Belt Balancing / Usage Issue

[Cyrikyty]

So, I tried to use a splitter system in my base to make sure every lane I pull off the main bus gets full throughput, as long as I got enough plates on my 12-lane iron and 12-lane copper bus. This worked relatively nicely, until I started using a majority of the resources, and I encountered 2 oddities. For starters, this is how I pulled off the resources:

1

Splitters.png (1.43 MiB) Viewed 7251 times

The first issue I encountered was that a handful of furnaces would not be producing due to the lanes being backed up. However at the end of the belt, I had actually run through my entire supply, leaving the belts completely empty. These were the furnaces in question:

2

Copper Furnaces..png (1.83 MiB) Viewed 7251 times

and this is how my belts looked like after all my assemblers had their fill:

3

Copper Lines.png (504.78 KiB) Viewed 7251 times

Something just doesnt add up, some copper just gets stuck somewhere, and my total consumption was around 250 per second, which isn't a total of 12 red belts, its a bit less. I have honestly no idea how it is getting stuck.

Secondly, despite my best efforts, some lines got heavily unbalanced. This happened mostly to my iron lines, and I would see literally all 12 of my lanes running at half a belt each, all on the same side. This eventually caused a shortage since it couldn't supply full belts. This is how that roughly looked like:

4

Iron halfes.png (1004.75 KiB) Viewed 7251 times

I got belt balancers after my furnace setups, and every time a belt leaves the bus, so I'm at a loss here aswell.
So, if anyone knows why these issues are happening and has a solution for them, I'd be happy to hear it.

[zOldBulldog]

I use a similar approach and I am happy with it. But there are a few tricks.

1) using the splitter priority output can work great. Especially as it let's you *see* when your resources are low, so that you can compensate by adding more supply or adjusting the transport. One caviat:. I rarely balance/shift more than 4 lanes at a time. I don't attempt 12... yet, even though I do have a dozen belts of the same ore because I split it into sets. There seems to be a limit as to how many belts you can left shift. Consider breaking the shifts into sets of 4 or even upgrading the splitters to blue.

2) In your first screenshot, I recommend doing the Cascade *after* the splitter that you use to pull resources from the bus instead of before. Helps your left belts to stay full.

3) It is normal to have idle furnaces if your belt is backed up in that area. I understand that you have empty belts at the end. The problem is that your copper is backlogged getting there. Look for the spot where the backlog starts. Or if you can't find the spot try upgrading your bus to blue bets/underground's/splitters.

4) The iron half belts are probably caused by some of your assembly lines pulling only from one side of the belt. Easily fixed by putting a right/left half belt balancer (like you do at the exit of the smelters) in the area where you see the issue start.

Try things out and please report the results. My instincts tell me there is some important know-how hiding behind your problem.

[malventano]

If items come into both sides of the balancer at the same time, it can't send both of those items to the same (output priority) belt - it has to output to both lanes since the single output lane can't handle the volume. There's a way around this by getting creative with input and output belts connected to the circuit network (intermittently stalling the non-priority output to force more items onto the priority output), but getting the numbers just right can be tricky.

[Cyrikyty]

Thanks for the advices, I have already been able to identify the copper issue, alltough I still don't understand why it happens. It turns out, when my cascade of splitters crosses the empty space between my 2 6-lane sections, thats where it clogs up. This I could actually fix this by setting an input priority on the lane closer to the top, where the belt goes out. Like this:

1

Copper Lines mk2.png (1.39 MiB) Viewed 7231 times

I still do not understand the actual mechanisms behind all of this, but from what I can tell, it works quite flawlessly, the furnaces are now all producing, and my consumption has gone up accordingly to almost exactly 12 red belts worth, so its not stuck anymore.

