Yep, and when it all comes back around, b/c the priority recycle is consuming at the same rate it outputs, it stops taking new inputs altogether
#Sushi Mechanics?
queen cairn
summer zinc
yea the priority must go to the recycler is the key thing
queen cairn
It's literally 1=1 lol
summer zinc
input priority*
it's easier when im doing it for just one lane im not sure how you did it when its all such a mess
queen cairn
I just didn't
Lanes weren't important to this build
I built an 11:11:10:8:4:4:2, and reduced it
summer zinc
when i did my recycling loop i realised i needed to do a forced sideload because i was only using one lane
because if the right lane is allowed to go into the loop, then there will be space for the input priority splitter to fill with taking in via the main bus
queen cairn
That only exists b/c of this but yeah
That would be risky if sideloading didn't prefer the left lane
summer zinc
without the forced sideload both lanes will go into the loop and clog
queen cairn
Shouldn't but ok
summer zinc
rather it will clog that part of the picture
queen cairn
Tbh you don't even need to have separate splitters
It looks to me like you have multiple splitters for the recycles
B/c you have multiple loops?
End of a 2 item loop is completely sorted
summer zinc
imma be real i dont even know how to explain it correctly. i only observed something was happening, tried bunch of stuff and realised sideloading fixes it. sec the clip is long to show the problem when i remove the forced sideloading
the clog on the top half lane of blue belt for the 4 recycled arrays when i removed sideloading (25% plastic each belt gets recycled /condensed into one belt and split back into 4 separate 25% belts)
the clog cascades all the way back to the loop and so messes up the spacing completely
its jammed full of plastic now so its 50% copper cable 50% plastic because the recycler jammed
might take a while for the video to process
got to run but the way i see it, the left lane will always fill up the spacer-loop and give no chance for right lane to enter
and because the left lane is always filled by recycler loop with no gaps the main bus belt has no chance to fit more plastic into the giant loop as a whole
by using both lanes the spacing is different and now instead of a saturated left lane we have 50/50 left and right lane so thats where the main bus belt is able to squeeze more plastics and overload the system
think im talking a bunch of nonsense words but there doesnt seem to be proper terminology to convey it except with clips and pictures lol 😅
gleaming quarryBOT
queen cairn
ngl I haven't mastered a->b sushi yet
Under some conditions you can change the yellowbelt that's rate limiting to a red but idk what
If the whole thing comes to a halt b/c of machine shutdown it'll be resilient
Seems like you have very long recycle lines and the lane thing might be doing something, like you said
Oh my jesus your recycle lanes are long
Out and back are both inputs
summer zinc
yeah there's a reason why i didnt want to do recycling loops...lol
for this efficiency meme is basically untenable for giga sized factory
queen cairn
Eff makes it much more viable than normal
summer zinc
thats only half the red chips needed
queen cairn
Consider that you can feed like 2 speed beacon BC machines on a bluebelt of GC
summer zinc
the loops are so massive and the spaghetti so messy i can barely follow how to connect the red wire to red wire (so to speak)
queen cairn
So an out and back loop is out 1 machine, back 1 machine
I'm under. That's less than a redbelt. gucci
summer zinc
oh hmm
so the recycling loop also doubles as the transport for the production output?
queen cairn
Yeah, I could get 2x more machines per side, 8 total, still under a total redbelt
This is not resilient to dump
B/c it's compression feeding
Same output, but throttled
summer zinc
what is dump in this context
i can recognise the throttled setup but don't know what is compression feeding
queen cairn
When the belts are in compression, the designs you're using to mix
There's a teeeny bit of space
I must've got my math wrong let me reduce one more step so it's clearer
Bro how do I forget this every day
It's the same thing I forget lol
christ, okay, I remembered how to do a rate limiter lol
I always forget this chucklefuck and go "hey, why isn't this working?"
queen cairn
Slightly fancier version, and why rate limiting can be sensible sometimes. I'm not going to 1:4 multiple times and loop, in practice. I'm going to do a 1:4 for each material once.
Rate is limited? I have the space to reintroduce
subtle meadowBOT
fair gull
The sushi maker stops to protect itself from running dry, causing only GC to be on the belt which fills the ASMs with GC only
which is why the second lane should be considered part of the sushi ratio, not just some supplemental thing
aka we need to force it into lockstep with the mixed lane:
gleaming quarryBOT
fair gull
I also found out something truly cursed:
the sushi maker operates differently with different belt types:
left is all yellow, right is all blue
currently verifying a blue build with yellow and red output speed to see whether it's the output speed or the splitter speed that makes the difference. takes forever though because the loops move slow af
summer zinc
is the ratio fucked when you say operates differently?
summer zinc
The sushi maker stops to protect itself from running dry, causing only GC to be ...
i'll leave the hard work of how to force lockstep to a smart guy like you 😉 but for now i got a pseudo solution that seems to work long enough without really being irrecoverable after removing inputs/blocking output belt so i'm gonna use it for the first practical use 😄
im just working on how to get a working build for a 100 machines setup, getting impatient to finally reach the coveted 1k SPM
feels like it's been weeks since i did anything in the "real world" save
fair gull
seems to work is not good enough
but you can block the second lane like in the bp I provided
summer zinc
ive been logging into the lab save for hours to figure out stuff
fair gull
perfectly normal behaviour
summer zinc
hang on maybe i can give a quick screenshot and you can give me suggestions on where / what i can block to see if the system recovers
or not
fair gull
is the ratio fucked when you say operates differently?
no, the ratio is fine*
but the bad news is that the "mode of operation"/item type on the loop depend on the output speed rather the the splitter:
queen cairn
ohhh fucking RIP
fair gull
left to right:
yellow build with yellow output, ratio repeats every 11 items
blue build with yellow output, ratio repeats every 22 items
blue build with blue output, ratio repeats every 11 items
blue build with red output, ratio repeats every 33 items
subtle meadowBOT
queen cairn
I mean, as long as it stays in ratio, but that's still WILD
fair gull
yeah, this is not good, because the ratio changes when for a while until the loops have changed
even stopping the red speed one to extend the output belt (wasn't long enough to verify the ratio) causes hiccups
so yeah, RIP
queen cairn
Oh yeah, did you peep the RC build I just cooked up?
summer zinc
i only tested 3 scenarios here, remove the 3 red circles separately and they all get back to the expected output after restoring any of the conditions back to normal, even after removing it
what oter criteria is there to make suretest it doesnt jam?
summer zinc
output block is just removing that belt in the right side red circle right?
queen cairn
Oh I missed that one, yep
fair gull
partial slowdown for both output and input, as well as slowly changing input/putput speeds, shock resistance, ...
I wrote some things down not too long ago
fair gull
Power failure
also brownouts
summer zinc
what is "shock resistance"
fair gull
you asked me that after writing it the first time
summer zinc
🤔 ok maybe i was on the way to bed and it didnt stick have to go back up
ah yea sorry it was 1am had a long saturday yesterday probaly want a good idea to try to cram more knowledge in at that time
fair gull
somebody else using your account?
summer zinc
no but i was pretty exhausted then lol seems almost like drunk typing
fair gull
I also found out something truly cursed:
the sushi maker operates differently wi...
implementing this change would be a good start
summer zinc
but not drunk just tired
summer zinc
i basically crashed on bed and made the bad decision to try to open discord and cram factorio to make up for lost saturday 🤣 i dont remember typing any of that
fair gull
unsolicited advise: bed should be used exclusively for sleep
summer zinc
normally i agree but factorio and this particular problem is too interesting and addictive to ignore for the entire day
took basically my whole self control not to open discord the entire saturday
does this count as partial slowdown
or what do we have in mind as "partial slowdown"
queen cairn
oh
That's a fail state
That's why thyme was saying it had to be in lockstep
You have altered your ratio
queen cairn
Yeah
So what's happened is, you're feeding GC on the bottom belt, until you run out of RC
Instead of 2 RC:20GC
Since you're using stackers, this is a very long time, many cycles
So you introduce a TON of GC that is out of ratio
summer zinc
this is eventually what happens but im not sure why this causes an irrecoverable failure
because im using RC at the toplane , once GC supply goes back to normal it will just go right back to where it left off?
i need to test RC though
i think if RC is slowed down it will be issue
queen cairn
Prolly not. Your machines are out of ratio
What I expect to happen is the first handful of RC get picked up by the first machine, then you get GC spam
And the end ~6 machines never light
fair gull
should not have written "partial slowdown" lol
slowdown would have sufficed
summer zinc
ok damn this is a new problem i did not foresee
of course loops doesnt have this problem lol
queen cairn
:D
summer zinc
well i guess this is why im stuck in labworld
queen cairn
lmfao
Hey I spent most of today in the lab too
You're making good progress on one of the toughest things in factorio
summer zinc
queen cairn
Oh that's funny
It's going to crash the entire line b/c you're still at partial feed
summer zinc
yea the longer at partial feed the line will eventually clog
and i havent even tested on RC
im 90% sure RC partial slowdown is worse
fair gull
why not implement the solution I shared?
queen cairn
was it a 1:10 dual lane?
fair gull
nah, the lockstep thing
summer zinc
actually i am not sure which is the lockstep one
i saw you shared two blueprints with the two belt rates but was that lockstep?
is this the lockstep mssage?
summer zinc
oh. sorry it wasn't clear to me
that colon was easily missed in the message i thought you were just stating an observation
hang on i need to copy and run it to understand your solution
fair gull
If things aren't clear, you ask for clarification.
