How Long Should DIY Sparkling Water Rest?
20 lb CO₂ Economics and a Mass-Loss Test

1. Economics first: even the first tank-full worked out cheaper

The original reason for moving to a 20 lb tank was simple: if you drink a lot of sparkling water, the small-cylinder exchanges add up. Here is what I actually paid for the CO₂-supply side of the setup.

SodaStream's U.S. site currently lists a 60 L spare cylinder at $31.99 and a one-time exchange at $16.99 (checked October 8, 2026). A 60 L cylinder contains about 410 g (14.5 oz) of CO₂. [1][2]

A 20 lb cylinder holds about 9,072 g of CO₂, or about 22.13 SodaStream fills. Buying the first cylinder and then exchanging it enough times to reach the same CO₂ mass gives:

$31.99 + (22.13 − 1) × $16.99 ≈ $390.93

Once the tank is owned, my actual local refill is $43 for 20 lb. Carbonics advertises in-person CO₂ cylinder refills, including 20 lb sizes, and says it uses high-grade food-quality CO₂ suitable for soda and beer. [3]

On a rough 700 L real-world basis, the gap vs. store-bought water gets even bigger

Up to this point, the comparison has focused on the CO₂-supply side. In day-to-day use, though, the more intuitive question is simply: how much sparkling water does this tank actually make? In my own use, it feels like about 1 L per day for roughly 2 years, so I treat one 20 lb fill as roughly 700 L. That is not a direct metered CO₂-yield measurement — it is a real-world estimate from actual use.

For context, SodaStream says its 60 L cylinders make up to about 60 L on average, while also noting that cylinder life depends on the carbonation level used. [11] A 20 lb tank contains about 22.1 times the CO₂ mass of a 410 g SodaStream cylinder, so a simple “60 L per cylinder” upper-bound calculation gives roughly 1,330 L. My rough real-world estimate of 700 L is about 53% of that figure. Because I normally carbonate quite strongly, and this setup also loses some CO₂ through headspace, relief-valve venting, connections and opening, I do not consider 700 L an implausible real-world number. Still, it is an estimate from use, not a directly measured tank-yield test.

On that 700 L basis, this DIY setup works out to about $0.37/L all-in, or about $0.06/L once I am only paying for the next CO₂ refill. By comparison, examples of store-bought sparkling water come out to $0.97/L for Walmart Great Value club soda, about $1.55/L for LaCroix 8-packs, and about $3.50/L for Topo Chico 12-packs. [8][9][10]

In other words, even against the cheapest store-brand club soda, the DIY route is already cheaper by the time I have made about 700 L on the first fill. Once the 20 lb tank is owned, the gap widens further. I am not including tap-water cost, tax, or shipping here, but those additions would not change the overall direction very much.

2. The key hardware feature: removing the bottle without releasing the pressure

This feature is what made the timing experiment possible. After injecting CO₂, I can remove the bottle with the infuser cap still attached, then let it sit for 1 minute, 15 minutes, 80 minutes or overnight without intentionally venting the headspace.

With the more conventional SodaStream-style arrangement I used before, I could not simply remove a bottle and park it elsewhere while preserving that same pressurized state. So this machine did more than save money: it let me make pressurized hold time an experimental variable.

3. First clue: the 1-minute sample looked wild, but tasted weakest

I first tested 1, 5 and 15 minutes of pressurized hold time after the same carbonation procedure. The sensory result was straightforward: 1 minute felt weakest and faded fastest, 5 minutes was in between, and 15 minutes felt strongest and lasted longest. Visually, though, the order looked almost backwards: the 1-minute sample foamed the most dramatically.

To quantify the visual behavior, I converted the glass-side videos to grayscale and averaged the frame-to-frame brightness difference, |I(t+1)-I(t)|, inside a near-surface region of interest. This “Motion Index” is not a bubble count and it is not dissolved CO₂; it is simply a relative measure of visible motion.

