O2 PSA Design Notes
O2 PSA Design Notes
O2 PSA Design Notes
1. There is a severe O2 shortage in India, and we see several groups/individuals trying to put together
homemade PSA units. Here are some guidelines. Please treat these as tips, not as rules. Registered
professional engineers who are skilled at making these systems should be the ones doing the work.
2. Just as it is anywhere else, people have worked on perfecting these systems for over 50+ years, so don’t
expect these tips will yield the best design. The intention is to demystify and explain certain aspects.
3. This document/tweet is NOT a recipe and is written with a bench-top system in mind when we wrote this.
We don’t take any liability. We are working on this project, so we may not respond to queries/ comments.
Adsorbents
4. Principle: Most Zeolites adsorb N2 preferentially over O2 -> When air is sent through an adsorption column;
containing a zeolite, the O2 moves through the column faster than N2. Hence, N2 is retained in the column,
and O2 is produced (at the feed pressure) at the column outlet. The production of O2 does not last forever,
and we have to stop the production of O2 (when N2 starts to slip through) to regenerate the column. Regen
is done by reducing the pressure in the bed. The gas that is vented from the bed is rich in N2 and lean in O2
and is vented to the atmosphere. Repeat
5. Think of a shopping street and where you send a group that likes to shop (N2) and the one that does not
(O2). The group that does not want to shop (O2) comes out pure on the other side. At some point, the shops
are saturated, so there is no more separation. Shops (and streets) are cleaned. Repeat the process. If
continuous operation desired, using two streets: One adsorbing; the other regen.
6. Zeolites used for O2 production: Zeolite 13X, LiLSX, 5A. Each zeolite works in specific operation modes.
Zeolite 13X- PSA (PH ≈ 5 bar, PL ≈ 1bar) ; LiLSX- VPSA (PH≈ 1.5 bar, PL ≈ 0.5 bar). Typical pressure ratios
are in the range: 1.5<(PH/PL)<5.
7. Typical particle sizes used are around ≈ 0.3 to 1 mm dia. Smaller particles: Increase throughput but also
increase pr—drop (Energy!).
8. LiLSX is much more expensive than 13X, but overall, it performs better (combination of capital and operating
costs). In crunch situations, this is perhaps not very important. You may be guided by what sieve is
available. So either might be ok but note that each needs a different process.
9. Zeolites are notoriously hygroscopic. Note that at 40 °C and 100% Humidity, air contains 0.05 kg of water/kg
DA. That is a lot. Guard beds are used to protect the sieves—typical materials: Silica-gel, Activated Alumina,
Zeolite 13X. A good rule of thumb is that these materials have ≈ 15-20 wt% capacity for moisture.
10. Typically, the feed end of the column has about 20-30 vol% of the desiccant, and on top of that, the Zeolite.
This config. allows removing some amount of water during the desorption step. A separate guard-bed might
just be a sacrificial bed. Calculate how long your desiccant will last!!
11. When you receive a zeolite, you may never know how it was stored or what it has seen. In labs, we assume
that the zeolite has taken up moisture. So this needs regeneration. Typically we heat zeolites up to
300-350C for ≈ 8 hours with a flow of dry gas. Dry means completely dry. <1 ppm moisture. You can use N2,
O2, Ar, Air, or He. When you are ready to pack a column, try or best not to expose the zeolite to ambient
conditions otherwise, it will rapidly adsorb water from the atmosphere and lose its N2 adsorption capacity.
12. How much adsorbent do I need? This is a complex interplay of how well the system is designed, how good
the sieve is, etc. Industrially, the bed-size factor, BSF, (lb of adsorbent per Tonne-per-day-O2) is used. Most
systems work in the 200 (excellent) to 800 (OK). So if you need a 5 LPM O2 with a BSF of 400 you would
need ≈ 2kg of adsorbent.
Equipment Selection
13. When you use a compressor, note that the compression increases the temperature of the gas. Higher the
pressure, the higher the temperature (for 5 bar pressure, the temp can be 150+ C). Many compressors
come with aftercoolers. At high temperatures, the desiccant and the sieve lose capacity. Make sure that the
gas is cooled. Without much sophistication, you can bring down the temp to ≈ ambient+10C. Ideal
temperature for a Zeolite= 30-40 C
14. There are just two moving parts in a PSA: The compressor and the valves. The valves undergo many
switches in a minute, so you need a robust valve. If there is robustness vs precision tradeoff, go for
robustness. The valves often are (typically) the weakest link (they switch 10/100s of 1000s times). If
possible, use pneumatic actuators or motorized valves over solenoid valves.