I'm going to assume you understand what's going on in the conversation otherwise.
summer zinc
ah yea i thought you were just sharing insight/observation about the lockstep problem, not offering the solution in the immediately following blueprint. my apologies 😭
queen cairn
Did we have the same solution?
idk but either lane is 2:20 so it doesn't matter which gets consumed, I think
summer zinc
am i missing something here with the setup? do i need to run it longer? it doesnt appear to be running in lockstep
lockstep = if one lane moves 1 item, other lane must move 1 item as well, if it doesnt move 1 item, the other lane also cannot be allowed to move, right?
gleaming quarryBOT
queen cairn
I don't understand what's going on during that thing's priming process but it black magic sorts green circuits onto the looper and once it does, it's ratio
So maybe introduce GC first
summer zinc
queen cairn
Oh, I mean, that's making sure it's got both sides before it goes at all
fair gull
it's almost if somebody has documented priming stuff earlier
works nicely for yellow belts, blue belts need to be primed RC first though (and then still have a significant warm-up phase)
queen cairn
summer zinc
why is the behaviour so vastly different from what i expect 😵💫
fair gull
what I did isn't actually forcing lockstep though
the lanes are still allowed to move independently. Both lanes are stopped when there's input starvation. But since GC on the mixed lane isn't enough, and the mixed lane is the near lane, the second lane is kind of consumption limited
summer zinc
whenever we say input starvation does this mean removing any one of the inputs entirely, or does slowdown of inputs also fall under the subset of "input starvation"?
fair gull
why is the behaviour so vastly different from what i expect 😵💫
what is the behaviour you're seeing and what is the behaviour you're expecting?
summer zinc
i never really thought about slowdown of inputs because the recycler loop in previous sushi build was like a bandaid fix
fair gull
input starvation is when there's less input than required
as long as you're supplying more than the ASMs need, it's fine. doesn't have to be 45 i/s
queen cairn
i never really thought about slowdown of inputs because the recycler loop in pre...
recycler is life
But also I think I like this solution? I still feel super iffy about what I did with the loop
It's just a 1:10 ratio balancer doing both lanes.
fair gull
what's going with the circuit belt in the loop?
queen cairn
That's the primer, I think it was better than not having it
queen cairn
It makes the loop prime with GC first
fair gull
so it doesn't read?
queen cairn
loop read, RC stops until GC 8
fair gull
bc as I have shown, the item type in the loops depend on the output speed. could be RC on that tile
queen cairn
I don't know what the control thing is doing either
sec, vid inc
It black magic sorts for some reason
It's me
I'm the dumb. That's a red loader
Anyhow I thought a GC dump of that magnitude was a lot better than an RC dump of this magnitude
fair gull
It black magic sorts for some reason
why is the output tile set to 9?
does the constant combinator add 1 for both GC and RC?
fair gull
I'm the dumb. That's a red loader
I'm using Express Loaders exclusively to prevent exactly that issue 
queen cairn
why is the output tile set to 9?
does the constant combinator add 1 for both GC ...
Yeah, it's an on/off switch as well, which is nice, but also it prevents having to pay attention during load
Here's the side by side so you can see what I mean by "magnitude" lol
summer zinc
#1277595827547799552 message
so this is the blueprint file and message which i had been happily using the 2:9 sushi ratio splitter (with red chips strictly being on one side) until i finally understood what testing for input starvation means
the top half picture is the latest one with lockstep that thyme sent today, and bottom half is the original blueprint that i have been using for testing. im trying to "spot" the difference but really struggling a bit - is the ONLY difference that the circuit cable is connected right at the very end of output belt? what else am i failing to spot the difference?
added circle for emphasis - blue circled was where the old circuit cable was connected to, red circled is where new lockstep implementation exists
fair gull
Here's the side by side so you can see what I mean by "magnitude" lol
I see, the 4 circuited belts aren't on the same network. That works then. Also your splitter arrangement looks better than mine (at least in that case).
queen cairn
Once I got the reduction method from your explanations, something clicked
It's like 2 steps from here lol
fair gull
https://discord.com/channels/139677590393716737/1277595827547799552/127910376542...
the sushi maker itself is unchanged
I made two changes around it:
- there's now a single GC input
this allows the circuit wires to protect the entire input - the output now also encompasses the second lane
this allows the circuit wires to block both lanes
summer zinc
ok got it
looking at ichaley method and going back to why we are only having RCs on one side. if we are doing lockstep / block both lanes when input starved. do we care that RC is now on both sides?
fair gull
added circle for emphasis - blue circled was where the old circuit cable was con...
the splitter between the blue and red circles was placed in case you wanted to use the other lane of the sushi maker for a 2nd stack of BC ASMs
summer zinc
the splitter between the blue and red circles was placed in case you wanted to u...
😮
i did not make that connection
fair gull
ichleynbin is doing 1:10 on both lanes - so 2 mixed lanes
you've been focusing on a single mixed lane and a pure lane, which requires 2:9 on the mixed lane
summer zinc
yeah im trying to think of the pros/cons of both approaches. i vaguely remember the reasoning of why i did just red on one lane but it feels like that reason is no longer needed if both lanes are moving together
fair gull
I'm not sure which one's better, it depends on how inserters pick up stuff and the resulting clotting, but I don't want to put any thought into it
queen cairn
I don't see offhand why 1:10 wouldn't work but I haven't tested lol
fair gull
having 2 mixed lanes is an extra variable
a single mixed lane with a supplementary pure lane feels more safe
summer zinc
i dont remember why it didnt work when i initially tested it, but i could have set the whole test up wrong because that was very very early in my understanding
like thyme said it depends on how inserter picks up stuff, but back then i didnt have a dedicated filter inserter grabbing only reds and I think that was the reason red slipped through and clogged up at the end of the belt (last machine, tip of the belt)
it tended to favour the inner lane nearer to inserter, the red chips on the outer lane seemed to slip through very frequently
queen cairn
I think GC dump at the start is actually desirable, given how many GC's the machines will take
summer zinc
but this was when i used only 1 stack inserter per machine and non-specialised (grabs both RC and GC) for all 12 machines
gleaming quarryBOT
summer zinc
also, does a constant combinator require power?
queen cairn
No
summer zinc
it doesnt look like it does in your pics
thats an interesting exception, i thought all of them required power
queen cairn
Nope, wire and constants don't, for whatever reason lol
summer zinc
thanks i will copy your blueprint and try to break it on my own and see what happens
good to see the different approaches anyway
queen cairn
You might get RC load at the end tile, but the chest defense should cover that
fair gull
You could simulate a constant combinator with a wired chest. A chest doesn't need power.
That a chest cannot do negative values is not important
summer zinc
hwat
queen cairn
You know what, my brain still thinks in loops
summer zinc
lmao when you put it like that
that is truly cursed. using a chest as a CC
never thought of it that way
queen cairn
My solution would for suresies work in a loop. I do not know how belt end works
If you +1 machine it should also work? You may need filter inserters.
summer zinc
here's your version
doesnt seem to have any problems on normal operations
gonna try the other tests
queen cairn
Yeah I think filter and chest should be sufficient
The other situations should be defended properly
queen cairn
I keep thinking about this image lol. Two item sushi is nearly trivial to do this way and I think this image is something of a proof of that. It's 3 items, you can arrange things with no crossings, ergo you can split any feeds. I feel like this reduction is already solved... but probably you don't feel that way lol. The group of 8 simplifies 3 times, to four, then 2, then 1. The group of four dead, simplifies twice. Both of those, have an input splitter, because they have an additional group. The group of 2 GC's simplifies once, the RC and the dead belt are paired still. So; Refeed + RC. That belt, + the once simplified GC. Combine with twice simplified dead belt. Combine with thrice simplified GC. On that splitter, take refeed and output.
These are probably the explanatory steps
summer zinc
These are probably the explanatory steps
ok this is way clearer to me. thanks a lot 
queen cairn
lmao, that sounds sarcastic, but you know, I'm def struggling to explain this in words that make sense XD
I know what I'm saying's not been getting through lol
summer zinc
no not really this one really makes it easy to understand. not much else to say except good job and many thanks i need to apply this so can't say too much
fiery portal
You're making good progress on one of the toughest things in factorio
laughs in circuits
queen cairn
*laughs in circuits*
Laughs back in ratio
fair gull
that looks like the output of a scrap recycler
fiery portal
Factorio, where L + ratio actually has meaning
queen cairn
So it's not there yet, but I am making progress on purple SushDI
summer zinc
These are probably the explanatory steps
Btw i feel like this deserves to be shared on Reddit with credit to you, thyme and maxreader of course. Irl finally caught up with me for a bit but i intend to do working example of using this pictorial example to make working splitter sushis for a couple easy to understand examples. 1:10 
, 1:2:3:1 flying robot frame, 1:2:3 yellow science, 1:1:2 engine unit (arguably also green chip but not everyone might be ok to belt copper cable), to name a few.
