Sushi thread

[sand estuary]

16 shouldnt be that much harder

[hardy fable]

there are 12 in total

[sand estuary]

Powers of 2 are better

[hardy fable]

you'd be surprised how easy it is to divide by 3 :p

[sand estuary]

Lemme get out my 3 belt splitter real quick

Working with belt sushi makes me annoyed vanilla doesnt have an 8th science pack.

Why

[hardy fable]

it makes things more challenging

like the uneven nuclear ratios

[hardy fable]

I figured out something neat

use the god damn belt instead of wasting it :D

[eternal olive]

smart

[sand estuary]

finally got it working to a level I'm happy with

belt side could be slightly tighter, but TBH I CBA

I could also have a half height one at the top, because there isnt really enough to fully saturate it

but oh well

[hardy fable]

good job 👍

did you go for a 3/3 ratio on the one? or 3/4?

[hardy fable]

because this sounds like an issue

like, I think you have a mismatch in spm

each belt has a lil bit more than needed, leading to the top one being starved

wait, that shouldn't happen

why is it not saturated

[normal gulch]

this is pretty cool, is that train how your going to have it ingame?

[hardy fable]

oh yeah it's 3/3

this change should do the trick

you have 3 belts going in, 4 going out, so having them each at 3/4 is advised

[hearty relic]

sushi is cooking

[sand estuary]

3/3

[sand estuary]

Yeah I was just guesstimating, madding and removing labs until it looked about right

Right ok I’ll do that later

[sand estuary]

Yeah

It’s going to go to each of the science grids and load the science into its respective wagon

Every train is going to be 1-8, apart from the science train, which only has seven cars and an extra locomotive

Oh and all the other trains are going to be from raw. (Ore train, crude oil train…)

[shut current]

Decided to just loop an outlet gate signal.

Has likely 0 correction mechanism for backpressure.

3/10 wouldn't recommend, but its simple and gets the job done
Not to mention vertically expandable, so you can make multiple lines of labs

Jesus, this thing is just, so damn clean

Nicely and cleanly divides a whole belt into 7 more belts. The 8th row is there to show there isn't even a single overflow or excess science going to it.

[sand estuary]

worked perfectly

I also took out the top row, because it couldnt be fed

7 inputs doesnt go to 8 outputs

I actually really like the red bands of deconstruction planners

Definitely easier (and smaller) than using belt rate limiters

and it works no matter how saturated the belts are

[hardy fable]

ah

you forgot one filter

[sand estuary]

yeah

[hardy fable]

luckily the belt should always be full but better have a failsafe 😄

[sand estuary]

Tbh you dont really need it

yeah its always going to be full of decon planners

i considered using fish, but the red is more vibrant, and absolutely free

[hardy fable]

right, how do you make so many of them tho?

do you have to manually pop them out one by one?

[sand estuary]

alt d is the shortcut

then press z to drop

[hardy fable]

I see

[sand estuary]

my networks always get full of decon planners anyway

I can just set up a logistic request and an inserter

It has a shorter footprint than my red and green print, which is cool. Imma come back and rearrange the red and green to make it shorter though

[umbral stump]

I was messing with stuff and got this lol

Sometimes when the BP is pasted it stabilizes into a 1-1-0 output (when both side loops end up containing only one color)

But idk why that happens seemingly randomly

[hardy fable]

uh, what are you trying to do exactly?

[umbral stump]

Well I was curious if there is an easy way without circuits to merge two items even if downstream clogs up -- like rate-limit both so there is always space for the other, even if downstream backs up

But regardless of that

I don't get this

They all always start the same way

These were all pasted on the same tick

Run them a while, and they're no longer the same:

Two of them even stabilized into the fabled 1-1-0 formation

[hardy fable]

ah, the issue is that you mix the item before trying to rate-limit

it is thus not possible

unless you recycle the sushi to build it back like you seem to be going for, but the "first sushi" would still be rate unlimited and thus prone to issues

[umbral stump]

yea I think I understand

[hardy fable]

unless it has enough buffer I guess

[umbral stump]

I have to rate limit and merge back to the start with priority

But what explains the above behavior?

[hardy fable]

the priority can be left before the rate limiting

but merging the different items together has to be done after

here's an example

(actually two)

[umbral stump]

What I couldn't figure out from your black box 12-1 solution is how to rate-limit 2s

Though ofc you don't have to answer :P

[hardy fable]

my guess is the fact that you connect both outputs of splitters to both inputs of other splitters, that sounds prone to chaotic behavior

[umbral stump]

I mean I considered simply having belt circuits before they enter the system and only allowing them to enter in pairs, but

That's not in your screenshot

I figured out how to get it working without circuits IF all input belts are saturated to begin with

But idk how to solve the initial pair merging without circuits, AND within 7 tile width

[hardy fable]

oh right, you're the madlad who's working on cracking my SA setup

[umbral stump]

Yea when I posted about "having cracked it" it was only when input belts are all saturated to begin with

[hardy fable]

what do you mean by rate-limiting "2s"?

[umbral stump]

I'll DM you what I have so it's easier to explain xD

[hardy fable]

ahah! I knew it wouldn't be that easy to figure out

[umbral stump]

Xaetral has taught me how splitter-based (no combinator) flow limiting works

I think I finally understand!

As part of my learning journey I built 1/5 and 1/7 from scratch, and then as a challenge I built 6/13:

Any requests?

Here's a rough outline of the method/thought process, in case someone finds it helpful (basically my notes as I was learning this):

[umbral stump]

Glossary:

• Recycle (R): items that are fed back into the Highest Point
• Discard (D): items that are fed back to the very start of the main input, and are prioritized (by using a priority splitter that uses the discarded items instead of main input)
• Highest point (H): a point that contains the highest throughput of the system -- everything goes through this point

Key concepts:

• Original Input = 1:

  • Initially, all of our input (1 full belt) enters the system

• Discard + partial Input = 1

  • Once the system stabilizes, we prioritize Discarded items over the main input belt, and so we only use up as much of the main input belt as we output (which is the rate-limited amount)

• Highest point = 1 + Recycle

  • We take the Discarded items and the partial input (which together sum to 1 belt), and together with the Recycled items we have the Highest Point

• Discard = 1 - Output

  • We split off Recycle from the Highest point and are left with 1 belt. We split the 1 belt into Output and Discard.

Generic 1/n method:

To get 1/n:

• Find p, lowest power of 2 higher than n
• We need a Highest Point p/n
• To get p/n, we must Recycle (p-n)/n and join it with 1 full belt, which comes from the prioritized input (where Discard is prioritized)
• Since the Recycled part is split off from the Highest Point, it means that we are left with p/n - (p-n)/n = (p-p+n)/n = n/n = 1 belt
• Since we want 1/n Output, we must Discard the remainder of the belt --> we must discard n/n - 1/n = (n-1)/n

Conclusion: to get 1/n:

• Find p, lowest power of 2 higher than n
• Establish highest point p/n by Recycling (p-n)/n and joining it with the prioritized input
• Split Highest Point into (p-n)/n (which is recycled back into the highest point), and 1 full belt
• The 1 full belt is split into 1/n Output, and (n-1)/n Discard

--

Generic a/b method:

To get a/b (a<b):

• p = lowest power of 2 > b
• H: p/b
• R: (p-b)/b
• D: (b-a)/b

[umbral stump]

6/13 Example, using a/b method:

• p is lowest power of 2, higher than 13 (the denominator): p = 16
• Highest point: 16/13
• Recycle: (16 - 13) / 13 = 3/13
• Discard: (13 - 6) / 13 = 7/13

Each component has to be made up from powers of 2:

• To get output 6/13 we will need (4+2)/13
• To get discard 7/13 we will need (4+2+1)/13
• To get recycle 3/13 we will need (2+1)/13

--> We need:

• 2 instances of 4/13

• 3 instances of 2/13

• 2 instances of 1/13

• Path:

Highest Point: 16/13

Only numerators below for simplicity (they are all actually /13)
16 split
|...8 split
|   |...4 to output
|   |...4 to discard
|
|...8 split
|   |...4 split
|   |   |...2 to output
|   |   |...2 to discard
|   |
|   |...4 split
|   |   |...2 to recycle
|   |   |
|   |   |...2 split
|   |   |   |...1 to discard
|   |   |   |...1 to recycle

output = 4+2 = 6 (/13)
discard = 4+2+1 = 7 (/13)
recycle = 2+1 = 3 (/13)

Now to build it, using the path above as a guide...

Expected output:

• Items per 10 minutes: 45*600*6/13 = 12461.54 items per 10 minutes
• Items /s: 45*6/13 = 20.769
  --> Checks out when tested with counter!

[distant gorge]

Have you seen the binary expansion method? It’s a lot easier to read and understand

https://www.reddit.com/r/factorio/comments/cehtbc/splitting_in_strange_ratios/

6/13=0. 011101100010 repeating, so you’d split in half, loop back that first half, then take the next three splits, then loop back the fifth split/power of two, etc

[hardy fable]

I wonder if that doesn't take more space

[distant gorge]

It should reduce down to the same thing I would think?

