A fermenter under CO₂ pressure feels sealed. The pressure pushes outwards, so how could oxygen get in? Quite easily, it turns out. To find out where it comes from, we built a small test chamber and put the parts of a fermenter through our oxygen meter one by one: gaskets, valves, sight glasses, spools and fittings.
In short:
- Over-pressure doesn’t keep oxygen out. It diffuses in through the gaskets while CO₂ leaks out.
- The gasket material matters most. Silicone let in nine times as much oxygen as EPDM, and EPDM almost four times as much as BUNA.
- Sight glasses with the original silicone gaskets let in six to seven times more than with EPDM.
- Valves differ a lot: a no-name 3″ butterfly with a silicone seal let in ten times as much as a Brewtools one with EPDM.
Why pressure doesn’t help
Oxygen doesn’t care about the total pressure. It moves through rubber and plastic from where its own partial pressure is high to where it’s low. In the air around the fermenter the oxygen partial pressure is about 210 mbar; inside, under CO₂, it’s close to zero. So oxygen works its way in through every gasket at the same time as CO₂ leaks out, and a couple of bar of over-pressure doesn’t stop it.
The test chamber


The chamber is a short stack of tri-clamp parts with a 3″ lid. The lid carries a valve, a pressure sensor and couplings for the temperature probe and the oxygen probe. For each test we clamped one part onto a reference assembly, filled the chamber with CO₂ to 2.8 bar absolute and left it closed – for anything from a few hours to eight days. The volume of each assembly, 0.34 to 1.33 litres, was calculated from the dimensions of its parts.
Before each test we flushed the chamber with CO₂: let it flow through, closed the outlet, built up pressure, vented and flushed again, until the oxygen reading was close to zero. The calculation only uses the rise from the starting point, so close to zero was good enough.
Temperature matters: the pressure follows it, and so does the conversion from oxygen reading to mass. After the first round of tests we built an insulated box around the rig to keep the room’s temperature swings out.
From oxygen reading to micrograms per hour
The probe measures the oxygen partial pressure in the gas. With the ideal gas law, the partial pressure, the temperature and the volume give the mass of oxygen in the chamber: m = pO₂ · M / (R · T) · V. Plot that over time and the slope is the ingress, in micrograms per hour. The lines are remarkably straight – R² is 0.99 or better in almost every test.

We also follow the over-pressure, corrected to 22 °C, as a measure of how well the assembly holds gas. The two don’t go together. The Brewtools PRV lost 8 mbar an hour without letting in any more oxygen than the reference, while the 3″ sight glass with silicone gaskets lost only 3 mbar an hour and still let in 67 µg/h. A part that holds pressure can still let oxygen through.
Gasket material
The first comparison was the simplest: the same parts – two 3″ gaskets, one 4″ and one 34 mm – with three different gasket materials.

Silicone let in nine times as much oxygen as EPDM, and EPDM almost four times as much as BUNA (nitrile). That fits what’s known about the materials: silicone is one of the most gas-permeable rubbers there is. A repeat with EPDM gave 7.6 µg/h, so the tests reproduce to within about 1 µg/h.
Valves, sight glasses and spools
Then we tested the parts that actually sit on a fermenter. Each one was clamped onto a reference assembly of the same size, so the difference is the part itself plus its extra gasket.