Think newbies in general when they see it in action and how to do it from scratch will be very happy to have this tool in their arsenal
The only question remains for me is deciding whether using one lane for 2 item sushi like rc on one lane for bc is a safer approach then ill have to do the forced sideload that thyme has
Tbh if you do those and put on Reddit yourself that would probably be faster and more accurate too 🤣
queen cairn
heh, well, I'll give the others a little bit longer to see if they can come up with a magical algorithmic breakthrough, and if not, I'll post it pretty soon
fair gull
I don't think sushi, especially the non-return kind, is suitable for newbies.
Sushi is a great tool for a few applications, but not something that everybody needs to have in their toolkit.
summer zinc
That's fair. Maybe I should have said intermediate players who have launched multiple rockets and are looking to do harder challenges like scaling up, or compact builds. I guess I still haven't disassociated myself as a "newbie" player.
summer zinc
why does the bottom half of original design managed to autosort greenchips in the loop but the top half with a minor modification always have red chips that mix in?
Let me take a clip with a clearer side-by-side comparison
queen cairn
lmao, timing issues. It has to do with turns, item spacings, when things hit the splitter and what order they get taken off. I wish I could explain more than that
Oh yeah max and I were talking about black magic 2.0
We def noticed the other day lol
queen cairn
#1277595827547799552 message this was when max and I noticed it was not-quite black magic, and yeah no worries on that one
summer zinc
My brain has a tendency to prune a lot of info that I cannot understand or use in the "right-now" to prevent info overload
This is truly bizarre
fiery portal
https://discord.com/channels/139677590393716737/1277595827547799552/127925897206...
Well, I knew it wasn’t from the start 
Invariants
summer zinc
I wonder if this is what thyme talked about too
He mentioned something about belt type having an effect on the pattern/rate or something
queen cairn
lmfao I mean, I know black magic has been removed too, and that this wasn't exactly it even if it hadn't been, but it looks like it
summer zinc
I didn't really catch or understood it but I remember him mentioning it
The two setups are supposedly identical right? Other than "timing" or turn or gaps
Im hoping I copied/made alteration correctly in terms of splitter junctures
If i altered it wrongly then yeah of course the rate will be wrong but lets hope I'm not that dumb
fiery portal
My brain has a tendency to prune a lot of info that I cannot understand or use i...
Always interesting to hear other mental processes 🤔
summer zinc
It's hard because I can hardly understand the terms yall use sometimes because I barely have the vocabulary to describe wtf im seeing or trying to do
summer zinc
Add in the fact I'm working at a glacial rate and have very little experience with these types of problems...yeah I'm a bit lagging behind in these discussions 
queen cairn
lol, I'm still doing some catchup work on theory developments I missed in the past year or two. I can at least explain why it doesn't matter
summer zinc
I made the alterations because I need to compact this loop to fit into the tiling of a bigger factory, alternatively I could rotate it to "fit"
i got about this far trying to replicate the several loops and trying to make a practical build out of all these days' worth of discussions
fiery portal
My big takeaways from this have actually been that learning physics means understanding tons of other things much better
summer zinc
hmm this is also pretty low resolution looking it looks like
ive no idea how to take a better looking screenshot. i typed /screenshot and the second pic was the output
fiery portal
Measure theory and symmetry/invariants/conserved quantities especially
summer zinc
i feel like i need to take these giga clear screenshots so i can zoom in to see the details, i still dont have the skill or experience yall have to guess or intuite what the blurry icons are
queen cairn
I'm trying to figure out how to apply Little's law to this purple sushi I'm doing. I feel like it's gotta be there
Like, if I'm consuming on the middle of the belts 10x, it's the rate between the two, not the overall belt
fiery portal
It is if you want to find total items on the belt
queen cairn
I should be so far under per
Yeah that's what I need rn tbh
I'm switching the iron to count sushi. that's the easiest way for me to spread it out
So I do need an item count
fiery portal
Count on belt = consumption rate * average travel time
summer zinc
Well at this this thread hasn't been a one-sided taking of knowledge by me at least im doing my part even though i haven't exactly contributed much except ask a ton of simple questions
Surprised yall got any takeaways from this
fiery portal
Count on belt = consumption rate * average travel time
If your consumers are spread out, you can average their location and use the travel time to that midpoint
queen cairn
I think I should do one belt length then?
fiery portal
No?
Not unless half of your consumers are done in one belt length
Define belt length
queen cairn
So I've gone up and down 20x
I guess it's at least two lengths, maybe four?
Average at half then so probably 2
fiery portal
Whatever length you choose, then length in tiles * 4 items per tile * consumption rate / (15 i/s) = count on belt
queen cairn
That's for lane I suppose, 4?
fiery portal
Well, if you’re using both lanes, stuff cancels out
queen cairn
8/30, yea
Hah speaking of cancelling, it doesn't matter what I choose
Four is the correct subset, but also, it's the same as whole belt.
fair gull
Measure theory and symmetry/invariants/conserved quantities especially
It's not just that. We've been told in the early semesters that the structured way of thinking we learn is what makes us valuable for all kinds of employers.
concepts such as symmetry/invariants/conserved quantities show up in lots of places though
also, being rigorous
summer zinc
not really related to sushi, but if i want to connect these individual chests to one speaker to blare an alarm /notification when item count = 9600 is there a way to have all the chests be wired up to one speaker, or the only option i have is to wire the chests individually to individual speakers to check for itemcount = 9600?
identified a new fail state where last belt tile is equal to 8 red chips so i decided to allow those into the chest at the end as well but couldn't replicate 8 red chip on a belt tile after priming the belts
fair gull
if the distance is too long, you can connect circuit wires via power poles
summer zinc
👍 guess if i have 100 chests i need 100 speakers
fair gull
You take a circuit wire of your choice, start at a chest, connect it to a power pole in reach (substation or whatever), connect it to the next power pole, and connect your second chest
rinse and repeat
summer zinc
Hmm but won't the circuit cables sum the contents of all chests connected along the power poles, I guess I didn't phrase the question properly. How do I de-entangle the contents of each individual chest all connected along one circuit network?
I want to be informed of when any single chest is completely full (think you guys call it a failure state) so I can go back and re-analyse the problem
fair gull
Yes. If you want to blare the alarm if any chest reaches GC=9600, you can use a decider combinator between the chest and the shared wire.
summer zinc
I'm trying to set up some extra test conditions like slowdown and simulate the train loading/unloading of 7 belts of green circuits to see what is happening
fair gull
Condition GC=9600
output GC 1
then connect all outputs and set off the speaker when GC > 0
summer zinc
Hmm. What if I need it to be Sum (GC,RC) =9600 instead?
Hang on I will play with the setups
fair gull
will you even reach that? requires all slots to be full?
summer zinc
Err that's the thing, I don't know
fair gull
How do you intend get a full RC stack when there's 9400 GC in the chest?
summer zinc
That's partially why I want this notification setting, there must be lots of scenarios I might not have thought to test although I might be under the belief I have tested it enough
fair gull
Just use Any = 4800 instead, should be good enough.
summer zinc
How do you intend get a full RC stack when there's 9400 GC in the chest?
🤔 thats a good question
So I need to check for slots available instead?
Run some sort of calculation for slots remaining perhaps?
fair gull
the answer is you don't, because the machine has likely stopped, causing the belt to stop
You could sum them with an arithmetic and then feed the sum into the decider, checking for >9400
summer zinc
Is there a way to check if a machine has stopped for a configurable X period of time?
fair gull
not in 1.1
summer zinc
I guess that's the actual thing I care about
fair gull
well, not directly
summer zinc
If my inputs have run dry for too long it is something I need to be aware of
fair gull
can always monitor the outserter
summer zinc
Especially as the scale of the base runs bigger and bigger and I have less manual oversight of every little subfactory all over the map
Right
I also need some sort of latch mechanism? To reset the timer/counter when outserter is reactivated implying that the machine is back to running
fair gull
If my inputs have run dry for too long it is something I need to be aware of
I do that by looking at the train stations. I'm using dynamic train limits, so if any station looks like this
"Copper Plate Requester 4/16" I know there's a Copper Plate shortage
summer zinc
Oh that is also a fair point, an empty station would also imply problems
OK thank you I think I can probably just monitor my logistic network at the unloading station for GCs
Since I need to be unloading 7 belts worth per minute anyway and I only know how to do that fast enough with bots
summer zinc
2 🤣 im doomed to fail
But, I am willing to plug some more time to overcome the challenges until enough is enough
For now, even the loopless factory seemed to be working, I need to run couple more lab tests then move on to the other stuff
I don't think I'm even 25% of the way to making bluprints for my 1k SPM but one step at a time i guess 😭 must be 2 or 3 weeks now
This is what I will probably import into the real save
fair gull
2 🤣 im doomed to fail
Not sure that's even possible without bots. can you use 4 wagons instead?
summer zinc
replace the infinity chests with belt lines to the unload station
fair gull
How fast do you think your next train enters the station?
summer zinc
Errr I'm not even sure how hard it will be to move everything to 4 wagons
my entire rail network has been signalled for 2 wagons
biggest problem might be intersections, i dont know. i havent even truly dabbled with rail optimisations
fair gull
how many SPM are you going for?
summer zinc
took all my time just to play with unloading mechanics (inserters have different unload speeds depending on belt angle, belt to chest/chest-to-chest, etc.)
