Print

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

[Kogan Onulsodel]

Measurement of Minecart Physics for a Precision Clock

Kogan Onulsodel, Dwarven Institute of Technology

I performed experiments to test the minecart physics of corners and ramps. Corners were found to subtract 1000 speed upon leaving the tile, and ramps were found to be 141420 to 141430 sub-tile positions long. These experiments are carried out in preparation for the creation of a perfect clock, which can generate a signal at any arbitrary tick in a relatively long cycle, with a cycle time which does not vary even over an arbitrary number of cycles. This device, the Arbitrary Time Oscillating Minecart Signal (ATOMS), is planned for release in the coming weeks.

INTRODUCTION:

Pioneering work by expwnent, et al., provided a basis for minecart physics shortly after its discovery. These experiments carefully quantified several aspects of minecarts, including the length of tiles, amount of friction, the speed loss for corners, the speed generated by rollers, and other essential aspects.

However, in spite of this effort, I identified two ambiguities in the published literature. First, in various places the speed loss for travelling around a corner is stated as being subtracted either at the center of the tile or upon leaving the tile. This ambiguity may lead to errors in the predicted position of a minecart after travelling through several corners. Second, it is stated that a ramp tile is 144000 sub-tile positions in length, as an approximation of the square root of 2. However, the square root of two is approximately 1.41421, and I felt obliged to investigate whether this was an error in the previous measurement, or simply an unusual feature of minecart physics.

These potential errors, while small, must be investigated and corrected before one can create the device which is the final goal of this work: a precision clock, which, after many cycles of the clock, never has a cycle which is longer or shorter than the stated length, even by a single tick. Such a precise minecart clock signal, with pressure plates appropriately placed, may be used to trigger precisely timed mechanisms. For example, an automated bolt splitting and marksdwarf training operation may be created by hooking doors to the pressure plates on this minecart clock track. This application is of particular interest as it cannot be simply performed by clocks relying on pressure plate delay, such as the classical n-step water clock, because it requires that the doors are opened and closed with time differences on the order of 10 ticks, while the fundamental resolution of the water clock is 100 ticks. While hybrid approaches may circumvent this problem, a clock whose delay is generated purely by minecart travel time is relatively simple, elegant, cheap to produce, and exceptionally dwarfy.

CORNER EXPERIMENT:

To determine the position at which the deduction in speed for traversing a corner is applied, I used a highest speed (50k) roller followed by 43 corners. This is followed by a straight track 12 tiles long and a downward ramp. I calculate that, if the speed penalty is deducted halfway through a corner tile, the cumulative effect from 43 corners will be such that a minecart will stop on the final straight track tile before the ramp, while if the speed penalty is deducted upon leaving the corner tile, the minecart will be able to reach the downward ramp. Thus, the question of whether the speed is deducted halfway through the tile or upon leaving the tile can be evaluated as whether the minecart goes down the ramp.

Spoiler (click to show/hide)

5CCCCCCCCCCCCCCCCCCCCCTTTTTTTTTTTTD
CCCCCCCCCCCCCCCCCCCCCC

KEY:
1,2,3,4,5: Roller
C: Corner
T: Straight Track
D: Downward Ramp

Upon performing this test, I found that the minecart did indeed traverse the full length of track and go down the ramp. Thus, I conclude that the speed penalty is deducted upon leaving the corner tile.

RAMP EXPERIMENT:

To determine the precise length of a ramp tile, a more carefully controlled experiment was designed. Assuming that the length would either be a rounded version of the square root of 2 longer times 100000 or 144000, as noted in previous studies, I needed to test several cases: Lengths of 140000, 141000, 141400, 141420, 141421, and 144000. To this end, I used a setup similar to the corner experiment to reduce the number of possibilities: With rollers, a carefully controlled length of track, and an upward ramp, I could test whether a ramp is longer or shorter than 141410 sub-tile positions in length. By repeatedly performing experiments with different pre-determined lengths of track, I could narrow down to a single one of the above possibilities.

I first tested whether the length of a ramp is 144000 or closer to the square root of 2 times 100000. The precise track for the first experiment consisted of a low speed roller (20k), 1 tile of track, a low speed roller, 1 tile of track, a low speed roller, 1 tile of track, a medium speed roller (30k), 6 tiles of track, a high speed roller (40k), 12 tiles of track, upward ramp and two tiles of track. I calculate that, for a ramp longer than 143640 sub-tile positions, the minecart will not be able to reach the top of the ramp, but will instead fall back down. For a ramp shorter than this, it will come to rest at the end of the track, where a wall blocks its further motion.

