"Heat Doubling" Bug Hunt

[mathmanican]

3 minutes ago, ghkbrew said:

It forms a droplet

Yep. You missed my ninja edit.  

With 4 contact points, you still get 1.5C gain.  So condensation (phasing down) doesn't appear to suffer from the extra 1.5C per contact surface. Just phasing up. 

[ghkbrew]

42 minutes ago, mathmanican said:

I wonder if by cooling gas on 3 or 4 sides with liquids or solids, can we can generate more heat than we lose.

Tested it, all tiles touching a gas gain the same amount of heat as a single tile would when condensation happens (just like we saw for freezing)

[mathmanican]

Here's another fun fact. The temp of steam will not drop more than 1.5C from this bug if doing so would drop you below 375.5 (3K above the phase change). So the drop in temp from the extra contact points seems to be clamped.  

Example, start with 377K, or with 380K, with all ceramic tiles.  Both get you this.  

Starting with 400K drops the temp to 394K. 

[ghkbrew]

10 minutes ago, mathmanican said:

extra contact points seems to be clamped.

Not exactly clamping.  The rebound only seems to happen in multiples of 1.5C

Starting with 377.1K water and 4 ceramic tiles:

Each contact point seems to check if the current temp is above TC+3 and only apply the rebound if it is.

[mathmanican]

1 hour ago, ghkbrew said:

all tiles touching a gas

I'm not sure if I see a 1.5C increase on the surrounding liquids. Instead, I'm just seeing the results of what conduction would yield, IF the gas were a liquid (so 625 multiplier). [Update: All my computations are clamped by total heat, the divide by 8 factor....] Liquid is 30g at 300K. Steam (30g) was painted at 310K in the middle tile. Conduction would yield 301.25K, which is exactly what I get from debug. 

The 310K steam should hence loose 3.75K from conduction. But it does not. 

The 60g water in the bottom is 305.75K. This means that 30g of 310.25K combined with the 30g of 301.25K water. 

One way to get this is to add 1.5C for phase change, and then subtract off a single 1.25C from conduction.  It seems completely messed up.  

 

1 hour ago, ghkbrew said:

Each contact point seems to check if the current temp is above TC+3 and only apply the rebound if it is.

These oddities we are finding should hopefully help them track the exact source of these bugs.  This stuff is pretty wonky. @Lifegrow, we need some appropriate words for all of this.  

[ghkbrew]

I think I have the algorithm figured out. 

1) All contact points calculate how much heat to send based on starting conditions.

2) In succession, each contact point modifies the temperatures of the tiles involved based on step 1.

3) after each modification, if heat was transferred, the tiles are check for phase changes.

  3a) If the tile is 3C above it's phase up point the tile is flagged to phase up (change material) and 1.5C is subtracted from it's current temp

  3b) if the tile is 3C below it's phase down point the tile is flagged for phase down and a new entity is created to represent the phased material (debris for freezing, droplet for condensation).  If this happens multiple times the entities over-write each other.

4) tiles flagged in 3a have their material changed.  tiles flagged in 3b are deleted.

 

Steps 3a and 3b are where the bugs happen, because they don't take into account the prior actions of contact points.

[mathmanican]

I repeated my experiment above with 330K steam (so I would not hit the total heat clamp EDIT: And failed....).  Sure enough, the 3 water sides touching all reached 303.75K (exactly what conduction predicts). The 60g of liquid (after the steam turns to liquid and joins) has temp 315.75K.  This is precisely the average of 303.75 and 327.75 (330+1.5-3.75).  My guess above seems validated in another setting. 

Possible conclusion: When gasses condense, they interact with ALL their surrounds as if they were already liquid, for a single tick. Then they fall (having gained 1.5C, but loosing only 1 of the conduction points.  Actually, I bet the tile they loose conduction to is the first one encountered in a top to bottom, left to right manner.  So top, then left, then right, then bottom. A ceramic tile on top should make this check easy (didn't work - no heat transfer). I tried with a 20C tungsten tile, there was heat transfer (only conduction assuming the gas were liquid, not steam, getting to 293.2362K), but the result of 315.75K was the same.  

