Sushi thread

[slim wave]

I see you have a full balancing system here

[tacit cedar]

yeah it's a 16:16 balancer

im thinking it might be better to have u-turns allowed like this

and I think it would be cool to have the recipes changing dynamically since you can do that in 2.0

[slim wave]

Imagine doing this in 2.0

[tacit cedar]

I chose 8 lanes because it was infinitely easier to make a power of 2 balancer

[slim wave]

Maybe I’ll make a sushi base out the gate for 2.0

I think that’s fun I think I’ll do a different fun thing for each planet

[tacit cedar]

you're doing this? I can try making some kind of a balancer

[tacit cedar]

oh that would be neat as well

the sushi base will launch sushi rockets that the other bases have to deal with ofc

[slim wave]

I think the sushi base would have “glands” for launching rockets

It would send a rocket that carries its own special blood filtered based on the demands of the planet it’s going to

[tacit cedar]

gleba is the fleshy planet so probably that gets sushi blocks

as a bonus, I remember reading that eggs can hatch on the belt to spawn enemies

so you can have an immune system

[slim wave]

I don’t know if I’m prepared to have stuff spoiling on those belts

[tacit cedar]

oh yeah it sounds awful

youd have to massively overproduce and filter the waste with kidneys or something

sushi but even laggier

[slim wave]

I think I would do it on Fulgora since it already has the sushi ore processing

For gleba I might do a logistic belt system just for spoiling items that routes them on sushi by request

[tacit cedar]

im thinking six 4:4s into four 6:6s

[slim wave]

I think I’ll be using splitter stacks for the most part and only use balancers in strained areas

What would be nice is a mechanism that will swap the two outer lanes with the two inner lanes

[tacit cedar]

like that?

or you meant in the intersection?

[slim wave]

That’s really good

Belt wizard

[obsidian roost]

Unnamed Blueprint

Unnamed Blueprint

[clever shoal]

that's so pretty

[shut current]

#930147588160782436 message - bloodbelts

#930147588160782436 message - hormones

"guys, new sushi just dropp-"

But tldr, basically just buffer-bundle sushi, and by hormones its just using the presence/concentration of buffered items close by a belt-reader to trigger stuff?

[plain hare]

Yeaaaa it’s been around but yea, items as information/ control system

[pulsar lance]

Real “hormones” feels like needing some items to trigger stuff, where those items aren’t actually part of the production like guns or sth

[plain hare]

Yea. If you just read the concentration of the same item you put on, I’m not sure it’s the same anymore

Naming gets hard tho

[pulsar lance]

It would be cool if you could make a factory (for Green circuits as an example), that converts a gun to a machine gun for every 100 produced circuits (whereas the gun and machine gun are hormones for that matter)

I mean, one could simulate that for sure

[slim wave]

I think that would be bad for the system because individual items get lost in the blood pretty quickly

There’s gonna be tons of belts that aren’t being read

[pulsar lance]

no one said there aren't supposed to be multiple guns

[distant gorge]

Which is still how a lot of the body works. You have to explicitly add "and is not a metabolite" to your definition of hormone to exclude say, glucose from the category

Otherwise "glucose is a hormone that induces the pancreas to release insulin" becomes a perfectly valid statement

[slim wave]

Is anyone good with combinators and could explain how to make a noise generator?

Looking to make a random belt blocking system for the blood belts

[distant gorge]

Why does that help you with blood belts?

[slim wave]

It would help stimulate more effective mixing

Balancers don’t mix if the belts are compressed

Of course the belts won’t be compressed but it would be more effective for mixing to push and pull on the bubbles

[dim rose]

You could use a balancer with "extra space".

[sand estuary]

Is this not a 1/11, it seems way more compact than the one further up

^ this one

what am I missing?

It has converged on 245 now, the image was taken just after it was finished

Its my first proper venture into rate limiting with splitters

[steep cloak]

yeah that's absolutely better, I knew about that design but wanted to see if it was possible to design one without doing that

[sand estuary]

oh ok

I was going crazy questioning myself

That makes my life easier. I might do a sushi-everything build with a ton of these

[steep cloak]

well the 1/11 is the only one where the repeated loop design is actually smaller

[sand estuary]

right ok

I kinda like the elegance of these tho

something about the way everything moves, and you can see it all (no undergrounds)

[steep cloak]

some of them might be the same size but eh, whatever

[sand estuary]

yeah

Can't wait to perfectly ratio satellite production with rate limited inputs

Whip out the 1/91 radar rate limiter

[dim rose]

This 1/11 limiter is actually a bit shorter, but wider.

[sand estuary]

yeah

Has anyone done rate limiting uranium? You could stop too much dark going out onto the belt

[distant gorge]

You can actually mix by having loopbacks

That’s what my nuclear sushi does

[slim wave]

Yeah but that seems like it would take up a lot of space

[distant gorge]

Not necessarily

[slim wave]

I think randomly stopping belts in front of a balancer would make tighter mixing

[distant gorge]

https://media.discordapp.net/attachments/231518135231053825/1249039144194146375/image.png?ex=66c4c5da&is=66c3745a&hm=ce05955b87d6854ba77f6026b5cccbe04ae39f86dce8771d89122f27dbd01420&

It sort of does here because I purposely want it very mixed

Mixing through time/forward back on the belt matters a lot more than mixing between belts

[slim wave]

I have an idea that I’ll call a humor where a belt loop has preferred items and will pick those items out while sharing the rest back with the main artery

And then deposit its own output directly onto the artery after this filtering system

[clear wedge]

Blueprint String

Yellow Kidney 12:6

[slim wave]

The kidney is moderately effective with only a single item filter but it really fails to make a difference after dividing the filters in half.

I think I’ll need to design a double or triple kidney for mixing multiple items

The job of this section is to suck up specific resources and flood them into a loop to make a high concentration of those resources

But it seems like it’s not sucking up enough resources on its own

[tacit cedar]

lmao what on earth
is that supposed to be a 18: 12 + 6 balancer, where the 6 tries to only have specific items (and only grabs about half from the belt)?

[slim wave]

I guess you could say that

If my math is right it will double the concentration of a resource on the 6-wide belt

I think I’m gonna reduce it to 3 belts so the concentration is higher

[hearty relic]

need help making sense of this; I want to be able to put these 2 items on a red belt. Fundamentally iron will come from a non full belt guarantee, and copper will also drain into here before feeding in an external copper source...

At which point, does this become correct?

ratios aint perfect by any stretch, but i think this is the right idea as far as priority is concerned

[slim wave]

So you are gonna have half a belt of copper and a mixed lane of iron/copper

[hardy fable]

if you want to make 1/3 + 2/3 ratio you should use 2 splitters to merge the 3 yellow belts into one blue

[tacit cedar]

or one splitter to merge a yellow and a red

into one blue

[hardy fable]

indeed

[slim wave]

I know jouster likes to work with low tech

I’ve seen him mix yellow and red belts but never seen him rely on blue belts

[tacit cedar]

i think im doing it wrong (2/3 RL)

[slim wave]

I have a challenge for you

Make a .98 limiter

Those are useful in niche cases where you want a belt carrying the maximum amount of items while still being uncompressed

[normal gulch]

easy: circuits

you can do any fraction with the same setup with circuits

[slim wave]

No circuits

The first time I ran into it conceptually was doing splitter circuitry so I wanna see it done without circuits

[normal gulch]

take off 1/128th with a ton of splitters, its close enough. then you have 99.21875%. if you want, you can do 1/64th instead and get a 98.4375%.

