Through my first year of sewing masks I struggled to understand how much protection my masks provided. After failing to find a satisfactory answer after reading dozens of publications, I finally submitted fabric samples for filtration testing to CTT Group. Here I present that data with more math and deeper analysis than my "Data for Sewists" write-up of the same data.

Topics include:

• Filtration after washing (10 machine wash/dry cycles)
• Woven cottons, non-woven polypropylene, and non-woven polyester
• Filtration and breathability
• Submicron NaCl vs BFE (Bacterial Filtration Efficiency)
• Comparison to ASTM F3502
• Calculation of single-layer filtration from multi-layer samples. See "layering" for why not vice versa.

Results

Overall, two layers of quilting cotton performed poorly, filtering only 26% of 0.8μm NaCl particles. OlyFun/OlyFun/flannel and sateen/OlyFun/flannel filtered best with nearly 80% filtration of 0.8μm particles. The OlyFun/OlyFun/flannel combination is a clear winner with breathability 33% lower than two cotton layers. Washing is fine: 10 machine wash/dry cycles did not impact filtration.

Results include data from 17 samples, each unique. The final test report typically has a single value: filtration at the MPPS (Most Penetrating Particle Size) for submicron NaCl, and average filtration over all particle sizes for BFE. Plotted below are the underlying data, with filtration plotted for each particle bin size.

Results: 3um BFE and submicron NaCl

Figure: The data on the left correspond the submicron NaCl filtration results, with solid lines for unwashed and dashed lines for washed samples. The bars near the middle of the graph show BFE average over all particle sizes, which is the typically reported result from BFE. The large X's correspond to BFE data for each plate in the Andersen sampler, with the ranges for each plate annotated at the top.

Observations on combined submicron NaCl and BFE results:

• Submicron NaCl vs BFE: Three of the four samples tested both with submicron NaCl and with BFE exhibith very similar filtration of 0.8μm NaCl as 3.0μm BFE. Two layers of Kona woven cotton are the exception with notably higher BFE than 0.8μm NaCl filtration efficiency.
• Non-wovens: Non-woven polyester and two different brands of non-woven polypropylene filtered nearly identically for the same weight in gsm (grams per square meter). (See analysis in the "Mathematical analysis" section below.)
• Washing: Three different fabric combinations tested in washed and unwashed versions exhibited nearly identical filtration before and after washing. The washed samples were washed and dried 10 times in laundry bags as part of regular family laundry loads on the default washer/dryer settings, representing typical usage.
• Repeatability: While testing only one of each sample type does not allow for quantitative assessment of variability, the high similarity in the results for washed and unwashed samples suggests very good repeatability.
• BFE individual plate counts: The dip in filtration for the two lowest BFE bins is curious and not understood. Typically filtration improves with particle size from 0.3μm to 10μm. Worse filtration for the 0.65-1.1μm and 1.1-2.1μm BFE bins than for 0.6μm-0.8μm NaCl is surprising. The leading guess at an explanation notes that particles exiting a fabric contain a higher fraction of small particles than the incoming flow, and if some fraction of particles deposited early on plates intended to capture larger particles, we would see exactly this dip (thanks to Steve Rogak).

Results: Submicron NaCl

Observations on submicron NaCl results:

• 0.3μm and smaller is a chaotic mess. The most difficult to filter particles, and the size specified by the new ASTM F3502 cloth mask standard, results in noisy data. Thankfully, by 0.6μm, the data is cleaner... and there is little evidence that 1≤μm particles contribute much to viral transmission.
• Best guess: Rewardingly, my "best guess" combinations sateen/OlyFun/flannel and OlyFun/OlyFun/flannel indeed filtered best.
• Washing: Washing and drying on the default cycles 10 times did not affect filtration. The dashed lines in the plot show washed samples, and they filtered nearly identically to their unwashed counterparts. The fabrics included in the washed samples include a fairly broad range of fabrics used by home sewists: Kona woven cotton, OlyFun non-woven polypropylene, Robert Kaufman cotton flannel, Pellon P40 non-woven polyester/viscose.
• Polyester vs polypropylene: Non-woven polyester and non-woven polypropylene seem to b equivalent with regards to filtration if we compare the same weight.
  • 2 layers of OlyFun (~130gsm total) and 3 layers of Pellon 930 (~125gsm) filtered nearly identically.
  • Comparing Kona/OlyFun/Kona vs Kona/930/Kona in the "Mathematical analysis" section, we see that OlyFun (non-woven polypropylene, 65gsm) filtered better than Pellon 930 (non-woven polyester, 41gsm) with the improvement being roughly proportional to the fabric weight (in gsm).
• Filter layer: A third filter layer is worthwhile. For 600um and larger particles, two layers of Kona cotton woven performed significantly worst.
• Kona vs sateen: As hypothesized, the sateen weave cotton filtered better than the standard weave Kona cotton. Perhaps surprisingly, the breathability (resistance) was similar.

