Train loaders that shouldn't work

Don't know how to use a machine? Looking for efficient setups? Stuck in a mission?
Post Reply
User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 232
Joined: Mon Mar 25, 2019 9:40 pm
Contact:

Train loaders that shouldn't work

Post by Khagan »

Here are two related loading arrangements that I don't recall seeing discussed previously (feel free to correct me if they have been) and that seem to work much better than they have any right to:
loader30_cropped.png
loader30_cropped.png (426.46 KiB) Viewed 2046 times
The one on the left completely eats both fast belts of input (i.e. 60/s), despite the fact that the apparent capacity of the (single-laned) belts adjacent to the inserters is only 45/s. Moreover, although there is no explicit mechanism to balance within each pair of chests, the number of items loaded into each is the same.

Its baby brother on the right is not quite so surprising. The throughput of 30/s (yes, it eats the whole input belt) doesn't actually violate the expected limit on the capacity of the last stage of the belts, but it does mean that the stack inserters are each pulling 15/s, compared to an expected limit of about 12/s from a fast belt.

Tertius
Filter Inserter
Filter Inserter
Posts: 650
Joined: Fri Mar 19, 2021 5:58 pm
Contact:

Re: Train loaders that shouldn't work

Post by Tertius »

This probably works, because items grabbed by the (first) inserter are replaced faster with sideloading than they would if items are just carried by the belt.

With sideloading, they appear from 2 spots (far and near) and can be grabbed by the inserter immediately from where they appear, while with just moving on a belt, items appear only from 1 spot and the inserter has to wait for them to move into its reach.

It seems this "immediately appearing" is significant enough to speed up the whole process by 3/s. After all, belt-to-chest speed is not limited by the inserter itself, it is limited by the inserter not being able to grab items fast enough from the belt to get its maximum theoretical throughput. Maximum theoretical throughput of the inserter is chest-to-chest.

For me, this is proof how painstakenly exact in every detail the transport belt and inserter simulation is. It is really simulating and handling every single item with all its movement, position and grabbing.

Nidan
Fast Inserter
Fast Inserter
Posts: 225
Joined: Sat Nov 21, 2015 1:40 am
Contact:

Re: Train loaders that shouldn't work

Post by Nidan »

In addition to what Tertius wrote:
When putting items on a belt, the gap needed at the insertion point doesn't need to be large enough to fit a whole item. Items already on the belt will be pushed back or held in place until there's proper spacing again. Thus, as long as the outer inserters can remove enough items to make up for the speed differential (e.g. full red to half blue or 7.5 items/s), there's no capacity issue.

Tertius
Filter Inserter
Filter Inserter
Posts: 650
Joined: Fri Mar 19, 2021 5:58 pm
Contact:

Re: Train loaders that shouldn't work

Post by Tertius »

Inspired by the example, I came up with this 6 chest train loader:
Screenshot_20221205_204945.png
Screenshot_20221205_204945.png (545.28 KiB) Viewed 1967 times



Consuming 2 45/s belts per 6 chests is not possible without sideloading - feeding the output from the splitters directly to the inserters is too slow. It's a miracle, but the chests are filled not much less equal than other "balanced" mechanical train loaders. And I guess it has a bit of integrated lane balancing for not completely compressed unbalanced input as feature.


Nonetheless I added active balancing with circuits, and to my surprise, I still get full throughput (usually not possible):
Screenshot_20221205_203838.png
Screenshot_20221205_203838.png (815.61 KiB) Viewed 1967 times


User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 232
Joined: Mon Mar 25, 2019 9:40 pm
Contact:

Re: Train loaders that shouldn't work

Post by Khagan »

Nidan wrote:
Mon Dec 05, 2022 11:18 am
When putting items on a belt, the gap needed at the insertion point doesn't need to be large enough to fit a whole item. Items already on the belt will be pushed back or held in place until there's proper spacing again.
I hadn't thought about that aspect. You are right that it will help to maintain a steady flow.
Tertius wrote:
Mon Dec 05, 2022 8:02 pm
Inspired by the example, I came up with this 6 chest train loader:
Impressive. I suspect that some of the express splitters and belts could be demoted to fast without loss of throughput.
It's a miracle, but the chests are filled not much less equal than other "balanced" mechanical train loaders.
Yes, it's fairly clear how the sideloading helps with the overall speed-up, but the natural balancing just seems to pop up out of nowhere.

User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 232
Joined: Mon Mar 25, 2019 9:40 pm
Contact:

Re: Train loaders that shouldn't work

Post by Khagan »

I don't think the loop-backs are needed for the 6-chest version. I can get full throughput with
loader90.png
loader90.png (150.75 KiB) Viewed 1931 times


It's a steady 90/s without the balancing circuit; with the balancing circuit it's still mostly 90/s but occasionally dips to 89/s.