Like, it’ll definitely be bigger to start, but there should be reductions possible

[hardy fable]

for sure, but the solution I have never uses more than 2 loops but it seems that binary stuff often does

[hearty relic]

#vanilla-chat message

[umbral stump]

for some reason my Discord thinks the video is 12+ days long

[umbral stump]

Oh that's cool ty! Lemme try it with 1/5, 1/7, 6/13 and compare the build with the ones I came up with using Xaetral's method

[distant gorge]

I expect we’ll be able to come up with some rules to be able to reduce the binary one down to Xaetral’s method

[umbral stump]

Here is using the discard/recycle formula on the left (which I shared before), and the binary expansion (right):

Binary expansion method:

• Successive splitters
• Discard 0s
• Output 1s
• Recycle (loopback) from last splitter

1/5 binary: 0.0011001100110011...
repeating: 0011

Path:
Layout:

• Splitter 0 input: main input & discard -- prioritize discard

• Splitter 0 output: discard & forward

• Splitter 1 output: discard & forward

• Splitter 2 output: output & forward

• Splitter 3 output: output & recycle

[umbral stump]

Discard/recycle formula method:

Target: 1/5
Lowest power of 2 higher than denominator: 8
Highest point: 8/5
Recycle: (8-5)/5 = 3/5
Discard: (5-1)/5 = 4/5

(Then you have to work out how to get 3/5 and 4/5 using binary representation, i.e. 3=2+1, 4=4, and route accordingly)

Trying to work out the numbers at each segment of the binary expansion layout

[hearty relic]

yeah, i dont really know why tbh

[umbral stump]

Hmm I struggled to figure out an easy way to work out what the binary expansion ends up doing, here is my best guess:

So binary expansion doesn't actually follow the same recycle & discard logic as the formulaic approach

It goes to higher granularity I think

Which is evident from the greater splitter count in this example

I mean the discarded amount always has to be the same

because it's 1 minus the output, and the output has to be the same in both (1/5)

but the recycle amount I think depends on the inner workings

Well, the recycle amount depends on what highest point you are trying to reach (Highest Point = 1 + Recycle)

And the highest point depends on what power of 2 you need to be able to split into the components (lower powers of 2) that build the numbers you need

So the binary expansion method I believe must have a highest point that is 2^repeat_length

Since: 1/5 binary: 0.0011001100110011...
repeating: 0011

The repeat length is 4

So the highest point has to be 2^4/d

And d is some denominator which I think can be calculated somehow, lemme think

Well our target 1/n in this case 1/5, so 5 is defo a factor of d

But we saw here d was 15 not 5, hmm

Oh

In the binary expansion we always recycle 1/d

And the output is built from the various 1 bits

Oh wait, so we have:

Highest point: 16/d
Recycle: 1/d

And we know Highest point = 1 + Recycle

-->
16/d = 1 + 1/d
16 = d + 1
d = 15

OK so finally I think I understand what numbers the binary expansion will always work with:

• r: Find repeat length of the binary representation of the fraction --> Highest point numerator: 2^r
• To find the denominatord: d = 2^r - 1
  -->
• Output: 1/n
• Discard: (n-1)/n
• Highest point: 2^r / (2^r - 1)
• Recycle: 1 / (2^r - 1)

For example, using the above with 1/5:

• r: 4 (repeat length of the binary representation of 1/5)
• d: 2^4 - 1 = 16 - 1 = 15
  Output: 1/5
  Discard: (5-1)/5 = 4/5
  Highest point: 2^4 / (2^4 - 1) = 16/15
  Recycle: 1 / (2^4 - 1) = 1/15

Which are the same numbers as in the screenshot

--
Just by looking at the screenshot, or at the numbers used by the two methods, it's not immediately obvious to me where the shortcuts are that would allow us to reduce the binary expansion solution into the more compact formulaic solution

Or wait

In this case I think I see it

I think this is the shortcut?

[distant gorge]

You still need to cut 16 by 4 somehow

[umbral stump]

So then when we do that, we are here I think:

[distant gorge]

I think it might actually be where the discard and recycles come back

[umbral stump]

Here I pruned the two discards into one

Instead of 16/2 + 8/2 = 8 + 4 = 12 (/15 --> 4/5)

I directly do 8/2 = 4 (/5 -- > 4/5)

But then it affects the required recycle amount

So we must destroy all subsequent logic and rebuild

[distant gorge]

So you divide 4/5 by 4, discard 3/4?

So like, I’m wondering if that means “if you have sequential 0’s, combine them into one 0 and flip the remaining digits” or something like that?

[umbral stump]

Hmm what do you mean by 3/4 sorry?

[distant gorge]

Recycle three of the four fourths, sorry

[umbral stump]

Ah yea I think you're onto something, that does sound like what this pruning idea does

[umbral stump]

Ah gotcha, yea fifths

Yea of the remaining 4/5, 3/5 must go to recycle, and 1/5 to output

I'm struggling to figure out how this pruning approach would work in general

I think 1/5 is not a good example due to how simple reducing 0011 is

I'll take a break for now though, but it was fun working on this! Ty~

Oh also I just realized now what you meant by 3/4 sorry xD

At first I thought it was a typo instead of 3/5

But 3/4 of 4/5 is 3/5 xD...

My bad

I think since the formula method starts by already minimizing the power of 2, it automatically takes advantage of all possible shortcuts, assuming you construct the numbers that it asks for as efficiently as possible?

Which ofc it is not necessarily trivial to do that as efficiently as possible, but since the power of 2 is minimized, the "main path" you start with is already smaller, so I believe in practice it would be easier to build towards a more compact solution using that as a starting point

On the other hand, I think the binary expansion is much easier to explain and execute, but will leave you with a wider initial solution (which you can make more compact by trimming sequential 0s / 1s, but at first glance it seems that whenever you do that, you have to rebuild most of the thing I think)

[distant gorge]

Right, you don’t build the whole thing at once

[umbral stump]

Hmm, I see now what you meant by the combine & flip idea

At first we are doing:

0011
--> two discards, two outputs, one recycle

but when we merge the two discards into one, we are doing:
--> one discard, one output, two recycles

So we have flipped what goes to output and what goes to recycle

[distant gorge]

You place the splitters first going either way then connect the loops

[distant gorge]

Yep!

[umbral stump]

But idk how this looks like when there are further discards down the line

[distant gorge]

That’s very intuition based and I haven’t sat down and rigorously proven it so

[distant gorge]

It would just keep going I presume?

[umbral stump]

So need to work through an example more complex than 1/5

[umbral stump]

But what gets flipped when there are discards later?

I understand why the recycle and output get flipped here

But not sure I understand what would happen if there were discards mixed in too

[distant gorge]

011101100010
01110110101

[umbral stump]

Oh right, there is no recycle

It's just discards and outputs

Recycle is just something you calculate at the very end

Once you have repeatedly done this pruning

Until everything is as compact as possible

THEN you calculate recycle

[umbral stump]

Which is what you meant here right?

So if I understand correctly, what you said originally would happen here? That is, the binary expansion's solution gets compacted into the formulaic solution

Will investigate that another time, by going through a more complex example ^-^ For now, gotta take a break

Thanks~

[distant gorge]

Sort of yeah

[limber dirge]

Is there a generic solution to mix 3 compressed belts (of different types of items), into 1 almost compressed belt of all 3 items combined in a accurate 1:1:1 ratio?

[plucky isle]

+faq codesushi | this may help

[empty compass]

Codesushi

#bot-stuff message

[plucky isle]

https://cdn.discordapp.com/attachments/334735604342325249/1076302300974415953/blueprint.txt (link refresh)

[limber dirge]

Thats 1:1:2 ratio isn't it?

[plucky isle]

You can set whatever ratio you want

[limber dirge]

Oh cool, I will try it tonight, thanks.

[plucky isle]

Enforced by the circuit logic

[umbral stump]

I'm no expert, but I have just learnt splitter-based rate-limiting

So if you only want 1/3 of each belt maximum (in other words, even if one of the incoming belts is not fully compressed, the system would leave space for its share of the 1/3rd final output), then you could use that

Though ofc, rate-limiting does not have to be done with just splitters (without wires/combinators) -- I think with wires/combinators it's prolly easier in general

[limber dirge]

I was trying to get a splitter based approach (one which doesn't require combinators), belt wire sensors are fine.

[obsidian roost]

CodeSushi

Instructions:
Wait for resources to fill up the control belts, then toggle the top constant combinator.
While running, rotate loaders to simulate input drought or output backups.

[limber dirge]

Not at gaming setup, just wanted to checkout the BP string using the BP bot.

Ah, I see, it uses combinators. Maybe there is no 100% splitter only method.

[umbral stump]

This is what no-wire rate-limiting might look like -- it requires much more space though:

[hardy fable]

what about yellow belts feeding blue ones?

[umbral stump]

true!

[limber dirge]

In the left most case, will the output belt be 75% full? (25% of each type)?

