90 degree 4-belt merge

[tengumai]

I'm trying to merge 4 belts to 4 belts at a 90-degree angle using as little space as possible

I haven't been able to come up with anything smaller than this (8x6)- does anyone here know how to improve it?


T

[Mehve]

I guess it depends on your definition of "merge". Based on your design, I'm guessing you're not worried about keeping the paths distinct from each other.

I used a 8:4 balancer design and turned half the inputs sideways to make this. Fits in a 7x6 plot of space neatly, or, if you don't mind a little stray side insertion, you can squish it further into a 7x5 plot of land.

[TruePikachu]

I guess it depends on your definition of "merge". Based on your design, I'm guessing you're not worried about keeping the paths distinct from each other.

I used a 8:4 balancer design and turned half the inputs sideways to make this. Fits in a 7x6 plot of space neatly, or, if you don't mind a little stray side insertion, you can squish it further into a 7x5 plot of land.

That design acts as a bottleneck, since it only has 2 belt input capacity on each of the two sides.

EDIT: And if two adjacent belts which are on an edge of the input are at 100%, it bottlenecks since it can only has capacity for one of them.

EDIT: With a small rearrangement, you can save a column wide:

Code: Select all

║║║║
║║╨/\
║/\╨╚═
║╨╚═══
║ ╥
/\/\
║║╨╚══
║╚╡╥╞═
║╥╥║
║║║║

[Mehve]

Potential bottlenecking s inevitable - the OP is specifically requesting a setup with fewer outputs then inputs. Without knowing how full each incoming belt is, there's no way to optimize further. If we know for certain that one direction is heavier then the other, then your solution (and the OP's) would be harder to bottleneck. But without knowing, the best that can be managed is to avoid any instances of fewer internal lanes then outputs, which my solution (and yours and the OP's) achieves.

[mergele]

But in your solution each input direction has a section with only 2 belt width. So if one of those inputs is more than 50% full it will bottleneck and a 75%-25% ratio will neverproduce a filled belt.

[TruePikachu]

Generally, merges should be designed such that it does not bottleneck if one input is 100% and the other is 0%, of course assuming the output width is greater than the input's width.

[PyroFire]

Nailed it

[TruePikachu]

Nailed it

Very clever.

Though technically my 5x8=40 is slightly smaller than your 6x7=42, if one uses total area as the singular measurement.

I'm almost certain the minimum width is 5 without becoming very long, so the next target dimensions are 5x7=35 or 6x6=36; the former would obsolete both our designs.

[PyroFire]

Revised 5x8.
No bottlenecks, no splitters feeding underground belts, perfect split, and one noteworthy improvement from the previously posted 5x8 splitter; you don't need to offset your entire belt line by 1 block.



I attempted the 6x6.
no matter what you do, unless you build outside that 6x6 grid it is impossible to achieve a perfect split.
This is the closest i could get, with only 2 of the incoming belts being bottlenecked on a splitter.



If you allow slight altering of the criteria (aka it doesn't have to be Exactly fitting inside the box) and minor changes to the input belt rows, this is slightly smaller than the previous 2.



All in all, i'd say the 8x5 is probably the best design out of all of them.

[TruePikachu]

@PyroFire:
While I didn't elaborate, technically, my 5x8 doesn't require the gap on the side input, or rather, the gap can be removed from within the footprint. Something slightly notable, however, is that the original merge from OP contains that gap as well. Your 5x8 can include that gap inside the footprint as well; tunnel the lower side input and move the 2nd-lowest side input down a tile.

For the 6x6, do you have an explanation for your assertion? I can state that the minimum practical width is 5 because a splitter on the edge of the bus would either need to go inwards (requiring extra tunneling stuff for the blocked belt), or outwards (creating the 5 width).

[AutoMcD]

Nailed it

Nice work on those!

[PyroFire]

@PyroFire:
For the 6x6, do you have an explanation for your assertion? I can state that the minimum practical width is 5 because a splitter on the edge of the bus would either need to go inwards (requiring extra tunneling stuff for the blocked belt), or outwards (creating the 5 width).

Screenshot or it's not true kthnxbai

[TruePikachu]

@PyroFire:
For the 6x6, do you have an explanation for your assertion? I can state that the minimum practical width is 5 because a splitter on the edge of the bus would either need to go inwards (requiring extra tunneling stuff for the blocked belt), or outwards (creating the 5 width).

Screenshot or it's not true kthnxbai

4x8 is not going to be possible because of the requirement that each belt needs to be merged:

Code: Select all

 +║║║║+
1|║╨╨║|
2|/\/\|
3|║?║╨|
4|║ ╚══
5|╨╥╥ |<-
6|/\/\|<-
7|?   |<-
8|╥  ╥|
 +║║║║+

The three arrows indicate where the other three side inputs need to come from, while the upper set of splitters is already as small as it can possibly be made inside a 4x8 area (and one of the splitters aren't even connected to the side input). No matter where you put the other two splitters, one of the required side inputs will be blocked and unable to be connected anywhere (in this diagram, row 6).

[MeduSalem]

Some of my approaches to the problem:


6x7, but I don't like how there's an infinite loop on the one lane of the lower left splitter:

test 1a.JPG (31.69 KiB) Viewed 25500 times

6x7, but I didn't like the 2 empty tiles... urgh:

test 2a.JPG (43.11 KiB) Viewed 25500 times

5x7, now that's what I could live with:

test 3.JPG (40.62 KiB) Viewed 25500 times

Seems like my 5x7 solution is the most compact one presented in the thread to date that still supports full throughput. 5x7 = 35 tiles, 6x6 would still be 36 tiles.


I somehow prefer symmetric solutions because I can't stand splitter-mess... Maybe I'll find another one... But I don't really think that 5x5 or 6x6 or a solution with 4x7 or 4x8 is possible without some trade-offs.

[AmbulatoryCortex]

5x7, now that's what I could live with:

Seems like my 5x7 solution is the most compact one presented in the thread to date that still supports full throughput. 5x7 = 35 tiles, 6x6 would still be 36 tiles.

I really like this design because it should easily fit anywhere where two lanes are otherwise reserved for underground belt crossing. It's a drop-in solution.

[Tnarg]

No where near as good as MeduSalem 's 5x7 my I made this to sit in the shape of the T-Junction.

T.png (229.41 KiB) Viewed 25418 times

Edit: Just looked at Tengumai's first post and spotted how close my design is to his. But this one is a little more compact.

[CLion]

5x7, now that's what I could live with:

i just shared this pic, it's so beautiful *_*

[PyroFire]

I attempted a perfect balanced X intersection.
Ended up rather large.
Would love to see this done better




And the inverted T-Junction.
Got it to 7x6 so far.




6x9 4-lane individual belt balancer (the one on the right)