Methods

Sample List: Following is my list of samples for NaCl submicron filtration and 3.0um BFE (Bacterial Filtration Efficiency).

Sample ID	Test @0.3um	Test @3.0um	Outer layer	Middle layer	Face layer	Washed?	Description
3a	Yes	-	Kona	P40	Flannel	No	My most common construction.
3b	Yes	-				Yes
3c	-	Yes				No
4a	Yes	-	Sateen	OlyFun	Flannel	No	My present best guess at the best combination
4c	-	Yes				No
5a	Yes	-	Kona	-	Kona	No	Community minimum: This is common in the sewing community.
5b	Yes	-				Yes
5c	-	Yes				No
6a	Yes	-	Kona	OlyFun	Kona	No	Non-woven polypropylene is believed by some sewists to be superior to non-woven polyester for filtration. This sample will allow a comparison, with 6a vs 7a comparing polypropylene vs polyester. Also, while some studies show non-woven polypropylene as filtering well, I haven't seen assessment of filtration after washing.
6b	Yes	-				Yes
7a	Yes	-	Kona	930	Kona	No	A standard construction in the sewing community.
7c	-	Yes				No
8a	Yes	-	Sateen	930	Sateen	No	Kona cotton has ~130um gaps visible in microscope photographs while this sateen is a similar weight, thinner yarns, and without such gaps. 7a vs 8a will allow comparison of Kona (quilting woven cotton) vs sateen (a woven cotton with fewer gaps between yarns)
10a	Yes	-	OlyFun	OlyFun	-	No	Compare similar weight of different non-wovens. Added circa March 31; results expected mid/late April.
11a	Yes	-	SmartFab Double-Thick	SmartFab Double-Thick	-	No
12a	Yes	-	930	930	930	No
13a	Yes	-	OlyFun	OlyFun	Flannel	No	Calculate Pflannel=P13/P10

Washing: The washed samples were washed and dried 10 times in an LG front-loading washer and LG electric dryer on the default wash and dry cycles, with Seventh Generatoin free & clear fragrance-free detergent, in a mesh laundry bag.

Fabrics: Many prior studies of filtration effectiveness of cloth provide inadequate details of the tested fabrics to allow sewists to purchase identical fabrics. The following include links to manufacturer websites for fabrics in this study:

	Link	Description	Weight
Kona	Kona cotton	A typical quilting fabric widely used in the sewing community and also tested by Lustig et al.	147.5 gsm (measured)
P40	Pellon 40 sew-in midweight stabilizer	A non-woven 85% polyester, 15% viscose interfacing, which Pellon incorrectly described as "100% polyester" on its webpage (!) and so I used it for a while before realizing the error.
930	Pellon 930 sew-in midweight	A non-woven 100% polyester interfacing.	41.1 gsm (measured) 0.511m x 2m piece weighs 42g
OlyFun	OlyFun	A non-woven polypropylene intended for crafting projects. Some sewists believe this makes an excellent filter; see makermask.org.	65 gsm (specified)
Flannel	Robert Kaufman cotton flannel		169.5 gsm (Specified) 5 oz per sq yard
Sateen	Spoonflower Organic Cotton Sateen	100% organic cotton, weave has smaller gaps than Kona.	130 gsm (Specified: 3.8 oz per square yard) 142 gsm (Measured)

Microscope photos of these (and other) fabrics are at: ofb.net/~ania/fabrics/

Comparison to ASTM F3502

MakerMask.org has a nice summary of this new standard for face coverings. Masks are rated for filtration of 0.3um particles, like N95 standards and unlike the European standard for community face coverings (CWA 17553) which tests with 3.0um particles. Masks are also rated for breathability.