About the point to put the cascade after the splitter, I actually preferred it this way since it wouldn't matter at all if the topmost belt is full. As long as all belts have more than 1 full belt of throughput combined, I got full throughput into my assemblers. So unless the input is so low that all 12 belts cant merge into one full belt anymore, it works at 100% efficiency. (After thinking about it, I DO have cascades after every splitter going off. Every Cascade coming before a splitter comes after the previous splitter. It literally doesn't make a difference)

When it comes to the unbalancing issue with the iron-half belts, I was thinking of that, but I wondered if there was some setting that was a bit cleaner. I actually already fiddled with all kinds of belt balancing after it leaves the bus, but couldn't ever get that to work. Guess I'll have to make a blueprint for an absolutely hideous balancer-cascade. Oh well.

Thanks for the advice!

[zOldBulldog]

A lot of the Factorio old-timers are so accustomed to balancer structures that they feel uncomfortable with the lane shifting concept even though nobody was ever able to say why. I didn't like the "balancer" approach because it allows all belts to dwindle to "less than full belt" conditions, and I like to feed my assemblers full belts.

I am glad that I'm not the only one enjoying this approach.

[Hedning1390]

The splitters in the first image are going in the wrong direction. They should start with the one pulling off, and then one by one fill the empty space left by the one before.

There's also no reason to go all the way across. Take a few lines at a time then cut some off when there's nothing left.

There's little point in lane balancing after an already balanced furnace layout. Balance before the assemblers using the line instead, and make sure it is input balanced. That said there's often no need to input balance. It is only needed if some furnaces are output blocked while some assemblers are without plates. Usually assemblers clear one line first, so it looks unbalanced, but it's not a problem because the reason they don't eat both lines is because they don't need it.

[ColonelSandersLite]

Secondly, despite my best efforts, some lines got heavily unbalanced. This happened mostly to my iron lines, and I would see literally all 12 of my lanes running at half a belt each, all on the same side. This eventually caused a shortage since it couldn't supply full belts. This is how that roughly looked like:

4

The attachment Iron halfes.png is no longer available

I got belt balancers after my furnace setups, and every time a belt leaves the bus, so I'm at a loss here aswell.
So, if anyone knows why these issues are happening and has a solution for them, I'd be happy to hear it.

This is caused by inserters taking more from one side of the belt than the other. It can have a sort of cascading resonance effect down the line. It's easily solved.

Consider this example array:

20180522091231_1.jpg (597.96 KiB) Viewed 7165 times

These gear machines pull roughly 3/4 of their iron plates from the right side of the belt.

Simple Fix:

20180522091327_1.jpg (596.73 KiB) Viewed 7165 times

This fix shunts gears from the left side of the belt into the newly formed gaps on the right side of the belt. In this example, it's not *perfectly* balanced, but it's pretty close.

There are ways to improve that ratio to perfection, but it's bulky expensive and yields little.

[Zavian]

That "balancing" is purely cosmetic. It does not improve throughput.

[Jap2.0]

That "balancing" is purely cosmetic. It does not improve throughput.

For output it does (if you saturate your belt), input it might help slightly if you're input starved, so the inserter has less time to wait between items.

[Hedning1390]

If you build enough to need exactly a full belt then it will consume a full belt and nothing will starve. If an inserter has to wait all that means is you are no longer consuming a full belt, hence things will back up slightly and the inserter will not have to wait any longer.

[ColonelSandersLite]

@Zavian
It stops the problem described in 4 above. If you want to assert that 4 above is purely a cosmetic issue, I can grant that point for the sake of discussion. That doesn't mean that the question actually asked should not be answered though.


@Hedning1390
Not only is that a situation that is not specified by the OP, it's clearly not the use scenario being discussed. Why is that incredibly obvious? Because that half belt cascade thing doesn't happen when an array is consuming the whole belt. It only happens when it is consuming less than a belt.

[Zavian]

By point 4 I think you are referring to half full iron belts. But unless you are deliberately blocking one lane, unbalanced load does not cause a throughput problem. Yes, some people think it looks awful, and it definitely looks like belt is unbalanced, but in practice it does not restrict throughput. Hence why I said "That "balancing" is purely cosmetic. It does not improve throughput". The only times you need to "fix" "unbalanced belts" is if you have too many inserters all loading onto the same side of a belt, and when combining belts. eg going from 2 belts to one belt, and at least one of the input belts is unbalanced.