Spoiler (click to show/hide)

2T2T2T3TTTTTT4TTTTTTTTTTTTUTT

KEY:
1,2,3,4,5: Roller
T: Straight Track
U: Upward Ramp

Performing this experiment, I find that the minecart traverses the ramp successfully, indicating that the length of a ramp is less than 143640 sub-tile positions

The second track created for this experiment consists of the following:

Lowest speed roller (10k), 2 tiles of track, lowest speed roller, 2 tiles of track, lowest speed roller, 2 tiles of track, lowest speed roller, 1 tile of track, medium speed roller (30k), 5 tiles of track, high speed roller (40k), three tiles of track, high speed roller, 1 tile of track, high speed roller, 24 tiles of track, upward ramp, and multiple tiles of track. For this particular track, if the length of a ramp tile is greater than 141410 sub-tile positions, the cart will not be able to make it up the ramp tile and will fall back down. However, if the ramp is less than or equal to 141410 sub-tile positions, the cart will be able to reach the top of the ramp.

Spoiler (click to show/hide)

1TT1TT1TT1T3TTTTT4TTT4T4TTTTTTTTTTTTTTTTTTTTTTTTUTTTTTTTTT

KEY:
1,2,3,4,5: Roller
T: Straight Track
U: Upward Ramp

Performing this test, I found that the minecart failed to make its way up the ramp. However, after coming back down the ramp, it rolled back to the final roller in this setup. The roller pushed it back, and the second time it reached the ramp, it successfully made it up.

Given these measurements, I can safely say that the length of a ramp L is given by 141410<L<143640. Given that the square root of two is equal to 1.41421, I therefore expect that L=141420 or L=141421. Strictly speaking, the correct value could be discerned by a similar experiment, creating a setup for which a minecart will be able to traverse an upward ramp if L is precisely 141420, but not 141421. Note that, in constructing such an experiment, because all speed values exist in integer multiples of 10, L=141421 is actually indistinguishable from L=141430 without producing a ramp multiple tiles in length. However, according to the literature, the checkpoint effect places a minecart at precisely the halfway point of the tile two tiles after a downward ramp (with a one tick delay while it traverses the intermediate tile), allowing a precise reset of the sub-tile position. As such, more precise characterization of tile length should be unnecessary.

CONCLUSION:

I have characterized the ambiguities of minecart physics which I originally set out to investigate to the required precision. For future work, I will consider the development of a clock, designed in accordance with what I have learned. However, preliminary efforts at this design have revealed a further small gap in our knowledge of minecart physics: the literature does not address the question of whether friction is applied to a cart while it undergoes the checkpoint effect. This question is quite subtle relative to the question of acceleration addressed by Larix in his seminal work on the checkpoint effect, as he addressed acceleration on the order of 5000, while the friction of a track is 10. As such, in order to achieve the perfect clock for which this work is preliminary, I may further need to investigate the subtleties of friction during in the checkpoint effect, since even extremely small effects will, over many cycles, lead to a missed tick in any clock design.

[Goatmaan]

Looks like the lofty goal is rolling closer!
interesting results so far.
Keep up the good work.

  Goatmaan

[Larix]

Nice work figuring out the nitty gritty details.

As far as precise-period minecart repeaters are concerned, those aren't a lofty goal for the future: these are models that produce absolutely regular periods that can be used for building clocks and i've tried them out extensively (one of them by building a clock that kept precise time for more than a year, i.e. the switch of day on the in-game calendar co-incided with the same clock "hour" and never shifted).

The main issue with minecart repeater irregularity is this:
Carts keep track of their speed and position with very high precision, so if e.g. you have a circular track with an installed roller to maintain speed, like so:

╔═════╗
╚══┤══╝

the repeat period will typically vary by +/-1 step between passes. Basically, you have a path length that's an exact multiple of 100 000, which must be divided by the "average" cart speed (calculated to a precision of ten units). That virtually never gives a precise integer result: a cart going through this cycle of fourteen tiles with an average speed of 27680 (number drawn from thin air, but should be in the range of what you'd see with a medium-speed roller; and it's only a rough model, the per-step calculation of speed makes averaging pretty zany) would have a circulation period of 50,578034 steps; in about 578 of 1000 passes, the cart would take 51, in the other 422 cases 50 steps to make the round.

