Air Filter Test

The Idea

After overhauling the top end of my KLR, I was trying to get a handle on the new jetting I would need.  While I had the head off, I had cleaned out and opened up the exhaust port, and cleaned up the flange weld on the headpipe.  I also opened up the top of the airbox, and went to a Moose dual-layer foam air filter after killing two K&N filters.  The K&Ns flow a lot of air, but I've had problems with them clogging solid with the fine silt found in the California deserts.  Filterskins help tremendously, but I've also had problems getting the K&N properly clean after getting it well-silted.  So, I've gone back to foam.

I had to jet richer than I had expected after switching to foam, but the porting and airbox work were also a factor.  To get an idea of the filter's contribution, I decided to run a simple flow test.

Disclaimer: This is not intended to be a full-up research project contending for a Nobel prize; it is only a quick and dirty way to compare air filters. I haven't spent a lot of time on the writeup, so let me know if I've missed something.

The Setup

I scrounged a 1/8 hp centrifugal fan from work, and a Setra digital manometer with a range of +/-15" of water (24" of water is roughly equal to 1 psi).  Using 3/4" plywood, I made two plates: One to fit the 4" inlet on the blower, and another to provide a base for the filter.  I cross-drilled the filter baseplate and inserted a 1/2" steel rod with a tapped hole in the center, to take the filter hold-down bolt.  I also added a 1/16" static pressure port tube between the plates.  I screwed the plates together, and used E6000 adhesive to attach the plates to the blower.

One concern was how close the blower airflow matched the engine airflow. Using the digital manometer, I measured a (roughly) average stagnation pressure of 1" H2O, corresponding to a 20 m/s air velocity. The blower outlet was 8.8 cm in diameter, so the total airflow is 7.3 million cc/minute. The KLR motor is 650 cc, and at 7500 rpm redline, runs 3750 intake strokes per minute, for an airflow of 2.4 million cc/minute. Thus, the blower pumps three times as much air as the engine.

Fortunately, centrifugal blowers can tolerate a high intake resistance, so even though the filters might not be able to handle the full flow of the blower, I could still run them and get pressure differentials.

I attached various filters to the base plate, and read the static pressure on the manometer. The pressure tap was connected to the low pressure side of the transducer; a higher number indicates a higher pressure drop across the filter. The results are presented in the table below. One thing to remember is that as the filter resistance goes up, the flow through the blower goes down, so the numbers aren't directly comparable as they would be for a constant airflow. However, the engine works in a similar fashion; even though it is a positive displacement pump, the compressibility of air means that higher filter resistance will yield lower mass air flow (even though the volume is the same).

The Results

K&N 1: Used for several thousand miles and cleaned
K&N 2: Clogged solid on LA-Barstow-Vegas, mesh banged up, cleaned
Moose 1: New foam filter
Moose 2: Foam filter, used for about 500 miles on street and 100 miles off-road
Twin-Air: New foam filter
K&N filters were oiled with K&N filter oil; Moose, Twin-Air filters were oiled with Bel Ray foam filter oil.

If multiple values are shown in the table, it means that I reinstalled that filter and took another reading. If the reading fluctuated, I took the average value. The "relative vacuum" readings are relative to 2.50" of water, corresponding to the pressure drop through the unoiled filters. If there are different readings for repeated tests in the first data column, I used the average for the reference value.

The "open intake" reading refers to the setup with no filter installed; it is only for completeness, but does show that the filter provides useful flow organization. Note that the high vacuum reading doesn't really mean lower flow in this situation. Without a filter in place, the flow rate is so high that the airflow is separating from the surface as it turns the corner into the inlet, and the pressure port is reading a localized low pressure. This is why I'd like to run these tests with an airbox, or a pipe connecting the filter to the blower, to let the flow reorganize before measuring the pressure.

The K&N and Moose foam filters both had the same pressure drop when unoiled. The K&N filter was hardly affected by oiling the element, but the Moose filter showed a big jump in resistance with oiling.

Photos of the basic setup, K&N filter in test, and Moose filter in test. Click to enlarge.

I was also surprised by the difference between the two oiled Moose filters. The freshly oiled filter had 0.30" more pressure drop than the slightly used filter! The only reasons I have come up with for this are: 1) the small amount of dirt on the used filter has drawn some of the oil out of the filter; 2) I over-oiled the new filter; 3) manufacturing differences between the two filters. I even tried to squeeze more of the oil out of the new filter and retested it, but came up with the same number.

The readings with filterskins were also interesting. Adding a filterskin to a K&N has a much greater effect than adding one to the Moose filter; enough that a foam filter with skin has less resistance than a K&N with skin. I think the reason for this is that the K&N only flows through the sides, while the foam air filter also allows flow through the top. Also, the filterskin has roughly the same surface area on both filters; and it negates the corrugated design of the K&N.

The Twin-Air was a little odd. First, it had no hole in the top for the post on the filter basket to come through. So, I cut an "X" in the filter for the post to come through. The Twin-Air was also a very tight fit on the basket. I had to stretch it a bit, which seemed to make it smaller and give it less surface area than the Moose filter. This may explain the higher flow restriction from this filter. There was no effect (or even a positive one!) from the filterskin! Again, I think the small size of the filter made it difficult to seal the base of the filter to the plate with the skin on. So, there was probably leakage through the base, resulting in lower pressure drop readings than we would expect.

As a result, I'm pretty happy with my reversion to the Moose foam unit so far; flow is better than the K&N with filterskin, and the filter should be easier to clean. If it survives this year's LA-Barstow-Vegas ride without the need for skins or cleaning, I'll be sticking with it. I'll also test a dirty Moose filter after the ride.

UPDATE: I tested the Moose filter (#2) after the 2002 LA-Barstow-Vegas ride. Wouldn't you know it, this was the first time in the 19-year history of the ride that it rained on both days! So, I didn't get a lot of dirt loading. One quarter of the filter was pretty heavily loaded, resulting in a slight increase in restriction. I'll post a picture of the filter/dirt if I get a chance.

Retest considerations

I would like to do these tests with a stock airbox, to better account for the airbox's effect on entrance and exit flow. This would also allow me to test things like the effect of the filter screen, and the inlet flow straightener. I could also place the static pressure tap further away from the filter (i.e., at the carb boot, going into the fan), to remove the effects of entrance flow. If anybody has an extra airbox they'd like to donate, let me know... (It would be irreversibly modified during the tests.)

It would also be very useful to see how the filter resistance changes as they accumulate dirt; however, this would require metered amounts of dirt, and a high-accuracy digital balance to measure the dirt accumulation of the filters. It would give the best answer in the foam vs. gauze (or foam vs. gauze with filterskin) debate, as filter flow over a long ride could be anticipated.

Finally, if I had a fan exit throttle and a way to measure the flow exiting the fan, I could compare the filter pressure drops at a constant flowrate. This would give the most comparable readings.