[umbral stump]

No, it'll be 100% full

Each of these contraptions limits the rate to 1/3

[limber dirge]

Oh nice, 1/3rd of each belt is perfectly good enough for my needs.

Who came up with this. If it works, its genius! (about the 1/3rd contraption).

[umbral stump]

Well, it'll be 100% full as long as each of the input belts has at least 1/3 compression

For example here is what happens if inputs are only 1/3 each:

It will still compress over time, just not as fast (though the delay difference is quite small, it stabilizes quickly in a system with so few belts)

[umbral stump]

Idk who came up with it, but just today Xaetral has taught me a method using which I can figure out a way to build any type of rate limiter ^-^

(using splitters only I mean)

I mean in practice you might as well just use a combinator setup ofc

[limber dirge]

Oh wow!

Can oyu link me to that discussion?

[umbral stump]

Since it's much easier to setup and configure

[umbral stump]

A lot of it happened in private, but I shared my "study notes" earlier, lemme link you

[limber dirge]

I love non-combinator based designs that involves splitters and maybe wires only.

[umbral stump]

#930147588160782436 message
Here, read from there

[limber dirge]

I will read and try it out tonight, currently at work. But this is exciting development. I might have questions later on this thread.

[hardy fable]

the og was from here #vanilla-chat message
it seems it was understandable enough so that Fryer had done his own understanding of it so I guess it's not that bad

[limber dirge]

I maintain my own splitter only (uses wires though) ratios.

But a generic solution can be a game changer.

[steep cloak]

I need to make a video on the sushi setup

[limber dirge]

I think it would be fun to have a dedicated sushi hour meeting every weekend.

[umbral stump]

I struggled to figure that out until you enlightened me regarding your thought process around discard and recycle ^-^

[hardy fable]

sushi stuff sunday

[umbral stump]

now I'm hungry

[hardy fable]

right, at least multiple people can explain differently

it's good to see knowledge spreading

[umbral stump]

Btw when you read through it, also check the later bits of convo with Maxreader, who shared a link to an alternative method (binary expansion) that may be easier to implement in general (but would usually result in bigger, less compact setups, if no further optimization is done)

[umbral stump]

I'm trying to think about this, but it feels to me that the blue belt solution would not be resilient to lack of downstream consumption right?

If the blue belt starts filling up, it'll clog I think?

although hmm, all systems would clog without a loopback discarding excess

so I guess the blue belt solution would require one as well, naturally

right, so nvm, it wouldn't clog, same principle, BUT! Apart from the fact that it requires all belts downstream to be blue to guarantee no clogging, it also must loopback from downstream

which is different from the splitter-based one, which instead will simply reserve gaps for each input type, if it is missing

[limber dirge]

I feel the setups that does not require blue belt are better. Because those hacky yellow belt setup are easier in justifying all the sushilogy.

For example this build.

[umbral stump]

or wait am I misunderstanding something?

[hardy fable]

no no, you're right

[umbral stump]

can you do the blue rate-limiting and then split off onto a yellow output?

and loopback the remainder?

instead of having to keep the whole thing blue downstream and then loopback after whatever potential consumption happens downstream?

[hardy fable]

the blue belts can't have a single frame of being halted

[umbral stump]

but a blue splitter can output onto yellow

from one output

and loopback on blue from the other

[hardy fable]

oh, trying to put a failsafe I see

how is this loop different from the regular loop?

[umbral stump]

or will this just fill the yellow output?

I think this will just fill the yellow output won't it

if one of the inputs is missing

yea, cos the blue contains 3 yellow belts

at max

so if any input is missing, the 1 output yellow belt will be flooded, nvm

[hardy fable]

I think this will give unpredicting patterns for the output, since the splitting will depend on the consumption pattern itself

[umbral stump]

well I was thinking that there are almost no splitters in the system

just 3 yellow inputs merging together via blue, and then 1 blue splitter outputs it onto yellow, and loopbacks via blue to discard excess back through each input

but yea I have realized now that this would just flood the yellow output

so the only way to stop that from happening is to have a big blue loop that goes through wherever the items need to go to, and then comes all the way back to discard excess at the input point

[hardy fable]

being able to recover from both a lack of input items and the output belt backing up must really be the endgame of sushi belts

[umbral stump]

or wait maybe I'm being silly

yea nvm xD

This system also has to loopback from downstream

So yea, just use wires 🤣

or if the items don't have to go very far, then looping back is reasonable

[umbral stump]

without loopback, circuitry is the only reasonable way that is resilient to output backing up right?

Oh wait

I saw someone post a setup that used dummy items to reserve spots

But I think that requires loopback too

[hardy fable]

it's definitely possible with dummy items, but I wonder if it's even possible without both them and circuits

I don't think so

[umbral stump]

Yea I cannot really think of a way to force a "clogged" belt to flow on demand without circuits (well, apart from removing items from it )

[hardy fable]

since having a missing item would allow the sushi belt to compress more than it should

[umbral stump]

(or loopback, but with loopback it will never clog to begin with)

[hardy fable]

unless there's a big enough buffer

because that's really what's happening when a regular splitter sushi clogs

it's when this tiny buffer is full

if you expand it beyond the sushi belt total capacity it becomes resilient

[umbral stump]

yea, I suffered through many of such incidents while trying to work that thing out

[hardy fable]

so do I

[umbral stump]

and when the splitters stop splitting (due to the clog), it literally sounds like the thing grinds to a halt lol

or do they actually have a special sound effect for that?

or maybe it's my imagination due to the sounds suddenly stopping

[umbral stump]

ah I hadn't thought about that regarding the buffer -- didn't realize that's what you meant earlier

wait so

if each item's personal loopback buffer is greater than the sushi capacity, then you could merge them without rate-limiting right?

because it would flow constantly regardless?

and it would flood the sushi with whatever is present

[hardy fable]

splitters make sound only when they split, that must be the reason why :p

[umbral stump]

but when a missing input arrives, it will be cycled in?

[hardy fable]

the issue with missing inputs is that if not rate-limited, the remaining inputs will take its place on the sushi belt

when this happens there's no issue yet

the issue appears when the missing item comes back because there is no space left for it

[umbral stump]

right but if that is a desirable feature for a particular sushi, then it can do that by expanding the buffers to be greater than the sushi correct?

that should cycle in the missing item when it arrives?

[hardy fable]

(actualy it really takes its space but there's no space for the item that took its place, if that's an understandable sentence )

[umbral stump]

because the bigger buffers force even full sushi to flow constantly?

I will try to build this to see if it does what I think it does xD

[hardy fable]

big buffer would just be that missing space

[limber dirge]

Slightly confused looking at this, it claims that the rate limiting is not based on combinators, but clearly conbinators are shown in screenshots.

[umbral stump]

Oh those are just measuring the flow

[limber dirge]

Oh interesting.

[umbral stump]

To prove that it is indeed flowing at the desired rate

[limber dirge]

so btw this is good to learn, although in my screenshot I was talking about mix ratio setup, which is different from rate limiting.
#930147588160782436 message

But good to know rate limiting is a science too.

[hardy fable]

ok TIL loaders work opposite to inserters

they output on the left lane first

[umbral stump]

Yea you can mix, such as how I showed here:

[limber dirge]

So any rate limit fraction can be devised? or only certain factors are workable?

[umbral stump]

You limit each item input to the rate corresponding to its desired ratio

In this screenshot each one is limited to 1/3, so the merged output ratio is 1:1:1

[hardy fable]

well a ratio is just two rate limitings :p

[limber dirge]

I feel Sushi talks should deserve its own channel here. Often we end up mixing threads of convos within this thread.

[hardy fable]

like if you have a 1/3 limiter and a 2/3 limiter going on the same belt you got yourself a 1:2 ratio

[umbral stump]

Yea any rational number fraction

[hardy fable]

yeah any a/b

with a and b being positive integer numbers

[limber dirge]

@hardy fable and @umbral stump Hope to sync up with you guys sometime. What time zones are you in?

PST here.

[hardy fable]

GMT+1

[umbral stump]

Hmm I just thought about successive rate-limiting

That might potentially yield more efficient setups than the formula method right? I'm not sure

[hardy fable]

k I'm ready for the test

I am not getting above 160

[umbral stump]

GMT+2

[hardy fable]

it recovered, just as planned 🙂

it's insanely resilient holy shit

[umbral stump]

right that's what I thought should happen -- in that screenshot with red only, there is constant flow right?

[hardy fable]

there is indeed

[umbral stump]

yea, so if buffer > sushi, flow is constant, so it floods with available input, but cycles in missing input as it arrives

pretty cool

[hardy fable]

now let's make that without loaders...