However, when a cart is subject to the checkpoint effect or collides with an obstacle, all remaining movement amounts for the turn are discarded:
If a cart checkpoints off a ramp at 100.000 speed, it doesn't matter whether it's at sub-tile 50.000 or 140.000 of the ramp at start of turn, it'll always be at the very end of the next tile at the end of turn.
If a cart hits a wall, it'll end its turn at the very corner of the tile, "touching the wall". Remaining movement rights are dropped entirely. Cart movement can be made utterly regular by bumping carts into ramps with blocked upward exit, with enough speed to travel the whole ramp length against the ramp's acceleration. In that case, the turn after collision will always start at the extreme end of the ramp tile, at zero speed. There are no increments which could carry over and accumulate into errors.

My clock-worthy repeaters all use "stop and restart" to prevent incremental aberrations, but when carts are accelerated to the ramp maximum (by impulse ramps), further acceleration will also be truncated, and you can reach precise cycles just by checkpoints. In the simplest case, the minimal top-speed accelerator coil

╔╗
▲║
▲║
▲║
╚╝

once it's spooled up to maximum speed (takes something like 400 steps), has a completely invariable period of five steps, just by "speed limit" truncation and checkpoint effect.

PS: the stop-and-restart method of removing speed irregularities on minecart repeaters can also be achieved through signal processing: stop the cart on a roller temporarily deactivated by the cart itself hitting a pressure plate, or hold it back for a bit by a floodgate/wall grate/raising bridge that once again gets opened with a fixed delay by the cart hitting a pressure plate before encountering the obstacle itself. I may be mis-thinking, but can't come up with a way the basic concept can pick up and accumulate irregularities.

[Kogan Onulsodel]

I was aware of the precise position problem. I did do some brute force calculation and found some precise track lengths which give a precise-period repeater powered by rollers, but the trouble is that it's hard to make one of a set number of ticks. I'd really like an exact multiple of 200 (so that it syncs up nicely with all of the various time delays), but I've yet to find one (you figure that 1 in 10000 track configurations will give you a precise period repeater, and 1 in 200 of those will give you a mod(t,200)=0 track, so that's about 1 in 2,000,000 tracks that will work... and I've found that you need to go to very long tracks for it, which is a bit inconvenient).

I do have an idea for a design that essentially does use the start and stop principle to accomplish my goal, I plan to write it up soon, I just haven't gotten around to working out the last couple of details and writing it up yet. Also, thanks for that link... I'll definitely be reading through to see if there's something simpler to do exactly what I want.

[Larix]

Right, i was mostly thinking out loud; i haven't spent much thought on cycle-with-forward-roller, i just more or less knew they're not perfectly regular, for the given reason.

Good luck spotting a setup that fits your specifications. That sounds much more exacting than what i'm used (and willing) to do...

A possibly interesting alternative is to get speed by "bouncing" a cart into a roller - i.e. instead of the roller working with the cart's movement, use a roller working in opposite direction (with a hook instead of a closed ring track). Interaction between carts and rollers in opposition often results in compensating for speed variations between individual cycles: a cart with very little distance left on the turn in which it moves onto the roller emerges with a comparatively large distance amount from the roller and a cart with a large remaining distance will emerge with a very small remainder (random example: a 40k strong roller, a cart rolls onto it with 10k remaining distance - it ends its turn 10k onto the roller, on the next turn moves off the roller with a speed of precisely 40k again, so leaves the roller and gets 30k into the next tile. On the next round, the cart will arrive with a larger remainder - presumably exactly those 30k - and thus travels only 10k off the roller the next time it bounces into it). I haven't looked at this thoroughly; intuitively, it should basically mirror itself exactly all the time and it empirically gives an invariable period quite often, but it may still be possible to run into +/-1 step oscillations with such a design.

[NW_Kohaku]

Sorry if this is a more basic question than you were answering, but why, exactly, does the setup of multiple rollers in your last setup matter?  I thought that a cart moving into them would simply be set to the speed of the roller, so having a high-speed roller go through one tile of track then into another high-speed roller will discard any speed past the original roller?  (Or is that purely a matter of making the sub-tile math add up to exactly what you were hoping to test?)

[Kogan Onulsodel]

(Or is that purely a matter of making the sub-tile math add up to exactly what you were hoping to test?)

