Key takeaways
- Gas leaves the cylinder near 850 psi and must reach the diffuser under 45 psi, which is the entire reason the regulator and needle valve exist.
- The high pressure gauge shows vapour pressure rather than contents, so it stays near 800 to 900 psi until the liquid CO2 is almost gone.
- Bubble counts are not comparable between tanks; only a drop checker or a measured pH drop tells you the dissolved concentration.
- Aim for roughly a 1.0 pH drop from the degassed baseline, reached over one to two weeks rather than in a single evening.
- If fish gasp at the surface, shut the solenoid first and raise surface agitation before doing anything else.
Between the liquid CO2 sitting in your cylinder and the dissolved gas your plants actually use, there are eight separate components. Each one can fail. Two of them can fail in a way that empties a cylinder into a 200 litre tank overnight and leaves you netting dead rummynose tetras before breakfast.
Most planted tank emergencies I get called out to are not plant problems. They are plumbing problems with fish living inside them. What follows is the full chain in the order gas travels through it, which is also the order you should buy it, assemble it, and commission it.
The chain in order
Gas leaves the cylinder at roughly 850 psi (about 58 bar) at normal room temperature. It has to arrive at the diffuser somewhere under 45 psi, metered down to a few bubbles per second. Everything in between exists to make that reduction controllable and repeatable.
- Cylinder. Holds liquid CO2 under its own vapour pressure.
- Regulator. Drops cylinder pressure to a working pressure you choose.
- Solenoid. Switches gas on and off from a timer.
- Needle valve. Meters the actual flow rate.
- Bubble counter. Makes that rate visible.
- Check valve. Stops tank water siphoning back up the line.
- CO2 resistant tubing. Carries gas without losing it through the wall.
- Diffuser or reactor. Puts the gas into solution.
The drop checker is not part of that chain. It hangs in the tank as an independent witness, and it is the only piece that tells you whether the other eight are doing their job.
Cylinders: size, thread, and why bigger is cheaper
A 5 lb (2.3 kg) aluminium cylinder is the smallest size worth owning for anything over 100 litres. Expect around $80 to $130 empty, and roughly six to twelve months of running on a 200 litre tank depending on your injection rate.
A 10 lb cylinder costs perhaps 30 percent more to buy, but a refill is only slightly dearer than a 5 lb fill, so your cost per gram of gas drops noticeably. The catch is height. A 5 lb aluminium cylinder stands around 46 cm tall, a 10 lb closer to 51 cm and considerably fatter, and plenty of cabinets will not take the larger one.
In North America the valve outlet is CGA-320. Across much of Europe it is W21.8 x 1/14, listed as DIN 477 No. 6. Buy a regulator that matches your local standard rather than stacking adapters, because every extra joint is another place gas quietly weeps out over months.
Aluminium cylinders sold in the US carry a DOT-3AL stamp and a hydrostatic test date on the neck. Fill stations generally refuse cylinders more than five years past that date, so read the stamp before buying secondhand. Transport them valve up and secured, never loose in a footwell.
The regulator, and what its two gauges are telling you
The high pressure gauge reads cylinder contents, and it lies to you in a useful way. As long as any liquid CO2 remains it will sit near 800 to 900 psi, whether the cylinder is 90 percent full or 5 percent full, because the reading is vapour pressure rather than quantity.
The low pressure gauge shows working pressure, which is what the needle valve downstream is actually fed. Ceramic diffusers typically need 20 to 30 psi to push gas through the disc, inline atomisers usually want 40 psi or more, and reactors will run happily at 15 to 20 psi.
Budget single stage units such as the ISTA or the Milwaukee MA957 sit around $90 to $170 with solenoid and bubble counter included, and they work fine for years. Their weakness appears only at the very end of a cylinder, which is a specific enough failure to have earned a name and a well understood cause you can design around. If you are still choosing, the differences that actually matter are set out in the trade offs between single and dual stage bodies.
The solenoid is what lets you sleep
A solenoid is an electrically operated valve that shuts gas off when the timer cuts power. Without one you inject twenty four hours a day, which wastes gas and, more to the point, drives CO2 into the water all night while plants have stopped consuming it and oxygen is falling.
Most run warm to the touch, drawing somewhere between 3 and 8 watts, and that is normal. A solenoid too hot to hold is not, and usually means a failing coil. Set it to open one to two hours before lights on and close one to two hours before lights off, so the gas peak lands inside the photoperiod rather than after it.
The needle valve does the real work
This is the component people under-buy and then blame everything else for. A coarse needle valve gives you maybe a quarter turn between nothing and far too much, which turns fine adjustment into guesswork.
Proper metering valves such as the Ideal Valve or Fabco NV-55 series give several full turns across the useful range. Fitted to a decent regulator, you can move a tank by two or three bubbles per minute instead of jumping in steps of ten.
Set it, then leave it alone for at least 24 hours before judging the result. Valves also creep, so a rate that was stable in March can read measurably different in June. That is worth knowing before you go hunting phantom faults elsewhere in the chain.
Bubble counter, check valve, and tubing
A bubble counter is a small water filled chamber that turns gas flow into something countable. Fill it two thirds, using distilled water and a drop of glycerine if you want bubbles that separate cleanly instead of clumping.