Watch https://www.youtube.com/watch?v=vWwYbf-CoYA . Ideally watch the whole video, but if you are short on time skip to the 4 minute mark and watch the "As unbalanced load as possible segment". Also the top comment by ShredGuy99 is a good summary.

Cliff Notes:

1) Balanced belts do not require Balancing.

2) A fully compressed belt will not become more compressed (allow more throughput) through the use of balancers.

If you still don't understand, then tomorrow I can try to explain in other words why fixing the lane balance of the iron plates is unnecessary. (There is some merit in balancing the gear wheels on the output belt. One full belt of iron plates makes exactly one full lane of iron gears, (without productivity modules), so without something the 8th assembler will probably be output restricted, but what you posted doesn't actually balance the output belt. You are better off just using a lane switch after the fourth assembler, then if needed you can add a lane balancer at the end. Hint: they probably actually don't need lane balancing).

Edit : The problem in the first post is a result of a horribly split-off strategy. Don't do that. It is just horrible. If you want to use the priority splitters then here is a much better pattern for splitting off, including a full lane balance and a full balance at the end. Note that I have deliberately setup the load to create as unbalanced a load as possible.

Split.png (1.86 MiB) Viewed 7088 times

[ColonelSandersLite]

But unless you are deliberately blocking one lane, unbalanced load does not cause a throughput problem. Yes, some people think it looks awful, and it definitely looks like belt is unbalanced, but in practice it does not restrict throughput.

As already stated, I grant the premise that it's entirely cosmetic. I did that specifically so that neither one of us would have to go to the work of arguing the point.

The guy asked how to fix a specific problem. I provided a simple method. Whether fixing the problem will actually improve throughput or just improve aesthetics is irrelevant to the question actually asked. My own guess is that it will just make his underlying supply problem more visible, but that isn't without merit.




Edit:
If you do *really* want to debate whether or not the simple lane balancing trick I showed can effect throughput, explain:

20180523055505_1.jpg (784.11 KiB) Viewed 7063 times

The circuit network shows the item's per second for each test setup. A is the top setup, with B being the bottom setup.

[mrvn]

Thanks for the advices, I have already been able to identify the copper issue, alltough I still don't understand why it happens. It turns out, when my cascade of splitters crosses the empty space between my 2 6-lane sections, thats where it clogs up. This I could actually fix this by setting an input priority on the lane closer to the top, where the belt goes out. Like this:

1

Copper Lines mk2.png

I still do not understand the actual mechanisms behind all of this, but from what I can tell, it works quite flawlessly, the furnaces are now all producing, and my consumption has gone up accordingly to almost exactly 12 red belts worth, so its not stuck anymore.

About the point to put the cascade after the splitter, I actually preferred it this way since it wouldn't matter at all if the topmost belt is full. As long as all belts have more than 1 full belt of throughput combined, I got full throughput into my assemblers. So unless the input is so low that all 12 belts cant merge into one full belt anymore, it works at 100% efficiency. (After thinking about it, I DO have cascades after every splitter going off. Every Cascade coming before a splitter comes after the previous splitter. It literally doesn't make a difference)

When it comes to the unbalancing issue with the iron-half belts, I was thinking of that, but I wondered if there was some setting that was a bit cleaner. I actually already fiddled with all kinds of belt balancing after it leaves the bus, but couldn't ever get that to work. Guess I'll have to make a blueprint for an absolutely hideous balancer-cascade. Oh well.

Thanks for the advice!

Without the input priority here is what happens:

You have a splitter with 2 inputs and one output. Items arrive on both inputs and initially both are full red belts. So half of each is put onto the output belt. That means the top belt backs up to the furnace. The bottom backs up to but that causes the previous splitter to keep the copper on the original belt it came from.

By setting an input priority you ensure that the top belt runs through and only gaps are filled from the bottom belt.

[mrvn]

But unless you are deliberately blocking one lane, unbalanced load does not cause a throughput problem. Yes, some people think it looks awful, and it definitely looks like belt is unbalanced, but in practice it does not restrict throughput.