That's it. I had to get it to add up just right. The friction in the tiles of regular track between the rollers gave the cart a little bit of lag in each case, and this eventually adds up to landing in just the right spot.

[kingsableye]

This is better written than a lot of lab reports I've seen at my Uni!
Nothing to contribute. I am currently recreating your experiments, however, so hopefully in the future.

[Kogan Onulsodel]

Implementation of a Precision Arbitrary Minecart Oscillator

Kogan Onulsodel, Dwarven Institute of Technology

I demonstrate a precision minecart oscillator, the Arbitrary Time Oscillating Minecart Signal (ATOMS). Utilizing multiple ATOMS, I demonstrate the Modular Oscillatory Logical Extreme Clock Unified for Labor-saving Enjoyment (MOLECULE) capable of producing signals at any arbitrary time relative to any other time in the oscillator period.

INTRODUCTION:

Minecarts have proven a powerful tool for dwarfputing and, in particular, the generation of oscillators and clocks. The ability to generate a time delay using the transit time of a minecart track allows for the possibility to generate arbitrary delays, rather than ones constrained by the time delay of triggers, generally allowing a resolution of 100 ticks. In addition, characterization of minecart physics has prepared the way for detailed, careful plans for minecart tracks.

However, oscillator designs so far, while powerful, have generally not exploited the theoretical potential of the field. Ultimately, it is possible with minecart oscillators to set arbitrary triggers, delayed by a well-defined time, chosen to the individual tick. I present an oscillator that is fully capable of producing this in a straightforward manner. The oscillator is a modular design, referred to as the MOLECULE, composed of several ATOMS. I first detail the ATOMS design below, then demonstrate the combination of multiple ATOMS to form a MOLECULE.

ATOMS:

The ATOMS design utilizes a combination of rollers, impulse ramps, and the checkpoint effect to produce a well-defined position and speed at every point in the track, which does not vary from one cycle to the next. A continuous loop is dug out to form a complete circuit. The loop begins with rollers and impulse ramps to accelerate the cart to a speed of approximately 100,000. For the rest of the loop, I alternate impulse ramps and level track such that the checkpoint effect is constantly at work.

The initial digging for the basic design with a maximum range for trigger timings T consists of a simple loop of dimensions 3 by (T+6)/2. In addition, tiles must be dug out for power distribution to the roller. These tiles must be dug from a corner parallel to the long axis of the loop. For example, for a very short version of the basic design, with T=14, a 3x10 loop is dug:

Spoiler (click to show/hide)

+OOOOOOOOOOO
+++++++++++O
O+OOOOOOOO+O
O++++++++++O
OOOOOOOOOOOO

The additional floor tile in the northwest corner is dug out to simplify the powertrain later in the design, particularly allowing two ATOMS to be placed side by side with a single gear driving them. For construction of the MOLECULE, pairs of ATOMS are always activated together, such that this design is particularly helpful.

For simplicity, I have assumed the use of constructed track ramps to create impulse ramps. However, as this can be much more labor intensive than carved ramps and track, it may be preferable to carve the ramps at this stage. If so, the design is as shown below:

Spoiler (click to show/hide)

+OO OO OO OO OOO
++▲▲▲▲▲▲+▲+O
O+O OO OO OOO▲O
O++ ▲+▲+▲+▲ +O
OOO OO OO OO OOO

Note that the design requires level floor on the corners, implying a long axis that is an even number of tiles in length. The shortest possible track for this design is 3x8, with T=10. A modified design will be demonstrated below which allows other possible lengths.

Once the initial shape of the loop is dug out, track must be carved or constructed on every tile. All ramps should be carved as impulse ramps, and 3 of the 4 corners should be carved as corners, with the input of the loop as the exception. The result for the T=14 design should be:

Spoiler (click to show/hide)

+═▲▲▲▲▲▲═▲╗
O║OO OO OO OO▲
O╚═▲ ═▲═▲ ═▲╝

with walls surrounding the whole track and all ramps impulse ramps which accelerate the minecart in the clockwise direction. A highest speed roller is then added with powertrain:

Spoiler (click to show/hide)

☼OOO OO OOOOOO
╟╢▲▲▲▲▲▲═▲╗O
O║OO OO OO OO▲O
O╚═▲ ═▲═▲ ═▲╝O
OOOO OO OOOOOO

I used a two tile long roller to simplify the powertrain, as there is virtually no additional cost to the use of the two tile roller. The roller, in this example, pushes east. A minecart is added on the roller adjacent to the 6 tile impulse ramp. When activated, the roller will push the minecart onto the impulse ramp, which will lead the minecart to quickly circulate on the loop. A pressure plate can be placed on any of the level track tiles, allowing a trigger every 2 ticks once the minecart leaves the initial stage of acceleration. Note that triggers are primarily sent at odd times, for the first pressure plate trigger defined as t=1, though a pressure plate on the final corner is guaranteed to be at precisely t=T. An additional pressure plate may be added on the final tile before the roller, though it is not guaranteed that this pressure plate will be activated precisely one tick after the previous one, if friction slows the cart sufficiently.