Tertius
Filter Inserter
Filter Inserter
Posts: 650
Joined: Fri Mar 19, 2021 5:58 pm
Contact:

Re: Train loaders that shouldn't work

Post by Tertius »

Yes, the loop-backs were my previous purely mechanical balancing. I just added the bottom horizontal belt line as initial try. Previously, in this place where underground entries that lead to underground exits on the opposite side of the rail with another line of chests. This was my original concept for feeding 2x 45/s belts per wagon. Unfortunately, it isn't possible to balance that setup for equal load of both chests of a pair, because both inserters of a pair are grabbing from the same lane, and as soon as I try to balance both with circuits to grab equally, one lane will stall, so no maximum throughput.

Your minimalistic solution is very surprising. It's ingenious. Although it looks as if some noobie player just put together some splitters and different belts to somehow feed some inserters. It's ugly and it should not work, considering the sophisticated, beautiful, symmetric, complex setup of other balancers, but it's working perfectly. I'm still lmao.
If you use it for 4 wagons, throughput will be approx. 358/s (from 360/s max).

By chance, I did thorough research on the internet for train loaders the last few weeks, because I'm in the process of building some megabase on my own. Wanted to know what other people use and pick the best concept. Not just copy the best blueprint, but make an own design with the best concept. But as far as I remember, I didn't find any concept with that horizontal single lane feeding. If there was any, I skipped it and did not recognize the potential.

User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 232
Joined: Mon Mar 25, 2019 9:40 pm
Contact:

Re: Train loaders that shouldn't work

Post by Khagan »

Tertius wrote:
Tue Dec 06, 2022 1:55 pm
It's ugly
My artistic soul is mortally wounded. What is wrong with elegant simplicity?

Seriously though, you want ugly? Here's ugly (and cheap):
loader90fast.png
loader90fast.png (369.24 KiB) Viewed 1814 times
No attempt at mechanical balancing. The splitters just stuff as much as possible into the upstream side-loading belts, while ensuring there is an overflow path that goes further downstream. It Just Works™.

Tertius
Filter Inserter
Filter Inserter
Posts: 650
Joined: Fri Mar 19, 2021 5:58 pm
Contact:

Re: Train loaders that shouldn't work

Post by Tertius »

"Ugly" only in aesthetic sense - functionality is top. The "cheap" one is not so ugly, as far as I see it. It's not so unsymmetric as the previous one, however I prefer the previous one: 1 tile smaller and simply consumes 2 blue belts, which is usually the output of 2 production lines.

I upgraded my lab setup with these, and there is really an advantage of having a circuit-enforced output balancing. Suddenly the input is balanced as well, which speeds up train unloading at the other side of the production line.

User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 232
Joined: Mon Mar 25, 2019 9:40 pm
Contact:

Re: Train loaders that shouldn't work

Post by Khagan »

Tertius wrote:
Wed Dec 07, 2022 12:34 am
I prefer the previous one: 1 tile smaller and simply consumes 2 blue belts, which is usually the output of 2 production lines.
I don't use blue belts much for routine transport. Even in a multi-megabase, there are only two ingredients that are both sufficiently valuable to justify blue belts for bulk carriage and sufficiently voluminous that one might want to load them into a wagon at 90/s, namely steel and red circuits.

But if you do want blue belt inputs, it's an easy tweak (note the small but essential change in splitter priorities):
loader90express.png
loader90express.png (365.07 KiB) Viewed 1779 times
Shortening is possible, but comes at the price of a (very) small reduction of throughput due to the loss of the 1-tile buffer after the final splitter:
loader90short.png
loader90short.png (361.25 KiB) Viewed 1779 times

mmmPI
Smart Inserter
Smart Inserter
Posts: 2673
Joined: Mon Jun 20, 2016 6:10 pm
Contact:

Re: Train loaders that shouldn't work

Post by mmmPI »

Those are really interesting to look at !

I have seen people using splitter instead of the horizontal lane, for the 2 first example with 4 and 2 inserters (grabbing directly from the splitter). It has the similarity to also present the items 2 by 2 to the inserter. I would think splitters achieve more consistently the balancing part when the belts are not full for the 4 inserter version, or when stack size is different. but I would also have said that your versions won't work due to throughput limit 2x1/2 blue belt is 1 blue belt that's what i expected, it challenges my established beliefs :).

I found the 2 subsequent designs even more puzzling as those using both the side-load and additional 1/2 belt with regular load on 1 inserter, it couldn't be replaced with splitters as 3 of them would not allow any material from that additional 1/2 belt thus making them their special category for me as fully utilizing the trick in a more visible way than just being faster than grabbing from the splitter.

