3 and 4 way intersections

[kazaakas]

I tried running some smaller 3-way intersections on the test bench with several lengths of trains, but noticed that for certain lanes/directions, trains end up waiting a very long time before they could pass, sometimes until at the end of the test when all the traffic cleared up. Clearly, when train length is adjusted, it has an effect on the prioritization of who is allowed to pass.

Funnily enough this results in a huge score, but it's not very fair as the throughput is very unbalanced. I wonder if there is a factor for this that's worth measuring for determining whether a result is actually useful.

[Avona]

I wonder if there is a factor for this that's worth measuring for determining whether a result is actually useful.

You need to change the tpm. It's tuned for 2-4 trains so that there is a small gap that allows other trains to go. In the picture, the left circled CC is for manually changing the options. The central display is for your information. On the right tooltip, I have indicated that the train signal is the appropriate signal to change. Modify the signal until you get the desired train behavior.

[Hovedgade]

Here is yet another roundabout. I know it is symmetrical but i just couldn't resist naming it this way.
Link: Totally symmetrical roundabout

Picture

Metadata: RHT, 38x38, No buffers, Allows U-turns, 4 tile gap.

Cheers.

[RunItsATrap]

I wonder if there is a factor for this that's worth measuring for determining whether a result is actually useful.

You need to change the tpm. It's tuned for 2-4 trains so that there is a small gap that allows other trains to go. In the picture, the left circled CC is for manually changing the options. The central display is for your information. On the right tooltip, I have indicated that the train signal is the appropriate signal to change. Modify the signal until you get the desired train behavior.

Thanks for that hint, very useful.

Speaking of which. I am currently working on a buffered intersection for 5+16 trains. It only supports these directions: W->N, W->E, W->S and N->W, E->W, S->W. The number of trains per unit of time is such that in the real factory, #(W->N) = #(N->W) and so on and forth, so in the testbench, there should be as many trains coming from the west as from the other three directions together. Can that be configured/hacked somehow by me, or would that require deep shenanigans with the circuits?

[Avona]

I believe that you would need to modify the circuits to reduce the tpm on the X->W trains, although they would even out anyhow as the outbound west line would only be able to take maximum trains. I don't believe that there's a built in way to do this with the test bench and I don't have time to mess around with factorio atm. Perhaps Hans can give a better response if he sees this.

[hansjoachim]

I wonder if there is a factor for this that's worth measuring for determining whether a result is actually useful.

You need to change the tpm. It's tuned for 2-4 trains so that there is a small gap that allows other trains to go. In the picture, the left circled CC is for manually changing the options. The central display is for your information. On the right tooltip, I have indicated that the train signal is the appropriate signal to change. Modify the signal until you get the desired train behavior.

Thanks for that hint, very useful.

Speaking of which. I am currently working on a buffered intersection for 5+16 trains. It only supports these directions: W->N, W->E, W->S and N->W, E->W, S->W. The number of trains per unit of time is such that in the real factory, #(W->N) = #(N->W) and so on and forth, so in the testbench, there should be as many trains coming from the west as from the other three directions together. Can that be configured/hacked somehow by me, or would that require deep shenanigans with the circuits?

Been thinking about adding that functionality, but thought nobody would use it.
You asking for it is perfect.
It's possible now with just seting one value. On The lane, you want to limit find the combinator that reads the oil signal. Replace the oil signal with the tpm value you want times 600.

[hansjoachim]

I wonder if there is a factor for this that's worth measuring for determining whether a result is actually useful.

You need to change the tpm. It's tuned for 2-4 trains so that there is a small gap that allows other trains to go. In the picture, the left circled CC is for manually changing the options. The central display is for your information. On the right tooltip, I have indicated that the train signal is the appropriate signal to change. Modify the signal until you get the desired train behavior.

Thanks for that hint, very useful.

Speaking of which. I am currently working on a buffered intersection for 5+16 trains. It only supports these directions: W->N, W->E, W->S and N->W, E->W, S->W. The number of trains per unit of time is such that in the real factory, #(W->N) = #(N->W) and so on and forth, so in the testbench, there should be as many trains coming from the west as from the other three directions together. Can that be configured/hacked somehow by me, or would that require deep shenanigans with the circuits?

Now you can set tpm per lane/per direction. Set auto off and adjust the value on the constant combinators