<shamelessly lifted from innovate>

In order for a device (like the ECU, display, datalogger...) to measure a voltage on a "sensor" (like the LC-1 Wideband O2 sensor) correctly, both devices must use the same ground reference.

Basically ALL ECUs I know use as sensor ground the engine block. They mostly have specific wires for that. They have additional grounding to the block also for high current signals like the injectors. The only time I have seen different is on race cars that are push-started and use only a small battery to supply the electronics and no alternator. The grounding the ECU housing has to the kick panel, or instrument panel and so on is just for shielding reasons (EMI), not for ground reference.

Most cars have the battery negative bolted somewhere to the frame, and then a ground strap connects the engine to the frame. The ground strap and its connection pads are the largest source of trouble.

During cranking the starter current has to flow through the ground strap and those pads. I measured up to 5V ground offset on older cars between battery/frame and engine block due to slight corrosion on the contact patches. If the ECU were grounded to the frame, this ground offset of 5V would add to the sensor output voltage during cranking, so, if a sensor is trying to output say 3V, a device grounded to the frame would see 8V from the sensor, which would be of course wrong and could be destructive.

A lot of ECUs have pretty powerful overvoltage protection that shorts the input overvoltage to ground or clamps it to 5V (which has the same effect). They rather sacrifice a sensor (by shorting) then the ECU itself.

If the LC-1 has part of its ground (through the ECU) to the block, and the heater ground to the frame, in the above case there would be a 5V difference between heater ground and system ground. This ground difference can be a path of least resistance for the starter current through the electronics of the LC-1 and blow it up. In addition the overvoltage protection of the ECUs will short the analog outs of the LC-1.

The reason the LC-1 has separate grounds for system and heater is to minimize noise, that's correct. But its to minimize ground offset noise created if there were a common cable for heater and system, or noise created by contact patches where the LC-1 cables bolt on. So the effect of minimizing noise is achieved when both are bolted to the same metal piece (like the blojust ck), not on the same lug.

Grounding them to radically different grounds creates the opposite effect. Even if the LC-1 survives the starter current, the noise from the radically different ground can couple through the LC-1.

"the back of the head, the ground behind the kickpanel, any other ground in the PCM harness" is wrong. There are good reasons for ECUs to have multiple ground wires. Some are for sensor ground reference, and some are to carry high current injector or ignition currents. But ALL of them, except the ones solely for shielding, typically go to the engine block as that is the best ground in a car. Not only for the above offset reasons during cranking, but also because in operation the alternater is the electrical power source for the entire car. And the alternator is usually securely grounded to the block. In any electrical system the best ground is always found at the source. In a running engine the battery is only a power user, just like any other. Most people think battery negative is a good ground. It is not. You might just as well ground to the ground of the headlight bulbs.



all of this means that it explains why my LC1 sensor was playing up at low voltages... and why my current wiring setup with a mass of earth tags on the chassis where the coil used to be ... is less than ideal.

just figured i should share this...