1000 only
i think after that i might run it for say 10hours to prove that i can indeed sustain it and start over with better principles
this save about 180 hours deep now its my first ever foray into post-rocket
fair gull
You should either use multiple train stations unloading GC, or use longer trains. 7 belts from 2 wagons is a bit excessive.
summer zinc
the pre 100 hours or so (about 10 separate rocket-launch-and-quit saves) have just been focused on doing the time-based achievements like TINS GOTLAP etc
i might test 7 belts from 2 wagons first, and if it fails i'll add another. im not too sure pros and cons of either because i intend to add like 4-5 stacker spots behind the station in series
so it might be issue to load 7 belts into 2 wagons that is bottleneck instead of unload. or producing 7 belts. or something else
fair gull
1k spm shouldn't be an issue UPS wise, but if you use bots whenever you run into througput issues, you might not be able to sustain 60 UPS in the end
summer zinc
its a bit hard for me to know how everything runs but i'll find out when i find out, i guess. i have been running everything in more manageable, discrete problems for now (design 8 belt worth of GC production factory and load into train, produce 2 belt worth of RC into train, produce 8 belt worth of iron plate, etc.)
fair gull
You'd also have to separate the logistic network for each (un)loading station from other networks.
summer zinc
weirdly bots seems pretty efficient when i did the testing of 10 belts loading into 2 wagons. only needed 50
yeah all my logistics networks are fully isolated
i learnt that lesson very early
i thought i could automate building by having a massive bot network but all it ended up was bots trying to cross a lake and neve rmaking it, and turning back and getting stuck on that loop. very annoying
now i just set up networks where i need them. have train that supplies the construction materials
just two spidertrons to lay some rail foundations to places of building
fair gull
its a bit hard for me to know how everything runs but i'll find out when i find ...
splitting problems into subproblems is a good method to go about it
summer zinc
It's just a way for me not to give up and quit entirely 🤣 i know i am prone to morale breaking easily so i learnt to manage it in my youth
Certainly bit off more than I can chew many times and just end up never seeing a project through
It's also why I might seem to ignore yall messages at times, I just need to ignore and learn to go back to your advice when the time arises but not try to take on extraneous info that doesn't help to solve a particular problem at hand. Cause it just makes me deviate off track
I dont really have any of engineering or compsci training, not even the formal maths stuff so it's kinda just...empirical observation and problem solving from there? I dont know what to call my 'approach'
fair gull
Train stations is the very first thing I designed when making my City Brick base. Everything else is rule be their specs.
summer zinc
If a problem seems indecipherable after a few attempts I guess I come here for help
Like when I tried to unload 4 infinity chests with 4 stack inserters and wasn't getting a saturated blue belt 🤣 apparently i needed to alter hand size. talk about niche info
I could tell the problem was when the inserters swung back to chest to grab items which causes gaps in the belt but i didnt know how to solve it
fair gull
You need altered hand size when directly loading the belt, but sideloading (or using a splitter) is fine with full hand size.
summer zinc
seems to be ok, removed the loading stations infinity chest to simulate GC production halting, and it still recovers to my calculator output levels. think i'll export this to main save and go to perhaps purple science first, then yellow science and try to apply some of the stuff learnt here
weirdly there can be some gaps seen on certain belts but im not sure why it happens
fair gull
2 stack inserters can only fill a blue belt if it's free flowing
if it occasionally stops, they need to sync up again (they will do so reliably) but it's going to cause a few gaps
summer zinc
Sadly something caused my green circuit production facility to break but i have no idea what
might take a while before i figure it out. its good to go back into the "real world" once in a while to test for actual "real-world" situations when my understanding is iffy
im not even sure if the broken state is recoverable, or just temporary because i switched it off with a power switch until GCs are needed again at the loading station
queen cairn
Yeh there's a reason I normally avoid powered sushi
summer zinc
im not even sure if it was plastics that overloaded, or GCs. by the looks of it, it was GC overloading the belt
although i couldnt make sense of it my hunch still lies in the way i configured the inserters to throttle/stop loading belt with GCs when it detects a certain state on belt (2plastics, hold reading)
inserters must have kept putting GCs on (how? power went out, it should be off) despite no plastics on the belt to alternate/sushi
this was not supposed to happen
my extremely primitive way to limit the sushi. also the very very first attempt that was at the start of this thread
might have to go back to the test world to see what happens
seriously have no clue how it happens
summer zinc
wat is dump GC and GC dump state
gleaming quarryBOT
summer zinc
im hopping off so i think i'll leave this here for now
queen cairn
lol I'm trying to figure out what you did still
summer zinc
me too i kinda forgot cause it feels ages ago
queen cairn
Oooo inserter sushi
summer zinc
yeah i made the rokkie error of not importing GCs
queen cairn
Yeah that's a tough combo. Okay so you did end up with too many GC on the belt, a dump of GC?
summer zinc
too many GC on the belt yes. i think. but it looks like also too much plastic
hard to tell
crap the autosave got overwritten so theproblem state is gone
queen cairn
lmfao rip
summer zinc
i manually fixed it by removing GC by hand
queen cairn
You still know it exists though
summer zinc
unable to retrieve it now
yes
i didnt fix it thats for sure
problem should crop up again if/when i resume production and stop production somewhere
queen cairn
At least you know where to start digging, too
summer zinc
right now production starts when there is less than 1 train load worth of RCs at the loading station
production stops when there is >=7 train load worth
using some SR latch to remember the state when it first triggers the stop production signal
no idea if that is the thing causing it though
i tried removing power source, blocking the red belt in test world, couldnt really get the GC dump state
queen cairn
It's possible that's the thing that causes it to fail
It only has to be off by a couple, I think it has to miss only 1 insertion
summer zinc
how does something "miss" an insertion
queen cairn
You're regulating the insertions
summer zinc
im not even sure reading hold on a belt would lead to an error
still not entirely clear on the mechanics of a hold read
queen cairn
4 items per belt lane on a full belt, anything ON that belt is read
Every tick it broadcasts the contents of that belt
summer zinc
yea i got that part, but what if the belt is constantly moving? in a looped sushi the stuff is ideally constantly getting rotated, thats the idea
is supposed to be 2 plastic 2 GC on any individual belt at all times
allow GC insertion only if GC <2 (1, or 0). i guess if GC=1 then this causes overload
since hand size is 2, or 3. depending on current belt situation
i use a CC to set hand size depending on the amount of plastics on belt
plastics=2 on a belt => maximise production state, load as many GC as possible, increase hand size. plastics <2, input starvation state, reduce hand size to 1, or 2, depending on detected plastic (CC= +1plastic, control signal = plastic. belt will read 0,1,2 plastic +1CC to alter hand size)
queen cairn
So I actually just developed a technique
"Delay Decay" is what I've been calling it, b/c lol. It's close to a linear moving average
I'm measuring two sections of belt, one on the loop, and the other is the feed line, both pulse read. Every time I see an item, add 1000 to the delay
For faster items, have a larger negative clock, I think iron was like -130 or something, 8 ticks between items
It comes very close to properly loading a belt without overloading it
The advantage over EMA is that it's actually fewer combos, and also it preemptively smooths. If there's a chunk of items, wait a long time, but during normal operations, it's spaced out very evenly.
summer zinc
So i need combinators for it to work?
This gc problem gonna keep me up at night ..smh
queen cairn
I mean, this is one solution lol
summer zinc
I need to maintain 25% loaded belt for 4 parallel belt, but they all merge into one full belt only to resplit again into 4 25% belts after the loop
queen cairn
You're inserting right onto the same belt
summer zinc
How do i implement this
queen cairn
You could do a thyme trick and just route the plastic in on another belt, then splitter mix it
Delay decay is kinda nice, because it automatically recycles without having to filter off. If there's an item on the belt, add the delay. If you put an item on the belt, add the delay
summer zinc
Im not sure where the feed line begins because its actually coming not from an infinity chest or an external full belt from train
But it is getting fed literally by 2 machines on each parallel belt
queen cairn
Oh also, that RC build is "out only." If you "out and back" you can deal with looped belts so much easier
summer zinc
I need a way to detect when there is a certain amount of GCs on an entire length of belt and halt gc outserters when there is that many
queen cairn
That's the easier way to say it. You can put machines on the return belt.
summer zinc
Reading just the preceding belt is insufficient, i believe
queen cairn
Delay Decay measures only one belt, pulse
It's only that section that matters for it
summer zinc
I dont even know "how many" is the limit for an entire length of gcs
queen cairn
idk you get a yellow lane
900-450-450
Oh sorry I'm doing redbelt math still
My brain is stuck in redbelts right now, 50% higher than that
You could use those little factoids but probably that's more mess than it's worth. It's exactly a yellow lane and a half you want.
summer zinc
Honestly i think I'll be clearer what i need to change when i can actually detect the problen state, lol
queen cairn
It's not even that bad tbh, you don't have to do the undergrounds even probably
also I guess it'd have to be this, but w/e
As long as the left side also drains, right side will have 1.5 yellow lanes
I would honestly suggest going over the whole thing and seeing how you would do it now, maybe even a fresh start with your new tools?
