Belt Designs for ColonelWill

[MadZuri]

I hate to rain on your parade psorek, but those designs will have some heavy losses under full load. They don't have any fast turns. You're trying to shove a full compression trough a turn without splitting it first. That's the hard part about these designs, getting a fully compressed output when there is a fully compressed input. I'll test one of them on stream real quick and post results.

[MadZuri]

As promised, I completed some testing for one of the designs while streaming, so there were plenty of witnesses.

results

The corners are not fast corners, so the inside edge of every turn loses about 35% compression. As the outputs for the right side have fewer turns on average than the left side, the compression losses are greater on the left side and the right side outputs a denser stream. Your design might work perfectly if it was built with fast belts with express belt turns. This is a LOT harder than it seems.

[dee-]

Hm. What so you exactly mean with "rebalancer"? That the input, even only from one lane of one belt, gets equally distributed across all output lanes?

[quinor]

Rebalancer takes X inputs and spreads them evenly on Y ouptuts.

MadZuri, you're completely right. I just forgot to mention that I use them before fully compressing the belts - so for example instead of 2 compressed belts I use 4 that I can turn freely. It's still 100% OK for my application, becouse I'm evening with them output from train station which is obviously not compressed.

[MadZuri]

That makes sense then. I tend to combine those 2 things, such as the 12 to 4 merger I posted earlier in this thread. It also balances the line. Pretty much every design I make is for full compression max throughput.

[dee-]

Rebalancer takes X inputs and spreads them evenly on Y ouptuts.

Assume X = Y = 8 which gives 8 belts in and out = 16 lanes in and out
Fill only one lane from the input. By what you say it should be evenly distributed through all 16 output lanes, each with a ration of 1/16.

I can't currently test it but from what it looks like to me it's not balancing this way?

8x8

[quinor]

Right, it doesn't balances lanes. It would have to be done separately, I haven't needed it.

[dee-]

Right, it doesn't balances lanes. It would have to be done separately, I haven't needed it.

Okay, balancing lanes is a bit extreme But does it balance belts because if you only feed items through the topmost input belt, how do items get to the topmost output belt?

[quinor]

It does. Items go through the very last vertical tunnel to the right - I'll draw a picture when at PC.

EDIT:
Here it goes:

top to bottom

8-8_evener_small.png (398.03 KiB) Viewed 10260 times

The balancer is built from two 4-balancers (this is roughly first 7 tiles from the left), then 6-tile shuffler (which alters belts in such way that first and second, third and forth and so on are from different halves of input) and finally last tile of 4 2-balancers that do rest of the job.

[dee-]

The balancer is built from two 4-balancers (this is roughly first 7 tiles from the left), then 6-tile shuffler (which alters belts in such way that first and second, third and forth and so on are from different halves of input) and finally last tile of 4 2-balancers that do rest of the job.

You're right! So it's just me being stupid and commenting without trying it out first

annotated

Actually it's pretty nifty
Thank you for sharing the nice design

I suspect throughout will suffer a bit because of the corner-speeds but that's how far one can get.
Especially the top right quarter with the wavy belt draw my attention. I'll give it a go when I find some time but I suppose you've already optimized it. Maybe swapping some belts that go to the output splitters... hm...
For sustaining maximum throughput I suppose one has to split all incoming belts into two, then have two balancers, then remerge afterwards to the original number of belts. That's corners.


Edit: forum logged me out making a mess with threads

[quinor]

What you just described (splitting each belt into two, then balancing and merging again) is exactly what I do - with the exception that then you'd want 16-balancer which I have only in big version yet . At least 4-balancer (first part of 8-balancer) seems to preserve full compression.

[Koub]

For sustaining maximum throughput I suppose one has to split all incoming belts into two, then have two balancers, then remerge afterwards to the original number of belts. That's corners.

As you won't be able to optimize throughput and space needed for the rebalancer, if you eliminate the space constraint, it might be mode "simple" to just double the places where the belts turn as shown there :

https://forums.factorio.com/wiki/inde ... KUcd6r.gif

[quinor]

If you eliminate the space constraint, the task of building perfect rebalancer goes quite easy. I can build you out of box even loseless 64-64 then.

[Koub]

sure, but if I have to make a space optimal 16x16 rebalancer to compensate for the compression losses of my 8x8 space optimal rebalancer, I'm not sure space optimization is very productive.

[quinor]

Still depends on your needs.

[MadZuri]

4from16side

2from16side

[MadZuri]

New train station design is up:

both

[OBAMA MCLAMA]

I knew there was beter designs.

and jeeze mad, just how much do you actually spend playing the game compared to optimizing everything.

keep up the work, with you around I dont have to even drop a thought about optimizing certain things. I love the updated unloading and loading.

[MadZuri]

8balancer

Credit to Tallinu for the base of this design!

loadingunloading

Anything I can do, I can do better.... (sing along) I can do anything better than me!

[Talguy]

Anything I can do, I can do better.... (sing along) I can do anything better than me!

You continue to astound us(if not then me at least).