Faster Belt Corners

[ssilk]

I watched now 1 hour of those videos. I like this one: https://www.youtube.com/watch?v=yU6OwsqETzI https://www.youtube.com/watch?v=_53ihfaESZ4
This is about what I think about "production streets" in Factorio: Delivering a "mixture of items".

[spacesloth]

Someone help me out. I was curious about the slow belt corners and did some experimentation. I found that the corners are actually faster rather than slower. For each corner there is a slower outside belt and faster inside belt which is what causes the loss in compression but if you have a left corner paired with a right corner then the compression loss disappears and the travel time is increased due to the corners moving items faster.

Anyway, below is a link to a video showing my experiments. If I am missing something please let me know. Thanks!

*Video removed due to invalid tests cases

[Boogieman14]

I think what you're missing is that you're not actually looking at compression. Whenever you say 'compression' in your video, you're actually looking at... let's call it synchronisation between the two lanes. When people talk about compression it isn't about comparing left and right lanes, it's about how close each item is to the previous and next item in the same lane. And that's what corners in belts actually mess up, causes the effective capacity of the belt to drop

[n9103]

Looks like you're not actually testing compression, since compression requires as little space between items (in the same lane) as possible. You just seem to be comparing one side's total travel time to the other's, which is more or less the total length actually traveled.
So, that renders all your tests in their current forms invalid for your intended purpose.

Good news is that you can make a few minor modifications and get these testing for compression.
For the most part, you can either set up a length of belt before the test, and have inserters load it up fully, and then connect the belt when ready.
I would recommend no less than 10 items per side for a decent demonstration of compression loss through turns.

Oh, and you'll need to make your measurements not on when the first item arrives, but when the last one does. In case that wasn't apparent.

[spacesloth]

Thanks for the replies! Okay I think I see how to really test compression now. Very helpful!

[User_Name]

And that's what corners in belts actually mess up, causes the effective capacity of the belt to drop

In other words, if belt is 100% loaded ("compressed"), it will deliver more items per second to the destination, if it has no bends.

Let's call this number x.

By the looks of it (just a rough estimation), single turn make belt deliver 0.85 x items per second, and 1 left plus 1 right turn drop belt throughput to 0.7 x.

"Fixing" corners with bits of faster belt helps, although I'm not sure if it is able to keep 100% compression.

If belt is not fully loaded (not compressed), turns do not matter. Use as many as you like.

Speed of delivery (where splitters shine) does not matter much.
Red/blue belts are used because of their throughput (ability to deliver X items per second to the destination), not for speed.

[sillyfly]

To add to what was already said - I think it is important to understand that what matters most is not speed but throughput. An interesting test would be to have a fully-loaded belt as suggested, and then have it go through either straight or curved belts, and see which option can transport more items in a given amount of time, or - what would probably be easier - have a set amount of items in either side (but more than just two!) and see which option can transport all of them first.

[Khyron]

I'm not sure if the slowdown in corners is actually realistic. I should probably know since I minored in logistics (esp. intra-logistics / material handling), but I don't.

It is realistic for any simple belt operating with two or more items in parallel and without item spacing, as is the case in Factorio. Items traveling around a corner belt move slower-than-normal along the inside lane and faster-than-normal along the outside lane. Without spacing, each item is speed limited by the item ahead of it. Therefore total throughput is limited to the minimum speed at any point on the lane. That's just geometry.

Perhaps a better question to ask is: How would a factory deal with this kind of througput problem. I can imagine two methods:

• Use geared (fixed-ratio) independant belts for each lane on the corner section such that each lane maintains the same angular velocity.
• Buffer at the corner entrance and exit, then pack items back-to-back on the outer lane, allowing it to carry exactly the missing throughput the inner lane couldn't take.

Bringing it back to Factorio and addressing this problem in terms of gameplay, (because it's not an absolute must to implement realism in a video game... gameplay > realism) I somewhat like that a player can observe the throughput at a corner and sense something is not right, stare at it for a while then start trying to fix it. The game does provide some ways to improve the throughput but they're not very obvious.

The burden on the player for not understanding or noticing belt throughput being limited by corners is not huge. It's like, 10% at a guess?

It is also in line with other gameplay elements, such as understanding the importance of load-balacing or how inserter preferences (such as which side of the belt it will take/put) etc. All these things fall in to: observing your factory and identifying areas of improvement. You can also take the alternate 'brute force' route and simply lay down a second layer of belts, or use faster belts everywhere (to a point).

