legacycentral bbs

Yes, the dreaded knock sensor. Here's the order of events:

First, the car was running great, but I soon fell victim to the common hesitation problem, so I figured I should replace the knock sensor. Took out the old one and it was cracked (big surprise ) Being broke, I went to the yard and pulled one from a my94 NA Legacy that wasn't cracked, and dropped that in. That worked for about 3 months until I noticed hesitation again. No CEL yet. But realizing, or rather hoping, that I could end my hesitation woes for good, I scrounged together the $50, went to the dealer, and bought the updated knock sensor with the white connector. This is where it all went to hell. When I went to put in the new sensor, I decided that I should also fix the fraying wire at the connector, coming from the ECU. So I cut the wire and fastened new connectors (standard 16ga car audio style, male/female) to both sides. Shined up the mating surface, bolted it down, and started the engine. Instantly I got Code 22 - Open or Shorted Knock Sensor circuit. This is a constant CEL, so it tells me that it's not something intermittant. So I pop it out again and find, as usual, that user error is the case. When I put in the new connectors, the shield was touching the center signal wire, causing a short. Sweet, I thought, easy fix. So I re-wired the connectors on both sides, made sure that the shield was not touching the signal anywhere, taped it up, and bolted it back down. No luck, same Code 22, same crappy driveability, so I realize that I have to get deeper into the diagnostics. I busted out my multimeter, and with a friend's help tested the continuity between everything I could think of:

signal from ECU harness to connector = positive
shield from ECU harness to shield at connector = positive
shield at connector to negative battery terminal = positive
shield at knock sensor to negative battery terminal = positive
signal from ECU harness to shield at ECU harness = negative
signal at knock sensor to shield at knock sensor = negative

The only things that I don't know how to test are the actual funtionality of the sensor, and the pins on the actual ECU. I believe that the B56.4 pinout is good because there is continuity to ground at the shield at the connector when the ECU harness is plugged in, but I don't know what to look for from the signal wire with the harness in the ECU. And as far as the functionality of the sensor, I've been led to believe that it creates a voltage or waveform, much like a dynamic microphone, and the ECU will initiate timing retard when the input signal peaks too high. I have no idea how to test if the knock sensor is emitting this waveform or not. I only know that it's not shorted within itself, and that it is brand spanking new, so I would hope that it is in working order.

Other factors include:
- Intermittant Code 35: canister purge solenoid, but who cares.
- At this point I can only torque the knock sensor bolt down to 13ft/lbs (not the Haynes recommened 15ft/lbs.) because I slightly stripped the block
- During the first intall of the dealer purchased knock sensor, I also built and installed vrg3's do-it-yourself FCD, but since then I have taken it back out. For those that don't know, the FCD is tapped into a completely different harness on the ECU, so I see no coorelation. Nonetheless, it's out now.
- I have repeatedly reset the ECU by both the negative battery terminal disconnect method, and the D-Check/Clear Memory method using the green connector and such.

Help!
-mike

Dont know how much help it is but having had the same intermitent pain in the ass code 22's I think Ive narrowed the culprit down to the wire touching the the block where there is a split in it (fook knows how it got split but I sorted it and it seems to work, or at least it might be doing lol).
Surely the sheild wire shouldnt be any polarity (im no electrician, they told me electrons flow in the opposite direction and I gave up lol) but Im sure the sparky where I work never runs shielding on cables that he uses to anything.

Kind of an obvious question, but did you check the continuity while jiggling the wires?

Yeah, everything's solid connection-wise. I just don't get it. Any way to test the ECU itself? It shouldn't matter that I'm using a different connector, should it? The OEM one didn't look like anything special.

peace,
-mike

What kind of connectors precisely did you use?

Did you just cut off both sides of the stock wiring (the connector on the harness and the connector on the sensor)?

If so, is it possible that the connection on the sensor side isn't solid?

Hmmmm. Typical crimp-on plastic 14-16awg male/female connectors often used for car audio. The female being the kind that would clip right onto a speaker terminal.
It doesn't seem possible to me that the signal on the sensor side isn't solid, I've been at it so many times. If anything, that was the splice that I was most confident in. But any way you know of to check this? It's not like I can check continuity to something from this end.
Also something to point out: I didn't start throwing Code 22 until I tried the brand new sensor with the "homemade" connections. Until then, the other two sensors i had in was just giving me the hesitation, no codes. Now, if I use the old sensors by splicing new connectors on them, I still get Code 22, so something's gotta be wrong because of what I did with the wiring. Admittedly, at one point I did short that circuit, but now I'm clean. It's like the ECU's not responding to that fact that everything's OK now, even though I reset more than once.

I don't mean to be disrespectful or anything; I'm just throwing out possibilities. I don't know how experienced you are with crimping but a lot of beginners have trouble making strong crimps.

For a good crimp you need the wire to be made out of a soft conductor so it can partly deform with the terminal. There should be no insulation at all under the portion of the terminal that you crimp. Enough wire should be stripped so that all the strands show through on the other end. And, of course, the terminal should be of an appropriate size for the gauge of wire being crimped to. Ideally you'd use crimp-on terminals that also have a spot to crimp to the insulation for strain relief, but these aren't as easy to find as they should be.

Try this: unplug connector B56 from the ECU. Measure the resistance between pin 5 (knock sensor signal) and ground. According to the service manual the sensor should show a resistance of 560K; it should hold steady. Then have a friend jiggle the wires at the connector and see if it changes.

Don't worry, no disrespect taken. It does seem the obvious answer, but I'm satisfied with the splices. I've done plenty of this in the past and have the right tools for the job, but I agree with your skepticism. (that's why I went to check my work first, and found the initial problem ) But that aside, I'll test the resistance of pin 5. Now, do you mean at the ECU or the harness?

Cool; I'm just trying to be thorough.

I mean at the ECU end of the harness. Unplug connector B56 and measure the resistance between the pin of the yellow harness connector and ground.

By including the harness as well as the knock sensor in the circuit you test the entire circuit that the ECU uses.