The audio changed too. The 4–18 kHz energy ratio moved from 27.5 → 43.6 → 47.8%, while spectral centroid shifted from 3.4 → 6.9 → 7.9 kHz. That is consistent with a change in bubble-event character, but the recordings also contain valve noise, surface effects and container resonance, so I treat these only as relative acoustic fingerprints.

4. Could a 0.1 g kitchen scale see the difference?

One volume of CO₂ corresponds to roughly 1.96 g of CO₂ per liter of liquid when the gas volume is referenced to standard conditions. For 800 mL (27 fl oz), that is about 1.57 g per volume. So a 0.1 g scale has enough nominal resolution to see meaningful relative differences.

Simply weighing a sealed bottle before and after carbonation does not separate dissolved CO₂ from gas in the headspace. Instead, I compared the stable mass before opening with the stable mass after opening and degassing had settled down.

5. 1 minute vs. 15 minutes vs. 80 minutes vs. 8 hours

Once 15 minutes and overnight were clearly different, the useful missing point was somewhere between them. I chose 80 minutes because it sits near the middle of 15 minutes and 8 hours on a logarithmic time scale. The hold temperature for this set was 0°C (32°F).

The 8-hour sample still lost 1.3 g less than the 15-minute sample. After the measurement sequence, the taste impression lined up with the same direction: overnight still felt strongly carbonated, 15 minutes was moderate, and 1 minute was noticeably weak.

6. The fit gives τ≈74 minutes. My practical answer is about four hours.

Fitting the four points to an exponential relaxation toward a non-zero asymptote gives:

L(t) ≈ 6.22 + 1.79 · exp(−t / 74.2)
τ ≈ 74.2 min, R² ≈ 0.974

Here, L(t) is the measured mass loss after opening, not dissolved CO₂ concentration itself. Also, this is a three-parameter fit to only four measured hold times, so I do not treat the high R² by itself as proof that the model is uniquely correct. I also tried an empirical logarithmic fit during the analysis, but I use exponential relaxation here because it maps naturally onto a standard gas–liquid equilibration model.

If the measured loss is acting as a proxy for how far the closed bottle remains from gas–liquid equilibrium, a standard lumped gas–liquid mass-transfer model has the same form:

dC/dt = k_La(C* − C)
C(t) = C_eq − A·exp(−t/τ)

The 74-minute value is therefore best described as an effective time constant for this particular setup: 0°C water, 800 mL, this tall bottle, this headspace/interface, the initial mixing from CO₂ injection, and whatever slow internal circulation remains during the hold.

The order of magnitude is not strange for small-scale forced carbonation. University of Florida IFAS gives an example where water can take hours to reach full carbonation and notes that agitation can greatly accelerate the process. [4] A 2024 review of CO₂ diffusivity in water also emphasizes the strong temperature dependence of molecular transport and covers data down to about 273 K. [5]

Technical note: why might the short-rest samples degas so violently?

I find it useful to separate the pressurized hold from the opening event.

1. During the hold: immediately after injection, the closed bottle may still contain more CO₂ in the headspace and in undissolved microbubbles than it will at equilibrium. Over time CO₂ redistributes into the liquid.
2. At opening: pressure drops rapidly, leaving the liquid supersaturated relative to atmospheric pressure.
3. Short rest: residual microbubbles and gas pockets can provide nucleation sites, so the decompression creates rapid visible bubbling and CO₂ loss.
4. Long rest: if those microbubbles/nuclei have diminished and more CO₂ has moved into solution, the opening can look calmer while the drink retains more carbonation.

Single-bubble experiments show that liquid-side CO₂ mass transfer depends on bubble size and exposure time, which is another reason not to treat “bubbly appearance” and bulk equilibration as the same thing. [6] I did not directly measure microbubble count or nucleation-site population here, so this part remains a mechanistic hypothesis consistent with the video and mass-loss behavior.