I still would prefer circuits see already how much cleaner :P

[slim wave]

1/64 takes 7 splitters

Then 1/128 takes 8

No it would need way more

Like 14

[normal gulch]

yeah 2 splitters per division of 2 if that makes sense, which is why again I dont use it anymore

instead of 2 flat for circuits

[slim wave]

None for circuits

[normal gulch]

right I was thinking of a standard limiter

[tacit cedar]

I tried this method (https://www.reddit.com/r/factorio/comments/cehtbc/splitting_in_strange_ratios/)
but I couldn't figure out 1/11

[distant gorge]

1/11=0.0001011101…
So discard first three, collect, discard, collect three, discard, collect, recycle the remainder

should be close to optimal if not optimal

[tacit cedar]

this is what I ended up with

[distant gorge]

Same thing, yep

[tacit cedar]

I just used my tried and true method of modifying a 1:16 balancer

ok i am eepy I will do more tomorrow

these fractions will probably be the death of me

[slim wave]

You could use a dummy item to make 1-x

Use a dummy item at 1/3 rate to make 2/3

[tacit cedar]

like this?

[slim wave]

Yeah

Then you can take out the input for copper

Actually you could take out a lot and just have a single loop of belt too

The plates on there would be like a program

[tacit cedar]

sounds annoying to copy paste

[slim wave]

Yea

If I designed it I would make a second blueprint to send a short burst of copper to the input that you’d make and then deconstruct later

[hardy fable]

so, a 49/50 limiter?

[dim rose]

When would you want uncompressed belts?

But maybe I mean: if it's a 49/50 RL, then the throughput is not maximal right? Is belt compression still "special" in current versions (1.0/1,1)?

[slim wave]

Yeah belt compression still affects a few things

[dim rose]

Which things?

[tacit cedar]

What do you use for the base 2 division?
Also I don’t really get the explanation
The image doesn’t look like you’re discarding 5 times

[distant gorge]

That’s a separate method than the strictly base 2 method

Still haven’t been able to sit down and fully translate the two

But to get the binary expansion you can just use wolfram alpha

[tacit cedar]

this looks different from yours?

[distant gorge]

I cut it off at the repeating part

But again, the spreadsheet version I showed was not created via the binary expansion method

[distant gorge]

Okay, so I figured out that the right thing to do is use the normal binary expansion of the values instead of the fractional binary expansion

e.g.

Numerator is the ones you "keep", extras has to be decomposed into "recycle" and "waste", which in this case can be done two different ways

The only "question" I have left is how to programmatically determine how to decompose your "extras" into recycle and waste with effectively only bitshifts and addition

[tacit cedar]

I still don’t understand 👍

[slim wave]

I don’t understand the math but I think I understand what youre working out

[distant gorge]

https://media.discordapp.net/attachments/930147588160782436/1277477976593666078/image.png?ex=66cea114&is=66cd4f94&hm=6b9c6d7afda3570e23798744dea520e7498591d87f5e4d32425ef1a92621d7ba&=&format=webp&quality=lossless
This sort of thing but in mathier form

[tacit cedar]

theres no way the top one is actually optimal, right

I mean i guess you could go red -> blue

[steep cloak]

@tacit cedar

[tacit cedar]

oh nice

[steep cloak]

I wish there was a way to get rid of the gap but I don't see anything

[tacit cedar]

meanwhile I came up with these which are kinda gross

[steep cloak]

top one is same length as mine so it's still pretty good

looks wacky though

[tacit cedar]

you could remove a space by adding a splitter I guess

[steep cloak]

yeah it's just weird

make it filtered as well just to save the items

[tacit cedar]

could try something like that

but if youre mixing belts theres easier ways for 2/3 lmao

well this ties it

(bottom)

or wait the priority is on the wrong side

for mine lmao

is this right?

[verbal osprey]

Test

[tacit cedar]

it is being tested

1800 is good

blue belt is 2700 items per minute
2700 * (2/3) = 1800

[verbal osprey]

Yes sorry i was directed to join this thread but i didnt know how to permanently follow it

Thanks for the help with the splitters btw i was able to make some really nasty spaghetti sushi with it but it works beautifully

[tacit cedar]

show the spaghetti

[verbal osprey]

There is some in the original sushi question i opened let me link

#1277595827547799552 message

Theres a few more showcase after that message

Scroll or look for streamable link

Cause discord has size limit

[distant gorge]

Took me a second but looks like it yes

[hearty relic]

Use it for power poles

[tacit cedar]

2/5 (very much in-progress)
my gut tells me there has to be a way to save a splitter (yellow splitters are discard)

[distant gorge]

One sec

So yes, you were right that there's a way to save splitters. You use the same splitter to make the discard as the total recycle instead of treating them separately

[tacit cedar]

this?

[distant gorge]

Generally, that will be the case whenever the discard and recycle need to be the same value (or the output and either value need to be the same)

[distant gorge]

yep

[tacit cedar]

does not look right

[distant gorge]

oh, whoops

hold on

derp

one too many splitters at the end

[tacit cedar]

very clean

[distant gorge]

nice and symmetrical

[tacit cedar]

im not sure how to make it 3 wide but this design is pretty neat

[stuck scaffold]

Hey, why does that particular interaction with the two splitters in the middle bother me so much? You're right about the goofiness you'd need to make it 3 wide, it bugs me a lot that that is the case

If you could put either one input to output then you'd be fine but nOoOoooOOoOoooo

It's fundamental splitter stuff, like, they're 2 wide and head in 1 direction. Why is that interaction so bizarre?

[verbal osprey]

its arcane sorcery

the loops always trips me up like it magically reduces the rate by existing

[stuck scaffold]

You would think this would be good. It is not lol

[distant gorge]

Scroll up for the math

[verbal osprey]

how far up though 🤔 im not sure if i should procrastinate

[distant gorge]

.

[verbal osprey]

i better get to setting up my chem plants tbh i just hopped on

[stuck scaffold]

I feel like once the fundamental clicks you're gonna level up

Balancer/splitter theory's not hard, just a bit annoying

[verbal osprey]

i feel like using a lower tier belt should also somehow reduce the rate? but it still requires a loop

[tacit cedar]

I mean you can do stuff like this

[distant gorge]

That’s just incredibly limiting with how much you can do

[stuck scaffold]

Hey, 2/3 + 1/3 is 3/3

[verbal osprey]

yeah i dont mean to use it exclusively

but like adding it to a bag of tricks

[stuck scaffold]

Yeh, it's in the bag

One of the weird things is that belts are thirds, and splitters are quarters. You can get a lot of combos that way, it's just funky sometimes

[verbal osprey]

like i already use one saturated red belt and one saturated yellow belt and combine it with a blue splitter to make 1 saturated blue belt, so the reverse should also work in theory

1 full blue belt, blue splitter, out to one red and one yellow belt and backpressure should take care of the ratio?

[stuck scaffold]

Backpressure can be very dangerous in sushi

Many recycled rate systems will seize the instant backpressure exists

The output loops to the input, so if you have too much output for the input to handle, and it backs up, the input doesn't flow. Crap I'm speaking gibberish again

[steep cloak]

I think this is correct

the splitter filter is unnecessary cause it was from a previous iteration but it saves like 2 items so I left it

[tacit cedar]

looks good I think

[stuck scaffold]

https://i.imgur.com/B0nBMJ5.png lmao this was before the belt patch where loop balancers got fixed, back when sushi was young and fresh and 1/7s didn't work right

The redditpost I got it from, I posted 7 years ago lol

Y'all with your splitter fixes, this was probably 0.16

[stuck scaffold]

slick

[tacit cedar]

it just feels so long

idk it might be optimal but it's still weird to me

you can get it a bit smaller with weaving

[stuck scaffold]

I haven't been toying with this design so idk if it's possible, but can you flip the middle section vertically? Might be able to save a tile that way

This is the tile you save, I don't know if you can jimmy the belts below it around

Seems like you have to braid one more of the returns

I think if this one can be YB'd it can be done?

