@Eradicator:
Pickup speeds from belts are counter-intuitive at the moment. Inserters always pick up from the nearest belt lane first.[EDIT] Inconvenient, when outer lane provides more i/s...
Example: (moving average contraption stolen from vanatteveldt here)
belt to chest timings.png (859.54 KiB) Viewed 8573 times
Re: How to load 4 blue belts per wagon?
Posted: Wed Sep 11, 2019 4:08 pm
by mrvn
A few cases are missing. How does a straight belt with just left or just right lane filled and curves with just inner or outer lane filled (both directions) compare?
Re: How to load 4 blue belts per wagon?
Posted: Wed Sep 11, 2019 4:20 pm
by eradicator
@disentius
How long of a timspan is that the average of? It seems very unlikely to me that an inserters grabbing from a curve with both lanes filled is 5% faster than one where only one lane is filled. The -30% of the "direct belt" case is pretty shocking too.
@disentius
How long of a timspan is that the average of? It seems very unlikely to me that an inserters grabbing from a curve with both lanes filled is 5% faster than one where only one lane is filled. The -30% of the "direct belt" case is pretty shocking too.
Remember: inserters chase items.
The straight case has the inserter chasing items along the length of the inserter so it has to shorten and lengthen. That's painfully slow. So those numbers didn't surprise me that much.
Another thing to test: yellow, red and blue belts in comparison. How much doe the belt speed affect the difference for each configuration?
Re: How to load 4 blue belts per wagon?
Posted: Wed Sep 11, 2019 7:40 pm
by disentius
@Eradicator:
20 minutes on 200x game.speed
Multiplying count by 6000 to get 2 decimals and output per second
Smoothing factor -> 8000
Please verify my measurements , and the setup. I could be wrong
[EDIT]
And because of this:
inserter pickup detail.gif (1.22 MiB) Viewed 8533 times
Its faster, i think.
@mrvn:
Correct, not all variations tested. That is why wrote: "Example"
Re: How to load 4 blue belts per wagon?
Posted: Wed Sep 11, 2019 8:16 pm
by disentius
@mrvn: it gets weirder.
( btw, these values fluctuate about as much as the value in the animated gif above.)
belt to chest timings2.png (1.53 MiB) Viewed 8527 times
Re: How to load 4 blue belts per wagon?
Posted: Wed Sep 11, 2019 8:22 pm
by eradicator
Just noticed that there's an extra belt on the 1533 one that the 1600 one doesn't have. Thus the 1600 has to do less work chaising the items because the possible area for items to grab within the source tile is smaller on the inner corner of a turning belt than it would be on a straight belt. That actually makes it *less* weird for me. It all looks very consistently like inserters spending more time on retract/extend are significantly faster (as rotate isn't as slow as rotate/extend).
Re: How to load 4 blue belts per wagon?
Posted: Wed Sep 11, 2019 8:40 pm
by eradicator
1600 quite obviously benefits from the second lane. 1333 refuses to grab from the inner corner.
screenshot_60812.jpg (466.78 KiB) Viewed 8524 times
Evil twist: The mere *existance* of more belts makes it slower.
screenshot_191853.jpg (114.05 KiB) Viewed 8524 times
Re: How to load 4 blue belts per wagon?
Posted: Wed Sep 11, 2019 9:05 pm
by disentius
Yep.
Thanks for taking the time to test the test setup.
Do you think we should we report this as a bug?
Yep.
Thanks for taking the time to test the test setup.
Do you think we should we report this as a bug?
Didn't test it so much as use it. Can't guarantee the numbers are correct. Also you're using the nixie tubes mod with the lower download count *gasp* :p.
Nah. Inserters are batshit crazy like that. It was also on the "controversion opinions" FFF about "disabling inserter chasing". They have all kinds of - literal - edge cases, depend on the direction you built them, etcpp.
Re: How to load 4 blue belts per wagon?
Posted: Thu Sep 12, 2019 10:15 am
by eradicator
Btw the wiki page on inserters has a blueprint for timing the inserter directly. Using that i measured that NWE inserters on the inner curve take 45 ticks per swing, so (60/45)*12=16i/s. While the S inserter takes only 43 ticks for a whooping 16.74i/s.
screenshot_400642.png (469.53 KiB) Viewed 8498 times
I also tried to abuse clocked inserters to get implicitly balanced loading, but misearably failed - only reaching 80i/s total throughput. With the old belts that would've been a success :P. With a double clock it might still work.
screenshot_401097.png (413.1 KiB) Viewed 8498 times
Re: How to load 4 blue belts per wagon?
Posted: Thu Sep 12, 2019 2:30 pm
by Vegemeister
I remember reading something from the Bob's Inserters discourse about inserters with configurable pickup and drop positions being fastest when both are as far as possible from the inserter base. Apparently the inserter speed limit applies to angular speed, or something like that.
The best vanilla inserter pickup configuration I was able to find was deadend sideload onto yellow belt, at 37 ticks/swing, like so:
deadend-sideload.png (151.81 KiB) Viewed 8484 times
The sideload lets the inserter pick from both lanes at the same time, and the yellow belt makes the items easier to chase. Either replacing the yellow belt with blue, or reversing one of the belts, slows the inserters to 38 ticks/swing.
The fastest dropoff I know, priority splitter from the side, is also 37 ticks/swing, so that's probably close to the limit.
With some tweaking, this method yields a 2-belt-per-side loader with only one splitter in the item path:
4-belt-loader.jpg (121.94 KiB) Viewed 8484 times
Belt buffer lengths between splitters and sideloads cannot be any shorter. Maximum chest imbalance is 1.546 to 1.
Just ramming a wall of blue belts into a yellow on the side of the train like this gets ~6821 items/minute, which is a little over 2.5 blue belts of throughput:
moar-belts.jpg (128.97 KiB) Viewed 8484 times
So theoretically 5 belts/wagon could be done without extra chests. But I woudln't try it with 5 actual belts. It'd work great if you have an infinite-productivity-powered ore patch with a wall of belts coming off it already, though.
So as far as i know the "trick" is to reduce the extention time by having the pickup and dropoff points be equally distant from the inserter base. Because it's not one motion but two:
Pickup -> Extend -> Rotate -> Retract -> Dropoff
Would be kinda nice if the inserter could retract *while* rotating...
During my trials i noticed that inserters actually change their "preferred lane" sometimes, when the belt becomes completely empty (or deconstructed) for a few seconds.
I actually tried my luck at circuits too, but was too impatient/stupid to find a working solution. It certainly looks neat how the inserters swing in perfect sync. BUT, when i attach my belt-speed-o-meter to that i measure the total belth throughput to an actual cargo wagon as "only" 105.1i/s, and not ~116i/s like the reddit post says. Furthermore if i *remove* the circuits i measure 110.5i/s. Though i wouldn't be suprised if the circuits actually make it slightly more UPS friendly. And they most certainly make it more balanced.
Ofc i can't guarantee that my counter is even correct, so if anyone wants to poke it. Number of active lamps = Number displayed. Averaged over 60s,15s,1s respectively top-down.
I suspect problems with your counter :D
Mine confirms the numbers VenditatioDelendaEst mentioned.
I just measured my counter on a double-loader full-throughput blue belt, and it shows precisely 45 as expected. I then went ahead and built 6 copies of the full blueprint. 2 Copies run at 116.7i/s, 3 at ~114i/s, 2 at ~112i/s. Some of the copies are rotated, but i don't see a pattern regarding speed. Somethings going on :p.