Khagan wrote:
Tue Dec 06, 2022 9:38 pm
loader90fast.png
No attempt at mechanical balancing. The splitters just stuff as much as possible into the upstream side-loading belts, while ensuring there is an overflow path that goes further downstream. It Just Works™.
this one seemed wrong to me because the material on the right belt can never go in the left most chest, which made me assume it is planned for use with already balanced belt, as otherwise the circuit balancing can get stuck, in which case i don't understand the difference with going for 1 red splitter per belt to feed a pair of inserter X3 , you could grab from the splitter directy, or have an additional horizontal sideloaded blue belt, or even both a vertical and an horizontal blue belt after each of the 3 splitters, all those requiring less pieces while potentially still using the same trick ? i feel like i'm missing something there.


I really like the last 2 as i found it more enjoyable to upgrade my 2 yellow belt to blue rather than drawing a potential 3+ lane wide bus of something when i can avoid it. Rich ressources settings and compact spagetthi help disregarding the cost, it's rather the number of pieces used that i tend to consider. the very last one being simple and elegant to me :)

for my personnal use i would add a splitter to couple the 2 input belt, to allow the material from each side to go to each chest even in case of unbalanced not-compressed belt but that's where my understanding stops, the design is simple but it's not so simple to understand the design.

Could you explain a little more why is that change of priority so essential, given a lack of understanding of the previous one ?

User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 232
Joined: Mon Mar 25, 2019 9:40 pm
Contact:

Re: Train loaders that shouldn't work

Post by Khagan »

mmmPI wrote:
Wed Dec 07, 2022 4:09 am
Khagan wrote:
Tue Dec 06, 2022 9:38 pm
loader90fast.png
No attempt at mechanical balancing. The splitters just stuff as much as possible into the upstream side-loading belts, while ensuring there is an overflow path that goes further downstream.
this one seemed wrong to me because the material on the right belt can never go in the left most chest, which made me assume it is planned for use with already balanced belt
Yes, all my six-chest versions assume that the belts are already balanced. They are loaders, not balancers.
you could grab from the splitter directy, or have an additional horizontal sideloaded blue belt, or even both a vertical and an horizontal blue belt after each of the 3 splitters, all those requiring less pieces while potentially still using the same trick ?
So try those out. You might find an even better version! But note that with the six-chest setup there is unavoidably some leakage from the earlier parts of the cross-belt to the later parts, so the earlier feeds must have more than their share.
Could you explain a little more why is that change of priority so essential, given a lack of understanding of the previous one ?
Nothing magical. The left-most feed must have more than half a red belt, to ensure the first inserter gets enough items, so needs the priority when the input is just a red belt. Conversely, if too much goes up the right-most feed, the central inserters will be starved, so it must be deprioritised when using blue inputs.

mmmPI
Smart Inserter
Smart Inserter
Posts: 2673
Joined: Mon Jun 20, 2016 6:10 pm
Contact:

Re: Train loaders that shouldn't work

Post by mmmPI »

Khagan wrote:
Wed Dec 07, 2022 6:30 am
So try those out. You might find an even better version! But note that with the six-chest setup there is unavoidably some leakage from the earlier parts of the cross-belt to the later parts, so the earlier feeds must have more than their share.
It was not obvious to me in designs not using the extra 1/2 belt to feed the first inserter in the horizontal lane, thanks for explanations.

No better version found, tested under those conditions ( no circuit balancing and 2 different stack sizes)

tests.png
tests.png (1.36 MiB) Viewed 1714 times
tests2.png
tests2.png (1.34 MiB) Viewed 1714 times
The bottom-middle 2 setup when slowed down to speed 1/4 seemed to show the version with the input priority on the top most splitter with a slightly better result, being more often showing 90 than 89 compared to the same setup without the input priority

User avatar
Khagan
Fast Inserter
Fast Inserter
Posts: 232
Joined: Mon Mar 25, 2019 9:40 pm
Contact:

Re: Train loaders that shouldn't work

Post by Khagan »

mmmPI wrote:
Wed Dec 07, 2022 1:18 pm
The bottom-middle 2 setup when slowed down to speed 1/4 seemed to show the version with the input priority on the top most splitter with a slightly better result, being more often showing 90 than 89 compared to the same setup without the input priority
As expected. The left-hand input to the last splitter is almost empty, but when there is something there it must be let through, since it has nowhere else to go. If the right-hand input to that splitter is refused admission it still has the option of going up the right-most feeder belt.

Post Reply

Return to “Gameplay Help”