summer zinc
not sure if i followed this step correct but lost now on how to reduce copper plates to just one input lane
0.7 copper, 0.2 steel, 0.1 plastic for 1 belt
like...not sure if this is the correct interpretation
just realised i didnt have to set up autodelete function of infinity chest to get a rough ballpark of ratios, left vs right chest
feels like somewhere deep in the lizard brain there is a recognition of pattern although i feel like i still got no clue wtf im doing
wonder if this is the fewest number of splitters i need for this setup, also if i can make the whole sushi maker more compact
summer zinc
just realised that chest defending against overload only works if there is undersupply of all materials
eg if i supply 0.1 blue belt worth of steel and 0.2 blue belt worth of plastic per minute, there will be 0.0297 BB and 0.0243 BB worth of steel/plastic respectively making its way to the end of the belt never to be consumed per minute
wonder why i never saw this problem crop up in the GC:RC sushi, or was it because the machines was perfectly consuming all inputs on the sushi belt
fair gull
3 ingredients is also way more complex to handle safely than 2
summer zinc
whats the ratio of this one i cannot puzzle it out
how does plastic and steel not build up over thyme (haha)?
fair gull
like...not sure if this is the correct interpretation
LDS ingredients are 20 Copper, 5 Plastic, 2 Steel
that rounds up to 32 unputs in the 2^n compressor
fair gull
whats the ratio of this one i cannot puzzle it out
5:2 Plastic:Steel, Copper is free flowing on the second lane
summer zinc
wont you be undersupplying copper if its at most copper occupying just one lane (50%) of a belt?
fair gull
Yes, copper is what limits the length of the assembling stack
summer zinc
i thought you need to top up copper into the 5:2 plastic:steel lane as well to make it equal to the 20;5:2 ratio
i see yea. i was wondering how im gonna fit 20 ratio of copper to 5:2 plastic/steel
ok i guess that partially explains why the assembling stack is so short then
whts the belt condition at the end of the line?
fair gull
it is so short because I was trying to make it short
that is from a RibbonMaze save where I can built at most 1 chunk wide
summer zinc
LDS ingredients are 20 Copper, 5 Plastic, 2 Steel
that rounds up to 32 unputs in...
yea i thought i could do some...rounding. but ended up being questionable. guess i have to match the ratio exactly, or undersupply all 3 inputs simultaneously or risk overload
fair gull
If you want to do non-return sushi, there cannot be any rounding. You have to supply the exact ratio. Even then there's more limiting things.
fair gull
whts the belt condition at the end of the line?
belt just reads.
inserter is enabled on "everything = 4" and has a permanent blacklist on copper
summer zinc
if calculator depends demands 0.0703 steel + 0.1757 plastic + 0.7028 copper usage on a blue belt, i was wondering if i could get away with something bad like 7:17:70 but its just a terrible terrible number
fair gull
uh, the ratio doesn't change just because you supply more materials
summer zinc
oh yea i think what i meant to say is 1blue belt worth of this ratio can feed 23.XYZ machines, but if i only put 23 machines in the assembling stack, the .XYZ will bite me in the ass and chest is not enough to defend because the excess 0.XYZ inputs will build up per minute
im thinking if i should scale down the amount (but still fill blue belt? how?) to feed exactly 23 machines but i think im just better off putting 24 machines in a stack and the 24th machine idling for 50-70% of the time
fair gull
All the excess does is create some backpressure. If your build cannot handle backpressure, it cannot handle outputs piling up.
Ok, that might not transport what I meant to say.
If your build cannot handle backpressure on the input belt, adding another ASM won't help you, because backpressure on the output belt propagates to the input belt.
I'll take a nap, I've been up since 0400.
summer zinc
ty appreciate the guidance im gonna work on the 20;5:2 sushi now
very cool, almost perfect 20:5:2 ratio but i have to wonder how stable this will be
will try to make it compact before i send it
some of the ugliest contraptions i have ever made
compressed and organised further
interestingly this interaction iteration the output belt loopback stays copper always
compressed even more
a shame about the rounding, only needed 23.188 machines so basically the 24th machine is only running 18% of the time and sleeping on the job the other 82% of the time
queen cairn
I'm not sure if you figured out the generic "decrease input belts to 1" but in general, grouping inputs at the very start is the way to go about it. Any time you can get a big group, that's better, so if you have a 10, try to make sure it goes into a block of 8, but the other 2 should be next to that block
summer zinc
i was wondering if the 20:5:2 ratio could be roughly approximated down to 7:2:1. Turns out, nope. Overload issues
well, i think if i loop it back to the start will be fine
queen cairn
Yeah, you can do that ratio on 32 and loop back
summer zinc
i just did a 20;5:2 ratio anyway and it looks more or less equal
7:2:1 on the left, proper 20:5:2 on the right
same amount of splitters weirdly
queen cairn
20:5:2 is weird. There isn't a trivial way to not underground
fair gull
well, i think if i loop it back to the start will be fine
need to unsort it or the error will accumulate
summer zinc
oh ya
loop the sushi belt back to the start, and filter splitter back to its constituent inputs to merge back into the main bus
queen cairn
I guess instead of undergrounding, you could add some extra splitters
summer zinc
meh. i kinda gave up on looping
queen cairn
lol l00ps are the easiest imo
summer zinc
my builds are too gdam massive to loop and idk what ichaley means by adding the output item back onto the loop either
fair gull
loop the sushi belt back to the start, and filter splitter back to its constitue...
don't need my fancy contraptions if you're ok with looping
summer zinc
i think they still needed to build a sushi from scratch though 🤔
it was way easier to build an accurate sushi this time its kinda crazy
queen cairn
my builds are too gdam massive to loop and idk what ichaley means by adding the ...
You can put machines on the right belt as well as the left.
It's a sushibelt. It's not a normal belt where you feed in one direction.
summer zinc
but where do beacons come in for this
summer zinc
if i have a 24 assembly stack it means i split it artificially half and half? 12 per side?
fair gull
You can put machines on the right belt as well as the left.
if the loop were a little wider, could put machines on the left, left-center, right-center and right
queen cairn
if the loop were a little wider, could put machines on the left, left-center, ri...
Yeah, I was mostly doing that to point out the belt's status 😛
queen cairn
if i have a 24 assembly stack it means i split it artificially half and half? 12...
Yeah this is the same number of machines
summer zinc
whats the difference between 12 +12 on this kind of loop vs 24 in a straight line?
queen cairn
Beacon sharing efficiency
You get sharing on everything except endcaps so the fewer endcaps the better
summer zinc
if i need a 24x10 row array?
queen cairn
summer zinc
my brain cant really comprehend what it looks like, guess the only way is to try and find out a side-by-side comparison
summer zinc
ah honestly feels like math homework took everything out of me to make that sushi maker accurate to 7:2:1 and rpeeating it again for 20:5:2
all in all im impressed with the result/accuracy/how easy it is with that 2^N method. i hope yall share it with the wider community
thyme's original tutorial might be good for the pro users but it feels a bit like draw the rest of the owl so i would use ichaleybin's amended tutorial to teach the newbs/intermediate players or non-software/math educated folks to "Make your own sushi" because that shit is damn delicious
queen cairn
lmfao fr it was a draw the rest of the owl
summer zinc
i mean u and maxreader could follow along just fine but it took a couple days teasing for me to figure out some of the jumps in steps
especially the part where wood/filler item gets removed and replaced with loop
now its almost laughably easy on the 2nd attempt (20:5:2). the 7:2:1 required a bit of alt tabbing to refer to ich's guide
make a couple more and i could probably do the algorithm by heart
queen cairn
Yeah it's a really simple one, the hard part for an actual computer algorithm is "Now make some educated guesses as to where the inputs should go"
Come to think of it, @fiery portal I know it's exponential scaling but how bad is that for the actual use cases? Is it so bad to enumerate all solutions?
summer zinc
how does one even go about proving that they reached the lowest number of splitters achievable for particular ratio anyway? i dont even know how to mathematically/logically approach such a task
everything has been done manually so far much like mitru's. there's some patterns you can use to reduce spaghettification but cant tell if i truly reduced it to its lowest possible size
fiery portal
how does one even go about proving that they reached the lowest number of splitt...
Now you’re finally at the part I’ve been at for many weeks now
I’m confident you can prove it with Thyme’s method and multiple items, but the one item slowdown is much harder to prove
summer zinc
i mean. the last time i even had what i'd call "formal" math training was GCE A Levels (UK Edition). I vaguely recall some shit like proof by induction, proof by contradiction. or something
i have zero framework to prove anything
i can roughly say yep thats the best i could reduce it to and that's it
with some of the terms you guys throw around it sounds ike you guys had much more formal and rigorous training to approach such problems
queen cairn
I've been holding off on finding an actual code implementation for some better understanding but it might be close to time. I'm also highly confident it's provable, and I think for actual use cases you could probably do a super naive implementation that searches more than it should
I had some formal training
fair gull
lmfao fr it was a `draw the rest of the owl`
the part where you remove the helper inputs and connect the second output of the last splitter is definitely like this
because when I was trying to generalise the method, it's what I tried on a hunch and it worked
fiery portal
Oh, you might be able to use higher up secondary outputs as long as you only use them on their respective sides… 
I’ll have to play with that
queen cairn
Oh, you might be able to use higher up secondary outputs as long as you only use...