[tacit cedar]

im not sure either of them can

try it out and see if it backs up

ok I tried saving splitters on my own, I think I got rid of two correctly?

well it's an improvement...

[distant gorge]

No longer at my computer to tell

[tacit cedar]

could un-weave like this i guess

[clever shoal]

thoughput issue

using redbelt

smhhh

2/7ths of a belt is insane though

just why

[distant gorge]

Only if you misunderstand the goal

[distant gorge]

Recipe with 7 total items in its recipe, two of which are the same

[clever shoal]

ah

[tacit cedar]

improved the 2/9 by just multiplying 1/3 and 2/3 lol

im not sure how good this method is (2/9)

[distant gorge]

I thought we went over that lol

[tacit cedar]

over the multiplying? wym
this book is a few weeks old, I'm just going through each design

[distant gorge]

#930147588160782436 message

#930147588160782436 message

[verbal osprey]

what in the world

never in my wildest imagination would i think my high school maths education could come back to bite me in the ass almost 2 decades later

prime factorisation? infinite series? jeez

[tacit cedar]

hard to improve

[distant gorge]

If conjecture is right, it’s actually impossible to improve

Conjecture being that if you follow the algorithm (which I still haven’t fully worked out), and apply it to your series of fractions with prime or power of two denominators, that’s optimal

[tacit cedar]

I meant in terms of space
but ty for checking splitter count

[clever shoal]

Leeks so beautiful

[tacit cedar]

I am suffering just as much on 3/5 lmao

so close but so far

it's just so loopy

i hate this

i guess this is fine

[distant gorge]

Looks right for fewest splitters

[slim wave]

Fewest splitters is good

Are you doing fewest splitters/most compact

As design options

[stuck scaffold]

Seems like Mitru's doing a bunch of them for 3 tile

[slim wave]

It’s looking to be the most effective way

For sushi it would be useful to combine two or three of them

[tacit cedar]

yeah I'm trying to minimize splitters and keep it in 3 tiles, then make it as short as possible

I guess I can go from 1/1 to 12/12

[tacit cedar]

my designs keep looking awful (4/9)

[stuck scaffold]

I don't have enough familiarity and my brain started tuning them out a while ago b/c it's like "yep splitter mess" lol. I'm only one theory jump away from actually understanding, I should put in the hour to catch up to y'all

[tacit cedar]

the way I came up with mine is that
a 1/X balancer will turn 1 belt into X belts, with 1/X of each belt filled
so if I want 2/X, I can just take 2 outputs as my output
but if the other outputs fill up, then the ratio would change, so they have to go somewhere
so I just route them back to the beginning with priority input

[stuck scaffold]

I was hoping it wasn't just a familiarity thing... I already know the theory and have to get my hands dirty before it clicks

lmfao it took me a solid minute to comprehend why this was necessary

Wow, okay, at least I can comprehend it as art now with a little more understanding. I'll peep some more designs as you post them and hopefully gain familiarity

[tacit cedar]

made it a bit shorter (bottom), not sure if it can get much better

I originally had the priority splitter sideways but it looks so bad

[tacit cedar]

@distant gorge this is (100011) repeating right?
I think your method can save more splitters than I can on my own

[dim rose]

Am I miscounting, or are those both 17 tiles long? The lower one does have the exit belt in the same row as the input, which saves space in some cases.

[tacit cedar]

oh yeah I think I misaligned the bottom counter gizmo and that made me think it was shorter

[slim wave]

You may want to put a belt between the belt reader and the exit hatch for scientific accuracy

[tacit cedar]

wym

[plucky isle]

How would that work??

[distant gorge]

ope, ignore the 6/5

wait, that won't work

I have to go but can look closer later

Maybe it will actually

[slim wave]

How is the 3/5 dividing

Unless I’m reading it backwards

[distant gorge]

It’s combining

[slim wave]

Does it flow top to bottom?

[distant gorge]

Gray is input

[slim wave]

I see now

[distant gorge]

The part I’m not 100% sure on is whether the numerator on that first splitter needs to be greater than the denominator… I think it works out so that everything “scales up” by 9/8 so it just works but I’m not 100% sure on that

[tacit cedar]

I can test it

[slim wave]

The input is 9/9 so how can that split into 4/9 and 4/9

[tacit cedar]

it definitely doesn't work but I can try to find out which fraction it is

oh the output is just max belt lol

i forgot to priority input?

[slim wave]

Yeah it’s not limited at the start either

Those two input belts are combining

[tacit cedar]

priority input changes nothing

[slim wave]

You’ll need an extra filter splitter at the top

[slim wave]

Yeah you’ll want to limit one belt instead of letting both belts balance at the top

But it’s still not gonna work out bc it will merge to 9/9 instead of 8/9

[tacit cedar]

this is 5/8

I have a solid design for that already

[slim wave]

This is the flaw

[tacit cedar]

clearly I need to implement the 6/5 on the white tile

5/8 * 6/5 = 30/40 = 3/4
maybe not

1/127 (for fun)

[distant gorge]

Nope, I said it was a typo lol

[tacit cedar]

ya ik

[dim rose]

Does the moving-exponential-average rate-calculator circuit work for this? I tried using it with a build that should output 337.5 i/min on half a yellow belt, and while the 10m production tab fluctuates from 337-339, the EMA goes from 331-342, which is not quite accurate enough for me. I'm measuring the output of a build that produces yellow sci, but the output is quite irregular. Is there any way to improve the EMA? The first time I tested it it didn't seem to converge to the right value, but maybe I just didn't wait long enough.

[distant gorge]

That’s what they’re using, yes

[tacit cedar]

idk how it works
this one fluctuates between 21 and 22

it takes a while at 60x speed to converge

[distant gorge]

You can tweak the time range to balance responsiveness and accuracy

The the more accurate it is, the less stable your value gets. You might also not have the input multiplication high enough

What does the first combinator say?

[tacit cedar]

3.6k
which I assume is the # of ticks in a minute

[distant gorge]

Add another zero to it and the last one

[distant gorge]

Nope, first combinator has nothing to do with time, technically

[dim rose]

So the middle combinator is for the time range? It's at -3.6k for me.

[distant gorge]

The division one sets the time frame, and technically affects the last one

[tacit cedar]

ok what did that do

[distant gorge]

That should be for 60 seconds then

[distant gorge]

Should increase the precision so it doesn’t get “off” of the correct value as much

There’s always going to be some fluctuation though

[tacit cedar]

it's still fluctiating about as much as before for the 1/127

probably since it's not a whole number

about 21.26

[distant gorge]

You can chop 0’s off the output combinator to “show more decimal places” and put an imaginary decimal point on the number displayed

[tacit cedar]

can i have a gap in the nixie tubes

to put a text plate .

21.44 idk

[tacit cedar]

can 8/9 be made with smaller RLs?
my first thought was 2/3 * 4/3 but I can't have numbers greater than 1

[tacit cedar]

honestly the display looks more fluctuate-y than before

[distant gorge]

OH, that finally clicked for me... you got 8/9 by making 1/9, then feeding that 1/9 back to the start.... which then slows down the input by 1/9 making it 8/9

Something like that anyway

Which is still the case for all the others, it's just another framing of it

[distant gorge]

5/9 but fixed to use 6 splitters:

pretend I connected the green all the way back

I'm still stuck on trying to find a better algorithmic way to generate these :/

[tacit cedar]

made some 2/11s

[tacit cedar]

👍

[tacit cedar]

i am cursed to only make ugly rate limiters

[stuck scaffold]

awww, it's not ugly, it's cute in its own way

[hearty relic]

You can probably optimize the belt tier here

[tacit cedar]

I mean yeah you could use slower belts on some parts without breaking anything
i think it looks ugly

[hearty relic]

Rainbow belts

[tacit cedar]

5/9 is done but I have another for the garbo belt nightmare list

[stuck scaffold]

lmfaooo good luck getting that into 3 tiles

[tacit cedar]

first i need to get rid of all the extra splitters

[stuck scaffold]

I'm sure there's a pattern for beltback and you know it

But also ew it's still going to be 40 tiles long

At least

[tacit cedar]

beltback?