15. The airflow (for your compressor) ≈ 8 to 15 x O2 flow rate required.
16. Try to look for oil-free compressors to avoid contamination.
Bed Design
17. Choose a column diameter (D) that is > 10x particle dia (dp). This avoids channelling. Small-scale units
have length to col—Dia (L/D) ratios of ≈ 3 to 10. Longer column = Higher pressure drop! When filling a
column, tap it as you fill the zeolite, avoiding bridges. Try to fill most of the column. Empty volumes allow
gases to mix and kill separation.
18. Try not to exceed superficial velocities (Actual vol flow rate /col cross-sec area) of ≈ 0.3 m/s. Too high
velocities could crush particles. Try to reduce the length of tubing (avoids pressure drop and mixing). Keep
tubing diameter at a reasonable size (too small a diameter will offer pressure drop).
Cycle Design
19. The simplest PSA process is called the Skarstrom cycle. Adsorption step: Introduce air at the feed end,
collect O2 at product end; Desorption step: Reduce pressure (open the valve to atm or pull vac) from feed
end; Purge step: Take some of the product O2 (from a purge tank or the O2 tank) and send it through the
product end at the low pressure; Pressurization: Pressurize with feed from feed end.
20. Definitions: PSA: Pressure Swing adsorption; VSA- Vacuum swing adsorption; VPSA- Vac. Pr. Swing
Adsorption; Purity=Moles O2 in product/ Moles all components; Recovery= Moles of O2 in product/Moles of
O2 fed to the system. Air ≈ 78%N2; 21% O2; 1% Ar
21. High pressure (PH)- Highest pressure the process is run, Low pressure(PL)- Lowest pressure at which
process is run. Typical pressure ratios are in the range: 1.5<PH/PL<5. Normally a PSA will operate from ≈ 3
bar to 1 bar or from 1 bar to whatever vacuum level can be reasonably generated. Larger pressure ratios are
good, and you typically want to be on the higher end. The choice of PH/PL is a compromise between O2
Recovery and Energy. High PH/PL requires either higher pressures or lower vacuum-> Higher energy
consumption. Sometimes pressure levels are limited by compressor/vac. pump
22. O2 Purity cannot exceed 95% (O2/(O2+Ar)=21/(21+1) ≈ 95). Most systems can reach ≈ 93%. The zeolite
cannot separate O2 from Ar, hence the limitation. Typical recovery on a well-designed system can be in the
40 to 60% range.
23. Skarstrom cycle can be implemented either in a single-column or a multi (2+)-col mode. If you use a single
column, then a tank downstream is used to ensure that the user gets continuous flow. It is customary to use
tanks ( also for multi-column processes. You can program the solenoids to achieve these.
24. The duration of each of these steps is critical. There are ways to estimate these times, but it does depend on
the material, your flow conditions, temperature, pressure etc. High-end systems work very short cycle times
(rapid PSA) ≈ 10-15 sec. Medium-scale units work with ≈ 1 to 2 min cycle time. This is something you should
play with. Note that these systems take time to respond. So do not expect to see the effect of your change
right away. Give about 5 to 10 cycles to complete.
25. The amount of purge is important. Use a valve between the purge/product tank, adjust the duration and the
setting to see what happens to your purity. Start with long purge times. Industrially once the optimal value is
found, these valves are replaced by a capillary to provide suitable resistance and to reduce fittings.
26. There is always a trade-off between Performance Indicators (PI): purity-recovery-energy consumption-size
of equipment. This is seen in almost all separation processes. When evaluating your results, don't just look
at one of the PIs. You might get outstanding purity, but recovery can be low! Optimizing the system takes
time and experience.
27. PSA systems could take up to 50 cycles to give stable purities. So do your testing long enough. While
seeing pure O2 come out in the first few cycles will (and should) bring joy, do observe for a long enough
period.
28. All the best
References:
1. Ruthven, D. M., Farooq, S, Knaebel, K (1994). Pressure Swing Adsorption. New York: VCH Publishers
2. Ackley, M. W. (2019). Medical oxygen concentrators: a review of progress in air separation
technology. Adsorption, 25(8), 1437-1474. [By far the best review on this topic]
3. Knaebel,https://www.adsorption.com/wp-content/uploads/2016/04/AdsorberDes2.pdf
Disclaimer:
1. This is NOT peer-reviewed.
2. The authors take no liabilities for whatsoever consequences that might arise from
the use of information provided here.