55 minutes ago, ghkbrew said:

I think I have the algorithm figured out.

This does not account for the fact that regular conduction is occurring with superchilled steam as if it were liquid. So the phase change check somehow gets noted before conduction happens. But I think you're onto something.  Gonna check to see if making the bottom tile tungsten changes things. 

[ghkbrew]

35 minutes ago, mathmanican said:

This means that 30g of 310.25K combined with the 30g of 301.25K water.

Is it duplicating water here?

Edit: nevermind, I see 30g condensed steam + 30g pre-existing water

15 minutes ago, mathmanican said:

loosing only 1 of the conduction points

Could it be conduction based on the starting gas? (still conducting just not getting the liquid-liquid multiplier)

 

Anyway... I need to head to bed.  I'll definitely be playing with this tomorrow. If you haven't figured it all out already

[mathmanican]

49 minutes ago, ghkbrew said:

Is it duplicating water here?

30g of water on 3 sides at 300K.  30g of steam in the middle at 330K. When the steam becomes liquid, it falls and combines with the 30g below (hence the bottom becomes 60g.  I'm using 30g because I know stuff won't flow at that size. Liquid metals would provide a larger volume to work with. 

Interesting new facts. (Originally there was 30g of steam at 330K in the middle tiles)

The sides still become 303.75 (conduction of 30g liquid at 300K with 30g liquid (Update: steam) at 330K).  The newly formed liquid above the insulation tile spits out at 327.75 (as observed above when the insulation was liquid).  However, the liquid that forms above the tungsten tile is now 322.2875K.  So the LAST tile encountered seems to affect things.  If the tungsten tile is ABOVE the steam, then the temp at which the liquid appears is 327.75K. 

How do we get the 322.2875 temp. Simple. The conduction temp from a 100kg at 293.15K tungsten tile, with a 30g at 330K LIQUID water (update: steam) tile is 320.7875K. Add 1.5K to that, and we get exactly 322.2875.  

Your idea on 3b above seems spot on.  This experiment suggest that it may be the LAST tile that interacts with the phase changed liquid seems to be the one that gives the temp of the newly created substance. Just to confirm, I tried the following as well, and got the predicted 315.75K that comes from the 30g steam condensing and combining with the 30g of liquid that was already in the bottom cell. The tungsten cell in EVERY experiment hit exactly 293.2362K (the value conduction gives if the steam were water for steam).  

At least for a gas condensing into liquid, I think that you will never add more than 1.5C  from the phase change. On the other hand, you DO gain the heat from conduction with each cell touching the gas that will phase change, but only loose the heat from conduction that affects the LAST cell interacting with the gas. 

I'm pretty sure we could rig a contraption to exploit this. Not sure yet how, but we can sleep on it. 

UPDATE: I increased the volumes to 100kg of both steam and water.  The Tungsten tiles interact with the steam as though it is steam (not liquid), so my conjecture is wrong (you have to have the 25 multiplier). Similarly, the liquid interacts with steam.

@ghkbrew, my entire set of computations are all impacted by the total heat clamp (the division by 8 factor).  

[mathmanican]

58 minutes ago, ghkbrew said:

Anyway... I need to head to bed.  I'll definitely be playing with this tomorrow. If you haven't figured it all out already

Same. The dang total heat clamp in conduction kept messing up my thinking, and I think I fixed all my comments above.  It's a mess, hopefully cleaned up, and the numbers are still valuable. Your algorithm might be spot on. If so, we just need to abuse what happens in the last tile to see how much we can affect amplifying heating or cooling.

[wachunga]

3 hours ago, mathmanican said:

The dang total heat clamp in conduction kept messing up my thinking

The total heat clamp is a pretty lass you unwittingly follow into a dark alley. Whereupon you are bopped on the head by a ruffian. Sometime later you stagger into the light of day with pockets empty and a skull full of aches. I've been waylaid on too numerous an occasion.