[umbral stump]

🫠

that's where that weaving pattern comes into play xd

what is it called

that pattern that visits each cell

[hardy fable]

back where it all started

[umbral stump]

wow it took me a few seconds for my eyes to properly focus and figure out what was going on here lol

readability has improved a lot

also what's this tractor

[hardy fable]

the good ol substation :p

changed quite a lot since this day of august 2017 didn't it?

also the underground belt shrouds

they were so boxy

[umbral stump]

haha yea

[hardy fable]

and now belts have wheels under their transporting surface

[umbral stump]

oh yea

I didn't even notice how pasted on the ones in the screenshot look lol

minimal shadows too

[hardy fable]

and of course the entire reason of this architecture, splitters weren't able to filter anything back then

[umbral stump]

hard to tell whether they're floating above the surface or what

the belts I mean

[umbral stump]

ah... xD

just like actual sushi belts then

[hardy fable]

right, looks like it's viewed from the top while everything else is at 45°

I need to make a splitterless inserter only sushi lmao

would be naturaly rate-limited

[umbral stump]

yea! and also, if you look at a random segment, if a shadow was drawn at a distance below, you'd think it was floating

and without a shadow, it's ambiguous what its height is relative to the surface

since it is so flat and dimensionless (and shadowless)

[umbral stump]

how about box explosion or machine deconstruction sushi? #vanilla-chat message

though the items would need to stick to belts in the vicinity instead of just piling on the ground

[distant gorge]

Not that hard :p

That’s what I do for nuclear in Angel’s

The only splitters involved are for getting rid of certain byproduct items and a “mixer” to break up groups on the belt

There’s also this shenanigans so I only have to run two belts instead of three

Looks like this in transport

[umbral stump]

oh cool!

So a place like this is where a dummy item could be used to reserve space right?

Though would have to loop it back at the end of this splitter section

[distant gorge]

Sure, but the point is more that I only need 84/min of the red

And 1.57 belts of the green

And their consumption is tied together

The moral is that if you need a tiny bit of an item along with a ton of another, you can mix the belts en route and unmix them at the end. Sort of like mixed trains that service a single drop off station

[hardy fable]

a signal on its carrier right

[umbral stump]

is no-wire rate-limiting going to die out once wires can read splitters and undergrounds?

will that make combinatorless belt wire rate-limiting dominate no-wire splitter rate-limiting?

[hardy fable]

combinatorless and no-wire are not mutually exclusive :p

[formal cipher]

wow, just wow, it's genious

took me a time to wrap my head around, and ki couldnt teach me how to converting decimal fractures in binary. but given the binary fracture, i can now build the sushi i always wanted 😄

just doesnt work. it should be 1/7 - fractional binary is 0. ( 0101 )

but even with 6 times it gets stuck

wait, i have build it wrong

[distant gorge]

One too many splitters at the end

And not prioritizing the recycle separate from the discard

[formal cipher]

did that priorization?

[distant gorge]

Try again with that sentence?

[formal cipher]

the left most splitter has priority input

still gets stuck

[distant gorge]

.

Not literal splitter priority

The last split off has to come in separately from the “0” split offs

[distant gorge]

See here

[formal cipher]

is this the correct implementation?

[distant gorge]

Should be good, now you loop back the discard’s to the input again before that first splitter

[formal cipher]

doesnt work, backs up

kann auch nicht, 0.(0101) ist offensichtlich nicht richtig, KI hat mir Mist erzählt ..

sry, AI lied to me about the exact result, told me was 0.(0101) but thats wrong

asked AI again, 1/7 is 0.(001001)

[distant gorge]

Well there’s your problem lmao

Don’t ask AI

Use an actual fucking calculator

https://www.wolframalpha.com/input?i=1%2F7+in+binary

[formal cipher]

if it just was that easy .. which calculator can do that

[distant gorge]

See link

[formal cipher]

finally ..

[distant gorge]

So massive

Flip your recycle the other way to eliminate the underground

[formal cipher]

a sushi of 9

[formal cipher]

sushi of 6 with loop

[umbral stump]

tru! I just meant like, in my mind there are three types of control:

• No wires
• Belt (and chest/inserter/etc.) wires, but no combinators
• Wires w/ combinators

No wires have the advantage of, well no wires -- and therefore it "just works" (assuming correct design for the intended goal) -- but will likely require more space

Belt wires have the caveat of not being able to read splitter or underground contents (or items inserted into machines)

Combinators can track everything, and accomplish more complex tasks with generally lower space requirement, but require electricity, and care

But then, in 2.0 belt wires can read splitters and undergrounds right?

So will that make no wire solutions even more niche?

Already currently it feels to me that no wire solutions are perhaps "overkill"? Or at least, they feel like they generally use more space and that you might as well just use wires and/or combinators?

I'm no expert in sushi and such though, so I don't really know

[distant gorge]

Entirely depends on the use case

Like, I think measuring the entire belt is usually a total waste of time and effort

Regardless of situation

You’re either better off tracking all the inputs and outputs or measuring what’s on the belt

Unless you’re in the scenario where you always need a particular mix of stuff

At which point it’s worth distinguishing blood belt sushi from “input sushi”

[umbral stump]

What do those terms mean please?

But for example, for rate-limiting, isn't it "much easier" to use a clock? If I am not mistaken, it will work even with non-saturated inputs right?

And so I am left with questioning what the benefit of no wires is in practice? It takes up more space, and requires more design effort -- whereas, if electricity is guaranteed, the clock method would be just as reliable right? and require a fraction of the implementation effort, and almost no space

So practically, is no wire just for very small tasks, or for situations where for some reason electricity supply is not reliable?

(though don't get me wrong, I love the no wire solutions, I think they're very cool and elegant and I have really enjoyed delving into how they work and how to approach designing them -- but here I am just thinking regarding practical problem solving, where minimizing effort, space and resource usage is preferable)

[formal cipher]

would love to do some sushi without the need to loop everything back

knowing entire belt contents could help with this

while eliminating the faulty memory cell

[distant gorge]

Blood belt sushi: tons of items going around in an infinite loop servicing many different machines or recipes in entirely arbitrary/random ratios, you add items when the rate of them on the belt drops. e.g. my nuclear isotope sushi belt

Input sushi: Only feeding one or two recipes in a particular predefined ratio. Might not even need to loop. e.g labs, or my example with uranium and thorium ore.

[shut current]

oof sorry for the ping

[umbral stump]

Like for example someone was asking about merging items in a 20:10:3 ratio, and it seems incredibly quick and easy to design and build using a clock:

Whereas to do it without wires would require designing 20/33, 10/33 and 1/11

Or 2:1 & 10:1 (--> 2:1 = 3 --> 3 * (10:1) = 30:3 --> [(2:1)*30] : 3 --> 20:10:3)

• 2:1 needs 2/3, 1/3
• 10:1 needs 10/11, 1/11

(I guess breaking the wider fractions up like this might yield simpler designs, but not sure if it would be more compact than directly doing 20/33, 10/33, 1/11)

[umbral stump]

---

Would a system like this in theory would be able to ratio merge (using rate-limit clock) without loopback and without reading all belts (only requiring a few belts to be wired)? I think it should be able to

• Only allow input & rate-limit segment & rate-limit clock to flow/tick when:
  • There is 1 full input belt segment of each item
  • There is space at output (belts are not full)

At input, we check to ensure there are 8 items of each type on 1 belt segment of each item type
--> If not, we pause the clock as well as the input & rate-limit belts

At output, we leave some belts of space (depends on belt speed, but say ~10) and then have 1 hold read belt directly followed by 1 pulse read belt. We connect the pulse to a memory cell that resets every ~60 ticks (again, depends on belt speed). If at any point the memory cell reads 0 for a full cycle, while the hold belt contains any items, then it means the hold belt items did not flow into the pulse belt --> belts are full downstream
--> If that occurs, we pause the input rate-limit clock as well as the input & rate-limit belts
(The reason for leaving some belts of space is because the memory cell system needs to run a full cycle followed by a few combinator ticks' worth of calculation before it reacts)

--

If I am not mistaken, this essentially mimics the fully saturated item composition of the simpler clock rate-limit & loopback system

• As long as there is no fully saturated input of all types arriving, the whole rate-limiting system freezes essentially
• And if the output is getting full, it also freezes
  --> It only flows at full saturation, during which it mixes items the same way as the simpler system would when the simpler system is at full saturation
  --> And it pauses when the output is getting full

[small sonnet]

a) thank you for doing the thinking for me :b
second, what parameter does the clock cycle adjust?

[distant gorge]

Can you give me a picture?

[umbral stump]

xD I created a demo of a simpler variant of it as part of some discussion here:
#combinators message

But that demo was not perfect, because:

• I was not pausing the clock, but only the belts, which can introduce inaccuracies due to variance caused by when the input belts are not saturated
• I was measuring the whole sushi belts for simplicity, rather than doing the hold->pulse idea which should be able to avoid needing to wire the whole sushi

[umbral stump]

I'll remake it later or tomorrow according to the specs I outlined above

[umbral stump]

The clock is used to rate-limit the belts to force the items to merge in the desired ratio (example below:)

[umbral stump]

Please see this message + screenshot for an example ^

---

(Btw on a different sushi topic, look at this abomination I made while trying to help someone -- if someone's interested and reads through the context, I'd be curious if more compact designs are possible [without filter inserters or long handed inserters])
#1254776516018180146 message

(Ignore the cars and wires, those are just to simulate the OP's mod's machines)

[glass needle]

Why no filter or long handed inserters?