A simple modification allows for T=12,16,20... designs which are out of phase (i.e., pressure plate triggers are primarily at even times) with the original version when they are activated at the same time. A second tile of level track immediately after the long impulse ramp is added. Otherwise, the design is virtually identical, with a length of 3 by (T+6)/2. I will hereafter refer to this as the 2nd type ATOMS design (the original design may be referred to as 1st type). For example, the above example might be paired with a loop of size 3x11 and T=16:

Spoiler (click to show/hide)

OOOO OO OOOOOOO
O╔▲═▲ ═▲═▲═▲╗O
O▲OO OOOOOOO▲O
╟╢▲▲▲▲▲▲══▲╝O
☼OOO OO OOOOOOO

In this example, the wall with the gear assembly is shared by the two ATOMS setups.

ANALYSIS OF ATOMS:

The oscillation for the ATOMS setup above consists essentially of two stages: First, an acceleration stage, which initializes the minecart to a high speed of approximately 100,000. The ATOMS setup acceleration stage lasts a total of 12 ticks. Thus, a trigger to activate an ATOMS setup will result in a first pressure plate trigger a minimum of 13 ticks later. This has both been calculated from known minecart physics and experimentally measured on my prototype setup. Second, the ATOMS setup has a trigger stage, where the minecart passes over alternating track and ramps, with the possibility of placing pressure plates on every level track tile. This allows an ATOMS in the trigger stage to send a trigger every two ticks.

It is clear that the trigger stage can be made extremely long by extending the length of the loop. As noted in my previous paper, there is some uncertainty about the presence of friction under the influence of the checkpoint effect. As such, I can only place a lower bound on the length of the trigger stage allowed by this setup. Assuming that friction is applied as normal for tiles with track, that is, 10 deceleration per tick, and that corners also cause a reduction in speed of 1000, as normal, it will take approximately 2500 ticks for this setup to approach the minimum speed of just over 70,000 required for a minecart experiencing the checkpoint effect to continually travel at 1 tile/tick. As such, I can safely say that the maximum T allowed by this setup is likely over 2500.

Considered as a simple oscillator, ATOMS is capable of producing clock-precision signals. Because there is no corner carved into the track at the start and end of the circuit, the minecart collides with the wall, cancelling all of its remaining forward momentum from the previous cycle. The roller then restarts it, and it travels down the impulse ramps, achieving the same well-defined speed and position on each cycle of the loop. As a stand-alone oscillator, therefore, ATOMS is quite powerful, as it is capable of producing clock-quality signal with efficiency (defined as the number of ticks in which a trigger can be sent divided by the total number of ticks in a period) approaching .5 for a long loop.

MOLECULE:

In spite of the power of ATOMS, the combination of multiple instances of the device can create an even more powerful setup. I here outline a method for combining multiple ATOMS setups to create a setup which can have arbitrarily long period and generate signals at any arbitrary time in that period.

First, consider the matched paired of ATOMS discussed above. Note that at every point in the trigger phase where one minecart is on a ramp, the other is on flat track. As such, it is possible to generate a signal at any arbitrary time in the trigger phase. Additionally, note that, if a similar pair was designed for T>90 on both setups, it could be activated by a pressure plate, and if that pressure plate is not triggered again until after both minecarts return to the start position, the pair of ATOMS setups will be kept synchronized.

Finally, consider two or more matched pairs of ATOMS setups, as described above. It is possible for each in such a series of pairs to trigger the next pair, until the final pair triggers the first.

For a concrete example, I have implemented a MOLECULE consisting of four ATOMS. Each pair consists of T=100 (2nd type) and T=102 (1st type) ATOMS. A pressure plate in each setup triggered at t=88 activates the other pair. This pressure plate is physically located on the 2nd type ATOMS at coordinates (2,12) for the location of the roller for the 2nd type ATOMS defined to be (0,0). That is, it is located 12 tiles away from the final corner of the 2nd type ATOMS setup. This leads to the other pair giving its first trigger precisely at t=101. Thus, the overall setup has a precise period of 200 ticks, and a pressure plate is activated at every single tick of the cycle.

