Thread by @GeePawHill

Oxygen Not Included, my gas gathering and separating rig.

This is one of those builds that looks way way harder than it is. Had you shown me this two years ago I'd've wet my pants. Now it's one of the earliest things I do.

Let's break it down a little.

The problem: We have a lot of gas, and anything that isn't o2 just slows everything way down, cuz dupes have to work a little, run to o2, breathe a little, run back. Further, most of this gas is useful for other things, if we collected & filtered it. That's what this is.

First, look at the frame. This is a WIP shot. We have a 3-tall that's a little wide, then a bunch of 4-talls in a row stringing out next to it.

Each one of these is an infinistore, a way to store gas at nearly-infinite pressure. We want this, cuz there's hella amounts of gas.

The infinistore technique for gas is simple: put the gas vent submerged under a teensy bit of liquid. The gas comes in, sees the liquid, says, excuse me, just passing through, then moves out into the space. Only *then* does physics realize its overpressured there.

There are couple of ways to make an infinistore. The annoying part is getting the right teensy bit of liquid. See the rightmost WIP box? It's a cup of insulated tiles and one airflow tile. We use a bottle-emptier to fill first the bottom of tile of the cup, and then level 2.

But it *won't* fill level 2. Instead, it spills over the side of that airtile. But it leaves behind the perfect amount of liquid to partially cover the vent. Once that happens, we can add the second airflow tile and start sealing things up.

All of these four infinistores are built that way. The wide one is built with two of those so we can have two inputs. It also has two pumps. The narrow taller ones just have one input each.

So that leftmost guy is a slop pit. We can pump any old gas in there any Look at the piping. WTF? That's all optimization. The thing would work just fine if we just started pumping gas into the partially covered vent.

There are three optimizations. 1) don't bother putting o2 in the slop pit, and 2) if gas is going out, and gas is coming in, don't pump it twice, once at source, once in the infinistore, just let the input flow straight to the output. 3) don't spend power on electric filters.

The easiest optimization: A bridge goes from each input to right before the pump doing output. Incoming gas will want to jump that bridge if it can, and head straight out. If it can't, it'll then go to the partially covered valve. That's done in all four of the ones you see here.

The hard optimizations are the mechanical filters. The best explainer is from Tony Advanced, here.

I give you fair warning: you can use this without understanding it, but it's easy to screw up. My first 97 times I stressed and watched and made sure, and hosed it a dozen or so.

But what it does is give you an unpowered filter that never fails as long as the "pass" output isn't blocked. If the pass line *is* blocked, it'll let passing gases go out the fail line. (It doesn't matter if the fail output is blocked, that just blocks input.)

I have two not-oxygen mechanical filters at the bottom of the slop pit. I have one gas-specific filter at the top of each of the narrow infinistores.

I can build left-to-right input mechanical filters for gas & liquid all day long. Every time I have to do right-to-left I have to pull a Winnie the Pooh and sit down to think.

Think think think.

I do usually get it, tho. ;)

A key insight: this works exactly the same way for gas or liquid either one. This is a huge boon to replace liquid filters, which run very hot.

So now you get the concept. Slop pit takes anything but o2, prefers to output its input. Each filtered store takes a custom gas, and prefers to output its input.

There's a priming thing that has to happen. I am taking my Molly to the ER. When I get home, I'll show you.

So now all four are built, none in active use. Time to bring at least two online, the slop pit has to be primed to acceot o2, and the first small one has to be primed to accept co2. To prime one of these, you have to pump a few packets of the pure gas w/o any other gases.

Note: The slop pit filters o2 *out* of its input. The co2 pit filters co2 *in* to its input. The deep secret in this pic is hard to see: the gas valve has a downward pipe on both ends going to the two green bridge outputs. This forms a tiny loop of 4 segments.

That loop has in it, *permanently* now that it's been primed, 2 grams of o2, just forever running around in a circle. Why does that matter? "One tile one substance" also applies to "One pipe one substance". No gas that isn't o2 can get into that loop.

Gas runs into the first white bridge input. If wants to jump into the loop. If it's not o2, tho, it *can't*. It goes to the second white bridge input, same story. It passes up into the infinistore.

What if it *is* o2, a full packet of 1kg? 998 grams jumps, to join the 2 already there. This new 1kg packet stays in the loop, and it hits the gas valve, which is set to 2 grams. 2 of those 1kg go across the valve, 998 skip the valve and go out the back of it, to the left.

That line, out the back of the valve is the filter's "pass" line. It will only ever receive packets of the filtered gas forever. Back at the first white bridge input, we took 998g, but left 2 behind. They hit the *second* white bridge input. They jump into the loop, too.

As long as the filter's pass line isn't backed up, this will perfectly pull full gas packets of the filter's gas out the back. Even if it backs up, it will only send false *fails*. That is, o2 will come out the far side of the two white bridge inputs, even tho it's o2.

Downside: takes practice to build one from scratch, is tedious to prime, can have false negatives.

Upside: No special materials. No power consumption. Indifferent to power brownsouts or glitches. Can never give a false positive.

Compare to the regulat gas filter, which pulls 120w for every packet, but can neither false positive nor false negative, just blocks instead, or if power goes out. Or the element-sensor/shutoff filter, pulls 10w continuous, but gives permanent false positive on power glitches.

People argue a lot about the merits of all three. Trust me, people argue. But I got used to using mechanical filters, and still use them for anything permanent.