Oh that message was a while ago haha

[steep cloak]

I mean half of that is already done
https://mods.factorio.com/mod/no-more-long-handed-inserters

[tacit cedar]

found a 3-wide 1/4 rate limiter where the splitters kiss
also what's the notation for belts contraptions?
I see 4x4 for balancers, but would rate limiters be 1/4? since only a fourth of the belt is filled?

[slim wave]

There’s no notation

I see people use : more than x

Rate limiters don’t have any common notation though

[tacit cedar]

also for balancers I've seen
1:1
1x1
1-1

the hyphen makes sense because its like "4 to 4"

whereas the x is like "4 by 4"

if that makes sense

[steep cloak]

dashes to me mean train notation, colons mean ratios in general which balancers are

[tacit cedar]

oh makes sense

[tacit cedar]

btw why does splitter sushi sometimes get backed up here?

and what's this called/what's it for? like a basewide blood vessel sushi belt

oh this is a good screenshot

[tacit cedar]

also I wonder if you could guarantee somehow that splitting the belt doesn't leave some without full sushi, idk

[normal gulch]

I’d have to have a blueprint to troubleshoot

[normal gulch]

Hm I’ve done this before but for rail blocks

[tacit cedar]

idk it just appears sometimes when I do splitter sushi, I'll try to remember to send you one next time I notice it

[slim wave]

It’s because of the item buffers and little transport line micro-gaps

Sushi belts have trouble totally compressing

[tacit cedar]

rip, should I be leaving a space?

like for 3 items, do them all 1/4, so there's a space of 1/4?

[slim wave]

Try putting priority on those splitters. Make sure your belts are recycling with unlimited throughput

[tacit cedar]

priority on which splitters? and it should be throughput unlimited, it's only one belt

[slim wave]

The merging ones. They might behave nicer with input priorities.

[tacit cedar]

so the bottom 3? priority input I assume? to the bottom belt?

[slim wave]

It’s just a limited throughput system, limited to just under 45 items/s because it’s a sushi belt. If you put less items on it you will only have reduced the maximum throughput further.

You can put less but if you put them all to 10.75 items per second you’re only losing a little throughput

[tacit cedar]

more annoying to do with splitters though
combinators would be easier

[normal gulch]

I get what your saying lizzy but at the same time i've done it at full capacity and it should work

3 signals wrong, and 20+ trains deadlock

to be honest this entire thing isnt designed too well, I never should've done the smelter on the opposite side of the big iron patch

kidding, its over 100 trains

[hardy fable]

it's items being stored in the little buffer spaces of the sushi

the first filling up are these

then then they cause micro lags in the sushi

causing the merger you mentioned to fill a bit as well

it should happen when you play around with the sushi running, any stop in the flow would cause that

the solution being to consume items, enough for the buffers to empty again

it should definitely not happen during regular operation

[steep cloak]

sushi blocks

[clear wedge]

Blueprint String

T1 Module Factory Mk.2

[hearty relic]

about 15 T1 modules a minute

[slim wave]

Have you ran this on high speed with variable inputs

I can see that sushi with the wires and plates getting clogged

[hearty relic]

Circuit control on the wire and plates stops the belt to make room for wires

[slim wave]

I see

[clear wedge]

Blueprint String

T1 Module Factory Mk.2[recipe=coal-liquefaction]

[hearty relic]

A coal liquification variant

[tacit cedar]

it keeps happening during normal operation

the recycler doesnt clog but the bit after the rate limiter does

idk how long a video I can upload lol

[hardy fable]

looks fine to me 🤔

@tacit cedar is it filling more and more?

because if it stays that way it is good

[tacit cedar]

I fixed the clog just to see if it would happen again

[tacit cedar]

it did happen again

i will leave it longer

[slim wave]

Try this

Might have marginal improvement

I don’t expect that to fix it though

And look at transport line gaps in debug

[tacit cedar]

the buildup is just here currently
left side of solder (gray spool of wire) and copper cable splitter
and it continues upstream

I wonder if it being all left turns means the left side clogs

idk what the lines mean tbh

oh you said gaps

[slim wave]

Have you posted this in the technical factorio discord?

Here you can see the belt stalling

[tacit cedar]

i dont think im in that discord

[hardy fable]

I'd like to mention that I wasn't able to reproduce the issue

it just works™️

[tacit cedar]

weird, and you left it running for a while with assemblers taking from it?

[hardy fable]

yeah I put the game on x16 speed

with and without inserters active

[tacit cedar]

maybe the bottom right splitter is magic ¯_(ツ)_/¯

[hardy fable]

oh you don't have it

shouldn't change anything

my 2 guesses are

1. you fiddle the sushi belt after it started

2. mods

tried with T1 belts as well

[slim wave]

Don’t think these mods or any mods will affect belt behavior

I’ve been thinking about this all day and I still have nothing

Literally all day I’ve been turning it over and over in my head

[hardy fable]

I've seen mods make inserters sometimes not wanting to put items in only my modded furnace at absolute random, so who knows

[muted patio]

oh

[foggy junco]

it's not getting backed up

it does stutter every now and then, but the inputs don't back up, it just stays the same way

if you had the input running before closing the loop, that can happen since there's items on t he belt already

and since the input + output is exactly a yellow belt, those extra items won't lead to it backing up, but also won't get cleared

to fix this issue, just run over the belt where the stutter is happening and manually collect the items

and then it should run perfectly smooth afterwards

[tacit cedar]

I swear I wasn't touching it, idk

[foggy junco]

just run over it, manually collect all the items where the clog is happening

[tacit cedar]

maybe I'm delusional

[foggy junco]

and it should be permanently fixed

[tacit cedar]

yeah but

it just happens sometimes
I haven't had anything clog to the point it stops working

[foggy junco]

it won't fix itself

but it also doesn't cause the entire thing to back up

and still keeps the belt perfectly filled with 1/3rd

e.g. if you have a chest where you input exactly 15 items / second, and output exactly 15 items / second

and input e.g. 50 items into said chest

it will always stay at around 50 items, right

since it cannot empty as it would require the input to be smaller than the output, right

[tacit cedar]

also I found someone with a smaller 1/4 rate limiter (right) so I stole it

[foggy junco]

that's the same thing here

[tacit cedar]

splitters do not kiss though

[foggy junco]

btw, if space is an issue, you might wanna try circuit based limiter designs

[tacit cedar]

more like

circuit cringe

[foggy junco]

all you need is one constant combinator + one arithmetic

[tacit cedar]

ugly

where are the sprawling splitter abominations

there could be like 4 times as many splitters in this

also I forgot to add lights

[slim wave]

You don’t need combinators for belt limiters

[tacit cedar]

ok I think this is a 1/7 rate limiter for sushi science

you can also just rate limit with different belt colors (yellow -> red = half full red)

[tacit cedar]

yeah that looks about right

[tacit cedar]

got a 3-wide 1/7 rate limiter design, could maybe be made shorter?

[obsidian roost]

Unnamed Blueprint

[tacit cedar]

circuitless sushi science^

[hearty relic]

Love this technique btw

[waxen urchin]

circuitless sushi leads to the question

circuitless circuits

[slim wave]

Oh I did that last year

I made a belt saturation detector with splitters that can be used to do computing

[tacit cedar]

can it run doom?

[slim wave]

I haven’t done a Turing complete computer with it yet

And data corruption is a major concern with this thing. I think I’d struggle to make memory cells

[hardy fable]

I have done belt math in the past

because I wanted to test my fuel cell factory without having to build and run a huge nuclear reactor

so I made a belt contraption that outputs a single used cell for every single fuel cell that gets into it

= 1 -
used cell = 1 -

[tacit cedar]

oh nice
so the iron plates are constantly looping, if something goes into the priority input, it'll replace the iron for that loop, get filtered out, and then there's a gap for the fuel cell to go in

i assume the top splitter is just a buffer so that there's a gap for it to back up a little bit in case the iron gets replaced for a loop

[hardy fable]

you assume very right

fellow factorio/kirby/noita enjoyer

[tacit cedar]

Hämis 👍

[steep cloak]

funny how different people jump to different solutions - I would have made a recipe that turns fuel cells into used fuel cells so it's just an assembler

[tacit cedar]

not really sure where else to ask
I tried #quick-questions, #blueprint-designing, and making a thread in questions: #1268449341870182544
I got this 3/5 rate limiter down to 3-wide with weaving, but I really want a 4-wide blue-belt-only one
also @distant gorge said that I'm using too many splitters? idk how many is optimal

I made this by taking a 1:5 balancer, sending 3 outputs to the end, and setting 2 outputs to the beginning with priority input
and removing doubled splitters

left to right
5:5
merged belts and removed redundant splitters
"waste" 2/5th of a belt goes back to input

[distant gorge]

So, step 1: find the highest power of 2 greater than 5, so 8. 8/5s of a belt will be the input to your "recycle" loop splitter. 5/5ths will come from your input belt, 3/5ths from your recycle loop.
Step 2: 8/5 is split into 4/5 and 4/5, both of which are higher than the needed recycle so discard one and keep going.
Step 3: 4/5 is split into 2/5 and 2/5, add one to the recycle loop and keep going.
Step 4: 2/5 is split into 1/5 and 1/5, let one be your output and the other go to the recycle loop.
Step 5: Take all of your recycle loop stuff and put them into your splitter from step 1.
Step 6: Take all of your discarded belts and send them back to the input belt

Should only need 4 splitters

I'm still trying to figure out the way to convert between the binary expansion method in this post and that method: https://www.reddit.com/r/factorio/comments/cehtbc/splitting_in_strange_ratios/

[waxen urchin]

you want a 4 wide with no belt weaving?

