Second Thoughts About Grounding

[NinerSevenTango]

In my last messages and diagram, I recommended putting a switch in the ground circuit of the LC-1 wideband O2 sensor to allow for narrowband simulation when the scanner is not connected, and to break the extra ground when the scanner is connected.

After thinking about it a little more, I have decided that the switch might be a bad idea.

The reason is, if I installed them this way, it would only be a matter of time (a short time, probably) until I forgot to put the switch in the proper position. If I did, I mean when I did, then if the scanner was not connected and the LC-1 was not grounded, then when I start the car the wideband sensor itself will be damaged, because it cannot handle being in the exhaust stream without the heater on. And if I left the switch on when the scanner was connected, I would run the risk of funky readings or blown scanner or blown LC-1.

So I came up with a slightly better idea, to use a relay driven by one of the outputs on the EIO to make or break the extra ground connection.

I made a drawing for a simplified hookup, while I was at it I figured out the neatest way to do it, considering that since I am going to be using the narrowband simulator, I have to get at the connector from the stock O2 sensor. So I figured I will buy an O2 extension, use the plug off the end that plugs into the factory wiring harness, and use the power and heater grounds on it as well as the O2 signal wire.

Here's the new drawing.

Comments welcome, especially if I made a mistake!

--97T--

[jimmyblue]

I'm thinking that it might be a whole lot simpler in the
end to do this with circutry. For the narrowband
simulator, a crappy old LM339 comparator with the
ground tied to PCM ground, a trim-pot for the IN-
that sets your pivot voltage (dividing from some
LC1-local reference voltage & ground) and the IN+
to the LC1 output, the open-collector output to
a resistor network that makes 1V open circuit
and gets pulled to ground by the comparator,
there's your switching sensor waveform with the
bonus of some substantial ground (common mode)
rejection. That part can take more than 30V on
supply so you can run it bare. Now you have a
0-1V output that requires no external support, parts
cost should be under $10 at Radio Shack including the
little project box.

[99hawk589]

I have one quick question about wiring up the calibration wire. Should it go through the EIO or to another chassis ground? I currently have it wired in-line with the button / led to the ground (pin 4) on the OBDII plug. Is this correct?

[NinerSevenTango]

Should be OK there.

--97T--

[NinerSevenTango]

JimmyBlue,

I've been thinking about your suggestion. I understand how it works, I believe. It will give some isolation between the ECM and the LC-1. (But if the LC-1 common gets noise compared to the ECM common, you could end up with more than -.3V on the input(s), which could cause flipping of the output at the wrong time.)

But what about when the scanner is connected? In that case, we still need to let the LC-1 get its common from the scanner to avoid a ground loop. And this connection disappears when you unplug the scanner.

Still thinking about it .....

--97T--

[S2H]

I'm thinking that it might be a whole lot simpler in the
end to do this with circutry. For the narrowband
simulator, a crappy old LM339 comparator with the
ground tied to PCM ground, a trim-pot for the IN-
that sets your pivot voltage (dividing from some
LC1-local reference voltage & ground) and the IN+
to the LC1 output, the open-collector output to
a resistor network that makes 1V open circuit
and gets pulled to ground by the comparator,
there's your switching sensor waveform with the
bonus of some substantial ground (common mode)
rejection. That part can take more than 30V on
supply so you can run it bare. Now you have a
0-1V output that requires no external support, parts
cost should be under $10 at Radio Shack including the
little project box.

you draw it out...get me some part#'s or at a minimum some part values...
and I'll build it and test it

[jimmyblue]

Here is a schematic of what I'm talking about. Last
I checked Radio Shack still carried the LM339
comparator although they seem to keep pruning
the inventory.

The threshold is referenced to the same source
ground & ref as the signal, and the output is
referenced and fed by the PCM side. So there
should be a high degree of common mode
(ground) error rejection, roughly the comparator
gain which is quite large. I put the series resistors
on the input to guard against gross ground offset
punching out the ESD protection diodes (since
this was mentioned as a weak spot on the LC-1;
current limiting goes both ways).

[NinerSevenTango]

Nice circuit, Jimmy.

In the case where the scanner is disconnected, and the LC-1 is feeding the ECM with simulated narrowband O2 signals, I hadn't planned on worrying about isolation for this signal, since it flips and flops anyway. Doesn't it? What am I missing?

According to the drawings I could find, it looks like the low side of the ECM O2 input is isolated from common, so it can take an arbitrary (but reasonable) ground. (At alldatadiy.com, demo vehicles, 2003 Bonneville V6, electrical diagrams.)

When the scanner is in use, logging wideband O2, you still have to switch the LC-1 ground from wherever it is to the ECM ground through the EIO, right? Where we need the isolation is on the wideband O2 signal, I thought. It should work fine if everything is referenced off the ECM ground through the OBDC2 connector, the problem is then you shouldn't also ground the circuit via the LC-1. An isolator would allow leaving both grounds connected.

--97T--

[jimmyblue]

I saw several permutations of wiring and an unknown
ground offset, so I figured something general-purpose
was the ticket. If the ground is jacked by more than
+500mV (LC-1 > PCM ref) of course you'd be hurting
if using the LC1 to simulate the NB output, because
you'd never swing below 0.5 (and in fact if you look
at the O2 thresholds vs airflow mode in low-flow cells
it's more like 300-350mV).

This little circuit was meant to simply produce a NB
signal from the main LC1 output (presuming it was
centered at 2.5V=15:1 or so; if you have a different
output transfer curve than 0=20:1, 5V=10:1 then
you'd have to revise the threshold, and if you had it
upside down like 0=10:1, 5V=20:1 then you'd have
to flip the inputs + for -).

Still think that an instrumentation amp buffer is the
way to go for eliminating ground offset as an issue
for EIO datalogging, but I need to know the range of
ground offsets to be rejected, before I can select
an single-supply amplifier IC (or determine I have to
generate a negative supply, to use a wider-input-
range one).

[NinerSevenTango]

OK, I understand now, I wasn't considering that we're talking 0 to 1 volt swing. Yep, an offset there would never let it see the low condition. And noise could drive the mixture nuts too.

Phooey! That means your circuit will have to be added to mine.

I looked at instrumentation amps, optical isolators, and little switch-mode power supplies. Per - chip prices are not cheap!

All this is making the extra bung look more and more attractive.

I guess I'll start thinking about it from a temporary / portable perspective now.

Hmmm, how about a cheapo inverter plugged into the cig lighter, and a cheapo 5 amp power supply to power the LC-1? And ground through the datalogger. LOL. Maybe that's not so funny.

On the other hand, if no scanner is connected and there is no ground loop, grounding the LC-1 to the cylinder head SHOULD give a pretty good ground without offset.

--97T--