How long will a CO₂ tank last in an aquarium?
Days = the CO₂ in the cylinder ÷ the CO₂ you use a day, and the CO₂ you use a day is bubbles × what is in one bubble. Every chart on the subject skips that last number, and it is the one the answer hangs on. For the most common planted setup — a bubble counter making 5 mm bubbles, a ceramic diffuser in the tank, CO₂ on 8 hours a day — a 5 lb cylinder at 2 bubbles per second lasts about 165 days, 5.4 months, using 13.8 g of CO₂ a day. A 24 oz paintball cylinder at the same count lasts 49 days.
Now point the same counter at a reactor instead of a ceramic disc. Same bubbles, same count: 314 days. Through an inline diffuser, 127. Nothing about the cylinder changed; what changed is how hard the gas in the counter is squeezed, which the next two sections explain. It is why one keeper's 5 lb lasts 5 months at 2 bps and another's lasts 10 at the same count, and why a bubble count copied from someone else's tank means very little. The pages that rank for this question never say what a bubble holds; work their own durations backward and the bubble they imply runs from about 0.03 mL to 1.1 mL — one calculator assumes 2 mg a bubble, which at room pressure is a 13 mm bubble, the size of a blueberry.
| Cylinder | CO₂ | 0.5 bps | 1 bps | 2 bps | 3 bps | 4 bps | 6 bps |
|---|---|---|---|---|---|---|---|
| 95 g disposable | 95 g | 28 | 14 | 6.9 | 4.6 | 3.5 | 2.3 |
| 12 oz paintball | 340 g | 99 | 49 | 25 | 16 | 12 | 8.2 |
| SodaStream 60 L (about 425 g) | 425 g | 124 | 62 | 31 | 21 | 15 | 10 |
| 500 g (disposable or refillable) | 500 g | 145 | 73 | 36 | 24 | 18 | 12 |
| 20 oz paintball | 567 g | 165 | 82 | 41 | 27 | 21 | 14 |
| 24 oz paintball | 680 g | 198 | 99 | 49 | 33 | 25 | 16 |
| 2.5 lb | 1.13 kg | 330 | 165 | 82 | 55 | 41 | 27 |
| 2 kg | 2 kg | 582 | 291 | 145 | 97 | 73 | 48 |
| 5 lb | 2.27 kg | 660 | 330 | 165 | 110 | 82 | 55 |
| 10 lb | 4.5 kg | 1,319 | 660 | 330 | 220 | 165 | 110 |
| 5 kg | 5 kg | 1,454 | 727 | 364 | 242 | 182 | 121 |
| 15 lb | 6.8 kg | 1,979 | 989 | 495 | 330 | 247 | 165 |
| 20 lb | 9.1 kg | 2,638 | 1,319 | 660 | 440 | 330 | 220 |
| 10 kg | 10 kg | 2,908 | 1,454 | 727 | 485 | 364 | 242 |
Days of CO₂ at 8 hours a day. One bubble = a 5 mm sphere (0.065 mL) at a ceramic diffuser's line pressure of about 1 bar (15 psi) and 22 °C: 3.65 g/L, 0.24 mg a bubble. Leaks, the purge at each change and a single-stage regulator's end-of-tank dump all come out of these days; your own last cylinder is the better number (the tool's "My last refill" tab).
| If your setup differs | Multiply the days by |
|---|---|
| Into a reactor, ladder or bell instead | × 1.90 |
| Through an inline diffuser instead | × 0.77 |
| Bubbles 4 mm across instead of 5 | × 1.95 |
| Bubbles 6 mm across instead of 5 | × 0.58 |
| CO₂ on 6 hours a day instead of 8 | × 1.33 |
| CO₂ on 10 hours a day instead of 8 | × 0.80 |
How much CO₂ is in one bubble?