[tacit cedar]

3/5 from my recycle loop?
if 3/5 goes out then surely only 2/5 comes back?

[tacit cedar]

yeah

[distant gorge]

No

[waxen urchin]

slide the blueprirnt blud

[obsidian roost]

Unnamed Blueprint

[tacit cedar]

iirc the top one is rearranged a bit, I think I swapped the order of two of the splitters

[waxen urchin]

wait but the bottom one is literally 4 wide

[distant gorge]

5 splitters, sorry

red is discard, green is recycle

[tacit cedar]

recycler gets 4/5 + 2/5 > 1, if I did it right

[distant gorge]

not setup right

[tacit cedar]

oh you put the recycler separate

[tacit cedar]

is that blue 1/5 the output?

[distant gorge]

yes

[tacit cedar]

output should be 3/5

this is my 1/5

which does only use 5 splitters

[waxen urchin]

this is closest i got at 1623

its actually going down

[tacit cedar]

yeah it takes a while to adjust

[waxen urchin]

itll probaly stabilize at 1620

[tacit cedar]

idk how to make it faster, i stole the combinators

[waxen urchin]

its fine

[tacit cedar]

iirc it's average over a minute or so

I just speed up time

[waxen urchin]

idk how to make it 3 wide without belt weaving

yeah its stabilized at 1620-1621

@tacit cedar have you tried using inserters?

[tacit cedar]

no sounds gross

[waxen urchin]

but it sounds promising...

[tacit cedar]

my brain was SO STUCK on it being 5 wide, trying to get it 4 wide

but it was 4 wide the whole time

wtf

i was working on this for days

[hardy fable]

would this do the job?

[tacit cedar]

it does seem to work

[hardy fable]

I can clearly see the pattern tho

[tacit cedar]

wym

[hardy fable]

I don't use your fancy counters 😄

I just check the ratios by hand

[tacit cedar]

oh idk

I plugged it into the item counter and got 1620 per minute, which is equal to 3/5 * 2700

I moved on to 4/5, this feels kinda close to being 3-wide without weaving, idk

and this 3-wide with weaving possibly

[distant gorge]

should be six minimum I think

[tacit cedar]

🤔

[tacit cedar]

which is the output?

[distant gorge]

blue at the bottom

[tacit cedar]

that's still 1/5?

[distant gorge]

oh, derp

hold on

got lost in the sauce

[waxen urchin]

the throughput of the top is 36 isnt it?

the most top belt

waitno

[tacit cedar]

idk but I can make it red without breaking anything

[obsidian roost]

Unnamed Blueprint

[waxen urchin]

the top throughput is 20

[obsidian roost]

Unnamed Blueprint

[hardy fable]

fun challenge, managed to do it 3-wide but not without making it stupidly long

[tacit cedar]

oh that's very clean

[tacit cedar]

nice

[tacit cedar]

bottommost underground too long @waxen urchin

by one tile

[tacit cedar]

look inside
same length

[tacit cedar]

the 5/9 also looks a bit wide

[dim rose]

You can definely make that three wide with enough undergrounds.

[tacit cedar]

yeah

honestly 2/9, 4/9, 5/9, 7/9 are all kinda ugly

and 8/9

I'll just finish up the book and send the whole thing

[clear wedge]

1/2 Rate Limiter
1/2 RL, Belt Only
1/3 Rate Limiter
1/3 RL, Belt Only
1/4 Rate Limiter
1/5 Rate Limiter
1/6 Rate Limiter
1/7 Rate Limiter
1/8 Rate Limiter
1/9 Rate Limiter
1/12 Rate Limiter
2/3 Rate Limiter
2/3 RL, Belts Only
2/5 Rate Limiter
2/5 Rate Limiter Weaving
2/7 Rate Limiter
2/9 Rate Limiter
2/9 Rate Limiter Weaving
3/4 Rate Limiter

[clear wedge]

3/5 Rate Limiter
3/5 Rate Limiter Weaving
3/7 Rate Limiter
3/8 Rate Limiter
4/5 Rate Limiter
4/5 Rate Limiter Weaving
4/7 Rate Limiter
4/9 Rate Limiter
4/9 Rate Limiter Weaving
5/6 Rate Limiter
5/7 Rate Limiter
5/8 Rate Limiter
5/9 Rate Limiter
6/7 Rate Limiter
7/8 Rate Limiter
7/9 Rate Limiter
8/9 Rate Limiter

[tacit cedar]

hm that did not send just the book

https://factoriobin.com/post/qD01iUh5kAyzYHs3-EXPIRES

the X/9 rate limiters are all ugly

tragic

[distant gorge]

Heh

1/9

[tacit cedar]

thats actually the only one I was kinda happy with its design
also hey, only 6 splitters

[distant gorge]

hmm, how did that happen

[distant gorge]

ooh, factorizations

I think I've got it now

[tacit cedar]

idk what that means

[distant gorge]

1/9 is just two 1/3's in sequence

[tacit cedar]

i dont do math i just steal balancers and remove the splitters i dont need

[tacit cedar]

iirc I made the 1/12 by doing a 1/4 and a 1/3
though I routed both recycle belts all the way back to the same starting belt

[distant gorge]

recycle vs discard are separate concepts

[tacit cedar]

wym

[distant gorge]

green is recycle because it's part of the "infinite sum", red goes all the way back to the beginning and isn't part of the infinite sum because the first splitter is always overfull for the output

like, the output of the first splitter is always a whole belt, never partial belts

[tacit cedar]

oh I see, so "discard" is the leftover that isn't part of the ratio, that goes all the way back
and "recycle" is just the normal loops

[distant gorge]

so the inputs to it don't affect the output at all

yep

The recycle is what corresponds to the repeating decimal expansion

[tacit cedar]

honestly i never really got how the loop exactly affects everything
I just know that I can loop belts from the end to the beginning to make an n:n balancer smaller

[distant gorge]

0.333333333333333333333 in binary is 0.01010101010101010101010101010101010101010101010101010101010101...
So you want to discard the first half, then keep the second half, then send the leftover back to the beginning to make the infinite loop

There are some tricks to reduce more complicated repeating fractions, but doing that will always work

If your decimal repeats in n number of digits, and m of them are 0's, you should need at most n+m splitters

[tacit cedar]

so 2 digits, plus 1 of them are 0s, so a 1:3 balancer needs at most 3 splitters?

[distant gorge]

yep

I'm still working on the full set of reduction rules

the post I linked earlier will always work at least, but can use too many sometimes

[tacit cedar]

i know the wiki page on balancer mechanics says For n → n balancers where n is a power of two numbers, n×log2(n)−n÷2 can be used to calculate how many splitters are needed.
and i think it's safe to assume for any balancer where n isn't a power of two, you just go up to the nearest

[distant gorge]

correct

That's what determines the amount going into the second splitter, which should have the first recycle loop

and that first recycle loop must equal (power of two) - (denominator)

[tacit cedar]

I think turning an n:n balancer into a 1:n balancer lets you remove half the splitters, then add 1?

[distant gorge]

so for 1/3, the power of two higher than 3 is 4, so you need a recycle of 1/3

[tacit cedar]

and same for n:1 balancer, since it's just 1:n backwards

[tacit cedar]

i guess remove half the splitters, round up, then add 1

[distant gorge]

How many did you get for 2/9? I think I'm getting 7?

[tacit cedar]

might as well make yellow belt versions

[distant gorge]

sequential 1/3 and 2/3

8* not 7

[tacit cedar]

i wonder if this can be 3 wide

slightly altered version

[tacit cedar]

that it?

[distant gorge]

ye

I think

[tacit cedar]

looks good

tried making it smaller

[dim rose]

Have you considered side-loading as a rate-limiter? Depending on the use-case you might want to combine it with a lane-balancer.
If the input is already balanced across both lanes, you can use side-loading to easily limit the throughput by another factor of 2. (It might not be that compact though.)
Combining side-loading with different belt levels can also be used to easily make a 1/6 rate-limiter for a blue belt.

[hardy fable]

I did that for my SA setup

to get 1/12 you just do 1/6 on a half belt

and you can use the red to green ration to get down to 1/3

[tacit cedar]

oh true

and yellow to blue is 1/3, I put those in the book

I will look into the side loading later, am taking a break

[tacit cedar]

what did you use to make these?