ASTM F3502			My samples
Filtration	Level 1	≥20%	All my samples filtered ≥20% @ 0.3um except the two layers of Kona cotton.
	Level 2	≥50%	OlyFun/OlyFun/flannel filtered 50.4% at 0.3μm, barely meeting the Level 2 threshold. The dip in the data at 0.3μm is surprisingly large, and some samples showed filtration >50% at 0.2μm and 0.4μm (Sateen/OlyFun/flannel, Kona/OlyFun/Kona), raising for me questions about measurement noise.
Breathability	Level 1	≤ 15 mm H2O	All my samples met the Level 1 threshold.
	Level 2	≤ 6 mm H2O	OlyFun/OlyFun/flannel met Level 2.

Mathematical analysis

We can extract the filtration improvement due to the addition of layers by comparing samples that differ in only that one layer. Assume for simplicity that:
       P_tot = P₁ * P₂,
where P_tot is the total penetration of a multi-layer fabric combination, and P₁ and P₂ are the penetration amounts for the constituent single layers.

This assumption is not quite correct as we can see when we look at others' data on filtration of multi-layer samples; see Layering for details. It seems that the first layer in a multi-layer stack-up provides greater filtration than subsequent layers.

For 0.8μm, assuming Ptot = P1 * P2
Calculated Filtration Efficiency of Single Layer		Source data (measured)
OlyFun	47.78%	Ptot=39.2% for Kona/OlyFun/Kona, P1=73.6% for Kona/Kona
Pellon 930	27.65%	Ptot=53.5% for Kona/930/Kona P1=73.6% for Kona/Kona
Flannel	43.60%	Ptot=21.4% for OlyFun/OlyFun/flannel, P1=49.1% for OlyFun/OlyFun
Sateen	23.02%	Kona/Kona Kona/930/Kona sateen/930/sateen
Kona	14.21%	Ptot=73.6% for Kona/Kona
OlyFun	29.90%	Ptot=49.1% for OlyFun/OlyFun
SmartFab Double-Thick	24.81%	Ptot=56.5% for SmartFab/SmartFab
Pellon 930	20.23%	Ptot=50.8% for 930/930/930

Non-woven polypropylene vs polyester: The first two estimates for OlyFun and Pellon 930 suggest that non-woven polypropylene and non-woven polyester filter similarly for the same weight (gsm) combination. Pellon 930 is 41.1gsm (measured) while OlyFun is 65gsm (specified). If we scale up the estimated Pellon 930 efficiency correspondingly (27.65%)*(65gsm/41.1gsm) = 43.77%, which is very close to OlyFun. This suggests that a similar weight of Pellon 930 (a non-woven polyester) may give similar filtration as a similar weight of OlyFun (a non-woven polypropylene). Indeed, the later measurements of OlyFun/OlyFun and 930/930/930 filtered nearly identically.

Test selection and particle sizes

CTT Group provided excellent, responsive, kind, knowlegable customer service.

Many studies of cloth masks use submicron particle sizes to test filtration by cloth fabrics and cloth masks, consistent with the 0.3um particle size standard for N95 masks. However, what particle size dominates Covid transmission seems to be an open question. Marr 2020 notes that NaCl and proteins are usually part of the viral droplet, so even dried out particles floating the the air could be expected to be much larger than a single virus. While some assume that large particles can hold more virus, large particles tend to be generated in the oral cavity which is not where Covid seems to occur. Covid is a lower respiratory illness, primarily, and small (1um or so) particles are generated there. Also inhaled particles <1μm tend not to deposit in the body but rather to be exhaled back out.

I suspect 1μm to 10μm particles dominate Covid-19 transmission, explained in more detail in my post on particle sizes. Unfortunately, this range coincides with a range of uncertainty for fabric filtration. Many fabrics block 5um and larger particles quite well. They block 0.3um particles quite poorly. The range where cloth masks start to be effective coincies with a range whose importance to COVID-19 transmission is under debate.

I don't know of a commercially available 1μm to 10μm filtration test. So I tested with both submicron NaCl and 3.0μm BFE.

Timeline

• 9 February, 2021 (Tuesday): Samples mailed to CTT Group.
• 6 to 10 business days expected for shipping.
• 2-3 weeks expected for testing by CTT Group.
• 23 February, 2021: Samples received by CTT.
• 1 March, 2021: Results received from CTT for 0.3um NaCl samples
• 8 March, 2021: Results received from CTT for 3.0um BFE samples
• 6 April, 2021: Samples 10a, 11a, 12a, 13a received by CTT
• 7 April, 2021: Payment processed for samples 10a, 11a, 12a, 13a
• 19 April, 2021: Results received for samples 10a, 11a, 12a, 13a