A bubble is a sphere you can measure. Hold a ruler against the counter and the volume is π/6 × the diameter cubed: a 5 mm bubble is 0.065 mL. The cube is the trouble. Judge it a millimeter small and the true bubble is 1.73× what you assumed; a millimeter big and it is 0.51×. So by-eye bubble counting is good to about a factor of two before anything else goes wrong, and the honest output of any bubble-count calculator is a range.
What sets the size? At a slow count each bubble leaves the nozzle when its buoyancy beats the surface tension holding it to the rim — Tate's law — so the nozzle and the liquid decide it. Turn the needle valve up and the bubbles grow as well as speed up: one counter maker gives 0.05 mL a bubble at 1 mL of gas a minute and 0.1 mL at 10, which is 4.6 to 5.8 mm. So doubling the count more than doubles the gas, and the bubble to measure is the one at the count you actually run. The liquid matters too: by the same law the nozzle makes bubbles roughly half the size in mineral oil as in water, so two identical counters, one topped up with oil, can show the same count while passing CO₂ at about two to one.
| Bubble across | Volume | CO₂, into a reactor | CO₂, behind a ceramic diffuser | Inline diffuser | 5 lb at 2 bps, ceramic |
|---|---|---|---|---|---|
| 3 mm | 0.014 mL | 0.03 mg | 0.05 mg | 0.07 mg | 763 days |
| 4 mm | 0.034 mL | 0.06 mg | 0.12 mg | 0.16 mg | 322 days |
| 5 mm | 0.065 mL | 0.13 mg | 0.24 mg | 0.31 mg | 165 days |
| 6 mm | 0.113 mL | 0.22 mg | 0.41 mg | 0.54 mg | 95 days |
| 7 mm | 0.180 mL | 0.34 mg | 0.66 mg | 0.85 mg | 60 days |
Sphere volume × CO₂ density at 22 °C: 1.92 g/L at a reactor's 0.05 bar, 3.65 g/L at a ceramic diffuser's 1 bar, 4.75 g/L at an inline diffuser's 1.6 bar (gauge). Density by the virial equation, 1.808 g/L at 25 °C and one atmosphere. mL × g/L = mg.
The counter sits after the needle valve, so its gas is at whatever pressure the rest of the line needs. A reactor or a ladder takes gas at the depth of its inlet — a few hundredths of a bar. A ceramic diffuser only lets gas through its pores above its bubble point, so the whole line from the needle valve to the disc sits at about 1 bar (15 psi) above the room — below the 20–30 psi working pressure diffuser makers ask for, because the needle valve needs headroom — and so does the gas in the counter. Inline atomizers need more, about 1.6 bar. Each bubble you count is compressed to 2 atmospheres and holds 1.9× the CO₂ of the same bubble heading for a reactor. That is the whole difference between the 165 and 314 days above, and it's why a ceramic disc that clogs and needs more pressure quietly makes every bubble heavier.
Work out your own bubble: the last cylinder, or two weighings
The number worth having is the one your own setup has already measured. A cylinder that ran out after a known number of days at a known count gives the CO₂ per bubble directly: contents ÷ (days × bubbles per second × 3,600 × hours). Say a 5 lb lasted 80 days at 2 bps for 8 hours a day. That is 2,268 g over 4.6 million bubbles: 0.49 mg a bubble. Everything the cylinder lost is in that figure — the leak at a fitting, the purge when it was changed, the night the timer failed — which is the point: it predicts the next one. At 3 bps the next 5 lb would last 53 days.
No empty cylinder to go on? Weigh the one you have, weigh it again a few weeks later on the same scale, and the difference is what your setup used. On the typical setup at 2 bps that is 289 g in three weeks — plenty for a luggage scale, and enough for a bathroom scale if you weigh yourself with and without the cylinder both times. Add the tare stamped on the cylinder and the same two numbers give what is left and the date it runs out. The tool's "My last refill" tab takes either route.