[stuck scaffold]

The undergrounds when you have to loop inputs back

There's some cursed technique with your undergrounds, I half recognize a pattern

[distant gorge]

I keep needing to scroll back to look at this so .

I’m still trying to figure out why that 8/9 works intuitively without hashing out math first

[distant gorge]

The math ain’t mathin

[slim wave]

I did the math

Just start from the front and go backwards

So you need 1 to loop back and the rest comes out

If you’re working backwards you can think of the loops as multipliers rather than dividers

[obsidian roost]

Unnamed Blueprint

[tacit cedar]

I compiled the ones I have left to do

not sure how to optimize splitter count lol

[pulsar lance]

which compiler did ya use?

[steep cloak]

manual

[tacit cedar]

yeah I built these by hand

unless it was a pun

[tacit cedar]

for a 8/9 rate limiter, my steps are

1. make a 1:16 balancer (pretty simple)
2. route 7 outputs back into the input to make a 1:9 balancer
   3: 8 outputs are merged into 1 for the output
3. the remaining belt is routed all the way to the beginning with priority input

thats how I make all of these, I just make the first "1:first power of 2 that is larger than the denominator" balancer

you should teach me how to do math so I can minimize splitter count
maybe if I'm bored before 2.0 I can make up to X/16

[stuck scaffold]

There's prolly a different pattern so you can do any even number?

I'd look at 10:X and 12:X balancers, but also, probably those are just 5:X/2 doubled

Or 1:10, mb

iirc a 1:7 is just a 1:8 with a l00p

[distant gorge]

Composite number denominators can be decomposed into smaller ones

1/10 is just 1/2 of 1/5

Sort of

There’s some contextual dependence there that I’m not recalling

[stuck scaffold]

I was gonna say, all odd numbers have loopers, but most 1:X are automatically TU by nature?

Things might get funky when you're hooking up composites, idr

[distant gorge]

Ultimately you’re just building up a 1-N/D belt to feed back into the input to slow it down

[stuck scaffold]

Oooo some refeeds have priority and some don't

That's funkadelic

[distant gorge]

My problem has been that there’s like three methods and they don’t directly translate and figuring out how to translate them is 😵‍💫

[stuck scaffold]

lol yeah... It can be weird for sure

[distant gorge]

If I could do that, I’m confident I could write down an algorithm for the simplest method for any given fraction but until then it’s at least partly guess work

[stuck scaffold]

That's why with sushi I'm like like "Well, here are the basics/tools, after that it's "complex sushi" lol

[distant gorge]

We do have algorithms for methods that always work, but individually no guarantees that they’re minimal

That’s still just conjecture

[stuck scaffold]

Urrr, at some point I'll peep it, I need to keep putting some factory down though

[distant gorge]

I just take like snippets of five minutes at a time at work

You can do it in excel if you know the mathy bits

[stuck scaffold]

Yeh, it's been a hot second since I dove into balancer theory is all

||4 years?!? oh god I'm getting old||

The theory bits are fun for me though so if there's a theoretical problem, I may be able to figure something out

[distant gorge]

So the two mathy ways to do it are taking the infinite repeating representation of the fraction in binary, and using that to create two loops, one for “recycle” and one for “discard.” That’s the most algorithmic method but is very far from minimal most of the time.

The other method is to choose the next highest power of 2 (P), then use P/D as your starting point. Then you split things apart so you get three belts: a discard belt equal to D-N, a recycle belt equal to P-D, and your output belt N. Using the binary representations of P, D, and N can make this easier. This method is not as algorithmic, but I conjecture can lead directly to the minimal design as long as you have a prime denominator.

Then there’s Mitru’s method of taking a generic P:P balancer and chopping it up first to 1:D, then taking N outputs of D and moving them back to the input. Also algorithmic, but not minimal by any means

[stuck scaffold]

Ah, Discard goes into priority input at the start, that's the difference with recycle. Recycle has specific spots it needs to go in the splitters

[distant gorge]

recycle is the part that makes your infinite fractional representations finite

Discard is the part that actually slows things down at the input

[stuck scaffold]

I must be missing a fundamental piece of this puzzle still

[distant gorge]

You and me both

[stuck scaffold]

LOL

[distant gorge]

I’m talking just fractional single belt slowdown, not multi belt balancing

[stuck scaffold]

Yeah, that can come a bit later for me, same principles, more complicated

I think it's the same core things anyhow >.>

[tacit cedar]

all of these are coprime
but I guess 5/12 can be 5/6 * 1/2
the rest dont seem to have any way to make them out of smaller rate limiters

[tacit cedar]

no you can pretty much just multiply them in my experience
1/5 * 2/3 = 2/15
you just have to remember that the discard goes all the way back to the beginning, so you can't just concatenate two designs

[tacit cedar]

you need priorities actually
I've been using TL 1:X balancers for my designs

[tacit cedar]

discard has priority and recycle doesn't, is the main difference

[tacit cedar]

for this, step 2 is the recycle and step 4 is the discard

[stuck scaffold]

lol yeah I'm treading in your footsteps as we speak

[tacit cedar]

math :(

[tacit cedar]

I start with a 1:P balancer, then make it 1:D

this has been my starting point for a while

I should read that splitter paper at some point

https://assert-false.science/guyslain/papiers/splitternetworks.pdf

[tacit cedar]

I do not understand it 👍

[mental spruce]

holy fck a paper with factorio

[mental spruce]

naturally

might get counseling from claude for this

[tacit cedar]

i tried doing it myself but this doesnt look very repeaty, and idk if I can request more digits
I can guess it's 0111010001 repeating, but I can't be certain

[stuck scaffold]

There was a dude who figured out a trick with digits to repeat with primes

https://youtu.be/DmfxIhmGPP4?si=6oV1nFxO9-uEAO_h

[verbal osprey]

Lmao doing this by hand is just...

[stuck scaffold]

So uhh, I like the Mitru method for fractions

Turns out 5/32 is actually pretty easy

[tacit cedar]

I may not understand math, but I know how to steal it from people

def find_binary_fraction(numerator, denominator):
    remainders = {}
    binary_fraction = ""
    remainder = numerator % denominator

    while remainder != 0 and remainder not in remainders:
        remainders[remainder] = len(binary_fraction)
        remainder *= 2
        digit = remainder // denominator
        binary_fraction += str(digit)
        remainder %= denominator

    if remainder == 0:
        print(f"The binary fraction of {numerator}/{denominator} is: ({binary_fraction})")
    else:
        print(f"The *repeating* binary fraction of {numerator}/{denominator} is: ({binary_fraction})")

find_binary_fraction(1, 7) # Output: "The *repeating* binary fraction of 1/7 is: (001)"
find_binary_fraction(1, 8) # Output: "The binary fraction of 1/8 is: (0001)"
find_binary_fraction(1, 9) # Output: "The *repeating* binary fraction of 1/9 is: (000111)"
find_binary_fraction(2, 9) # Output: "The *repeating* binary fraction of 2/9 is: (001110)"

[stuck scaffold]

So when I watched that video, I was like "Hah, that's neat, too bad there's legitimately zero use for it ever" and I am glad to be proven false

