Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[flameaway]

The Problem

Spoiler (click to show/hide)

So, a few days ago I started trying to figure out ways to load minecarts with water as quickly as possible.

Water loading is a very slow process that seemed to quicken only when industrial scale processing was applied.  Like flash filling a pit with a bottom covered with mining carts.   But then you have to get the things out and onto the correct path.

Sucks when you are trying to develop a fast inline system for cart filling.

http://mkv25.net/dfma/movie-2511-thisisveryinteresting

This movie is very long because I was single stepping through the action and writing down the result.  Flash to changes and then carefully examine what the movie shows. Or just watch the second movie, it shows the same thing.  I'm leaving the first in as verification that the same path is repeatable - until you change something.

This was the result.

1. Push
2. Directly to the bottom of the pit
3. Fill
4. -- ?
5. -- ?
6. Up and south.  (this little dodge is why I went to this particular setup.  You get a planar direction movement, a z-level shift and another planar direction movement onto the new level.  I was trying to cycle through the water as quickly as possible.  Just bullrushing through gives you a skip or a water coating.  This way the cart loads in two ticks - as this video demonstrates.)
7. --?  (really like to know what is going on in these pause spots.
8. --?
9. Poof!? Back down to the bottom sitting in the aquifer.  Sorry for the long delay.  I had to find the cart.  It seemed to just disappear.
10. Back to the dwarf in the original starting position.



The first eight steps are standard.  The cart always fills and exits to the south.

This is the level just above the aquifer.  What you are seeing is a 1x1 hole with a SW track ramp at the bottom.  Just to the south of the hole is a NE track ramp with a channel directly above it and a path leading left into the gun tube on the level just above.  So the cart should path up to that level.  Instead it bounces right back to the starting position.

Is it repeatable?

http://mkv25.net/dfma/movie-2512-trial2

Yes.

Ten click cart loader.

Dont' ask me why.

Lets mess with the Geometry a bit.

I removed one track ramp on the gun level.



The down ramp to the direct left of the yellow circle leads down to the the only track ramp on the transport level.

Let's put it back.

http://mkv25.net/dfma/movie-2514-trial3

No dice the magic path is gone...

I'm starting to wonder if this stuff isn't connected to the deconstruct bug.  I deconstructed a ramp and built the same type of ramp in the same place. (*edit* Upon further reflection, I think that the number of deconstructions might cause things to get progressively wierder.  Doing this type of research means building and tearing down a lot of times.  Perhaps, these strange paths that work in other ways have simply been 'tainted' by deconstruction. *edit*)

On the other hand, the behavior of the loading pit itself is constant.

So at this point I'm a bit baffled.  The tiny loader is useful (loads a cart with water and moves it out of the pit in six ticks, like clockwork.

Does anyone have any idea what's going on with the weird pathing?

Tiny Loader Explained

Spoiler (click to show/hide)

"A dwarven push starts the minecart at 19990 in the middle of the next tile, in that the first nonzero speed that the cart has is 19990."
http://www.bay12forums.com/smf/index.php?topic=112831.msg3536975#msg3536975 (emphasis mine)

The "skip" into the middle of the next tile is why the tiny loader works when a dwarf pushes the cart, and fails to work when the cart is moved through on a more 'normal' pathway.

Therefore, you can run a mining cart through a trench but until it sits on the bottom for a tick it will not fill. 

This means that a typical cart loading procedure does something like this:

1) Control input speed carefully to avoid skip.
2) Allow the cart to deaccelerate until it completely stops and hits the bottom.  If the cart doesn't stop and hit the bottom, you get a water coating on an empty cart.
3) Set the system up so the stopping point happens on a roller which accelerates the cart out of the pit.

This simply cannot be done quickly without industrializing the process, which uses a lot of carts and track.

Until the TL

When a dwarf pushes the mining cart into the Tiny Loader, the cart starts in the middle of the tile on the bottom.  Then it fills the second tick.

That's why the tiny loader is so fast.

After loading, you still have the original dwarf push; and the only path for that vector to go is back up and out of the pit.

So when you want to cycle a cart through the tiny loader you need to have a dwarf pushing or (this is the beauty part) you need to mimic the effect of a dwarf push.

Keep in mind that simply getting a cart to roll into the trench with 19900 velocity will not work (at that speed the cart will load, but you drop a whole bunch of time transiting through water.  Very much faster and the cart will skip and sink and you'll get a nice puff of mist and a headache.)

Nope, you have to get the cart to start in the middle of the bottom of the loading trench and it needs to start with enough velocity to bump back up and out of the pit.

Code: [Select]

EE SW
ww OO
EE is the entrance path into the tiny loader
SW is the track ramp at the bottom of the pit
ww is a wall
OO is the outlet from the tiny loader

You can indeed mimic a dwarf push (And build a power free fluid gun) by dropping the cart exactly where the dwarf would stand when it pushes the cart into the pit.  But, you have to put a corner track ramp (a NE in this example) that leads into the pit on that particular spot.  So, for the above example:

Code: [Select]

NE SW
ww OO

This works beautifully.  It is impossible for the cart to have too much velocity to enter and exit the tiny loader, because in this setup, velocity only matters upon exit (because what you wanted to happen, water filling the cart, happened the first thing after the cart hit bottom, literally).  So, we should be able to shove carts through the Tiny Loader just as fast as we like, with no jams due to the Tiny Loader.  In a gun setup like I detailed, the carts will likely stack up in the recovery chute and politely wait their turn through the loader.  (not tested yet).

