Addendum to WPS LPG automotive fuel conversion


This is the addendum to the original How-To...LPG booklet, written in 1990. A lot has happened since then, though amazingly what's in the booklet is still correct. This addendum covers some of the hard parts that followed, such as fuel efficiency.

This addendum contains followup technical, operational, and subjective data that I simply didn't have when I wrote the booklet -- obviously missing was fuel mileage data, long-term operational and tune up issues, reliability, etc. While these are not really adequately dealt with here, it's probably all I'll ever do on it.

The real problem with this whole project, from a documentability point of view, is that it wasn't simply a science project -- it was also my daily transportation, a fun toy, an art project (I'll cover that part of this car in a separate, much more-fun document, in the next few months XXX years). When I started this in 1987--1989, I was quite poor, using scrounged and occasionally inadequate parts for the rest of the car, necessitating doing some things twice. Everything done on this project during that period happened in extreme slow-motion, if at all.

The fundamental drive behind this project remains: if gaseous fuels are so good for internal combustion engines (mainly in terms of overall fuel efficiency and vastly reduced emissions) why isn't it common? In essence, is there a catch? The answer remains no -- there is no catch. It's industrial laziness, and lack of vision on a national level. And yes, this is not a final solution to any transportation problem, only one component in the existing system.

-- Tom Jennings, January, 2003


Fuel mileage on LPG for my car is, considered as a system, fairly good. A factory-stock 1963 Rambler Classic Wagon, outfitted with manual overdrive transmission and 196 ci. (3.2 liter) engine, got 19MPG best case on the highway, which works out to 6630 BTUs of energy per mile. Mine has a much-newer 232 ci. (3.8 liter) engine, a non-overdrive automatic transmission; total 1996 average fuel mileage was 14.22 miles per gallon LPG, about 6444 BTUs per mile, equivelant to 19.55 MPG gasoline. Considering my larger engine/less-efficient transmission, I'm netting a decent improvement over the best this car ever could do on gasoline.

(I started recording fuel consumption/mileage data in 1994; with a few gaps due to screwups on my part, I have a record of every LPG purchase, odometer reading, date and usually location since.)

(September 2002 -- a recent trip from Los Angeles to Santa Fe and back, with a lot of meandering, some 3000 miles, the mileage remains over 14 mpg LPG, with a peak (300+ miles) at 15.6 mpg LPG, approximately 21 mpg gasoline equiv.)

These improvements are due to not to a magical fuel, but to a vastly but simply improved fuel and ignition system. Hand in hand with improved mileage goes exhaust emissions; while I don't have complete data on before/after emissions, the emissions improvement is mostly due to LPG fuel:


Relative fuel usage, compared in BTU per mile. Assumes 60 mph, as mileage
is measured at more or less that speed on long trips; presumably BTU/mile
is lower at lower speeds due to wind resistance and internal friction.

BTU::fuel data from

LPG     Gal/    BTU/    equiv gasoline
MPG     Mile    /mile   MPG
11      0.091   7664    14.88
11.1    0.090   7595    15.01
11.2    0.089   7527    15.15
11.3    0.088   7460    15.28
11.4    0.088   7395    15.42
11.5    0.087   7330    15.55
11.6    0.086   7267    15.69
11.7    0.085   7205    15.82
11.8    0.085   7144    15.96
11.9    0.084   7084    16.09
12      0.083   7025    16.23
12.1    0.083   6967    16.36
12.2    0.082   6910    16.50
12.3    0.081   6854    16.63
12.4    0.081   6798    16.77
12.5    0.080   6744    16.90
12.6    0.079   6690    17.04
12.7    0.079   6638    17.17
12.8    0.078   6586    17.31
12.9    0.078   6535    17.44
13      0.077   6485    17.58
13.1    0.076   6435    17.72
13.2    0.076   6386    17.85
13.3    0.075   6338    17.99
13.4    0.075   6291    18.12
13.5    0.074   6244    18.26
13.6    0.074   6199    18.39
13.7    0.073   6153    18.53
13.8    0.072   6109    18.66
13.9    0.072   6065    18.80
14      0.071   6021    18.93
14.1    0.071   5979    19.07
14.2    0.070   5937    19.20
14.3    0.070   5895    19.34
14.4    0.069   5854    19.47
14.5    0.069   5814    19.61
14.6    0.068   5774    19.74
14.7    0.068   5735    19.88
14.8    0.068   5696    20.01
14.9    0.067   5658    20.15
15      0.067   5620    20.28
15.1    0.066   5583    20.42
15.2    0.066   5546    20.56
15.3    0.065   5510    20.69
15.4    0.065   5474    20.83
15.5    0.065   5439    20.96
15.6    0.064   5404    21.10
15.7    0.064   5369    21.23
15.8    0.063   5335    21.37