[tacit cedar]

oh i didnt understand the video i just googled someone else's code

[stuck scaffold]

That's the idea from the video

[tacit cedar]

ok well it's working... but it's using more splitters than the old method 💀

I thought i was onto something lmao

[tacit cedar]

also I forgot to label one of the rate limiters
the 2209.0909 underneath the 9/10 is supposed to be 9/11

==========
hm
if I take half a belt, and halve it again (1/4)
and run the other half through a 1/4 (1/8)
then I can recombine them (preferably with a lane balancer lol), then they are added to become 3/8

oh I was overcomplicating

1. split into 2
2. 1/2 + 1/4 = 3/4

it's certainly a method, but I don't think it'd give very many minimal solutions (in terms of splitter count)

[slim wave]

2024

[tacit cedar]

this example uses twice as many splitters as the normal version

[slim wave]

Balancer of the year

Rate limiter

[tacit cedar]

💀

[distant gorge]

Yep, bigger prime, bigger repeat

[tacit cedar]

this is so sad

[tacit cedar]

btw do you have any advice for this

I was just about to start rereading this entire rate limiter conversation from the beginning

[distant gorge]

I still haven’t figured out how to get that second version from scratch

[tacit cedar]

starting here i guess

[distant gorge]

Oh it’s 7/9

And I realize the issue partly

[tacit cedar]

both are 7/9
top is the binary method
bottom is the deconstructed balancer method

[distant gorge]

7/9

[tacit cedar]

top one has 12 splitters
bottom one had 9 splitters
and yours only has 6
I wish I could do that lol

[distant gorge]

Part of the "trick" is recognizing when the recycle, discard, or output have the same value

I recognized that I needed a recycle of 7/9, which matches the output, so I actually want to sum to 14/9 and then cut it in half

[tacit cedar]

so you actually build it and try to look for shortcuts?
do you start from the front or the back?

[distant gorge]

I start from the 16/9

[tacit cedar]

how do you know it starts at 16/9?

[distant gorge]

Because 16 is the next power of 2 above 9

[tacit cedar]

oh I see

did I mess up

oh I see
the lower dark blue belt is 12/9 on a single belt?

[distant gorge]

yeah

that... looks right?

hmm

that's funky

[tacit cedar]

i think i fixed it

[distant gorge]

You need the other output of that splitter and merge it in to another splitter between the second to last and last ones

[tacit cedar]

by putting it into 2 belts of 6/9

[distant gorge]

yeah

[distant gorge]

yup

[tacit cedar]

ty for all the math help

[distant gorge]

okay my 8/9 still doesn't match yours and I'm lost on how to reduce it further

New conjecture, the "high point" can be either a power of 2, or a power of 2 times a factor of the denominator

[tacit cedar]

idk i forgot how I made it
probably started with a 1:9 balancer instead of a 1:16 -> 1:9

[clever shoal]

how do you deal with recursive

🤕

[tacit cedar]

alright I turned it into a block of splitters, not exactly what I was going for

okay 3 belts forwards 2 belts backwards might be an issue for 3-wide

redesigned it in hopes of needing less parallel belts, saved a splitter

[distant gorge]

[tacit cedar]

hm almost 3 wide

[distant gorge]

Okay so that’s getting back to my original idea of calculating back from the discard amount

Which works but sometimes takes more splitters

[tacit cedar]

oh yeah I think I was calculating the discard amount first

which worked well here since the recycle and output were equal

unless theyre always equal idk

[distant gorge]

They aren’t

[tacit cedar]

this one is rough

maybe I need a new design after all

[distant gorge]

7/9 can share the recycle and output all the way to the end

Unless…

[tacit cedar]

recycle and output is more than a belt no?

14/9?

[distant gorge]

Combine the 8 and the 6 last?

Away from computer

[tacit cedar]

okay i think i somehow managed to spaghetti 7/9 out, but it's already over 6 splitters, and I have to turn the remaining full belt into 2 and 7 for discard and recycle, i dont think thats gonna work

[tacit cedar]

like that? idk

this one in particular is very rough

[tacit cedar]

isnt that what i had

[distant gorge]

probbaly

[tacit cedar]

here

or wait

maybe not idk

[distant gorge]

I have one fewer

[tacit cedar]

I had 6 at some point

[tacit cedar]

this one is 6

and it looks pretty similar to yours

[distant gorge]

yeah same thing

[tacit cedar]

maybe got something

it came out pretty clean

[distant gorge]

effectively just moving splitters from the end to the beginning?

[tacit cedar]

yeah

I tried it before but this time it worked nicely

I take 2 out of 8 at the front, and put 6 to the end

I tried making one on my own

oh wait, my old one only has 6 splitters

[slim wave]

Now try 21/55

[tacit cedar]

it is now my turn to sare at it

[tacit cedar]

im gonna finish the ones I have queued up first

[distant gorge]

Now you understand lol

[tacit cedar]

reposting this

I guess I just started at 12/9 instead of 16/9

[slim wave]

Will 21/55 be easier or harder than 24/59?

[tacit cedar]

why do you need either of those

[slim wave]

I’m thinking purely about the maths

[distant gorge]

harder

24 is 8 plus 8 plus 8

[tacit cedar]

I think I have a 3/5 and 7/11 you can multiply to make 21/55

[distant gorge]

I'm not convinced multiplication always works yet

[tacit cedar]

i tried it a couple times and it's probably fine

i'll make one

ah, I did not yet make a 7/11

I will do it later

[slim wave]

I’m enjoying watching this channel but I feel like you two are 5 steps ahead of me mentally

[distant gorge]

[tacit cedar]

so anyway

12/9 instead of 16/9

to save a splitter

[slim wave]

So 8’s are hard to make?

[distant gorge]

easy

21/55 would be harder

[tacit cedar]

I mean it's trivial to make this 3-wide

but it looks ugly and there has to be a better way

[distant gorge]

Billy Mays here

[clever shoal]

Why are you making these

Are these for modded

Or

[distant gorge]

To scare off people who are afraid of sushi

[clever shoal]

For fun

[clever shoal]

I just can't wrap my brain around balancer math

But sushi is fine somewhat

I think PySE would benefit from sushi

[tacit cedar]

I think I made it worse

I can probably get rid of the loop at the beginning sec

well I don't have any better ideas

it sucks because this is SO MUCH CLEANER but it's one splitter extra

[tacit cedar]

so this one starting with 12/9 is a power of 2 (4) times a factor of the denominator (3)?

[distant gorge]

yep

[tacit cedar]

okay 7/11
I guess 16/11 then, since 11 is prime

[distant gorge]

Yep

[tacit cedar]

hm, both methods ended up about the same

most of the others have 6 splitters or fewer, only 1/10, 1/11, 1/12, 2/11 have 7
but this is 9

surely 9 isnt optimal

@distant gorge 6 splitters :)

I should revisit some of the 7 splitter ones later and try to reduce them further

[stuck scaffold]

Okay mitru I'm hot on your heels lol

I'm catching up anyhow

0.375 of a redbelt in copper first, 0.4375 of a redbelt for GC

Well, red lanes. 337.5 copper/s and 393 GC

I think the 0.4375 could be better?

0b0.0111

[tacit cedar]

0.375 = 3/8
0.4375 is 7/16, which I don't have a RL for

[tacit cedar]

im not exactly sure what ur doing though

[stuck scaffold]

I'm applying the theory

Just trying to get some throttlers in place, and I finally caught on to how you organize them

Keepers towards one side, returns towards the other

I think I can switch both to yellowbelt though and just drop the first splitter, I'm under yellow lane on both?

