solutions?

That was just meant as an example. Even if you have 8 belts going in and 8 going out you can run into situations where some of the belts will block or empty and potentially cause a throughput problem.

I haven't found a better solution than putting 2 balancers in series to fix the throughput problems. There are clos networks which are theoretically the "right" way to do it, but the balancers I've found based on clos networks have more balancers in them than the doubled balancers I've found elsewehere and come up with on my own.

papercrane wrote:I haven't found a better solution than putting 2 balancers in series to fix the throughput problems. There are clos networks which are theoretically the "right" way to do it, but the balancers I've found based on clos networks have more balancers in them than the doubled balancers I've found elsewehere and come up with on my own.

2 balancers in series and a clos network are equivalent in their properties. When implemented properly both should have the same amount of splitters.

d4rkpl4y3r wrote:

papercrane wrote:I haven't found a better solution than putting 2 balancers in series to fix the throughput problems. There are clos networks which are theoretically the "right" way to do it, but the balancers I've found based on clos networks have more balancers in them than the doubled balancers I've found elsewhere and come up with on my own.

2 balancers in series and a clos network are equivalent in their properties. When implemented properly both should have the same amount of splitters.

I was going to ask for an example but I looked back at the clos network balancers and it's the 4-2-4 one which has more balancers (28) vs. the 20 that the 2-4-2 balancer and the "standard" doubled balancers have. That does raise the question, though, why does one version have more splitters than the other when they're supposed to be equivalent?

WOW
This ridiculous, but amazingly ridiculous.
Keep your hopes up, someone might find a use for it.

olafthecat wrote:WOW
This ridiculous, but amazingly ridiculous.
Keep your hopes up, someone might find a use for it.

As soon as someone decides they want to balance their 512 belt wide iron bus for their 30 RPM factory, sure...


I would use bots and trains though when building a 30+ RPM factory.

I didnt want to make a new thread for such a simple question so here: I know that each imput needs to be able to reach each output, so whats the point of doing it like the one on the right when the one on the left is smaller and uses less splitters?

Blastit wrote:I know that each imput needs to be able to reach each output, so whats the point of doing it like the one on the left when the one on the right is smaller and uses less splitters?

Without the additional splitters the output of the balancer would not be balanced.

Tested the smaller one, yeah, didnt work

Blastit wrote: I know that each imput needs to be able to reach each output, so whats the point of doing it like the one on the right when the one on the left is smaller and uses less splitters?

I think you've touched on the fundamental pointlessness of balancing

TL:DR - Balancers, as defined by the community, aren't actually what people need in most cases.

In my mind, the ideal "balancer" solves the following problem:
" Given n inputs and m outputs, no input backs up unless all outputs are backed up".

Most people seem to be obsessed with the definition that:
" Give x items on n inputs, each output gets the same number of items"

But what circumstance is this relevant for?
If each of the outputs feeds a different assembly line, in most cases some of the outputs will back up because they don't need more materials, directing any excess to the other outputs. If none of the outputs are backing up, that means all of the downstream assemblers are waiting for this material, in which case you would typically want to prioritize some of them over others - normally "balanced" distribution to those assemblers isn't what you're going for.

If you are balancing to feed a smelter area, each smelter line will typically be designed to be able to process a compressed belt. If there isn't enough ore at the input to feed all the smelters at full capacity, it doesn't really matter whether all of them drop exactly to 90% of their peak throughput, or if one of them shuts off completely and the other 9 keep going at full capacity (unless the smelter outputs go different places, in which case you have the scenario in the previous paragraph).

The biggest "balancing" type problem I see is between trains and smelters. My current setup uses an 8-way balancer to distribute ore from 4 train stations to 8 lines of smelters. Train routing logic means that one station is used somewhat more frequently than the others, which means that during an ore shortage one or more of the input belts will still have items on it while the others are empty. But, with the commonly used 8-way balancer (belt balancer compendium thread), I find that certain combinations of full/empty input belts will cause one or more inputs to slow down slightly, even while multiple output belts are left empty.

So, my question to everyone is - in what cases do you truly need balancers (as they are defined by the community) and in which cases do you actually want "balancers" by my definition instead?

Selvek wrote: So, my question to everyone is - in what cases do you truly need balancers (as they are defined by the community) and in which cases do you actually want "balancers" by my definition instead?

You need balancers when you want to make useless but pretty and huge fractals in your base.

The ones by your definition isn't that useful either in most cases. Each assembly cluster can only take up to its full capacity, limited by belts and amount of assemblers. If you have a big bus and 90% of the factory is idling then the 10% will in many cases just work at 100%, which is probably exactly the same 100% as it would work at if all of the factory was running if your base is built to to handle that.

All you need is a good way to split off from your bus (which can be done without splitters for some designs), separate and complete production units for things you need constantly with high throughput and lots of ore.

Selvek wrote:

Blastit wrote:
So, my question to everyone is - in what cases do you truly need balancers (as they are defined by the community) and in which cases do you actually want "balancers" by my definition instead?

lets say a 4-wagon Train station is unloading supplies, with a 4 belt balancer, you can make items leave the wagons at equal rates without the use of circuits. This solves a problem wherein a train wait condition on "wait until inventory is empty" is slowing the unloading down because one wagon is not unloaded yet when all the others have finished unloading because products are not leaving the 4 wagons at the same rate.

If any output is backed up, all inputs enter at equal rates regardless.
Edit: According to Testing, a 4 belt balancer will only have equal input rates when you use 1,2,4 belts. if you use 3 belts, the input rates wont be the same

Basically you choose. complicated circuitry, or a basic 4 lane balancer

However i do like to point out that you are correct and "ideal" belt balancers are not needed in most cases.