Until recently, the biggest annoyance of using this car as a daily-driver was due to the combined factors of fuel availability, vehicle range, and tank-filling ambiguity. The problem wasn't miles per gallon, per se, but the real-life vagaries of availability.

To illustrate: in a gasoline car, in general around-town driving, you might drive until you know you have only a few gallons left, then fill up at your leisure at one of the plentiful 24-hour filling stations. You might put off filling up until the next day if you're in a hurry.

For example, you have a 14 gallon fuel tank, and get 20 MPG, and stop at a filling station when you have two gallons remaining; when you add fuel to the tank, you easily get within 1 gallon of full. These factors combine to give you an average-worst-case practical range of (11 gal*20mpg=) 220 miles between fillups.

I originally built this car with an 18 gallon LPG tank (at 80%), and get 14 MPG, with decent efficiency, as noted above. Now most of my filling-stations are closed after 5pm and on Sundays; in general I don't like being under 1/4th full, which is about 4 gallons (though I'll certainly run to the grocery store without worry). When I add fuel, or more accurately, when the poorly trained, slightly fearful low-level employee at the LPG filling station pumps fuel, I frequently do not get a truly full tank; I can't count on having more than 16 gallons in my 18 gallon tank. These factors combine to give me an average-worst-case practical range of (12 gal*14mpg=) 168 miles between fillups. If I don't fill up Saturday, I can't do much driving on Sunday without a search for fuel.

(The above applies mostly to local driving; on long road trips, the distances covered generally means you find LPG pretty much anywhere, though lessened on Sundays.)

The solution to this is obvious -- since I can't change availability, increase tank size! In February 1997, I installed a 35 gallon/28 gallon liquid capacity tank that increases my effective range to 300-350 miles, which utterly transformed the car -- I buy fuel every few weeks, and on long road trips, buy fuel once per day. For local driving, I sometimes drive to a dealer in Gardena (from Hollywood) and fill up -- it's worth it when you're buying 20 gallons at $0.81 per gallon. If I'm lazy I pay $1.40 per gallon at a tool rental place three blocks away.

I can't emphasize how much this has improved the car. The downside was, I had to gut the trunk space under the rear of the station wagon to mount the tank; until I find or make a "continental kit" for the spare tire it sits on the roof (ugly). As mentioned in the LPG Book, commercial LPG tanks come in standardized sizes and shapes, and are therefore not contoured to the chassis like OEM gas tanks.

Take this as a BIG HINT -- you want absolute maximum fuel capacity for a daily driving LPG car, I would suggest my rule of thumb, a real 300 mile range that leaves you with a few gallons for peace of mind. And you must do this safely -- no tanks or lines in the passenger compartments, etc -- with off-the-shelf parts.


I've gone through too many iterations of LPG hardware to get where I am. A lot of it is certainly my own inexperience, but a too-large portion of it is incredibly poor advice from so-called experts and a complete lack of interest and documentation from equipment manufacturers.