[tacit cedar]

if red belt is 1 then yellow lane is .25

which you're above on both

[stuck scaffold]

I was using only lane in both balancers anyhow

Yeah this is just a 75% and a 7/8 now, gotta love the simplify

That sure looks like 6/16 and 7/16

[tacit cedar]

@distant gorge 16 is overrated

[tacit cedar]

3 wide

16 💀

[stuck scaffold]

SushDI is gross

[tacit cedar]

14 :)

[stuck scaffold]

Oh, oh no, I know how to improve this I think

I should be able to use Thyme's trick

[tacit cedar]

another funny loop

with the two vertical splitters on top of each other

[tacit cedar]

what's that

[stuck scaffold]

The overfeeding and reduction method

2:9 ratio, start at 16, reduce, and refeed into dead inputs

[tacit cedar]

that looks scary

like it'd break if an input runs dry

[stuck scaffold]

Yeah, you can defend that with just wire tho

[slim wave]

You can also use dummy items

Fill in the gaps in the inputs

[tacit cedar]

wire icky. only splitters are real

[stuck scaffold]

That engine ratio hittin just right for compression sushi tho

Also I am impressed by any ratio of inputs

If it were combos I'd probably shy away, but it's just wire and wires are free on BP's

asdfkadsfadskh ratio THEN rate

Holy shit tomorrow I am going to make such a splitter mess

base is 30. 2-4-4-8-11-13-13

55 of 64

[stuck scaffold]

I went and reduced a couple of the numbers based on factoriolab

[tacit cedar]

it happened again
these are rough to get 3-wide

(this one is 6/11)

[stuck scaffold]

Yeah the backward facing splitter is tough

You could go one longer

Face the backwards one in

omg it doesn't even save the tile does it

[tacit cedar]

those splitter stacks, where one feeds into the other, but vertically

are horrible

[stuck scaffold]

Yeahhhhh

[tacit cedar]

they look amazing in 4 tiles wide

but 3 tiles wide they will destroy you

it's so hard

i hate this

[stuck scaffold]

Yeah turning them horizontal doesn't seem any better lol

[clear wedge]

Blueprint String

Simplified Blue Science Sush

[stuck scaffold]

Ratio is supposed to be 11:11:10:8:4:4:2, dividing by sulfur so it's showing up as the baseline of 1, 2 extra significant digits.

[normal gulch]

are you open to circuits?

thats a really neat solution though

[stuck scaffold]

I'd prefer to avoid combos

I'll have to wire guard the inputs

[normal gulch]

wdym combos?

[stuck scaffold]

Combinators in general are vulnerable to power failures

(I have a bad habit of removing poles)

[normal gulch]

that's fair, but a single solar panel and accumulator can fix that

[stuck scaffold]

lol, true

Why, did you have something in mind?

[normal gulch]

I can whip something up real quick yeah

[stuck scaffold]

😮 for circuit ratios?

[normal gulch]

oh yeah circuits can go in any ratio very easily and the limiters are tiny

[stuck scaffold]

lol hold on, don't tell me the wiring on that one

That's a fun puzzle

I posted that BP thinking nobody would respond and somebody would find it in the morning, I was about to head to bed

[normal gulch]

The combinators in that image are not the one ill be using, its a puzzle even I barely understand

I wont be on in the morning most likely so Ill just share a solution and spoiler it if you want?

[stuck scaffold]

My suspicion is faux clock

HAH it is a faux clock that's why there's 6 of them

Cute

[normal gulch]

faux clock?

the one in the image is || a combinator that counts to 7 then resets ||, so yeah that makes sense

[stuck scaffold]

It's ticking negative, belts are pulse read

As well as control

Hmmmmm no I need more than one combo to do that I think...

[normal gulch]

no reading, belts just do something like this. L goes to 6 or 7 then ticks back down

[stuck scaffold]

oh it's literally just a clock lel

[normal gulch]

yea thats the simple version, the more complex is for different ratios of items than 1:1:1:1:1

[normal gulch]

this ratio of 11:11:10:8:4:4:2, si that just left to right inputs?

[stuck scaffold]

If you can work out the tick times for each item it's not so hard

Yeah, I'm doing modulo 1k to see the rest of the digits

And it's in hundreths

[normal gulch]

I see just trying to get the right numbers inputted in the constant combinator

[stuck scaffold]

Oh wait did you see the other sush I did? I got the blue done up

[normal gulch]

I dont think it was finished when I saw it last

[stuck scaffold]

It's not defended, but it's stable w/ inputs

[normal gulch]

oh dang thats a lot more compact than it was before

completely direct insertion from engines/reds it looks like too

[stuck scaffold]

Yep, they never touch belt. I don't know if all those undergrounds are worth it, but... I guess it's working? XD

[normal gulch]

probably not but its cool looking!

reminds me of a ratio ive always wanted to do but never have done: 1 rail maker to 4 purple science

[stuck scaffold]

Oh noooo I was like "purple sounds silly I shouldn't SushDI purple"

But if there's a ratio, there's a way

[normal gulch]

where theres a will theres a way

I did a decent DI but the ratios were completley whack

I completed the circuit sushi if you want to see it

[stuck scaffold]

For this ratio? Did you do a bunch of clocks? XD

[normal gulch]

1 clock

5 combinators

[stuck scaffold]

Modulo

[normal gulch]

It used to be 1 clock per item but then codegreen and some others made it

any ratio you want just input in that

[stuck scaffold]

That's nifty, I'll have to think about that one tomorrow

[normal gulch]

yeah its my favorite way of doing it jsut because its so minimalistic

[stuck scaffold]

Yeah, I still have to recycle, but also....

[normal gulch]

Recycling is just built in as you see.. it used to require a chest but I realized if the constant combinator is static you dont need a chest

[stuck scaffold]

18 splitters, 6 undergrounds and a handful of belts

Oh wait I alraedy pulled another pair of undergroudns that's right

[normal gulch]

for yours?

[normal gulch]

close, im honestly surprised that yours is less

[stuck scaffold]

I haven't recycled yet

[normal gulch]

ive always done the limiting of those belts individually

ah right, that adds atleast 7

every time I do sushi it makes me want to go into a new game, use some BPs to make an offensive line straight thru the biter nests and wage war

make some recursive design that goes to the end of the world.. offtopic but the mall BP you showed me can accomplish that very very easily

[stuck scaffold]

Which one? The repeater station?

Mall bp...

[normal gulch]

repeater station, the mall repeater

yours was for solar iirc

and then I can just leave my game running while it uses recursive BPs to run to the end of the world

[stuck scaffold]

Ahhh yee, that was very easy to make generic

[normal gulch]

I love and hate this motivation to play factorio at 1am

[stuck scaffold]

ikr?

So I took thyme's idea and threw a constant on it

B/c 1) it's nice to have a literal on/off switch for your belt, and 2) it defends the loading problem

[normal gulch]

thymes?

[normal gulch]

oh I see, cuts off in any input is starved or CC is off

what kindof crack was I on

so this mall exists, I just need to add some inputs, recursive BP the rails, and then I can expand infinitely after a couple hours

[stuck scaffold]

lmao that'll be something to see

[normal gulch]

this is proof that I had way too much time on my hands

i can remove 95% of those combinators and get like 90% of the results so ill probably do that

[stuck scaffold]

loool

[normal gulch]

fuck it, im making a fortress to the ends of nauvis tonight

[stuck scaffold]

TONIGHT

I am not

lol it's way too late and I got good work done

save and quit while I'm ahead

[normal gulch]

yeah good point

I just hopped on though I got energy!