As already stated, I grant the premise that it's entirely cosmetic. I did that specifically so that neither one of us would have to go to the work of arguing the point.

The guy asked how to fix a specific problem. I provided a simple method. Whether fixing the problem will actually improve throughput or just improve aesthetics is irrelevant to the question actually asked. My own guess is that it will just make his underlying supply problem more visible, but that isn't without merit.




Edit:
If you do *really* want to debate whether or not the simple lane balancing trick I showed can effect throughput, explain:

20180523055505_1.jpg

The circuit network shows the item's per second for each test setup. A is the top setup, with B being the bottom setup.

Unbalanced consumption has no effect if all you ever do is use half a belt. At each split half a belt is consumed and the priority splitters will fill back the used up lane. After 12 splits all 12 belts are half filled. From then on you can only get halt belts split off. If a later station needs a full belt then you are out of luck.

So you need to draw balanced inputs from the bus or balance the lanes on the bus. With the bus split in the middle you can side fill one lane from the splitters going across the gap to refill the empty lane. Later it helps to split from the top lane, tunnel it under the bus, side balance it and feed it onto the bottom lane.

[ColonelSandersLite]

I'm not sure what any of that has to do with the price of tea in china or why you're directing it at me.

If it's meant as an explanation as to why Array B is a full throughput array, while array A is only getting about 95% throughput, you have made multiple mistakes in your reasoning.

For starters, those splits do not draw a half a belt. They draw very nearly full belt, which is why the whole 8 lane bus is nearly exhausted after 8 splits.

If you're talking about something else, I just don't know what it is.

[mrvn]

I'm not sure what any of that has to do with the price of tea in china or why you're directing it at me.

If it's meant as an explanation as to why Array B is a full throughput array, while array A is only getting about 95% throughput, you have made multiple mistakes in your reasoning.

For starters, those splits do not draw a half a belt. They draw very nearly full belt, which is why the whole 8 lane bus is nearly exhausted after 8 splits.

If you're talking about something else, I just don't know what it is.

Your arrays do not consume full belts. The inserters preferably take only from one side and you don't have enough inserters to consume the full belt. So one side is consumed more. Accordingly after a number of splits the remaining items on the bus are all on one side.

For array B the splitters and side loading mitigate this since it sorts the consumption somewhat. That explains your extra 5% throughput. It still consumes more from one side of the belts than the other as you can see at the end of your bus.

[dog80]

When you take off from the bus like that with splitters you must make sure that you also set the "input" priority. So, when you want to take of to the left, on every splitter you set INPUT AND OUTPUT to left. otherwise when the lane on the very left is full and the 2. too it will limit the 1. lanes throughput to 50% overall...

[ColonelSandersLite]

Your arrays do not consume full belts.

I said nearly.

Notice what's left on the belt between split 1 and 2? 4 plates. That's not even close to just using half a belt.

Did you notice the seconds timer in that circuit? 1.2k seconds. 20 Minutes. The arrays ran for about 10 minutes prior to starting the circuit in order to fully stabilize. The accumulation you see at the end is just a few stray plates that accumulated over the course of about half an hour. A minor tuning issue that was overlooked in favor of an expedient example. Array A took just a couple of minutes to fully back up the way it did.

I'm still not particularly sure what you point is. The point of the demonstration was to show that simply shunting iron from one half of the belt onto the other half can help bus throughput in some situations. You don't seem to be contesting the point.

Honestly, I suspect this is just a language barrier issue.

[Hedning1390]

Here's a tip: Cycle and Merge. What I mean by that is if you have 8 belts first take from belt 1, then 2, etc. For tiny needs pull from belt 8. When all your belts have been used merge them pair for pair, then shift the items with splitters. Perhaps you notice you can merge even more before shifting if most are pretty empty. You'll need about 0 splitters compared to what you use now.

Here's an even better tip: Turn and terminate. Instead of splitting off just turn it straight into the production line. What is your deal with having 8+ belts but anything pulling off only needing fractions of a belt anyway? Where's those belts on the bus even going? To just end at nothing? What's the point of that?