The check valve stops water climbing back up the line when pressure drops at night. Many bubble counters have one built in, but they perish. Fit a second, separate check valve close to the tank and replace it yearly.
Standard silicone airline leaks CO2 straight through its wall. Use tubing sold specifically as CO2 resistant, usually an opaque polyurethane, or you will spend the year refilling cylinders to feed the room.
After assembly, brush every joint with washing up liquid and water and watch for bubbles that grow. A leak too small to hear will still empty a 5 lb cylinder in a few weeks.
Getting the gas into solution
A ceramic diffuser sits in the tank and shears gas into a mist of fine bubbles that dissolve as they rise. It is cheap at around $12 to $30, visually obvious, and loses a fair fraction of your gas out through the surface.
A reactor sits inline on a canister filter return and holds gas in a chamber until it dissolves almost completely. It costs more, restricts flow slightly, and vanishes from view. The efficiency gap between the two is real and measurable, laid out in the head to head on dissolution methods.
The drop checker is your only feedback loop
Bubble counts tell you nothing about dissolved CO2. Bubble size varies between diffusers, dissolution varies with flow and tank depth, and two tanks running the same bubble rate can sit 15 ppm apart.
A drop checker filled with 4 dKH reference solution and bromothymol blue reads the actual result, with a lag of an hour or two. Reading it correctly, including what a yellow one at 8pm actually means, is covered in what each shade is telling you about the water.
Commissioning: the first fortnight
Assemble everything dry, pressurise, leak test, and only then start gas. Begin far lower than you think you need and climb in small steps over one to two weeks, watching fish rather than plants.
- Leak test every joint before the first overnight run
- Measure degassed pH as a baseline before injecting anything
- Start at roughly half your estimated final rate
- Increase no more than once every two to three days
- Run strong surface agitation overnight, every night
- Check fish at lights off, when dissolved CO2 peaks
The working target for most planted tanks is a pH drop of about 1.0 from the degassed baseline, which corresponds roughly to 30 ppm. Getting there without harming anything is set out step by step in how to raise the injection rate over days rather than hours.
A 180 litre tank that nearly went wrong
A customer brought me a fresh setup on a 180 litre planted tank: new 5 lb cylinder, ISTA regulator, ceramic diffuser, all correctly plumbed. He had set 3 bubbles per second on day one because a forum post quoted that figure for a similar tank.
By the second evening his cardinal tetras were hanging at the surface and the drop checker had gone distinctly yellow. We shut the solenoid, dropped the spray bar to break the surface hard, and ran an air stone overnight. Nothing died, which was luck as much as judgement.
We restarted at 1 bubble per second three days later and worked up across two weeks, finishing at 2.2 bps with a solid green checker and a measured 0.9 pH drop. Same equipment, same tank, completely different outcome, purely because of the ramp rate.
When it goes wrong, act in this order
If fish are gasping at the surface, cut the solenoid immediately, then increase surface agitation as hard as the tank allows. A 30 to 50 percent water change with well aerated water helps, but agitation acts faster because it drives dissolved CO2 out through the surface.
Do not wait to confirm with a test kit. Fish gasping at the surface while gas is running is evidence enough, and the behaviours that show up before an actual loss usually appear an hour or more ahead of it. Carbon dioxide is an asphyxiant in air as well as in water, so if you keep cylinders in a sealed cabinet, read the NIOSH exposure summary for carbon dioxide and ventilate the space.
Once the tank is stable, find the cause before restarting at the old rate. Nine times out of ten it is a needle valve that drifted, a cylinder close to empty, or a diffuser that was cleaned last week and now passes far more gas than it used to.
Questions readers ask about this
Fit the solenoid. Plants stop consuming CO2 in darkness while fish keep respiring, so overnight injection pushes concentration up exactly when dissolved oxygen sits at its lowest. A solenoid adds around $30 to $60 to a regulator package and pays for itself in saved gas within a year, since you cut roughly eight to ten hours of injection every single day.
On a 200 litre tank running around 2 bubbles per second for eight hours a day, six to twelve months is typical. The wide spread comes from leaks, diffusion method and bubble size. If you are refilling every two months on a modest tank, you almost certainly have a leaking fitting, or you are using standard silicone airline instead of CO2 resistant tubing somewhere in the run.
Physically yes, and plenty of people do, but check two things first. The extinguisher must be a pure CO2 unit with no dry powder, and the siphon tube has to come out so you draw vapour rather than liquid. Liquid CO2 reaching a regulator can freeze it open. Some fill stations also refuse extinguishers presented for aquarium use.
Match it to your diffusion method. Ceramic disc diffusers generally need 20 to 30 psi to overcome the resistance of the disc, inline atomisers often want 40 psi or more, and reactors run fine at 15 to 20 psi. Set working pressure first, then tune the needle valve, because changing pressure afterwards shifts your bubble rate every time.
For most tanks it is optional, and for heavily stocked or expensive stock it is cheap insurance at around $80 to $200. It cuts the solenoid when pH falls past a set point, limiting damage from a needle valve that drifts open or a regulator surging as its cylinder empties. Treat it as a backstop, never as a substitute for correct setup.