Behind a ceramic diffuser, 0.49 mg is a 6.4 mm bubble — believable. Into a reactor it would have to be a 7.9 mm bubble, and counters don't make bubbles bigger than about 7 mm. If the tool's "My last refill" tab says your bubble would have to be bigger than that, CO₂ left the cylinder somewhere other than the counter: brush soapy water on every joint from the cylinder valve to the counter with the gas on, and watch for growing foam.
How do I tell how much CO₂ is left? Weigh it
Not with the gauge — not for most of the cylinder's life. A cylinder holds liquid CO₂ with gas above it, and while any liquid is left the pressure is the liquid's vapor pressure, set by the temperature and not by how much is in there: about 56 bar (820 psi) at 20 °C whether the cylinder is full or two-thirds empty. Move it to a cold garage and the same cylinder reads less; nothing has been used.
The needle only starts to fall when the last liquid boils off, and at room temperature the gas left behind is dense — 0.19 kg per liter of cylinder at 20 °C, against the 0.68 kg/L a US cylinder is filled to. So the gauge starts to move with about 29 % of the CO₂ still inside. On the 5 lb at 2 bps that is 47 more days, and the time left after the needle moves scales with the bubble count like everything else, so no fixed "two weeks after it drops" rule can be right. From there the needle is a fuel gauge, falling as the gas is used. The forum belief that the last tenth of a cylinder can't be used is wrong the same way: a regulator feeding a ceramic diffuser at 1 bar keeps going until the cylinder itself is down to about that, and 1 bar of gas in a 5 lb cylinder is 12 g — 0.5 % of a fill.
| Cylinder at | 10 °C · 50 °F | 15 °C · 59 °F | 20 °C · 68 °F | 25 °C · 77 °F | 30 °C · 86 °F |
|---|---|---|---|---|---|
| Gauge reads while liquid is left | 44 bar · 640 psi | 50 bar · 720 psi | 56 bar · 820 psi | 63 bar · 920 psi | 71 bar · 1,030 psi |
| Still inside when the needle starts to fall | 20 % of a fill | 24 % of a fill | 29 % of a fill | 36 % of a fill | 51 % of a fill |
Vapor pressure and saturated-vapor density of CO₂ from the NIST Chemistry WebBook saturation tables (Span & Wagner), gauge = absolute − 1 atm, psi rounded to 10. "Still inside" is the vapor density over the US filling density of 0.68 kg/L (49 CFR 173.304a); European cylinders filled to 0.76 kg/L start to fall a few points later.
So weigh it. Every cylinder is stamped with its empty weight — "TW" in the US, usually in pounds, and in kilograms elsewhere. Take the regulator off, hang the cylinder from a luggage scale or stand on a bathroom scale with and without it, and subtract the tare: what is left is CO₂. The tool's "How long" tab takes the weight and the stamp and gives the days from there. A bathroom scale is good to about half a pound, which on a 5 lb fill is 10 %; a luggage scale does better.
There is a safety reason to bother. Once the liquid is gone and the cylinder pressure starts to fall, a single-stage regulator's working pressure rises as it does, so the needle valve passes more and more gas — and near the very end it can let the rest go at once, the end-of-tank dump. Overnight that can push the tank well past what fish tolerate. A dual-stage regulator holds its output steady; a weighed cylinder you swap on schedule never gets to the last stretch. The tool gives a "check it" date at three-quarters used, which at room temperature is about when the needle starts to move anyway.
Citric acid and baking soda CO₂: how much does a batch make?
The two-bottle DIY system runs C₆H₈O₇ + 3 NaHCO₃ → sodium citrate + 3 H₂O + 3 CO₂. Each gram of anhydrous citric acid makes 0.687 g of CO₂ and uses up 1.31 g of baking soda. The common recipe — 150 g of citric acid, 200 g of baking soda — makes 103 g of CO₂, about 56 liters of gas, and the citric acid runs out first with 3.2 g of soda to spare. The other recipe in circulation, 200 g of each, makes 105 g — barely more, because now the baking soda runs out first and 48 g of citric acid goes down the drain with the spent solution. If the bag says monohydrate, 150 g makes 94 g: 8.6 % of its weight is water.