[tacit cedar]

I think bottom looks better
I still think it could be smaller though

[tacit cedar]

wait im stupid lol

should probably not only draw from one side though

lane balancer costs two splitters though...

top: 1:1 lane balancer
middle: input-imbalanced rate limiter (only draws from right lane)
bottom: input-balanced rate limiter

but the thing is, I can already do it with less splitters

[tacit cedar]

wait a second, i made my yellow belt only 2/5 rate limiter smaller than my blue one 🤔

and with blue underground length it's so close to being 3-wide

[tacit cedar]

I got these 2/7 ones that are shorter, but not 3-wide

several more designs for 2/7

I wonder if there's some kind of "collaborative blueprint book"

oh wow, the bottom one I just made uses the exact same number of splitters, belts, undergrounds, and is the same dimensions as the top one that I made a few days ago

[tacit cedar]

saved a tile with belt weaving and copy tool splitter

this looks kinda cool actually

[tacit cedar]

a couple 2/9 designs

[distant gorge]

That was just excel to color in squares lol

[tacit cedar]

made a smaller 1/7 with one less splitter (top), the belts seem kinda weird still, maybe more space can be saved

I thought the belt weaving felt more compact but it's the same size...

[tacit cedar]

hm, the stacked version (which can go up to 7 high) seems kinda tall

compared to combinator sushi

different color for the starting splitter on 1/2 can make 1/3 or 1/6

I feel like this is kinda getting out of hand

[tacit cedar]

what do you call it when you use different colored belts, but not undergrounds so it's not "weaving"

[hardy fable]

I don't really think it has a name

[tacit cedar]

there are so many different ways to make a rate limiter
I prefer the pure blue belts, but there's also sideloading, belt weaving, and using lower tier belts to rate limit

[hardy fable]

btw using the sideloading trick you can do 1/4 and 1/3 for science instead of 1/7, makes the system smaller

e.g.

[tacit cedar]

oh having two different ratios on the same belt sounds neat

couldn't get up to the full 2700 spm though could it?

[hardy fable]

yeah any trick can help

like passing two different items through the same rate limiter to further make use of that precious real estate

[hardy fable]

you are right, because half of the belt is at 1/4 (thus 1/8 of the whole belt)

but it's not that bothering, as the lab chain is still quite long

[tacit cedar]

here's what I got

[hardy fable]

gud 👌

[tacit cedar]

what if I sorted the Rate Limiter blueprint book by X/1, X/2... instead of 1/X, 2/X...
that way, you can make sushi easier, like if you're making advanced circuits, you can just open X/6 and do
sulfur: 1/6
engine units: 2/6
advanced circuits: 3/6

though tbf currently they simplify and you can just open 1/X and do 1/6, 1/3, 1/2

[tacit cedar]

https://factoriobin.com/post/1VndQ76Z
break from this bp book for now

[hardy fable]

that 1/3 is very big

you can get it down to 3x6

[tacit cedar]

how

[tacit cedar]

this? looks fucked tbh

[hardy fable]

but under the hood it still uses the same discard/recycle method

[tacit cedar]

yeah it's topologically the same

btw anyone know what raynquist uses to make these nice graphs?

the Contraption I am working on is beyond my comprehension for the moment

and I may need to write it down

[hardy fable]

I don't see anything that couldn't be done on draw.io

[tacit cedar]

that looks cool ty

hell yeah, 4:4

ok i am getting kinda tired lol

[tacit cedar]

https://factoriobin.com/post/84hST2T-
head hurt, sleepy

todo: remove redundant and useless splitters
can probably replace the middle bit with a normal 8:8 balancer
i hope changing the loop back from two right lanes to just a normal belt is fine

[dim rose]

mixed-level

[slim wave]

Looks like something that’d be easy in ms powerpoint

[tacit cedar]

I wonder if there's a way to have circuitless/wireless sushi that outputs onto the sushi belt

[tacit cedar]

like I figure you just leave some room
but that removes throughput
so you'd have to leave gaps
like for advanced circuits could be something like 2/5 copper wire 1/5 plastic 1/5 green chip (1/5 red chip)

red chips are 4:2:2 -> 1 so maybe you could get away with only a ninth of the belt empty?

[tacit cedar]

1/2 rate limiter?

1/6?

well 1/2 works

OH WAIT 1/6 HAS SIDELOAD

[tacit cedar]

put all 3 loopy designs together

[tacit cedar]

ok thats a big clog, the right lane stopped moving completely

[tacit cedar]

holy shit i can just keep adding loops

[shut current]

Mitru's discovered a personal god today

Fun fact, if you connect the correct back-loops via tunnels to certain points, you can make all sorts of fractional ratios like 3/14 or 7/12
You can also connect back-loops from either side, not just the top

[hardy fable]

wut

no way

[distant gorge]

You could’ve just moved the underground exit one left

[normal gulch]

woah

[tacit cedar]

looks nicer this way imo

[tacit cedar]

wtf it clogged again

idk whats going on with it

the right lane is most of the problem

[slim wave]

Merge the circuits and robots first

[tacit cedar]

I'll try it, but why would that change anything

[slim wave]

Shouldn’t these be recycled all the way to the back instead?

Or at least not merged back with themselves. I think that’s the issue

It gives you 3/8 instead of 1/4

[distant gorge]

Inserters don’t pull evenly from both lanes

You need to lane balance before your recycle loop

[tacit cedar]

yeah I know the left lane gues pulled from more frequently

but I didn't think it should matter

since each lane is basically its own separate sushi belt

[hardy fable]

you are right, it should not matter

u and ur weird game again, I'm sure I wouldn't be able to reproduce the issue like the last time

[dim rose]

Is it possible that the ratio's aren't perfect due to the simulation limitations (discrete ticks)? In this case the ratio would sometimes be 1 off due to rounding.

Only some of the ratios would have this problem. You could try adding some blue belts so there is more space for the items to merge.

[hardy fable]

if the ratio is off by anything above what it should be you'd get a clog

but they already had a similar issue that I wasn't able to reproduce using the exact same setup so idk

[pulsar lance]

Theory based on that: the left lane gets pulled more frequently, but depending on the frequency of the items, this occurs more often, so say all of the blue chips are taken from the left lane, but robot frames are taken 80% from the left lane and 20% from the right?
Though that shouldn’t break something should it?

[tacit cedar]

it shouldn't

also check out this loopy sushi

[hardy fable]

that's beautiful :D

[sand estuary]

I cant lie, that might be a good way to make it viable for all the space age sciences

just add more bends

[tacit cedar]

I did already make this

probably could be improved now

[hardy fable]

just do 1/3 on half a red belt into your green sushi belt 🙄

I guess I don't need to be shy, here's my planned SA setup

[formal cipher]

gonna learn this again when sa hits 😅 #930147588160782436 message

are 12 science packs confirmed?

[hardy fable]

dunno but that's how many they told us about

and how many we've seen actually, since the last one is on the SA art

[tacit cedar]

https://factoriobin.com/post/yMbMfBrc

[slim wave]

IR3 sushi, I reformed one of these rate limiters into a convenient shape

[tacit cedar]

breaks if copper runs dry?

[slim wave]

That’s true it could

I’ll rate limit the tin to 1/2 so it can re-balance

But that’s for tomorrow

[slim wave]

By the way that limiter book is really handy

I think that the underground belts obfuscate it a little bit though, and it would be helpful to have a set with minimal crosses for reference

Maybe I’ll make my own

[tacit cedar]

wym underground’s obfuscate it?
I made the rate limiters by messing with balancers
For an X/Y rate limiter, I took a 1:X balancer, combined Y outputs for the output, then routed the rest back to the beginning with priority so they’s never clog, then got rid of excess splitters

If you find better solutions for rate limiters lmk
I am taking a break from factorio for now to try out Stormgate, since it became f2p

[slim wave]

These ones that help make it skinny also make it more confusing to look at

[tacit cedar]

My only real design criteria were “try to make it 3-wide”, “try to not belt weave”
I figured raynquist’s balancers are also hard to understand, have you seen the 9:9?

[slim wave]

Balancers are harder because they have to cross frequently whereas rate limiters won’t have to cross as much

[tacit cedar]

not 1/3?
copper is 2/3 right?

[slim wave]

That would take up some extra space but 1/2 will leave enough belt gaps for the copper to refill and rebalance

[tacit cedar]

true I guess 1/2 would be sufficient

[shut current]

I've found the simplest solution for copper-age bronze is to have a full belt of copper on the outside each, as well as having each sides output.

A full belt of tin and coal respectively, side by side run up the center between the pair of columns.

Use long and normal inserters to interface with the resources.