Butterfly valves. The no-name 3″ butterfly let in 98 µg/h – over thirty times the reference and almost ten times as much as the Brewtools 3″ butterfly in the same test (10 µg/h). We only tested one of each, so this says more about those two valves than about every valve on the market, but the difference is far too big to be noise. The Brewtools 2″ butterfly, closed, added about 2 µg/h.
That fits the seals: the Brewtools valve seals with EPDM and the no-name one with silicone – the same two materials as in the gasket test, with about the same ratio between them.
Sight glasses. Both sight glasses came with silicone gaskets against the glass. With those, the 3″ sight glass let in 67 µg/h and the Brewtools 2″ sight glass 48 µg/h. With EPDM gaskets that we made ourselves, the numbers dropped to 11.6 and 7.1 µg/h – six to seven times less, and close to the references. A later test of a small 1.5″ sight glass with EPDM showed no measurable difference with or without the glass.
How we made the EPDM gaskets is in Swapping silicone for EPDM in our sight glasses.
The polycarbonate spool. The Brewtools 2″ × 200 mm polycarbonate spool we use with our inline filter kit let in 75 µg/h. Polycarbonate is far more permeable to oxygen than glass or steel, but we haven’t tested whether it’s the plastic or its gaskets.
Small parts. On 34 mm, the Brewtools pressure relief valve made no measurable difference, and the sample valve added about 3 µg/h.
What it means for the beer
A few micrograms an hour sounds like nothing, but a fermenter stays closed for weeks. 10 µg/h for two weeks is 3.4 mg of oxygen. If all of it ended up in 60 litres of beer, that would be 56 µg/L – 56 ppb. One sight glass with silicone gaskets at 60 µg/h would make it more than 300 ppb. Not all of it ends up in the beer, since active yeast uses up much of it, but it keeps adding up after fermentation, when the beer is most sensitive.
After these tests we use EPDM everywhere on the oxygen-sensitive parts of the fermenter.
What we learned
- Over-pressure doesn’t keep oxygen out.
- Avoid silicone gaskets on anything that stays closed under CO₂ for long.
- Check what sits against the glass in your sight glasses.
- Check what the seals in your valves are made of – a silicone-sealed butterfly let in ten times as much as an EPDM one.
The gas line – tubing, ball locks and what happens when you connect them – got its own series of tests and its own post: The gas line lets oxygen in too. All the tests are in the table below.
All the tests
| Test | Parts | Volume (L) | O₂ ingress (µg/h) | Leak (mbar/h) |
|---|---|---|---|---|
| 1 | 3″–4″ spool, 4″ cap; gas line push-in – 2 m EVABarrier – push-in – valve | 0.70 | 25.4 | 6.4 |
| 2 | As 1 + T-piece and ball lock | 1.03 | 29.9 | 3.7 |
| 3 | As 2, gas connected | 1.03 | 81.9 | – (gas on) |
| 4 | As 2, EVA tube replaced with 316L pipe | 1.03 | 12.8 | 0.5 |
| 5 | 3″–4″ spool, 4″ cap, T-piece, ball lock – EPDM | 1.03 | 8.2 | 0.4 |
| 6 | As 5 – BUNA | 1.03 | 2.2 | 0.2 |
| 7 | As 5 – silicone | 1.03 | 73.8 | 4.4 |
| 8 | As 5 – EPDM, repeat | 1.03 | 7.6 | 0.3 |
| 9 | As 5 + connecting a purged, pressurised gas line (Camozzi valve – PTFE – ball lock with Festo push-in) | 1.03 | ≈ 600 µg at connection | – |
| 10 | As 5 + closed Camozzi valve – short PTFE – ball lock with Festo push-in | 1.03 | 9.0 | 0.2 |
| 11 | As 5 + closed DuoTight valve – short PTFE – DuoTight ball lock | 1.03 | 8.6 | 0.4 |
| 12 | As 5 + closed Camozzi valve – 10 cm EVA – ball lock with Festo push-in | 1.03 | Leaked too much to measure | 47 |
| 13 | As 5 + closed DuoTight valve – 10 cm EVA – DuoTight ball lock | 1.03 | 8.4 | 0.1 |
| 14 | As 5 + closed DuoTight valve on a ball lock with a welded 8 mm stub | 1.03 | 4.5 / 3.8* | 0.3 / 0.9* |
| 15 | As 14, valve open and gas on, connecting a bleeding ball lock | 1.03 | ≈ 20 µg at connection | – (gas on) |
| 16 | As 14, opening the valve after the line had been closed for a long time | 1.03 | ≈ 1,700 µg at opening | – (gas on) |
| 17 | As 14, valve open and gas on | 1.03 | 23.4 | – (gas on) |
| 18 | Reference: 3″–34 mm lid, T-piece, ball lock, 34 mm lid | 0.35 | 5.8 | 1.6 |
| 19 | As 18 + Brewtools PRV | 0.36 | 5.7 | 8.0 |
| 20 | As 18 + Brewtools sample valve | 0.37 | 8.5 | 0 |
| 21 | Reference: 3″–2″ spool, T-piece, ball lock, 2″ cap | 0.37 | 5.3 | 2.0 |
| 22 | As 21 + Brewtools 2″ × 100 mm sight glass, original silicone gaskets | 0.57 | 47.9 | 5.1 |
| 23 | As 21 + Brewtools 2″ butterfly, closed | 0.45 | 7.7 | 0.9 |
| 24 | As 21 + Brewtools 2″ × 200 mm polycarbonate spool | 0.74 | 74.8 | 9.5 |
| 25 | Reference: T-piece, ball lock, 3″ lid | 0.34 | 2.9 | 1.3 |
| 26 | As 25 + no-name 3″ butterfly, open | 0.66 | 98.3 | 10.2 |
| 27 | As 25 + 3″ × 165 mm sight glass, original silicone gaskets | 1.17 | 66.9 | 3.0 |
| 28 | As 25 + Brewtools 3″ butterfly, open | 0.66 | 10.1 | 1.7 |
| 29 | As 27 with EPDM gaskets against the glass | 1.17 | 11.6 | 0.3 |
| 30 | As 22 with EPDM gaskets against the glass | 0.57 | 7.1 | 0 |
| 31 | As 29 + Brewtools 3″ butterfly, closed (reference for the dry-hop tower tests) | 1.33 | 15.5 | 1.4 |
| 36 | T-piece + 34 mm–1.5″ adapter, 2.4 bar abs | 0.36 | 9.9 | 0.9 |
| 37 | As 36 + 1.5″ sight glass with EPDM | 0.36 | 7.8 | 1.2 |