It should be possible to draw up to N-1 belts, where N is number of input belts. Have to pay attention to individual feed rates, but at that point multiple inputs per material type are acceptable
fair gull
I don't think that works.
queen cairn
As long as the state stays compressed as a whole, theoretically I don't see why you couldn't draw more. Not that i have a solution, but I think
fair gull
The 2^n compressor get's reduced to a weighted x:1 balancer. Taking from splitters inside the "balancer" skews the input speeds and thus' the output ratio.
queen cairn
On the rate limiting side, the solution is actually almost trivial, you just... do it lol
queen cairn
The 2^n compressor get's reduced to a weighted x:1 balancer. Taking from splitte...
But in principle, I could feed say, engine sushi in as "one belt," yeah?
fair gull
what do you mean with that?
queen cairn
My goofy blue science mixer, had a ratio for gears:steel:pipe baked into it. I could've mixed those first, and then included only one input belt of that mix, at the proper rate, no?
Actually now that I look, I'm going to just confirm that that is how it works
As long as you're not decompressing, you could draw engine mix cleanly, I think
fair gull
You'd have to do the math on the basic ingredient ratios. 2/6 blue science ingredients are engines. the 2 engines have 8 ingredients, the 3 RC have 18 ingredients (or 27 when also making GC) and the sulfur doesn't expand
that's 27 total, or 36 when also making GC
queen cairn
fair gull
so the engine ingredients together would have to take up 8/27 of the sushi belt
queen cairn
Yeah, you have to stay under on rates, that was the 8:4:4 part of my ratio
I had to do it over 64
11:11:10:8:4:4:2
fair gull
I think at this point it's better to do it in stages. Mix the engines and RC separately, then mix those together with sulfur 8:18:1
queen cairn
Those solutions should pop out of the algorithm naturally....
Having the engine ratio mixer be a part of the greater mixer is the best solution
RC ingredients are mixing together too
fair gull
the best solution under what metric?
ability to debug? no
queen cairn
Min splitters lol
fair gull
I wouldn't say that's a given until I see a formal proof
queen cairn
Yeah it's definitely not at "given" state, I'm just teasing out things I think are provable to see what's useful
Breaking down things into sub-mixers seems useful
8:8 is 1:1 so 2:2:4:8 is 3 splitters
You can def multi-draw at ratio as long as you stay under on total rates, every input belt backs up at some point
Why can't you draw two mixes out of the output mixer even if you're way over?
Different speed belts works... it's a timing thing.
summer zinc
itslike some sort of beautiful chaos whenever i look at it
fair gull
perfectly normal phenomenon
summer zinc
anyhow. enough homework done for the day. will try this "sushi looP" 12 + 12 tomorrow
queen cairn
perfectly normal phenomenon
Yeah, actually I think we should take this to the main sushi thread anyhow lol, don't need to overload Darian
summer zinc
remind me again whats the upside of this looped sushi vs straight line? i have the memory of a goldfish and forgot what i was doing
20 beacon saved perhaps
queen cairn
Squareish is optimal for beacon savings, iirc?
But also you don't have to run a return belt back.
The return belt is an input belt
Then you recycle extras and it never jams
summer zinc
The return belt is an input belt brain hasnt fired up yet can you remind me again why i dont have to run return belt back
beacon savings is an idea i can definitely get behind although it means the build becomes taller for the benefit of being less wide
right now it's more rectangularish than squarish unless i need a 24x24 array
queen cairn
You do have to run return back for loop sushi. That belt does not need to be extraneous
Since you have to do something with the end of the belt anyhow, loop sushi is incredibly resilient and easy to setup as a defense to the end of the belt. Just don't have one. That you can loop it back is why it's potentially advantageous to cut long builds in half, and I think that's even more so the case for you. Most builds running speed beacons will not have such numbers of machines to play around with, but for you, squaring the builds up could save a lot of materials
That RC sushi I posted for you, it's sharing rate limited outputs, and the recycler
I'm not rate limiting four times, and recycling four times. I'm splitting one redbelt into four half yellowbelts. Simple as
One redbelt out is one redbelt in.
summer zinc
yeah i guess i instinctively went for squarish builds most of the time by choosing a square number based on the length of one particular stack and squaring it but didnt really force the issue by cutting it to a more appropriate length and looping
it does save a lot of materials though so i think i might have to do it more. will be interesting to look at the older builds now in this lens
also its never not ridiculous to me to see these pollution numbers
miners are one of the lowest in quantity per machine wise but still are the predominant polluters
200 miners produces 400 pollution / min, meanwhile 750 furnaces produces less than half that amount. ridiculous
🤣
264 machines just results in yet another longish rectangle
need some obscene 16x16 to resemble a square
lets see what 264 machine looks like beacon count wise for a 24length stack
i really wish i can see the exact number instead of this stupid 1.1k rounding
598+572 beacon = 1170
650+500 = 1150
summer zinc
do i even want to force a 17x17 square build 🤔 i think no
left version gives me a bit more space to fit those sushi makers off the main bus though compared to the right version
queen cairn
I find these the fun parts, figuring out how to make stuff work with the tools I have, trying to weigh my options
summer zinc
im trying to figure out if its worth the hassle but with the current effbeacon constraint 1 belt can only feed 23.18 LDS machines and its either squaring this 23.18 or trading off and making a slightly oblong square
think ill just set up a rectangle and move on, ive been dragging my heels way too long with these
queen cairn
Getting it done is a value all its own
Getting it done inside meme constraints like sushi, even harder
summer zinc
im not even sure if sushi is meme constraint for 12 beacon builds. it seems like 12 beacon kinda mandates sushi for the complex intermediates
whats the alternative? 3 input belts from 3 cardinal directions and 1 output belt in the 4th?
summer zinc
this is for up to 4 input materials?
summer zinc
oh i guess the output belt, one lane is also used for material
what are the crafts that requires 4 or more inputs anyway, i only can recall satellite i havent done the complex stuff in ages
queen cairn
Couple of odds and ends, spoder prolly
summer zinc
and the typical complex sciences? max 4?
plain vault
robot frames need a few i think?
queen cairn
Frames might need 4 but one's a liquid
Oh it's 4 and liquid is for electric engines, my bad
queen cairn
There's enough room to belt every intermediate here
You can do 12 beacon wagon locked but only for 2 item goods
Don't get distracted by this lol, but if you can get a third input into this, that'd be pretty massive. I don't think it's possible at all because of the wagon tiling, I don't have access to critical tiles.
This fails for two different lane reasons, super impressive fail tbh
summer zinc
This fails for two different lane reasons, super impressive fail tbh
how does the outserter load into the bottom right underground 🤔
direction is wrong?
queen cairn
It doesn't, that's one of the lane fails I can't figure out how to do anything about. It's not an input belt, that's for damn sure
summer zinc
a lot of it i feel is constrained by the beacon# choice from the outset
sushi is a nice way to boil it down to one belt in one belt out. i havent given much thought to just one belt for everything yet
not sure what the benefits would b
queen cairn
I mean, you could use sushi to do 12 beacon wagon lock pretty well
Can re-feed down the line if you wanted even
Can even get a splitter in on this side I guess lol, multiple refeeds and clean out on bottom
summer zinc
i have no idea wtf is a 12 beacon wagon lock and it sounds like made up words at this point
what does the word lock imply
queen cairn
Designs are locked to wagons
Wagons are locked to each other. If one wagon activates, they all do
That way I don't have to balance anything ever.
That's how I did the 20k. It's not my technique, but it's a rare one
I only did 10 beacon
summer zinc
no idea why there are transport line gaps when i try to extend beyond 1 sushi belt
input belts all look full
queen cairn
So I'm not 100% on this but I'm 99% on this, gaps = wrong ratio
It might be the undergrounds
Their mechanics are different than belts
Still shouldn't be either set of undergrounds, but they don't behave quite the same
This could potentially be squeezed anyhow
queen cairn
Welp I got close with a 20:5:2
Yeah I'm looking at yours and we have same splitter count, but as it turns out, the draw the rest of the owl actually starts all the way at the beginning, and where you group the inputs lol
Oh maybe you have fewer splitters than me? Wonder how
OH I get it now, it's stutter stepping because of the input block wire I think. Watch the control belt?
summer zinc
like...not sure if this is the correct interpretation
my steps here already has 9 splitters 'only' and that was the lowest i could get it to
everything else after was just reshuffling to compress it into less wide space at the expense of being longer
closed the game for now but looking at the picture again left most and very first sushi belt seems to have no gaps weirdly
queen cairn
Yeah I did a trick to organize it differently that would get +2 splitters but they should end up equivalent, okay
I think it's input starvation but not 100%
summer zinc
i have been adding supplementary copper plates where the calculator suggests i need to be adding. basically every sushi belt consumes about 0.75 blue belt worth of copper. i use the 0.25 to supply the next sushi belt, and 0.5 from the main line
ah fuck it i'll re open it real quick to get a better picture that was horrible i cant see what is happening
subtle meadowBOT
queen cairn
The actual input belts are gapless....
