Autotronics 4046 Close-Loop Controller

Autotronics, makers of the MSD aftermarket ignition systems, makes all sorts of weird little boxes for internal combustion engines. Some time in the 1990's I bought the 4046 (since replaced with the smaller and better but plug-compatible XXXX). In short, it's great -- it solved all of my problems working out a correct state of tune, and improved mileage and power. It took a little work however to get it to cooperate, partly because the install documentation sucks.

A closed-loop controller like this turns your engine into a big feedback loop -- by measuring the amount of oxygen left in your tailpipe after combustion, it adjusts the fuel::air ratio dozens of times per second, following the needs of the engine and your throttle foot.

"Closed-loop" is simply the term for systems that employ feedback -- adjusting some input (tweaking carburetor fuel::air ratio) in response to some measured output (tailpipe oxygen).

The 4046 works by applying a weak (very!) vacuum signal to the vent port on the LPG converter; the vent comes plugged with a little screen and filter, but as per the Autronics docs, when you pry it out, surprise! it's a threaded hole ready for the plastic nipple provided by Autotronics.

... and it turns out that little 90-degree elbow is special indeed. Read on!

Installation isn't that hard, but the printed documentation leaves out some incredibly important -- but easy to satisfy -- subtleties. Basically, install as they say:

• Have a muffler shop install a threaded bung for an oxygen sensor in your tailpipe, as close to the manifold as you can practically get (it wants to stay hot!).
• Pick a cheap, common oxygen sensor; they're all functionally the same. "Three wire" sensors have a ground wire, a wire to switched battery for the internal heater (see note below), and a wire for the sensor signal. Don't ever apply any voltage to the sensor wire -- you'll be out $40 real quick.
• Bolt the 4046 box to the inner fender (it's electronic, don't get it too hot).
• Wire it as required (it's not power hungry).
• Attach the funny solenoid to the LPG convertor.
• Follow *MY* suggestions on plumbing the vacuum signal!

Properly setting up the Autotronics 4046 vacuum signal

You want to pick the vacuum signal from the LP carb chamber that's above the throttle plate, but below the diaphragm. Vacuum here is fairly constant 6 -- 8" water column -- very weak! Impco carb's have two ports up there, just under the air filter flange, plugged with #12-24 fillister head screws (on my CA-125). Impco makes small, brass threaded nipples to convert them to signal pickoffs (I maintain a good junkbox; they're not standard parts, you'll have to improvise if you don't have the right parts.)

Route this vacuum signal to the solenoid, which is in turn routed to the VERY SPECIAL elbow screwed into the LPG converter.

The elbow has a small hole drilled in it, opposite the vacuum inlet. Atmospheric air leaking into this hole weakens the vacuum signal even further, as required.

Making the basic mixture adjustment

Therefore, you start out by setting the mixture on the rich side of normal, so that the 4046 can pull it leaner in response to the sensor's tailpipe reading.

There's hardly anything to adjust, really. You should have an average-reading voltmeter applied to the solenoid as described in the documentation; see the next section for more info on this.

My recommended way to check this step by step is as follows. I assume you have everything installed properly.

• Make sure the voltmeter is attached to both leads of the solenoid.
• Turn the car off.
• Turn the ignition on but don't start the engine. For about 20 seconds, the 4046 will pulse the solenoid on/off, making about 6V average. You can hear and feel the solenoid pulsing. If it's not, it's installed wrong. If it is, you should read about 5V -- 7V across the solenoid. If you don't your voltmeter setup is wrong. After about 20 seconds the pulsing will stop, it's expecting the oxygen sensor to be warm now etc. Turn the ignition off and on to re-start the 20-second test period.
• Proceed only when you have success with the previous step!
• Start the engine and bring it up to operating temperature.
• Now the average voltage across the solenoid reflects the dynamic adjustment the 4046 is making to bring the engine to perfect tune; it could be 0 Volts, 14 Volts (full battery), or somewhere between. If it's 6V - 9V you just lucked out (fat chance).
• 0V=LEAN (or cold sensor) ------- 6V=in range ------- 14V=RICH
• Set the idle to 1000 - 2000 RPM.
• Turn the low-speed mixture adjustment screw IN to richen, out to lean it out. You should see the average solenoid voltage change in response.
• Note that the average voltage mught hunt back and forth; it almost certainly will NOT sit solidly at one voltage. It never will, it's a dynamic adjustment. However, if the excursions are really large, see below for LOOP PHASE ADJUSTMENT.
• If it remains too lean, you might have a vacuum leak, too big a PCV valve, etc. Pinch off vacuum accessories one by one to see if it changes and go fix those. If it's too rich, well that's weird, start searching.
• Set the idle back to something sensible (600 - 600 rpm, whatever). If you get a rough idle, or single-cylinder misfires, don't despair, read below for the simple fix.

Adjustment of the 4046 is like herding cats; as long as it wanders around the recommended average voltage (6V - 9V) you're done.

Rough idle

On the intake manifold, right over the #6 cylinder, is a vacuum pickoff, that I used to feed the VFF-30 filter/shutoff. When I piched off the hose, the engine idled perfectly. Neither the hose nor the VFF-30 was leaking. I moved the pickoff for the VFF-30 to the base of the carb (PCV feed) and this solved it. I believe the problem is that the VFF-30's substantial internal chamber was messing with the pressure pulse in the intake, which is quite weak at low speeds, and since the mixture is not grossly over-rich as in the stock system, the weakened #6 mixture wold not fire.

I am going to look at resizing my PCV system to see how it affects phase response.

Loop phase adjustment

However, all engines respond differently, depending on many subtle factors (even the length of the hose supplying the vacuum signal to the convertor).

To make a long story short, you can change this response time by tweaking that little bleeder hole in the back of the plastic nipple screwed into the convertor! (I don't want to admit how long it took to work this out; it's in the documentation, sort of.) The stronger the vacuum signal, the quicker the system responds to errors in the feedback loop; you'd think faster-is-better here, but it's not that simple (we're getting dangerously close to servo system theory). Too quick a response tends to cause overshooting the goal, causing wild swings in the error voltage (the voltage across the solenoid).

Basically, the rule of thumb is this: if the voltage across the solenoid swings back and forth more than a few volts when the engine speed and load are steady (eg. at idle), you can try increasing the size of the hole in the elbow. This weakens the vacuum feedback signal; more signal has to be applied to have an effect, dampening it's response. If you make the hole too big the system will eventually stop working...

Mine oscillated from 5V to 10V, about twice a second, at idle. I increased the hole two drill sizes (go slowly, it's hard to drill holes smaller) and got the excursions down to about 1 volt, and a little more slowly. Quit while you're ahead, don't try to get it flat and steady, it's a waste of time.

(I'm embarrassed to admit that I actually ran mine from manifold vacuum, not carb diaphragm vacuum; I never got it to work right. One day, years after installation, I read the installation docs line by line, and found the obscure reference to the vacuum source. I worked out the details the hard way, and this page is the result. Live and learn.)

Voltmeter setup

I ended up making a slow-response voltmeter expressly for this purpose; I bought a 15V full-scale meter movement from Radio Shack, mounted it in one of their project boxes, then put a big capacitor across the terminals, and a resistor in series with the + wire, the time constant is about 1 second.