Print

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

[epdtry]

I've come up with a design for a perpetual motion machine which I haven't seen used before:


The left side shows the upper layer; the right side shows the lower.
[Edit: The pump pumps water from the north]

After building this, fill the lower level to exactly 7/7 and wait for the top level to dry out (so no water will enter the system once it's been started).  Have a dwarf run the pump manually, and it should start producing power immediately.  This design produces 100 power and consumes 25 to keep itself running, so its net power output is 75.  The power output seems to be reliable; I checked every tick for 1,000 ticks and it never stopped running.

This machine has some interesting properties:

• Once it starts running, the pump's input tile always appears to be empty, while the output tile and the other tiles on the lower level always appear to be at 7/7.  Despite the apparent lack of motion, there is (almost) always flow in the two tiles below the waterwheel.  (Testing a similar design with only one tile beneath the waterwheel showed that a single tile will lose flow for one out of every 10-30 ticks.  I have never seen the waterwheel lose power with two tiles beneath it, though I'm not sure why this should be the case.)
• I currently believe that water will never evaporate from this system, which should be useful for low-water maps.  Testing shows that when water spreads from a 7/7 to a 0/7, the source becomes 3/7 and the destination becomes 4/7.  Since no tile is at 1/7 even temporarily, water should be unable to evaporate.
• This machine produces an extremely small amount of lag.  Running a single instance showed no significant drop in framerate (Pentium 4 3.2GHz; 32 FPS idle).  Running six machines using similar designs caused the framerate to drop by only 1-2 FPS.

As of 2 Mid-Spring, 47, my fort's power plant is using similar devices to produce 597 net power from a 15x15 room, set up in such a way that it will never need to be restarted even if the output is overloaded.   I also have two replacement designs to test: one based entirely on tested principles which produces 964 net power, and another based on one untested principle which outputs 1186 net power.  After I test these, I plan to post them here (assuming they work).

If you have any questions, comments, or clever perpetual motion designs, please let me know!

[ein]

Looks useful.
One question though, what direction is the pump going?

[epdtry]

Looks useful.
One question though, what direction is the pump going?

*facepalm*
Yeah... I forgot you can't tell that from the picture.  It pumps from the north.

[ein]

Thanks, that's what I thought, but wasn't sure.

[Viprince]

assuming you did extra testing with the large scale power production, How does the system react if you mirror it, a waterwheel on each side, but keep the one pump? I'm not sure if both waterwheels will work with half the water going both ways each cycle or if they'll alternate...

[Grimlocke]

Thats a desing pretty similar to the one I usualy go with for smaller amouths of power required, though mine is slightly smaller. Like this:

1st  2nd

####  ####
#77#  # w#
##7#  #PW#
#7##  +pw+
#R##  +h++
####  ++++
Where # are walls, + is floors, R is water covered ramp, 7 is 7/7water, H is a floor hatch, P is a pump, W is waterwheel and where the capital letters indicate an impassable building tile. The pump pumps to the north.

It lags minimally, can have another waterwheel without losing power or stability, and is easy to set up and connect to power.

[epdtry]

assuming you did extra testing with the large scale power production, How does the system react if you mirror it, a waterwheel on each side, but keep the one pump? I'm not sure if both waterwheels will work with half the water going both ways each cycle or if they'll alternate...

It (somewhat mysteriously) works just fine if you mirror it.  The only problem is, once both waterwheels are in place, there's no way for a dwarf to get to the pump to start it.  You just have to add the second waterwheel after it's already running, and then be careful not to overload it so you never have to restart it.

Thats a desing pretty similar to the one I usualy go with for smaller amouths of power required, though mine is slightly smaller. Like this:

1st  2nd

####  ####
#77#  # w#
##7#  #PW#
#7##  +pw+
#R##  +h++
####  ++++
Where # are walls, + is floors, R is water covered ramp, 7 is 7/7water, H is a floor hatch, P is a pump, W is waterwheel and where the capital letters indicate an impassable building tile. The pump pumps to the north.

It lags minimally, can have another waterwheel without losing power or stability, and is easy to set up and connect to power.

Great - this is the kind of thing I was hoping for when I requested "clever perpetual motion designs!"

It looks like your generator and mine work pretty much the same way, though I had never really considered using the impassable square of the waterwheel as part of the containment for the pump's output.

Have you checked frame-by-frame to see if the power output is consistent?  I would be a little surprised if it was, since this similar design I tested loses power for about one tick in 20:

(pumps from north to south)
Losing power occasionally isn't much of a problem for running pumps, but for millstones, losing power for even 1 tick will cancel all the jobs.

[MrFake]

How many instances of this setup have you tried, epdtry?  That would be fantastic if that consistency in power generation is repeatable.  Usually, pumping a 7/7 system like that only randomly creates flowing tiles in the 7/7 water.  Maybe you've found a way to force them to flow every time.  It should converge to a consistent flow no matter what, but DF doesn't always play fair like that.

Actually, it should be possible to put in any number of waterwheels in the same system, so long as they're all over the same body of 7/7 water.  Just add a pump between every, say, 4 waterwheels.  Manually start one or two pumps and that should be enough to kick the whole thing off.  Getting all that water to flow consistently, though, would take a lot of trial and error.  Once it's going, that's a nice compact power generator churning out thousands of dwarfwatts.

[happydog23]

Thats a desing pretty similar to the one I usualy go with for smaller amouths of power required, though mine is slightly smaller. Like this:

1st  2nd

####  ####
#77#  # w#
##7#  #PW#
#7##  +pw+
#R##  +h++
####  ++++
Where # are walls, + is floors, R is water covered ramp, 7 is 7/7water, H is a floor hatch, P is a pump, W is waterwheel and where the capital letters indicate an impassable building tile. The pump pumps to the north.