summer zinc
these were my steps
in case verification needed
good to keep a backup here on discord anyway in case i accidentally override or delete save
queen cairn
Yeah I almost did it that way
I decided to do a funny trick to keep locality for underground reasons
I didn't group, in order to add a splitter
summer zinc
the rightmost one is the one i am using to copy paste to replicate 11 sushi belts to feed 264 machines ...technically its 23.188 AMs * 11 stacks
not sure what that means but im pretty sure i referred to your pictorial guide exclusively to do these steps
queen cairn
I'll dig into it tomorrow b/c it works in single but not in chain, and max and I were interested in chain
summer zinc
this one
queen cairn
I think you did the grouping well
I specifically split one of your groups
So that all returns were localized together
More clearly, starting spot
summer zinc
i basically try to keep the position static so i could track visually what was happening. even maintaining the splitter levels along the same position on the y-axis
queen cairn
Yeah, when I saw your second grouping, I saw that you were using many of the same things I was in my design. Copper grouping is the big one
Hello wood lol
summer zinc
only when i was more or less certain the splitters are in its lowest possible number then i started moving stuff around
queen cairn
I don't think you missed a step is the thing
summer zinc
its really the only way to not get lost for me, since tacking input to input across spltiter level its easy to get lost
queen cairn
If it's not input blocking, it's not simple
summer zinc
tbh im not sure what happen when i try to scale up to 5 sushi belts
queen cairn
It was the scaling
summer zinc
when i tested for 1 sushi belt it produced exactly the amount the calculator says
queen cairn
Max and I are theorizing on that right now tbh
summer zinc
scaling is never straihtforward in this game
not so easy to see what is happening. input belts are all full (can see - no transport line gap) and yet somehow there is line gap on the output sushi
queen cairn
The way you did your sharing looked correct to me, and no input belt gaps should be sufficient
summer zinc
but in the first sushi belt if all input belts are full, i tested individually into a chest and ran it for couple minutes and the ratio was perfect
queen cairn
This is actually weird aF
summer zinc
3000 copper, 750 plastic, 300 steel. 20:5:2.
right on the money. replicated it a couple times too. i havent even gotten around to test for stuff like slowdown/input starvation, just the basic block output belt, remove certain input chest and see what happens and it looked kinda fine
this should be what is happening on a full sushi belt and completely consumed by 23.188 machines
this should be the leftover copper plates due to backpressure, and going the other end of the splitter. 0.2593** not 0.2597
queen cairn
This is what I'm looking at tho
summer zinc
i merge the two with a full copper infinity chest it should still have plenty of copper
queen cairn
Those dudes
If they flash red
They shouldn't but they might
This is like that thing I did the other day. "I don't see a train in that block. But there's a train in that block."
summer zinc
yea thanks. i'll slow it down to 1/4 speed in about 4hours and record the clip and observe
summer zinc
maybe relying on backpressure causing some sort of gaps
not sure
supposed back pressure in action
summer zinc
after priming and letting it run for a bit and checking the contents of the individual chests and their ratios
primed belt and their chest contents after unpausing and letting it run for a while
two sushi belts: 2 input copper belt, 1 input plastic, 1 input steel belt.
Furthest sushi belt chest (From left, Chest1) contents: 1919 
, 479 
, 192 
. Ratio - 20 : 4.992 : 2
Chest 2: 2002, 501, 200
one sushi belt only Chest 3: 2058, 514, 206
removing the lockstep protection from input starvation allows the belt to run full send. i am not sure what is happening to cause the belt to occasionally not read as item=8 on a full input belt
some ghosts in the signal reading? suddenly freed up space on the belt ahead causes a signal to read as non-8?
ok somethings fucky here
belt starts off with zero transport line gaps
connect a circuit cable, and gaps start appearing
remove the circuit cable, and the GAP PERSISTS
fair gull
probably because there's a transport line break
summer zinc
deleting and re-constructing a brand new belt, and the thing goes BACK TO NORMAL
fair gull
belts that have the circuit connections removed remember that they had a circuit connection at one point
summer zinc
why is there a transport line break though? the inputs are exactly the same
the weirdest part is the transport linegap persists after removing the circuit connection
fair gull
I don't know, but it might be that circuit wires force a transport line break
fair gull
the weirdest part is the transport linegap persists after removing the circuit c...
belts that have the circuit connections removed remember that they had a circuit connection at one point
fair gull
I said I don't know whether that's what happens.
The how is easy though: Belt tells the belt manager to insert a transport line brake (that's a different thing than transport line gaps)
summer zinc
now heres the most bizarre part. after i did all these experiments with the circuit cable connection/reconnection i have somehow fixed my transport line gap problem despite doing nothing
fair gull
I have seen transport line breaks sometimes causing gaps.
summer zinc
#1277595827547799552 message this is the problem scenario with the line gaps observed and sushi belt shortage
this is what im seeing now^ with exact same circuit conditions. only difference is i tampered with the circuit connections but i put them back to its original configuration
fair gull
You could enable "show-transport-lines"
I bet there's a transport line break at the input side of the circuited belt
summer zinc
are you able to decipher this visual mess? my eyes doesnt know where to look. took screenshot of the whole thing
left vs right setups for comparison
fair gull
are you able to decipher this visual mess? my eyes doesnt know where to look. to...
each arrow is a transport line
and the circuited belt tile is it's own transport line, there's a break before and after it
not sure what the blue colour means though
for some reason the transport line break introduces gaps in the copper belt
which are passed onto the output belt to maintain the ratio
summer zinc
The break doesnt occur on input side of circuited belt right?
It occurs on the circuited belt itself
And only on copper one. Idk why
fair gull
Me neither. I don't understand transport lines well enough.
summer zinc
gonna experiment on the existing sushi belts by tampering with circuit cable connection and see if anything sticks. two of the five sushi belts are working as intended but again i dont even know what i did right to know how to fix what i did wrong
😵💫
removing this particular belt seems to fix whatever is plaguing the sushi belt directly north of it
nope nothing i do seems to solve it
nvm gonna wait for the experts to weigh in
subtle meadowBOT
summer zinc
seems to be blueprint dependent but im not even sure what i changed to fix it
this particular version of sushi maker seems to avoid the transport line gap problem that occurs in other variants. i dont even know whats the difference.
summer zinc
it is kinda weird innit? where the gaps show up
and even weirder that it doesnt seem to have an impact on the final output sushi belt
fair gull
It's not supposed to happen.
It'll desaturate the output via the circuit condition.
Ratio should not be changed.
summer zinc
is this from the latest clip?
fair gull
yes
summer zinc
the output sushi beltS doesnt seem to desaturate but i'll run the simulation for another hour and see the output LDS
fair gull
those clips aren't helping me btw
there's too much going on and I don't know what to look at
summer zinc
this is just one module that combines 3 input belts into one sushi for 20:5:2 ratio using yours+ichaley's method
the goal is to duplicate this module to have a total of 11 sushi belts
since only a tiny proportion of plastic and steel is used for the sushi i split it off to feed into the next few modular sushi belt makers until the calculator says i dont have enough from the original belt to make one saturated sushi belt
tile it leftwards, and top up the sushi input belts with more infinity chest/blue belt of input where needed to saturate the belt
should i record a clip of the single sushi belt maker to show the result?
i documented how i got this final design here with my "workings" if that helps
oops that one was for 7:2:1 workings...heres the 20:5:2 workings
subtle meadowBOT
summer zinc
if i rely on every signal being 8 to protect against input starvation, is it possible that if one of the 3 read-hold belts becomes empty (0 input items by chance or idk what scenario), the output belt will be enabled because 2 of the 3 inputsignals are reading as "8"? and the last input signal reading as 0 is counted as no input signal?
making sure im reading this line correct
queen cairn
Yes, that's a possibility. It's why I used a constant combo and set it to 9
It's also an on/off switch, but my primary motivator was to defend against the loading problem
fair gull
No, it's not possible once the sushi maker is filled.
As soon as any of the 3 read belts drops below 8 items, the output belt stops. When the sushi maker is filled, this immediately stops emptying the read belts any further due to backpressure.
For filling the sushi maker it's best to delete the first belt after the output belt until all read belts are full AND the sushi maker is filled up.
summer zinc
where does the looped extra go in a case like this at the end of the belt? currently the input end of an available loop goes to that purple looking loop
or its like this?
fair gull
not sure I understand the question
the marked area has twice as many purple science as you want in the output (because there are effectively two purple science inputs in the 8:1 compressor)
luckily, the splitter at the end separates exactly half of the purple science to the secondary output, which is loop back to the input of the compressor
causing the problem of having twice the amount of desired purple science in the system
If that looks like thinking in circles, it's because it is thinking in circles.
summer zinc
i think the loop being purple is purely a coincidence and not reflective of the reality
if i make the loop longer it will change the ratio, or the colour, etc
i think its hard to explain what i want to achieve without a proper mockup without that infinity chest
will try to do a propery setup wheni got the chance
fiery portal
not sure I understand the question
the marked area has twice as many purple scie...