On the whole, I would probably prefer that the belt corners continued to behave the way they currently do.

[ssilk]

What a Factorion normally searches is throughput not absolute speed. As Khyron mentioned, this depends on the smallest bottleneck in a "way" and not the fastest. And the items themselves behave quite different on the belts, if watching single item flow or on a full belt.

I made my tests with big mining fields that guarantee the needed number of items as steady flow, compress them up to one belt and then make my experiments with them and they show, that corners are slow yet. It is of course much more complex to make tests like so (that's one reason why I suggested belts, which can be turned on/off).

I recommend reading all below https://forums.factorio.com/wiki/inde ... _.2F_Bends

[User_Name]

The burden on the player for not understanding or noticing belt throughput being limited by corners is not huge. It's like, 10% at a guess?

According to this calculator, corners steal 35% of your throughput.

That means, if you have two lines of 9 steel furnaces each, feeding off the fully compressed yellow belt in the middle, you'll get 10 plates per second.
But if you have corners, you'll get only 6 plates per second from this line, because furnaces on the end will not be getting the ore.

[ssilk]

Yeah, and those numbers are taken from here: https://forums.factorio.com/wiki/inde ... ts/Physics

[User_Name]

According to the article, yellow belt with "fixed" corners still not as good as straight belt.

674-683 vs 719 respectively.
I wonder if splitting and rejoining after making the turn gives better or worse results.

[User_Name]

Did some tests.

Yellow belt with fast turns vs yellow split & rejoin.

Competitiors:


First run:
183:186
214:222

Split & rejoin wins.

Test track
http://imgur.com/1LpGKqd
All inserters receive power at once and start packing coal as fast as it arrives.
I cut power before end of the queue reaches last turn.

Obviously, the queue has to be mile long to calculate the exact difference. I don't have resources for it in the current game.

Maybe someone will do it properly.
It would be nice to include straight belt in the comparison too.

[User_Name]

By the way, just noticed that common belt rebalanced causes BIG loss of compression. I didn't make tests, but by the looks of it ~30+% of the throughput is lost at the rebalancer.

[Lee_newsum]

Did some tests.

Yellow belt with fast turns vs yellow split & rejoin.

Competitiors:


First run:
183:186
214:222

Split & rejoin wins.

Test track
http://imgur.com/1LpGKqd
All inserters receive power at once and start packing coal as fast as it arrives.
I cut power before end of the queue reaches last turn.

Obviously, the queue has to be mile long to calculate the exact difference. I don't have resources for it in the current game.

Maybe someone will do it properly.
It would be nice to include straight belt in the comparison too.

your test is invalid as the long of the beats is not the same

fist one 16 the oner one has 24

[User_Name]

your test is invalid as the long of the beats is not the same

fist one 16 the oner one has 24

I measure relative throughput, thus the length does not matter.

See here for the follow-up

[fluffy_5432]

I think that there's nothing particularly wrong with the fact that corners slow down throughput. The fact that "going straight" is faster encourages more streamlined designs and is also somewhat more realistic.

The fact that you can work around this with splitters could be considered a problem instead. It indicates that splitters are too fast. It would not be unreasonable for a splitter on a 100% full belt to cause a minor loss of throughput, something like 20%.

I think that fixing this with corners of a higher speed is ugly. Maybe there should also be an intentional throughput loss when changing belt types?

[n9103]

The fact that you can work around this with splitters could be considered a problem instead. It indicates that splitters are too fast. It would not be unreasonable for a splitter on a 100% full belt to cause a minor loss of throughput, something like 20%.

20% isn't exactly minor, but I do agree that it's silly that splitters are faster than the belt they're based on.

[User_Name]

The fact that you can work around this with splitters could be considered a problem instead. It indicates that splitters are too fast. It would not be unreasonable for a splitter on a 100% full belt to cause a minor loss of throughput, something like 20%.

20% isn't exactly minor, but I do agree that it's silly that splitters are faster than the belt they're based on.

It has nothing to do with the speed of splitter.
Split & Rejoin trick works because you split your belt in two before making turn, thus compression drops by the factor of two.
Corners reduce compression only if it's close to maximum, it is totally safe to make turn with belt which has 50% compression.
After making turn you join two belts with 50% compression into one with 100% compression again.

[Khyron]

According to this calculator, corners steal 35% of your throughput.

I was thinking 10% per lane, but even if it's closer to 20% per lane it doesn't influence my position. The corner slowdown seems like something that should remain part of the game.