It lags minimally, can have another waterwheel without losing power or stability, and is easy to set up and connect to power.

My brain is short of dwarfwatts today, but do the ramp and hatch actually serve a purpose in this design?

[epdtry]

How many instances of this setup have you tried, epdtry?

My fort's power plant as of 47/02/01 is running 6 pumps and 10 waterwheels.

That would be fantastic if that consistency in power generation is repeatable.

I've set up a waterwheel in a configuration similar to the one in my first post and checked its power output each tick for 1,000 ticks.  It maintained consistent power output throughout.  This does not guarantee 100% consistent output, but it should be pretty close.  (By the way, does anyone know of a program that can read from DF's memory whether a given tile has flow or not?)

Usually, pumping a 7/7 system like that only randomly creates flowing tiles in the 7/7 water.  Maybe you've found a way to force them to flow every time.  It should converge to a consistent flow no matter what, but DF doesn't always play fair like that.

The system creates mostly consistent flow in the tiles that are diagonally connected to the pump's input tile.  (A waterwheel over a single diagonally-connected tile will stop producing power for one tick out of every 10-30).  Here's my current theory regarding how this works (spoiler-tagged for length):

Spoiler (click to show/hide)

First of all, make sure you understand the three rules for water and the three rules for pumps on Kanddak's Hydrodynamics Education post.  I refer to the rules for non-pumped water as W-1, W-2, and W-3; the rules for pumped water are similarly P-1, P-2, and P-3.

The designs I've come up with (and Grimlocke's design, I think) take advantage of the restriction on diagonal movement in rules W-2 and P-2 as well as the fact that water moving according to rule W-3 seems to create flow in the source tile almost all the time.  Here is my hypothesis regarding the process used in my designs:

• On manual startup, the pump moves the 7/7 water in its input tile on the lower level to its output tile on the upper level. (Rule P-1)
• The water in the output tile cannot move under pressure because there are no orthogonally-connected empty tiles on the lower level (Rule W-2), so it stays in place.  There is currently no water in the pump's input tile, so the pump does nothing.
• At some point, the water in one of the tiles under the waterwheel spreads out diagonally into the pump's input tile (Rule W-3).  This creates flow in the source tile, which appears to persist for an unknown number of ticks.
• The pump now collects the water from the input tile and moves it according to Rule P-2.  The output tile is full, so the water is teleported through the 7/7 tiles until it reaches the source tile from step 3.  This refills the source tile back to 7/7.
• The process repeats from step 3.  There is now 7/7 water with flow beneath the waterwheel, so it powers the pump from this point on.

I believe that steps 3 and 4 happen in that order during a single tick, which is why no water appears in the input tile.  A pressure plate in the input tile never stops detecting water, which supports this hypothesis.

Furthermore, I suspect that using a single diagonally-connected tile per waterwheel fails because during the tick when the water in that tile spreads out, the water in the tile under the waterwheel is at 3/7 when the waterwheel performs its checks.

Actually, it should be possible to put in any number of waterwheels in the same system, so long as they're all over the same body of 7/7 water.  Just add a pump between every, say, 4 waterwheels.  Manually start one or two pumps and that should be enough to kick the whole thing off.  Getting all that water to flow consistently, though, would take a lot of trial and error.  Once it's going, that's a nice compact power generator churning out thousands of dwarfwatts.

As mentioned above, the only tiles with flow are those that are connected diagonally to the pump's input tile.  There can be at most four of these per pump, and each waterwheel must be placed over two of them to get consistent power output, so you can have at most two waterwheels per pump.

[MrFake]

One of us missed a key point, I think.  You can set up a flow without worrying about whether water actually moves from one tile to another.  It's how waterwheels over rivers work.  I guess it's not really a clever PPM, though, since the design is based on a glitch.

[gtmattz]

One thing which I have done in the past to get a good non-interrupted flow is to use a PPM to power a pump or 2 which circulate water in a larger system which you use to power water-wheels.  Sort of a power multiplier system, albeit an overly complex dwarfy one.

[epdtry]

One of us missed a key point, I think.  You can set up a flow without worrying about whether water actually moves from one tile to another.  It's how waterwheels over rivers work.  I guess it's not really a clever PPM, though, since the design is based on a glitch.

Sorry, I seem to be missing the point of this post.  Can you please elaborate?

One thing which I have done in the past to get a good non-interrupted flow is to use a PPM to power a pump or 2 which circulate water in a larger system which you use to power water-wheels.  Sort of a power multiplier system, albeit an overly complex dwarfy one.

Yep, that's how my current power plant operates.  In fact, I even used the exact term "power multiplier" in my notes while developing it .  I could never get the normal "pump water around a loop" system to work reliably, though, so I built a few multipliers based on the same diagonal-flow principle as the generator in my first post:

 
Pumps pump from the north.  Red arrows mark input axles; green arrows mark output axles.

[freeze]

Brilliant! But more importantly, which tileset is that you're using?*

*I kid, I kid. I really do want that tileset, though.

[epdtry]

Brilliant! But more importantly, which tileset is that you're using?*

*I kid, I kid. I really do want that tileset, though.

It's a combination of tiles from two different sources, made so long ago I don't remember the details.  It looks like it's mostly the default tileset with the line-drawing characters taken from Plac1d's tileset.  I've uploaded the result here.  The colorscheme is, I think, the "Natural" scheme from the wiki.