Yeah the loop being all purple is entirely inconsequential
For every purple that goes into the loop, another gets inserted
Your math is only right if you exclude the purple loop from your thinking (specifically, that there would be twice the amount of purple as needed)
It's not "luckily", it will always cancel out perfectly
summer zinc
i had a stroke re-reading my original question
this was what i was asking - does this protect me from all potential overload/jams?
was wondering how to make a loop at the exit end of the sushi belt, where to feed the exit end back instead of using chest defence (because there will be times i randomly research military)
black science will always stockpile since im not researching military most of the time
🤔 seems like this loop wont work, i need to split the sushi belt back into its constituents and merge with their respective input science?
i havent fast forwarded yet but im guessing it will eventually fill with black sciience only
sure enough 🤣
how many splitters does mitru's 1/7 rate limiter * 7 use? wonder what is the difference cost wise
breaking down the output end of lab sushi belt into its constituent/remainder science packs and feeding back into their respective input belt with input priority from this sushi separated science packs would protect against overload of black science/all other science right?
fiery portal
sure enough 🤣
Go back to thyme’s original with the filter splitters on the loop
You should filter everything off and back on again if it won’t always be consumed at exactly the needed rate
summer zinc
hmm which one is it
seriously need ability to pin some key messages here i have no way to retrieve them. theres a few important ones
queen cairn
Keyword searches are helpful
fair gull
For every purple that goes into the loop, another gets inserted
I should have put the "luckily" in quotation marks.
I like your explanation more, especially the line I reply to.
summer zinc
if i can rephrase my original question with a picture. im confused on where i should loop this red line
queen cairn
Return belts are input belts
summer zinc
Return belts are input belts
so basically i need another 7 splitters like i mentioned right? to filter it back into its individual constituent part
queen cairn
Yep
That doesn't look filtered
Thyme's I mean
This is filtered, you can combine the input splitter with the filter in some instances
summer zinc
the upper right of your picture is where the sushi maker is?
queen cairn
ye
fiery portal
This is filtered, you can combine the input splitter with the filter in some ins...
Yeah this is the (kind of) image I was looking for
summer zinc
wondering how i can make it (a) less bulky (b) is it truly the least amount of splitters, or am i better off using rate limiters
fiery portal
Rate limiters still need the same filter and recycle
queen cairn
Yeah, loop sushi with rate control, you're putting X items/sec on that belt, come hell or high water
summer zinc
whats the difference in terminology
queen cairn
The rate limiters release a partial belt which then you can mix at will, compression mixers don't
summer zinc
so what's the pros and cons of either approach? im thinking rate limiters generally require a lot more splitters
queen cairn
summer zinc
no idea if my intuition is correct
queen cairn
If I need 1/4th of a belt, I don't need to do it four times. I can feed four instances with one rate limit
https://cdn.discordapp.com/attachments/930147588160782436/1279871529789816893/image.png?ex=66e090bf&is=66df3f3f&hm=2c9b00f77062d8e4f95010031b1f0eaa1b1ea94e21a2c4e4e3ef23a1d8a20532& So with rate limit, if you starve a resource, it's just missing
fiery portal
You can even mix and match them
summer zinc
sorry can u zoom the part where you break 1 saturated belt into 4 instances
i guess that makes sense
queen cairn
Both methods of rate limitation are very resilient, though it takes a wire and a hard stop for compression. When you rate limit individual resources, you just put them all on the belt regardless
You won't overfill the belt. It's only 75% full
summer zinc
ok thanks will download the bp string to investigate on my own
queen cairn
Yeah, the belt fuckery in the middle is some true fuckery, but it does what it's supposed to, poking it should hopefully show you a thing or two about rate limits
summer zinc
21 splitters and a pretty ugly space footprint
a saturated blue belt of each should be able to feed 1000spm and three blue belt worth of sushi though that will need more splitters
queen cairn
So the thing about rate limiting that always caught my attention was that if I'm feeding one whole assed belt in, and getting 1/7th of a belt out, I'm doing something wrong
if I have to do it 7 times, that's worse
summer zinc
🤔 thats what i noticed but im still failing to grasp what you did here
feeding a whole ass belt just to throttle it to 1/7 is just dumb
gleaming quarryBOT
summer zinc
hmm thats interesting
i need 36 labs to consume 1k SPM and trying to come up with the least beacons for 36, not sure why, but for 8beaconed build making a square 6x6 array of labs somehow requires MORE beacons than a straight line of 12x3?
i also tried 7x5 +1 protruding and it requires 1 extra beacon
fiery portal
i need 36 labs to consume 1k SPM and trying to come up with the least beacons fo...
longer stacks mean a higher beacon sharing efficiency
Somewhere around 1/4 height/width should be optimal, I would think?
summer zinc
ichaley mentioned that square builds are most efficient but apparently not for all cases
does it differ # of beacon wise? 12 beaconed maybe favours squared?
fiery portal
four rows of 9 labs would be my guess
summer zinc
whats the pattern behind the beacon sharing efficiency depending on layout, i cant really see why
fiery portal
If you add a row of labs and becaons onto the "top" there, then all of your new beacons are only affecting the labs you added. If you add them to the "sides", they're affecting the new ones AND the last column on the sides
summer zinc
9x4 vs 12x3 - both requires 60 beacons. the 6x6 required 63
the 7x5 +1 required 61
course this isn't a big deal for labs but i reckon for something like 400 machine 
production the layout matters a lot more
fiery portal
Meh, not really imo
summer zinc
no? i thought the difference could be close to double digit difference in beacons needed
fiery portal
What's the actual impact of that?
summer zinc
If you add a row of labs and becaons onto the "top" there, then all of your new ...
if it's like this, wouldn't it just making a very long 36x1 array be the most efficient? but obviously that wont be the case. im not sure where is the tipping point
fewer beacons? less t3 modules (they cost idk maybe 60 pollution per t3 module to produce right now). and also power of course. seems like just another one of those tiny optimisations
fiery portal
So, starting with one lab and its beacons, ask yourself "is it more efficient to add another lab and set of beacons to the side, or the top?" and work your way up from there
summer zinc
theres probably a lot of middle ground but its important that i dont accidentally walk into the worst possible beacon configurations
fiery portal
Intuition says it should end up somewhere around "add four columns to the side, then a row on top" because of the 8 beacons per lab and 8 labs per beacon thing
fiery portal
fewer beacons? less t3 modules (they cost idk maybe 60 pollution per t3 module t...
Sometimes tiny optimizations get in the way of playing the rest of the game 🤷
queen cairn
Oh yeah, that's a good point. It trends towards square builds, but at the smaller end, that last row does make a diference. Also, 12 beacon builds are affected differently I think?
summer zinc
took me several re-reads to understand what max said about 8 beacon per lab and 8 lab per beacon thingy
screws with my head that the "neighbour bonus" if u can call it that is very low for the first machine/lab that is added NOT in series along the array. so the beacon cost for the first is high. but then adding extra along the the new parallel array is very cheap after the high cost of the first one
queen cairn
Yep
The ratio changes for 12 beacon b/c each new column is signifcanly more, b/c sharing is less
summer zinc
Sometimes tiny optimizations get in the way of playing the rest of the game 🤷
sometimes i feel like tiny optimisations IS the rest of the game because what else is there after rocket? its just more of the same
got to add some curveballs to make it a bit more interesting at least. ive seen some responses to sushi being why bother? just use 3 input belts and its not wrong but also not very interesting
queen cairn
There's a lot of places to find your challenges
It's 2 yellow sci machines with beacons to do 120 SPM. I could've been done an hour ago
Except, SushDI
summer zinc
i want to do those beaconed 5MW/10MW challenge after this. i mean im finally reaching lab delivery phase so im getting close to the end
of course this is assuming that everything doesnt juts fall apart like dominoes the moment i turn it all on at once, which is also a real possibility lmao
fiery portal
got to add some curveballs to make it a bit more interesting at least. ive seen ...
More complex mechanics ala space age/mods are my cup of tea
Give me better reasons to use sushi instead of shoe horning it into vanilla :p
summer zinc
i look forward to 2.0 and space age for sure. i feel like the endgame side of things, military is grossly unbalanced so its a bit sad to even have it exist
summer zinc
i wouldnt call it shoehorning tbh. at some point sushi is gonna make more sense than having 3 input belts for every complex recipe. especially after getting past the initial ramp to understanding how to make a sushi belt
i'd call 3 input belts "bruteforcing" a problem
fiery portal
Single complex recipes? Still straight feed unless they have 7+ ingredients. It’s interleaved complex recipes where sushi really shines
summer zinc
with spoilage being a thing in 2.0 or is it space age im excited to see what new problems i have to solve
queen cairn
Single complex recipes? Still straight feed unless they have 7+ ingredients. It’...
Honestly, breaking down the whole chain to sushi-able inputs, and DIing intermediates, has actually been pretty neat
Purple sci got two ores and two circuits input, DI'd a bunch, and put steel and iron back on the belt, it was something. Idk if it's good, but it was compact
summer zinc
god just realised follower robot count is actually just 30seconds researchtime so i misused calculator again and need 2x the labs
also think i might stick to just 5 sciences sushi'd and have purple+black in a sideload situation, since 8beaconed labs can have two parallel input belts and still be in range of 8 beacons unlike 12
now you mentioned the whole splitting full belt 7 times to make 7 sushi being nonsensical im wondering how i can achieve the same result but better with 2 wagons, 1/7 of a blue belt isnt even 1 stack inserter working full time
to unload from a wagon
summer zinc
unfortunately i think i softlocked myself into 7 x 1-2 trains carrying one of each science
gleaming quarryBOT
fiery portal
Hmm, factorio school mobile is dumb