I'll cya tomorrow

[stuck scaffold]

omg I shouldn't have taken the last look

But I'm glad I did

SHoulda just gone to bed. Subtle difference

It was right the first time

[normal gulch]

haha

the factory cannot wait for sleep

[stuck scaffold]

#vanilla-chat message Aight I think I'm happy with it

Wire defended inputs and output blocking

[hearty relic]

yeah, the higher tier sciences are probably gonna be much more difficult

[stuck scaffold]

Apparently there's a 1 rail:4 purple ratio

Which shouldn't be too bad, 4 in seems to tile relatively well

Should be able to DI sticks as well even

[tacit cedar]

I looked into using code to find a graph, given the numerator and denominator of the output
it knows that the discard would be denominator - numerator / denominator (so 4/11 would discard 7/11, the numerators add to the denominator)
and all it has to do is check every combination of splitters, up to MAX_SPLITTERS (which could == 6 for now)
but turns out I don't know how to do that lol

I did make the connection that a splitter needs 3 connections to be useful though (2 inputs, 1 output "adder", 1 input 2 outputs "splitter", 2 inputs 2 outputs "balancer)

[tacit cedar]

like the dosh bore?

[stuck scaffold]

lmao I love it when you come across the little details that sound silly but there's actually something meaningful there

Like "of course splitters need >=3" but it's actually something super useful for what you're doing

[stuck scaffold]

It's like my twobelts realization @distant gorge . It took me 4,000 hours to realize that two belts is two belts. But yeah I have a question about bloodbelt sushi. When you're doing EMA, is it possible to measure at the right place so that when the EMA dips below a certain value, you know that's the right time to add items for that section of the belt?

[distant gorge]

You could probably do it that way?

I don’t really see much value in that

[stuck scaffold]

Trying to avoid the smoothing

EMA is already smoothing

Oh right, I remember the specifics of your use case, I see why it's not relevant for you, chunk size

[distant gorge]

At some point, over engineering your abstractions starts removing the value of the abstraction

[stuck scaffold]

LOL, oh... yeah....

Good point

Why not just measure the belt there

[distant gorge]

Essentially yeah

For mine I purposely set it up so everything gets introduced just before the EMA measurement and is consumed after, so the belt is near empty on one side and fairly full on the other, so I only need one measurement circuit

You could have more circuits and not do that, but again, meh

[stuck scaffold]

Okay so for the "reduction" balancers, if you wanted two outputs, does anything specific change? Is it the fact it's a return belt and therefore choked to demand, that makes the difference, IE I could have two outputs but they both would need to have returns?

[tacit cedar]

I'm getting decent at 6-splitter designs
but I keep making perfect neat 4-wide designs

[stuck scaffold]

4 wide seems like cash money

This is the specific situation I was talking about. Without the filler item, I know I can take two off no problem. It's when the recycling happens on other ratios

Do I have to recycle off both?

[dim rose]

You want two belts coming out, but you are only using one output of the ghost splitter, so it is impossible.
If you use the other output of the ghost splitter as well, you have exactly what you want. Only two splitters needed.

[stuck scaffold]

Yeah the specific situation isn't an appropriate use of that, I'm jsut trying to understand the principle

These two belts are NOT the same

[tacit cedar]

IT'S SO UGLY

[clever shoal]

ẞ what the f*** why would you make this stop cease

[clever shoal]

You could say it's a 7 out of 10 haha

[stuck scaffold]

So... how do I get two belts out? 😛

[clever shoal]

You don't.

[distant gorge]

Well, your example is flawed because the wood belt is unnecessary

[stuck scaffold]

I can get two out of the simple version without a loopback is why I ask

[stuck scaffold]

Yeah I kinda didn't want to distract with a bunch of unnecessary stuff is all, my bad

[distant gorge]

Remove the extraneous belt and it becomes clear that no loopback is needed for that case

[stuck scaffold]

Yeah, hypothetical like the real thing that scarred Lemon

I think the difference between the example that DOES work, and the one that Lemon thinks can't work, is the loopback, IE Dead item?

I don't comprehend why it wouldn't, but also, I'm not sure how it would work if it did

[stuck scaffold]

This should work if I rejoin with a splitter? Idgi

I'm taking mix, and recycling

[tacit cedar]

what is going on here
mixer?

[obsidian roost]

mixer?

[stuck scaffold]

I don't know why it's stopped

The missing belt

I also don't know if it's properly mixed?

Balancers don't mix items necessarily I don't think, at least they didn't used to

[tacit cedar]

looks mixed to me

[dim rose]

It's not a simple 7-7 balancer, that's for sure.

[stuck scaffold]

It's not clean, but it doesn't look ratiod. I couldn't tell with that little belt info though

[tacit cedar]

i sent the bp try it out

looks like a modified 10:10 TL belt balancer to me

with 3 loopbacks to make it 7:7, but one is broken intentionally

[tacit cedar]

https://factoriobin.com/post/wPCziGqXyO1j9uC4-EXPIRES
i should post this here too

still finishing up to x/12

feel free to look through and find better solutions

[tacit cedar]

my rate limiter has a suspicious growth

[stuck scaffold]

it's nah ah tumah

https://www.youtube.com/watch?v=Oj-xjDzghz4

[normal gulch]

yeah, just a straight line to the end of the world

[tacit cedar]

may your UPS be high

[normal gulch]

hopefully

[normal gulch]

so theres this thing called a sushi balancer that literally balances each belt. I have a 4-4 version but nothing bigger

I think it was codegreen who made it

[tacit cedar]

i dont know what it means from that description

you mean like how a normal 4:4 balancer will kinda split the lanes normally, but this will actually mix?

[normal gulch]

it will equally mix all the belts

[tacit cedar]

oh so it is 4 adjacent 1:4s and 4 4:1s

neat

[slim wave]

Ah I see this

[tacit cedar]

also this is so hard to fit in 3 wide

[slim wave]

So this is different from a rate limiter

It provides a ratio rather than a limit

[normal gulch]

I mean it does limit rates by 1/4thing it. maybe some issues if it jams though

[tacit cedar]

alright done for now
https://factoriobin.com/post/Zo-hsi0p

the name kinda bugs me idk

[obsidian roost]

Rate RC Sushi

[stuck scaffold]

Takes just under 2 redbelts wires, 1 redbelt each of platic and GC, outputs just over a yellowbelt of RC.

Will stop if one lane fully backs up, but recovers fine.

No control mechanisms whatsoever. Just belt rates.

[hearty relic]

@stuck scaffold uhhh excuse me, wtf

omg, youre sideloading them

thats trippy

[stuck scaffold]

😄

Yeah they get split so the two halves use different lanes

Does look pretty incomprehensible though huh

[mental spruce]

bro was supressing the spaggeti inside of him for so long that he now needs to get it out somehow

[clever shoal]

What is the point of the deconstruction planner splitter

[hearty relic]

its a way to say "dont send anything down this path".

[clever shoal]

But there's a splitter right after it that defeats the purpose

[hearty relic]

its for rates

[clever shoal]

Whatever you say 🤨

[hearty relic]

im not the one who made the bp, ask @stuck scaffold

[stuck scaffold]

It's I need to ensure that only one belt goes out of that splitter

That way recycled goods are prioritized and I can't dump more than one full redbelt

[clever shoal]

Priority input when???

[stuck scaffold]

It doesn't show up on the BP for some reason

[clever shoal]

Still a waste of a splitter imo

But it works™️

[stuck scaffold]

It is not, I assure you

[clever shoal]

Can you show the difference when removing and using priority input on the next splitter?

[stuck scaffold]

Yeah sec I'll just unfilter and crash it

Let me show you the normal crashed mode first

No GC

No filter

... .am I a loon and it was fine?

