Novak water vapor injector

This page details some experiments I've made with Ron Novak's water vapor injection system, the design of which is available here. A local copy is here.

This system is from back in the late 1970's, when oil prices took a big jump and got people thinking about mileage. While billed as a water injection system, it's really an adjustable vacuum leak which admits moist air in proportion to the vacuum level.

The amount of 'bypass air' this device allows in is highest at idle (maximum vacuum), and lowest at wide open throttle (minimum vacuum). The device is powered by intake manifold vacuum, and has no moving parts. Air is drawn through an aquarium airstone to break it up into tiny bubbles, maximizing surface area and encouraging water evaporation. The moist air is then admitted to the intake manifold at a point below the throttle butterfly valve.

This system was designed back in the day when carburetors were still used on many cars. In my humble thinking, what it accomplishes is to allow a leaning out of the mixture at high vacuum/low throttle conditions, while adding some humidity that may help both lower chances of detonation and clean carbon deposits out of the combustion chamber. At higher loads, the lower vacuum pulls less bypass air in, and the mixture richens. This prevents detonation by fuel cooling effects and brings the mix back to the factory set point.

If the system does indeed work that way, it's pretty slick. The problem of course is that no modern vehicles use carburetors anymore. To meet emissions and mileage standards, all modern cars are fuel injected. Setting up this vacuum leak would lean the mix, since you're introducing air downstream of the mass airflow (MAF) sensor, but the oxygen sensor in the exhaust would detect the extra air, and the computer should bump the fuel map to compensate, negating any effect the system might have other than the possible cleaning of the carbon deposits. Even so, the system looked so simple I had to try it.

Note: I don't think they got the design quite right in the article. In the diagram, they have the valve in-line between the carb and whatever vacuum device you're sharing the line with. In this configuration, the down-stream vacuum device will get almost no vacuum due to the bubbler being hooked to it and its vacuum supply line limited by the valve. Instead, I think it would be better to put the valve between the bubbler and the T connector. That way the only thing affected by the valve will be the vacuum on the bubbler, and the original vacuum device at the other end will still get nearly full vacuum. Just my thoughts.

My injector system was built from a 2 quart plastic bottle, aquarium airstone, aquarium air valve, and clear plastic tubing from a builders store. I installed it on my 1993 Ford Explorer, tapping an unused vacuum fitting on the back of the intake manifold. The reservoir was tucked behind the driver's side headlight, and filled with a mix of 75% water and 25% denatured alcohol. After warming up the engine, I opened the air valve and adjusted it for a vigorous bubbling of the airstone.

I drove the truck for about 50 miles in town and highway, then used my G-Tech pro to do some horsepower tests. The test involved accelerating from a stop to 65 mph, recording the horsepower indicated by the G-Tech, stop, repeat. All runs were the same day, on the same road, in the same direction. The vehicle has a 4.0l V-6 engine and automatic transmission.

Here are the results:

Injection off

Run	Horsepower
1	100
2	99
3	113
4	97

Injection on

Run	Horsepower
1	110
2	105
3	105
4	104

If we discard the highest and lowest reading for each category, and average the remaining two readings, we get:

Injection off:	100 HP
Injection on:	105 HP

Considering that I wasn't expecting there to be any difference with the bubbler on or off, I was quite surprised at the results. A 5% horsepower increase isn't huge, but it is significant. The absolute horsepower reading is probably off, but that could be due to me entering 5280 lbs for the vehicle weight into the G-Tech. I actually don't know how heavy the truck is, but that's what it said on the sticker on the doorjamb. The stock horsepower for that engine is listed as 210. In any case, what we're interested in here is the change with and without the bubbler running.

After the performance tests I drove the truck for a combined total of 139 miles. The average mileage was 11.5 mpg, which is not very good even for this vehicle. Time in the fall ran out and I didn't get a chance to do a proper mileage run both with and without the system. I removed it for the winter, since it gets VERY cold in North Dakota, and the last thing I wanted to risk was an iced-up throttle valve.

Note: I later learned that the oxygen sensor was faulty, causing the computer to go 'open loop' and run a rich mixture. The bubbler in this case would slightly lean this mix and most likely give more power. A test with a proper O2 sensor is needed.

September 29, 2009
It'd been several years since I last experimented with this system, but I decided it was time for a bit more testing. The original system suffered a bit from the bottle not being as strong as it should have been, so I constructed a new system. The bottle for this system was a sturdy plastic storage jar with a silicone gasket. Holes were drilled in the lid for the intake air hose and a plastic plumbing valve, which was secured to the lid with JB Weld epoxy. The intake air hose was fitted with an aquarium air stone, and the output valve was plumbed to an unused vacuum port on the intake manifold. A sponge was also added to the bottle to act as a baffle to try to discourage the intake of liquid water (vs. water vapor). The system was installed and filled half way with a mix of about 1/3 isopropyl alcohol and 2/3 distilled water. Here are some images:

Bubbler jar

Vacuum port hookup

Whole system

Instead of acceleration testing, this time I made some runs to measure the mileage of the truck. First I topped off the fuel tank, then drove to a town about 25 miles away and back. I used cruise control to keep my speed as constant as possible. This was done with the bubbler valve closed, to deactivate the system. The tank was topped off and the amount of fuel added recorded. The valve was then adjusted such that the water/alcohol mix bubbled and foamed vigorously, and the whole thing repeated. Since the road, weather, and speed of the two runs were virtually identical, any differences should be due to the injection system. The weather conditions of the test were:

Air temperature:	50 F
Winds:			ESE at 8 mph
Humidity:		64%
Barometer:		30.08"

The results:

Control run (system off)	17.4 miles per gallon
Experimental run (system on)	18.7 miles per gallon
Difference	+1.3 miles per gallon
Gain	7.5 %

Unlike the power tests above, this time the Explorer had a functional oxygen sensor, and the computer never flashed a 'check engine' light at me. I'm happy with the result, though the skeptic in me is still looking for mistakes in the experiment. A proper study would entail many more than two runs. Longer runs would help, on a test track to negate differences in wind and weather that overland travel might introduce. This was simply what I could easily squeeze in after work, over two hours, and expending about $14 worth of gas.

Still, a 7.5% gain is very close to the 6% gain that Ron Novak noted in his original system back in 1979, and I find that rather remarkable. More testing is certainly warranted.