[steep cloak]

@tacit cedar I've found it fun to work on compacting and yellow belt-ifying some designs (1/1 to 1/12)

the 1/11 rate limiter is mega ass but hey it works

I didn't try very hard on it, just wanted to get one done

there's probably a design where sideloading and a lane balancer is used for the return belts

[slim wave]

Have to keep in mind they must be able to empty after a backup

Not only to keep up with the rate but to empty back to stable as well

[steep cloak]

these have literally exactly the same topology

what you're talking about is outside the scope of these designs

and generally sushi shouldn't be backing up, let's say you have a 1/3 rate limiter and a full input belt, when you recycle there's up to 1/3 of a belt being brought back, which gets input priority, and since you're putting out 1/3 of a belt then it just gets put right through the rate limiter and everything is fine

[slim wave]

I see what you’re saying now

I was referring to your sideloading thing but I misread it

[tacit cedar]

wow ur right this is fucked up

modifying a 1/12 seems like a good enough solution

[distant gorge]

If my conjectures are right, the higher prime numbers should get progressively worse to do

12 is nice because it’s just /2/2/3

[tacit cedar]

luckily, just like balancers, you can kinda just loop back some outputs

which is what I did for the 1/11 in the most recent screenshot

[sand estuary]

If the primes are a problem, could you not do a 1/22, then just add another one

Not my area of expertise tho

[dim rose]

22 has 11 as a factor, so I think it'll be at least as bad. I think that the large prime factors are more specifically annoying.

But #930147588160782436 message I think this technique will let you do 1/n pretty easily for any (integer) n.

[tacit cedar]

the reason non-primes aren't a problem is because you can create them out of primes
so the easiest way to make a 1/22 is to start with a 1/2 and 1/11

[sand estuary]

right ok

[tacit cedar]

like the 1/12 in the most recent image is a 1/4 and a 1/3

[sand estuary]

yeah

I wonder if theres 2 fractions that could be added together to make 1/11

1/132 + 1/12 works

[dim rose]

132 is divisble by 11, so you'll probably need to make 1/11 anyways.

[sand estuary]

hmm

[sand estuary]

looking for a combo of 3, but obviously the searchspace is much bigger

[tacit cedar]

what exactly are you trying to do

[sand estuary]

I'll make a python program

find 3 fractions that add to 1/11

or 4, or 5

solutions must exist

[tacit cedar]

so to add to it, the denominator must be divisible by 11 already right?

i do not think there are solutions

[sand estuary]

Hmmm maybe

[dim rose]

Yep, for any number of fractions that you want to sum to n/m, one of the denominators of the fractions in the sum must be divisble by m. (n/m must be a fraction in it's simplest form).

It's a pretty simple proof. (Max 6 lines of algebra.)

[sand estuary]

Right ok

[normal gulch]

I do have a question for yall: why do you devise all these ways to do splitter sushi when circuit sushi works with any ratio in a smaller footprint?

[sand estuary]

Because we can

And because it looks really elegant

[distant gorge]

ooh that's a really good way to reduce high prime factors, good catch

[sand estuary]

Nah

[sand estuary]

^

[distant gorge]

You can definitely reduce composite numbers by putting the prime factorization of "slow down" belts in series

ah, I see

I still need to sit down and figure out how to algorithmically reduce the binary decomposition to the smallest form

[sand estuary]

yeah

[tacit cedar]

because nuclear is too hard and combinator sushi cant handle blackouts

[slim wave]

Circuit belt sushi does do it

[tacit cedar]

too many types of sushi

i spent too much time learning balancers and such to not use it

thats why all my mines look like this

[normal gulch]

then place literally 2 solar panels and an accumulator and you're fine

[tacit cedar]

i dont have that kind of money

i spent it all on splitters

[sand estuary]

Splitters are the elegant solution

So pretty

[normal gulch]

with circuits, you can on the fly adjust the rate of items. cant do that with splitters

[hardy fable]

who cares, we want beautiful things

[dim rose]

You don't need sushi for simple setups for kovarex enrichment + nuclear-fuel reprocessing. With a single belt that loops it's pretty easy. You just need the right priorities on the splitters.

[sand estuary]

Why would you need to change the rate of items? Usually its for labs or something, where the ratio is fixed

[dim rose]

if you're adding more sciences...

[sand estuary]

Eh, if you know theres 14 sciences total, just set it to be 1/14 each

even when theres only 2 packs unlocked at the minute

thats also goo, you dont need circuits unlocked, you can do it much earlier game

[hardy fable]

your labspeed change anyway

are there really players that don't change their lab setup along the way?

[tacit cedar]

I do change it

at least adding prod

[hardy fable]

usually I do the first 4 packs sequentially and spaghettily, then I refactor evereything by adding yellow and purple at the same time on a sushi belt + prod modules in science pack assembling machines

and I never made space packs

[normal gulch]

for anything uncertain. EG a building belt so you can build the factory from a single belt

or a ondemand mall

[sand estuary]

Yeah but thats a different use case

using splitters for labs it totally logical

[normal gulch]

sure but it can be smaller 🤷

[sand estuary]

no power draw, it looks super pretty

[normal gulch]

ive done splitter sushi for a while and personally like circuit more

[sand estuary]

I'm not going to stop you doing circuits, but blatently going "Oh nobody should do splitters, because their bad, you can just do circuits" is a bit arbitrary

[hardy fable]

different people with different tastes

[sand estuary]

exactly

[normal gulch]

I never said that I was just questioning thats all

[hardy fable]

you don't like splitter sushi, I don't like circuit sushi, why not

fine, I made splitter sushi because I didn't like the randomness of circuit sushi

plus I like making big things from restricted stuff

like circuit has a lot of wires and functions all around, it's physically clean but conceptually messy

splitter sushi in another hand is physically messy as there's lot of belts and stuff, but it's conceptually very simple, just belts and splitters

[normal gulch]

thats fair

I use both approaches I just feel that splitter is more limiting, which can be good

[hardy fable]

also I like the idea that this solution is the most energy efficient, event if it's down to a few watts (first because of the splitters, but also because sushi only require 1 inserter per lab)

it pleases my ocd

[tacit cedar]

isnt there only like one infinite tech that takes 7 sciences?
rest are militaryless or purpleless

so you can reuse 1/6 sushi

[distant gorge]

Other way around, only two of them don't take seven

but they're the ones you're actually going to find useful to keep researching

[normal gulch]

only 1 takes all 7

[distant gorge]

ooh okay, got that mixed up with which ones take military

[sand estuary]

Going to start working on my own rate limiter stuff. Can someone send me the combinator display blueprint string for the rate output?

[tacit cedar]

#vanilla-chat message
this is what i used for the combinator math (I stole it)

and I just hooked up the output to santa's nixie tubes

[sand estuary]

Ok thanks

which one is the output?

[sand estuary]

I'm useless enough with circuits, someones just going to need to send me the full blueprint, nixie tubes and all

[obsidian roost]

Unnamed Blueprint

[sand estuary]

thank you

[tacit cedar]

I want to try blood belts
I watched dosh's sushi base and thought "what if sushi blocks"

I made an "intersection" (cross-shaped TL 16:16 balancer)

but I'm not really sure how to control the ratios
I'm pretty sure it'll end up mixed enough that I can just sample any belt to get a good idea of what's inside, though an "organ"'s output will be more concentrated near the organ ofc
but even if I get the data, I wouldn't really know what to do with it tbh

I could also consider setting it up like a real body
for example, a central computer, food (ore) gets processed before entering the bloodstream, etc

[tacit cedar]

put down normal beaconed furnace stack
it appears that stack filter inserters only have one slot, so I can't do the "no steel/stone brick" thing dosh did

also I'm getting more than a belt of output due to prod I should probably fix that

lmao

[distant gorge]

My way of doing blood belts is to mimic the way actual blood does it and only output product if the local concentration of it is below a set value, or the input is above a set value

So if you’re doing it in block form, each block is measuring the output of it’s sides and using that to determine what to do

[tacit cedar]

yeah I was considering "sample a nearby belt, turn off the block's production if the concentration of its product is too high"

[distant gorge]

Yep, basically that

[tacit cedar]

I also heard something about “hormones”

I wonder if the new fluid system allows fluid sushi

i should go to bed before i have any more ideas

[slim wave]

This is interesting what would this mean

It makes me think of using an item like a pistol to stimulate bulk production of something that is soon to be in demand

For example if you suddenly use a lot of red circuits you’d want to overproduce plastic to prepare for high demand

[slim wave]

I’m gonna try and freeplay this blood belt thing

[slim wave]

The belts hold so much resources, you can really switch it manually in the early game

[tacit cedar]

yeah that’s probably it

[tacit cedar]

lmk if you want a funny balancer like those 16:16 crosses

[tacit cedar]

hm. I like having the 4 belts going one way and 4 belts going the other way
this balancer is TL though so doing only the u-turn is 50% bottlenecked
but if it's city blocks it shouldn't matter?

how tf do you grid align this

[tacit cedar]

alright i just needed 30x30

[clever shoal]

oh god

[tacit cedar]

actually this wont work because the fluid will get stuck in the pump and never leave huh

[clever shoal]

this is cursed

[tacit cedar]

also all the trains are stuck on the left side because they cant actually turn left lmao

i tried setting the station to limit 1 and removing signals but shockingly trains crashed

[slim wave]

Ill be doing 12 lanes but im trying to decide how to merge and split them