LOL It's defended b/c of lanes

[clever shoal]

[stuck scaffold]

lol yeah I don't normally do the lane filtering

h/o let me show you the normal fail state w/o the lane nonsense

[clever shoal]

This surely is a machine of all time

[stuck scaffold]

Okay so here's the not crazy part. Fully loaded rate sushi to start. Drain plastic, reintroduce plastic(3rd image), nbd

Jesus christ apparently I defended it like 6 times and didn't need to. Upgraded the splitters in the middle so I wasn't choking them with yellowbelt input, too

Reintroduction is instant dead stop b/c of the missing decon filter.

[tacit cedar]

I did the same thing in my 1/2 rate limiter

[stuck scaffold]

To be clear, the extra rate choke with those yellow splitters wasn't necessary for function. It was just for cheap

One redbelt, is four, half yellowbelts. twobelts is twobelts.

[clever shoal]

is there a modular belt balancer

I doubt but I really need

N to 3 belt balancer

💀

I just need a modular rate limiter now

no more need for that

sushi is funny

[stuck scaffold]

I mean, I think there's just two types of rate limiter. Starvation, 1 belt into N, with loopbacks, and compression, N belts into 1

Compression rates

[slim wave]

It’s a form of bottle neck. The filter reduces the throughput through the system from 2 belts down to 1 belt. Only 1 belt can get between the splitters. Normal splitters have a total throughput of 2 belts.

[clever shoal]

oh okay that makes sense

[tacit cedar]

https://alt-f4.blog/ALTF4-27/

[clever shoal]

universal balancers actually have a use case for me

[obsidian roost]

4:4 Circuit Universal

[clever shoal]

I need N to 3

mechanical

:)

[tacit cedar]

im not really sure how it would be modular

I can try to make an 8:3 universal

[clever shoal]

that'll do

thank you if you do

[timid cosmos]

i made a little bot assembler using the easy belt method

[tacit cedar]

okay looking into it and trying some stuff, it might just be that an 8:8 universal is easier

[stuck scaffold]

So, potentially silly application of blood belts, but I think it's possible to use an item based delay system? If you know you need X items/sec, that means you know how many ticks between items. Counter with a -1 clock, each >0, enable input when counter for that item is negative. Each time you detect an item, input or already on the belt, add the correct number of ticks to the delay

[distant gorge]

Feels like it’s equivalent to the EMA method in a way but I would have to sit down and math it out

[stuck scaffold]

It should be relatively close to EMA, just more directly measuring belt state I think? If you get a pulse it responds to that pulse by waiting longer

i tried to make this purple science build work with blood belt but I don't think it's an appropriate application. The delay is way too huge on belts this long

[distant gorge]

On the contrary, I think that method might be farther from the belt state because you’re baking in an assumption about the flow rate you want into the measurement instead of just measuring

Especially because ticks/item is very unlikely to be an integer

[stuck scaffold]

Yeah, I have some leeway to play around with though I think. I think what I should be doing, is trying to maintain a belt state? Not full of anything, but has enough of everything

If the belt section I'm measuring hasn't seen the requisite iron ore, give it

[distant gorge]

For my case, I explicitly know that at steady state, I should have 700ish ipm of U238, a handful of ipm of most of the rest, etc

So I toggle inserters on and off directly from whether their item is higher or lower than that steady state throughput

Exception is plutonium that I explicitly want to buffer more of so it’s at like 60 ipm instead of 20

[stuck scaffold]

Yeah that's what I was trying to do here too but I'm getting massive bursts b/c I have like 130 iron furnaces all turn on for a long time until the average comes back up

[distant gorge]

Which is where you break them up when you have such a disconnect

[stuck scaffold]

lmao I already have them broken into 10 groups based on vertical belting

[distant gorge]

Ideally you have close to the exact number you need so they rarely need to toggle off

If they’re toggling off constantly, you have too many

[stuck scaffold]

Yeah I'm using the excess from iron furnaces, that are DIing to steel

So they spend 90% of their time off

Well, 90% of their iron needs to go to the steel furnaces, or more, the belt outserter is off

[distant gorge]

Yeah that’s actively detrimental to blood sushi

[stuck scaffold]

I noticed lol

Not what I wanted lol

[distant gorge]

I get the opposite problem of having a massive item drain that then turns off

So U238 being very low on the recycle then suddenly having a big (apparent) excess

[stuck scaffold]

Hmmmm

This is a weird one. I don't know what to do about this.

It's very difficult to get only some of that iron out of there

[distant gorge]

Clock the belt, change the time period or the threshold to be determined by the gap

[stuck scaffold]

I have the furnaces there. I just have to take about 6%

I can't use belt output blocking I don't think

I have 130 furnaces b/c increments of 10.

[distant gorge]

Hmm, I wonder if there’s a technique to easily take 1/16 of an input belt.

[stuck scaffold]

lmao I have to take it from this furnace

I get the longboi. I could fuck around with a shortboi

That's it

Ignore the extra fast inserters, they got copy pasted for some reason and will get cleaned up

This is the only thing I get to control, that furnace has to provide for its steel bro too

I'm reconsidering many of my life choices, particularly the belting/general layout

[stuck scaffold]

I have the loaders relatively well fine tuned with delay. I can remove most of this eventually, testing purposes for now

They've got the input belt mostly full but aren't themselves backing up

Iron's the worst one to figure out. EMA is working well enough for steel and I think it'll be fine

Idk I'm close with blood

[stuck scaffold]

Okay so blood is sortof functional but the bursts are weird. I should have plenty of space on the belt, and I thought the resources shoudl be spread out, but iron is concentrated enough I think the stick machines are what's causing the burst

[stuck scaffold]

Pulse feeding seems to work pretty well

I'm just not sure about that iron

[clever shoal]

is there a cheatsheet for all of these new terms

I don't know what a blood belt is 😭

[stuck scaffold]

Idk if it's Max's term but he's certainly been championing it, that I've seen

#930147588160782436 message

Basically sampling rather than measuring the whole belt

This delay thing is different from blood belts specific, but the iron and steel are both done blood belt style. EMA for control value

Delay is measuring recent past, but it throws away old events completely at a linear rate rather than exponential

I'm basically measuring two belt chunks, the one at the very top, pulse read, and the four at the very bottom, which are also control. Multiply all pulses by ticks between pulses, store that value. -1 all values

This decider is "Ticks until next item released." When the value is negative, release an item

[slim wave]

I like the name sushi but I think it would also be good to call it a triple meat deluxe belt

[stuck scaffold]

You can increase precision on delay decay by not decrementing by 1. If you need 5.5 ticks, you can send 11 and decrement by 2, I think? The implementation of this might be 12 kinds of fucked up tho b/c I was dropping negative values, and that won't work

[distant gorge]

Can you give an example of the delay setup again?

[clear wedge]

Blueprint String

Blueprint 1

[stuck scaffold]

Ignore the total spam, it also just pulse reads the belt it's feeding, top left. If you sync distances correctly it'll put the item on the exact right spot

There's probably simplifications there too, I'm just tryina figure it out still

[distant gorge]

that's a shit ton of combinators lol

[shut current]

@tacit cedar , would this be smallest a 1:1:1 can be?
Ideally this would be even narrower.

[stuck scaffold]

lmao so for sure, I don't need that many right here. It's one per resource per feed area, unless I can find a trick to each the delays

[stuck scaffold]

There's an underground trick for 2 tiles, maybe? Certainly 3

[tacit cedar]

2 wide might be impossible with only blue belts

though you can use yellow belts to rate limit of course

yellow belt = 1/3 blue belt

[slim wave]

Looks like it could have an efficient corner variant

[steep cloak]

having inputs inside the thing shouldn't count for total size

because an underground or another belt on the side is necessary

[slim wave]

But this extends naturally from braided belts

[steep cloak]

which also require setup at the start

[slim wave]

Yes