Finally, to facilitate maintenance, a pair of levers are used to disable the entire setup and to activate it. The powertrain required is shown below:

Spoiler (click to show/hide)

Pair 1:

+☼☼
☼☼☼

Pair 2:

☼☼☼
++☼

Levers:
Disable oscillator
Starts oscillator
Sets polarity for pressure plates, also linked to pressure plates

The green lever must be pulled once simply so that a pressure plate activates, rather than disables, the appropriate ATOMS. The black lever is pulled to start the oscillator, and must be pulled again immediately (within approximately 110 ticks) to prevent the first pair rollers from being always active. A latch could be implemented to simplify this operation. The red lever can be pulled to disable the setup for maintenance.

A different period setup could be implemented simply by changing the positions of the pressure plates. Longer tracks or more pairs of ATOMS could be used to make much longer periods. However, this system does have a limitation in that it will require considerably more engineering to allow much shorter periods.

CONCLUSION:

I have demonstrated a minecart oscillator that can be of arbitrary period, with triggers at any arbitrary time within that period. This setup allows perfect control for automated bolt splitting and marksdwarf training, danger rooms, traps, and any other operation which the community may choose.

[taptap]

Props for the original post, but I think your SPECIAL NAME (TM) circuit is needlessly complicated. With minecarts being deterministic isn't any circuit that at any point reproduces a state incl. subtile position clock-precise? As far as I understand your circuit works, because the cart leaves the roller tile only after hitting the wall, i.e. after each round discards any subtile movements that might otherwise aggregate with a simple roller circuit (that just accelerates the cart up to speed). because at some point each round you discard any subtile movements.

[Kogan Onulsodel]

That's absolutely true, if you just want a clock, you can make the setup much simpler. However, I also wanted to be able to send arbitrary triggers within the cycle, and while there are simplifications that require less work from my dwarfs, this is something that I can guarantee for arbitrary cycles, so I don't have to calculate a new track every time I want a different sized loop. You really have to fine tune if you want that capability for arbitrary signals. But for a simple clock... you could really simplify this a lot, just make sure the loop is the right length, and perform a quick calculation to make sure you don't get off by a step from the length of clock cycle that you want.

[Kogan Onulsodel]

In short, arbitrary function generator is more complicated than a simple clock. I'm also certain that you could do it with less dorf labor, but alternative designs that I've thought of require much more care.

Truth is, you don't generally need an arbitrary function generator. However, I decided that I wanted one, so I built it. I can think of a few traps where an arbitrary function generator really would be advantageous (for example, you could imagine a trap that opens up a gate, leaves it open just long enough for gobbos to path inside your killing chamber, seal them in, do all kinds of nasty things to them, clean up, and repeat, and the arbitrary function generator would be great for getting the timing to work well with the delay in their pathing behavior), but it's definitely overkill.

But I'm proud of the overkill. This is Dwarf Fortress, after all.

[taptap]

Imo the main problem is size and the wrong assumption that you have overcome furniture delays. You work with pressure plates, so you need to give them time to turn off before you run the next minecart across. This also forces you to a minimum size, which has very little in common with the small circuits shown here.

Have you built anything working on that principle in fortress mode? Mind showing it? (I always use imgur. Don't need an account to put up an image there.)

[Kogan Onulsodel]

Imo the main problem is size and the wrong assumption that you have overcome furniture delays. You work with pressure plates, so you need to give them time to turn off before you run the next minecart across. This also forces you to a minimum size, which has very little in common with the small circuits shown here.

Have you built anything working on that principle in fortress mode? Mind showing it?

Yes, you're right, it does have a minimum overall size. I did actually finish building it, but I don't have an image hosting account anywhere (maybe I'll get around to it someday...), so I just sort of drew my own mini versions of it in the spoilers. It works exactly as designed (I didn't post until I'd tested it), and it really just has to be much bigger than the version I originally posted. If you want, I can break down and get myself image hosting later on, actually post a screenshot, but it'll probably need to wait until tonight.

[NW_Kohaku]

FYI, tinypic.com doesn't require signing up for anything. I post things onto that since imgur doesn't show up on this forum.