7. Topo Chico tasting: same “strong carbonation,” very different feel

While I was running this test, Topo Chico Mineral Water began a phased return to Texas after being difficult to find for months in 2026. Houston-area H‑E‑B and other retailers were among the stores reported to be restocking it. [7] That made it a perfect real-world benchmark.

Plain

• DIY: stronger overall carbonation, sharper acidity.
• Topo Chico: still strongly carbonated, but the bubbles felt finer and smoother; the flavor was mellower with less sharp acidity.

Highball

I also made two highballs with Dewar's 12 Mizunara. Topo Chico felt finer and more delicate but gave me a slight bitter note. The DIY water had very similar carbonation strength, and its sharper acidity worked well with the whisky and made the sweetness easier to notice. I preferred the DIY version in this round.

8. Bottom line: 20 lb for cost, ~4 hours for retention

• Even including my used tank, the initial CO₂ charge and the connection hardware, the first ~20 lb of CO₂ still cost less than getting the same CO₂ mass through one SodaStream spare plus one-time exchanges.
• In my real-world use, one 20 lb fill feels like roughly 700 L of sparkling water, which works out to about $0.37/L all-in or about $0.06/L once I am only paying for the next refill.
• That means the DIY route is already cheaper than budget store-brand club soda by the first 700 L, and dramatically cheaper than premium sparkling water like Topo Chico.
• The removable pressurized infuser cap is what made hold time testable in the first place.
• The 1-minute sample looked the most violent on opening but tasted the weakest afterward; visible fizz was not the same thing as retained carbonation.
• Measured mass loss fell from 8.1 g → 7.5 g → 6.9 g → 6.2 g for 1 min → 15 min → 80 min → 8 h.
• The effective exponential time constant was ~74 minutes; 3τ≈3 h 40 min, so about four hours is my practical target.
• If I do not want to think about timing, I leave it overnight.
• Topo Chico remained the smoother, finer-bubble drink, while the DIY version was sharper and — in this highball test — my preference.

What started as a cost-saving project ended up connecting CO₂ economics, gas–liquid equilibration, opening losses and tasting. The next useful step would be repeated runs under the same conditions, ideally with pressure and temperature logged at the same time.

References

1. SodaStream USA, CO₂ cylinder refill / exchange / spare (one-time exchange $16.99, spare $31.99; accessed Oct. 8, 2026).
2. SodaStream, 60 L gas cylinder product specification (CO₂ content approx. 410 g / 14.5 oz).
3. CMR Carbonics, CO₂ cylinder refills in Houston / 20 lb cylinder sales and refills (food-quality CO₂ and in-person refills).
4. University of Florida IFAS Extension, A Guide to Carbonating Beverages at Small Scale (forced carbonation can take hours; agitation greatly increases the rate).
5. Polat, H. M. et al., Diffusivity of CO₂ in H₂O: A Review of Experimental Studies and Molecular Simulations in the Bulk and in Confinement, Journal of Chemical & Engineering Data 69 (2024), 3296–3329.
6. Chemical Engineering Science, Experimental study of interfacial mass transfer from single CO₂ bubbles ascending in stagnant water, 287 (2024), 119684.
7. Houston Chronicle, Topo Chico mineral water is returning to Houston-area stores (Oct. 3, 2026).
8. Walmart, Great Value Low Sodium Club Soda Sparkling Water, 33.8 fl oz Bottle ($0.97; accessed Oct. 9, 2026).
9. Target, LaCroix products at Target (8-pack 12 fl oz cans listed at $4.39; accessed Oct. 9, 2026).
10. Target, Topo Chico Mineral Water - 12pk/12 fl oz Bottle ($14.89; accessed Oct. 9, 2026).
11. SodaStream USA Support, How often will I need to replace my SodaStream CO₂ cylinder? (60 L cylinders make up to about 60 L on average; actual cylinder life depends on carbonation level; accessed Oct. 9, 2026).

← Back to Cooking & Life