In early April this year (2022) my beloved shitbox left me on the side of the highway. Despite cruising at a mere 60 mph (take "cruising" to be a figure of speech as cruise control was a premium, non-default option you'd buy from the BMW factory) the check-engine lamp began flashing and the car began to lose power. It always ran a little rough but never rough enough to trigger a critical failure like this. I knew this was a serious issue and I immediately took the car out of gear then coasted to a stop and turned the thing off. For once reading old, near-dead car forums paid off as I had learned from the experience of others that a flashing check-engine light (CEL) on an OBD-II car (on OBD-I this is only a minor fault) was bad bad not good and indicitive of something that had become seriously fucked.

Not willing to leave it overnight on the side of the road where my beautiful windows could be shattered I called a friend and we loaded it onto his trailer. We parked the car at his garage and I went to work on it the next weekend. When we rolled it off the trailer with the key in position 2 a strong smell of gasoline appeared; further it seemed some gasoline had dripped onto the concrete below. This was the second indication that something was fucked.

Clearly this was the immediate problem I needed to fix and may have been the cause of that loss of power. Because fuel leaked from the car I decided to start by diagnosing the car's Fuel Delivery system. The fuel lines running from the rear of the car to the front were dry. I lifted the car up, crawled underneath and traced the wet fuel to the upper intake manifold; I probably could have found this without lifting it up but ¯\_(ツ)_/¯

It appeared fuel was leaking from the top of the first cylinder injector and dripping down onto the lower intake manifold and finally spilling onto the floor. To get to the injectors on the 318ti you need to first remove the upper intake manifold.

After pulling the injectors I found that the rubber o-ring that seals the injector to the rail was torn so I ordered a rebuild kit and replaced it. Perhaps the injector was not seated well, causing fuel to push past the o-ring. In any case it was a shame as I already replaced the o-rings two years before when I bought the thing. I had recently removed the rail to try swapping in a new set of injectors I bought from eBay to cure my rough idle (spoiler: they weren't all four from the same car so the thing terrible) so the theory I hadn't reseated them correctly makes sense. Plus the injectors are held in place by metal clips that I probably installed wrong.

After I fixed this problem I tried to run the fuel pump manually to check for leaks by shorting pins 30 and 87 of the Fuel Pump Relay; unfortunately I shorted the wrong two pins and then the pump would not prime after turning the key. Without pressure in the fuel rail the car would not start.

It was around this time that I was to travel to Israel to meet a handful of people at Levana in-person, and with that only flight only a few days away I decided to have the car towed to the house I just bought, quite a good drive away from the garage where it presently sat. The car stayed in the garage at my house even before we'd stayed a full day there.

As an aside, having the car towed twice in a month convinced me to buy a truck, but unfortunately I've not at this time acted on that conviction; there are very few small US trucks made in the last 15 years and I don't care for a very large vehicle, just one that's equipped to tow a trailer with a car weighing a US ton or two on top.

The E36 sat for a while until life settled down. We visited Israel, Terra collapsed, we moved in to the new house, welcomed a 3 month old puppy into our home over the Summer and finally work became a regular cadence some time around September. I finally began to work on the car seriously again around this time. When you have at least one "reliable" car you can afford to do this instead of rushing to fix it and potentially making things worse.

I began to diagnose what I had fucked up by shorting pins 30 and 87. Logically I started at the fuel pump in the back and worked my way forward. I tore out the back seat cushio (surprisingly removable on my car) which gave me access to the fuel pump. There are two devices stuck in the fuel tank: on the driver side there is a fuel-level sender unit, and on the passenger side is the fuel pump, which also has a sender unit attached. For some reason the fuel tank is physically separated into two sides with a pump between them; each sender measures the fuel level on its corresponding side.

Pulling fuel lines was a reminder that the OEM parts were manufactured and installed before I was alive, and the car was rolled onto a lot in November '96, probably a few days after my birthday.

Anyhow, I did not realize the sender unit is behind the driver and the fuel pump is behind the passenger. Pictured above is the top of the sender unit behind the driver. I had started working on the wrong side, treating the sender unit as if it was the fuel pump. I probed voltage at the sender unit and saw voltage; I therefore figured I must have blown out the fuel pump somehow, because the unit was getting adequate voltage and should be running but it was not. I ordered a new fuel pump, dusted off my hands and called it a day.

When the pump arrived I tore out the old sender unit and went to replace it with the fuel pump only to find there was no pump on that side.

It was at this moment I realized I fucked up for what felt like the 10th time in this arc. I muscled the sender back into the tank (with the o-ring it is a very snug fit) and probed the unit in the other side of the tank. As it turns out the fuel pump was not getting voltage at all so it's hardly surprising it wasn't running.

I tracked the problem back to the fuel pump relay. Here is its electrical schematic:

  • Pin 30 connects to battery voltage at all times
  • Pin 87 connects the relay to the fuel pump in the rear of the car.
  • Pin 86 is battery voltage when the key is in position 2 and the antitheft module is not locked down
  • Pin 85 is a switched ground coming from the DME (engine control unit)

The first thing to do was to make sure the relay functioned correctly when tested in isolation. I took the relay out and bench tested it by applying a 9v battery to pins 85 and 86, then testing the resistance between pins 30 and 87. As expected when the +9v was applied 30 was shorted to 87, ie 0Ω, and when the voltage was removed the circuit was open, ie ∞Ω. It's interesting that when the relay is toggled I hear a physical "click" as if there is a literal switch inside the relay being opened and closed.

Now I tested the voltage inputs to the relay. These are ideal conditions:

  • Pin 30 should be battery voltage at all times
  • Pin 86 should be battery voltage when the key is in position 2
  • Pin 85 should be ground when the pump should be running
    • That is, when the key is initially put into position 2 and when the engine is turning over

I found that all of the above were satisfied except the last point. When I turned the key pin 85 read 6.5v which is far higher than ground voltage, ie 0v. Therefore the voltage being applied to switching part of the relay was less than the 9v I had tested the relay with; it was therefore not guaranteed the relay was being switched closed when this meagar voltage was applied. I found only a few people on forums with this problem and 2 out of 3 times they resolved it by replacing the DME. That's a very complex procedure on a '97 E36 because of the antitheft EWS module which needs to be synced with the DME else it will not prime the fuel pump or fire the spark plugs. Without fuel and spark the car just has air which is not even enough to run a hot air balloon.

To test this theory I decided to short the pin myself; I stripped the ends of a piece of wire, wrapping one end around an exposed bolt near the fuse box and securing that by tightening the nut, pinching the exposed copper between the nut and chassis. The other end of the wire I ran into the bottom of the socket that holds the fuel pump relay.

The wire was a bit loose on the relay socket end so I used a bit of pomade to hold it in place. I finally reseated the relay and fuse then turned the key; the fuel pump began to prime and I saw 40 psi in the fuel rail!

Shorting to ground (ie shorting it to zero) essentially removed the DME from having say as to when the fuel pump should be on: in this configuration the pump will run all the time that the key is in position 2 or the engine is spinning, which for my purposes is almost all of the time as I don't often idle in my car if the engine is off.

This is not a permanent solution however or one that will really get me farther than the inspection station (my vehicle registration lapsed in July), but it's a problem I enjoyed solving because it taught me a lot and reminded me my EE degree is not entirely pointless. I will probably replace the fuel pump in due time anyhow as the original one from '96 is still in there. Likely I will also pull the DME and inspect it for damage, potentially replacing the transistor that switches on the ground for the fuel pump relay.