I'll skip all the excruciating detail. Suffice to say I have a nice normal inline-six of 232 cubic inch capacity (about 3.8 liters) that produces only about 90 horsepower but tons of nice twisty torque. The engine is now dead-stock except as follows:

        Impco model CA125 LPG carburetor
        Impco model 'L' LPG convertor
        Autotronics MSD-5 electronic ignition
        Autotronics 4046 Close-Loop Oxygen Controller
        1979 AMC distributor/electronic ignition
        Two-wire oxygen sensor in tailpipe

This stuff is not expensive. The CA125 carburetor is probably the most common LPG carburetor, has two moving parts, and requires a $50 tuneup every 50,000 to 100,000 miles. Mine was new in 1989 and has been tuned up (replace diaphragm) once, in Jan 1996. (Impco's patents ran out on the CA125, and it is now cloned; in fall of 2001 I bought a brand new Nolff CA125 mixer for $59.00, less then the cost of the Imco brand rebuild kit!) Before I installed the 4046 Oxygen controller, in 1995, I constantly adjusted both low speed and high speed mixtures trying to optimize mileage, to no avail. Since the 4046 was installed, and carburetor adjusted, very little adjustment has been required; these were small adjustments to put the closed-loop in the center of it's range (see raw notes, above) or to change idle speed for subjective 'feel' improvements. Please see the carburetor selection nomograph to select the right carburetor -- you would-be hot-rodders, don't try to select a larger carburetor, they don't work like venturi-based gasoline carbs.

The Model L convertor was also purchased in 1989 and tuned up (diaphragms, valves replaced), also in 1996. It has no adjustments.

The MSD-5 ignition I purchased in 1985 for another car; I simply installed it on this car, originally on the old mechanical-points distributor, then later on the replacement 1979 distributor. It has no adjustments of any kind.

The 4046 O2 controller required one go-round of carburetor low- and high-speed mixture adjustments at installation time. The 4046 dynamically sets the fuel/air ratio at the carburetor, so the manual carburetor adjustments consist of setting the carburetor slightly rich so that the 4046 can lean it out on-the-fly by measuring tailpipe oxygen content. (It applies a modulated engine vacuum signal to the atmospheric vent on the regulator; it can lean the mixture but not richen it, hence the static setting is rich.)

(September 2002: I'm embarrassed to say I just fixed a major problem with the installation of the 4046. The 4046 modulates a vacuum source applied to the convertor to vary the mixture; I used manifold vacuum. WRONG! It *must* be picked off from a weaker source, such as the recommended mixer, below the diaphragm, above the throttle butterfly. The installation sheets are very vague about this -- and it's critical. It worked on manifold vacuum, though not well; symptoms were motorboating (hunting) as the slighest "on" pulsing of the 4046's solenoid would radically lean the mixture; the weaker vacuum source slows the loop response.)

The 1979 AMC distributor was chosen for a number of reasons, after a lot of agonizing "research". There seems to be very little acknowledged "science" in setting ignition timing curves; suffice to say here, the 1979 distributor had a much more agressive advance curve than the original distributor, which is basically what I was after. I plotted measured distributor advance curves by hand. It also had the advantage of being all-electronic, eliminating annoying and hard to adjust points, and all but eliminating shaft wear due to points-spring tension (a major source of hard-to-tune older engines, by the way!).


To be blunt, I now consider a closed-loop fuel system, on any car of any kind, to be entirely non-optional. I am now quite convinced that all older carbureted, open-loop fuel systems have essentially the same problem I had with mine, regardless of fuel, but most people are not measuring just how awful it really is.

Without the O2 controller, I simply could not get an optimum mileage/performance configuration. I'd end up with good low-end performance, but high-speed fuel consumption was terrible, or vice-versa. Most of my research was done on the road, through detailed and copious notes, but I also paid for four hours on a road dynamometer in 1991 which resulted in little new data (essentially I found less-non-optimum settings for the see-sawing adjustments).