Not all of it reaches the tank. When the bottles' pressure falls to what the diffuser needs, they stop delivering, and what is left — gas in the headspace and CO₂ dissolved in the liquid at that pressure — stays put: about 7 g with a liter of each against a ceramic disc. Of the 96 g that does get out, at 1 bps for 8 hours a day, a batch runs 14 days behind a ceramic diffuser and 28 into a reactor. The forums' "two to three weeks" sits between the two, which is the bubble-mass problem again. A 5 lb cylinder holds about 24 batches.
DIY yeast CO₂: how much sugar, and why two cups is too much for bread yeast
Yeast turns each gram of table sugar into at most 0.514 g of CO₂ and slightly more alcohol; in practice about 90 % of the sugar goes that way and the rest becomes yeast, glycerol and acids. A cup of sugar (200 g) in 1.5 L of water makes 93 g of CO₂ and finishes at 7.6 % alcohol. The popular two-cups-in-a-2-L-bottle recipe should make twice that, but bread yeast stops growing at about 11 % alcohol: the bottle quits at 141 g with 96 g of sugar — half the second cup — never fermented. With champagne or wine yeast (18 % tolerance) the same bottle makes 185 g.
The rate is the yeast's, not yours. It runs around the clock, fastest in the first days and slower as the alcohol builds, and faster in a warm room. Spread over 14 days the one-cup bottle averages 6.6 g a day, about 0.6 bubbles a second of 5 mm bubbles all day and night. A yeast bottle can't build the 1 bar a ceramic disc needs without leaking, so it belongs on a ladder, a bell or a reactor. For scale, a 5 lb cylinder is about 25 of those bottles.
How many bubbles per second does my tank need?
There is no number that transfers between tanks. Two bubbles a second behind a ceramic disc is 1.9× the CO₂ of two bubbles into a reactor; a 4 mm counter's bubble is half a 5 mm one's; and how much of the gas dissolves depends on the diffuser and on how much the surface is stirred. The usual starting point in print is about one bubble a second per 10 gallons. On the typical setup that is 6.9 g of CO₂ a day for every 10 gallons, so a 40-gallon tank at 4 bps empties a 5 lb in 82 days; the same rule into a reactor delivers 53 % of that gas. Start at the rule, then set the count by what's dissolved in the water — a drop checker going lime green by lights-on, or the pH dropping about one unit from degassed — and the CO₂ calculator turns either reading into ppm. Then bring the count back here: this page answers how long the cylinder lasts at that count, not what the count should be.
Run CO₂ on a timer: on an hour or two before the lights, off an hour before they go out. Plants don't use it in the dark, fish breathe all night, and CO₂ left on overnight is the classic way to find a tank of gasping fish in the morning. Unstable CO₂ is also on the short list of suspects in the algae guide.
Common questions
Is a SodaStream "60 L" cylinder 60 liters of CO₂? No — it carbonates about 60 liters of soda water. It holds 425 g of CO₂, which is 233 liters of gas at room temperature, and on the typical setup at 1 bps it lasts 62 days — not the 16 that reading "60 L" as gas would give.
Does a cylinder last longer in a cold room? No. What is inside is a mass, and the temperature only changes what the gauge reads. The gas in the counter is a little denser in the cold — 3.5 % more per bubble at 15 °C than at 25 °C — so the same count uses slightly more, not less.
Should the bubble counter have oil in it? Oil stops the water evaporating out of the counter, which is real. But the same nozzle makes bubbles roughly half the size in oil, so after the switch the count roughly doubles for the same gas. Count again after changing the fluid, and leave the needle valve where it was. Glycerin is thick enough that its bubbles don't follow the same rule, so the same advice goes double for it.
Refill or exchange? Either; the CO₂ is the same. An exchange hands you a different cylinder with a different tare, so read the new TW stamp before you weigh it, and check the hydrostatic test date stamped beside it — a shop won't refill one that is out of date.