The game simply places the cart at the bottom exactly as if a dwarf had pushed it.

This may seem odd but it is exactly the same behavior we see when a cart bumps up and over when proceeding through a 90 degree change and up a z-level.  In this situation the flow is simply going down, not up.  For the tiny loader, it's down and bump over, up and bump south.

The cart bumps down a z-level and shifts to the bottom of the pit, then the velocity is applied to lift it out of the pit moving south.

So, it is absolutely possible to build an energy free water cannon, that both loads and cycles very fast.  I'll post a working model later on today.  I have tested all the elements of one of the machines.  Including having carts cycling through a dry machine. But I haven't put the whole thing together yet. (I'm setting up good flow for the Tiny Loader from an aquifer.  The TL only a single tile hole, and if you can put five or six carts through it in forty ticks, you'll end up keeping the water level just at 5/7 a good bit of the time - which means empty loads.)  I have no idea what the rate of fire will be, but I have successfully had multiple carts running through the path right behind each other without jostling.  (All the carts accelerate in the same line, so bumping should be unusual, except in the loading area where the carts are dry.  So the TL should be able to handle several carts, especially if you set up a big one.

Here's one model.

Code: [Select]

EE SW ww
ww SW ww
ww SW ww
ww SW ww
ww SW ww
ww NE ww [x]


• = channel above this track ramp

EE = Spot the spent payload cart recovers to. This spot has a NE Track ramp on it

This is one arm of a multi level cyclotron.  The rest of the diagrams rotate around center while bumping up a z-level each time.  So, the next leg would look like this.

Code: [Select]

ww ww ww ww ww
XX NE NE NE NW
ww ww ww ww ww
            [x]

This design would move up two more levels and shoot to the left.  Then fall back onto the NE track ramp at EE.

This design exploits those extra shifts you get when bumping around z-levels on track ramps in different ways.  First the three card monte shuffle in the tiny loader, and then again through direction changes and bumping up to new z-levels in order maximize velocity (reduce ticks) in the gun barrel.  A lot of time is lost through corners when progressing on the flat.  This doesn't seem to be the case when going up and around.  (I'm not sure of this, but I'd guess a four level cyclotron with three acceleration ramps per level will be similar in power to an 11 tile linear accelerator, but with three extra levels of z.

I'd also like to point out, simply because it caused me no end of confusion, that track ramps that exist on the tiles where the game shifts a cart to during a z-level bump act only as direction ramps, not as acceleration ramps.

So, for example, in the design I've included above, each leg of the multi-level cyclotron has three track tramps that accelerate the cart - not five as it might appear. You can test this assertion by dropping a cart on a track ramp that is setup to provide impulse from a wall.  If you do so the cart won't accelerate.  However, if you fire a cart through the same path from the same level, it will.

Code: [Select]

ww PP
ww SW [x]

PP is the push through the accelerator from the same level.   That will accelerate the cart.  Dropping from a higher z-level onto SW will not.  The cart will be translated to the position south of the SW track ramp with whatever velocity it last had when moving on the level.

Keeping direction change track ramps and acceleration track ramps straight prevents a lot of weird results, not all of them though.[/code]

UPDATE:

So I built a full scale gun that rises through four zs.  Has 56 acceleration ramps, I think, So it accelerates a cart to the max speed. I've verified some of my wild assertions and am trying to pull my foot out of my mouth on others.

Doing the double shift.  That is up a z and over, does in fact retain velocity and shift direction during the move.  Here is why I think this:

When I drop a mining cart onto a NE track ramp and into the the Tiny Loader from one z level up, the cart loads.  When I run it through my big gun, the cart loads on the first trip and then the system runs dry.  The cart has built up max speed during it's first trip through the machine, using the double shift trick on the corners.  And the payload cart breaks the speed limit through the Tiny Loader and just gets a water coating on each successive trip.  Disappointing.  Now I have to rethink things at a pretty fundamental level. 

<sigh>

I need a way the speed limit the circuit during the recovery phase. 

Too bad I can't fill the recovery chute with packing peanuts.  It works for zero-g experiments. I saw it on "How the Universe Works"  How come I watch that show all the time and still don't know how anything works? Maybe I should stop watching T.V. and read science fiction. I bet that'll work. Or I could just switch to the Sci-Fi channel.  That seems more efficient.)

[rhesusmacabre]

(You haven't actually linked the first one.)

It may be important that in your first two trials the dwarf vacates the tile the minecart has to reach, but in the third trial he doesn't (or doesn't appear to anyway). A slow moving cart will in theory bump against a creature and stop dead.

[Di]

In the successful experiments when cart moves out of water we can see 5/7 water at the tile, in failed one a pair of 6/7's is formed exactly at the moment of cart ascension. Maybe it's lower initial speed didn't allow it to get as high on the ramp as in previous testing, and it didn't get as much energy when going down. Have you changed carts?
Dwarves are definitely different. It has been stated before that dwarven attributes don't have any measurable impact on the cart impulse, but maybe you've found a way to measure it.

[Urist Da Vinci]

...It has been stated before that dwarven attributes don't have any measurable impact on the cart impulse, but maybe you've found a way to measure it.

The "push" impulse on a cart gives the cart a defined velocity of 19990 in the internal numbers. It's the exact same regardless of the dwarf.

[flameaway]

***Moved to original post for clarity

[flameaway]

@RhesusMacabre

"(You haven't actually linked the first one)"

Thanks.

I've included the first movie.