The closed-loop system simply solved all problems. I had a few gyrations to go initially with optioning the carburetor, see notes, below. Essentially, you tune slightly rich with the loop open, eg. solenoid disconnected etc; the 4046 leans the mixture when the O2 sensor in the exhaust says it's rich. drops). With these settings, running open-loop, low-speed performance is reasonably OK, high-speed performance OK but mileage not great due to the tendency to run rich on throttle-up and hard accelleration.

Closing the loop, by enabling the 4046, causes the Controller to measure oxygen content at the tailpipe, through the oxygen sensor installed there, and modulate the fuel/air ratio produced by the carburetor by pulse-width-modulating a vacuum signal applied to the LPG Convertor. A small solenoid quietly buzzes as it dynamically adjusts the vacuum signal; measuring the average voltage across this solenoid is a direct indication of the loop status; 5-10V means the system is able to maintain the loop, with higher voltages indicating that it's having to apply more vaccuum to lessen the fuel to lean out the mixture; this is called the "error voltage" because it is directly proportional to how much the controller has to change the mixture (eg. how far is is from perfect.)

While a decent description of closed-loop servo systems doesn't seem appropriate here, you should note that the carburetor isn't controlled with a smoothly-varying adjustment provided by the Controller; instead, it pulses rich/lean/rich/lean... dozens of times per second, and measures the average result at the tailpipe. This works just fine because the motor is a big, slow hunk of iron, and the fuel/air mixture moving into the cylinders is also big and slow, relative to the nimble rich/lean adjustments; the big/slowness averages out the pulses into an effectively smooth mixture. You will notice that the average voltage across the solenoid varies slowly up and down somewhat; this is called "hunting", and a small amount is quite normal. If the engine speed "motorboats" fast and slow you've got a problem; this is covered in the Autotronics documentation. The varying of rich/lean/rich/lean... is called Pulse Width Modulation.

[The oxygen sensor by itself is also useful as a tuneup indicator. Please note that it is decidedly non-linear; it CANNOT IN ANY WAY be coerced into giving you a direct percent-oxygen measurement! It measures one specific thing: the presence/absence of oxygen on the hot side of the sensor, ie. inside the tailpipe, relative to the cold (outside) side. I installed the O2 sensor a few months before the 4046, and was able to closely set low speed mixture by enriching until the sensor indicated no O2, then leaning out til O2 is indicated.]

One device I found to be an utter waste of time and money was the Autotronics 4052 LPG Ignition Recurve Computer. It essentially recalculates a new ignition curve based solely upon engine speed, and while it might be useful for very heavy vehicles it generally just caused pinging on my car. It's a pain to adjust too.



The secondary effects in this area are all good. As is well known, reciprocating engine piston rings are far from a perfect seal; some fuel/air mixture always leaks past them, and worsens with age. While this is true for LPG as well as gasoline, when the cylinder walls are relatively cold, liquid fuels condense on the sylinder walls and wash the oil film off, ruining lubrication and vastly increasing wear. The liquid gasoline carries with it unburnt fuel components (most visibly black carbon) which ends up in the engine oil -- a quick glance at the dipstick on any engine will show you the result. New engine oil, a dull yellow color when first installed, becomes predominantly grey/black after 1000 miles or less.

While gaseous fuels also leak past the rings into the crankcase, they do not wash oil from cylinder walls, and except for extreme age or mechanical trouble don't generate bulk carbon, and hence the lubricating oil stays very nearly uncontaminated. Because they are gaseous, they are easily recirculated into the engine via the PCV system. (Probably some trace amount of fuel is also dissolved in the oil.)

Fact: my engine now has over 110,000 miles [Nov 98], consumes no (zero) oil, and engine oil stays bright yellow after 4000 miles (when I change it). I have gone as far as 8000 miles between changes, and the oil only darkens slightly. (September 2002 note: at 180,000 miles there is a little seeping from the rear seal; now at 3000 miles the oil is a greyish yellow.) It seems to need the valves rebuilt, based upon comparative annual cylinder-compression data from 1991; as mentioned in the original edition, this engine doesn't have hard valves, and the squirt-of-oil test seems to indicate the dropping compression (115-125, from 140-145 in 1991 [140-130 in 2000]) is valve wear.

Oil needs to be changed regularly anyways, as the molecules sustain shear damage and chemical changes due to heat, acid buildup from unburnt fuel, impurities, etc.


In 1987 rebuilt my engine, from a bare machined engine block on up, in the standard manner, and ran it on gasoline for two years/27,000 miles. In 1989 I installed this engine in the curent car, 87,000 miles on the speedometer, at which point it was converted to LPG fuel. It now has 170,000 total miles on it [Nov98], 90,000 on LPG. It does not seep or drip engine oil from any gaskets (except an annoying valve-cover gasket that I poorly sealed). I believe this to be due to the lack of gasoline and its combustion by-products, a very active solvent, in the engine oil. I certainly didn't do anything extrordinary when I assembled the engine.

One 'undesirable' (sic) effect is that the underchassis of my car has rusted, because the engine oil that would normally spray and drip under the engine is not "undercoating" the chassis. If you think I'm exaggerating, casually examine a half dozen or so gasoline cars over 10 years old and you'll see that their undersides are coated with a sticky black, tarry substance made of engine oil and road dust. Engine oil drips and mist is also the source of the darker section or dark spots down the center of the driving lane on roads.


These are random notes from my files over the years, in no particular order.

1995: One on-road failure: whiel cruising at 60MPH, car suddenly acted
fuel-starved (depressed throttle, engine would 'bog', let up, depress
only slighlty, engine cruised OK; could not maintain speed). Pulling
over to the side of the road and lifting hood revealed LPG convertor
had turned into a giant snowball -- water flow ceased. I unbolted it
from inner fender (two 1/4-20 bolts), removed water access cover, about
100mL of sludge fell out. Convertor, mounted at slight angle on inner
fender, had accumulated sludge from cooling system, eventually plugging
flow. Dumping out solved on-road problem. Months later [it took a few
years to plug up, no hurry here] moved convertor, made level.
Jan 96: Convertor rebuild. Wear was reasonable; stretched diaphragms, wear on valve faces, etc. Wax buildup. Cleaned with Dr. Bronners peppermint soap. Rebuild took 30 minutes.
Convertor runs fine on it's side; however wax will build up instead of running out the gas outlet. Is there a simple way to drain it? How often? With convertor lower than carburetor, wax did puddle insignifigantly (2-3 mm) in the bottom of the vapor hose nipple on the convertor after many months of operation. [September 2002: this is a false worry. Ignore it.]
CA125 throttle ass'y not symmetrical; cannot be rotated 90 degrees! However adapter Impco #A3-32 mates a CA125 mixer to a T2-4 throttle body (fits CA225) which is symetrical.
The CA125 has had the lean air dam assembly (AV1-14-4) installed since July 1993. It adjusted reasonably well for idle and high speed operation prior to Oxygen Computer installation. With the 4046 installed, it was difficult to get it adjusted for both idle and high speed; idle was fairly simple, but high speed required +2 notches rich (R) on the mixer assembly, derived by cruising at speed (60mph), accelleration/decelleration, etc, and adjusting the mixture control until the error voltage stayed within range. It tended to run generally too lean, and the error voltage was erratic. When I installed the rich airdam assembly (AV1-1447-2), I set the mixer mixture to centered (0), and assumed I would have to repeat the above process. The idle required only 1/4th turn leaner with AV1-1447-2 vs. AV1-14-4. When I set out to road test, I was surprised to find the 4046 error voltage stayed in range with the default mixture setting (0) -- under all conditions, including heavy accelleration (50mph up to 75mph), except for lean-out on high speed throttle-up (same as before).