steering gear jitters

That has been on my to-do list for a while. It's easy enough to get to the connector (at the ECU end - so far I've yet to succeed in even seeing the gearbox-end connectors - probably requires cowl removal to see them from above, or lifting the car to see them from below). What I'd like to do is put a torque wrench on the steering shaft and try to make a nice plot of resistance v. torque - which probably means removing the airbag some lazy day and putting the wrench on the shaft nut.

Just not today though....

-Chap

 

That would be an alternative to the torque wrench on the nut, yes. I could, for example, wrap some protective paper around the wheel rim, secure a hose clamp over the paper, hook a spring scale to the hose clamp, and apply a tangential pull. Spring scale reading times radius from wheel center to hose clamp would be torque. In fact that's pretty much exactly the procedure in my old Mazda manual for measuring steering friction (with the wheels off the ground, which only needed a small spring scale - since here we want to measure real-life operating torques, it would have to be a heavier scale).

Only, cobbling together the hose clamp and a hefty enough spring scale, etc., would probably take me as long as removing the airbag and slapping on the torque wrench, which I've already got.

Yes.

Ehrm. Really the important property of the rubber element (if rubber's what it is) is its elasticity under torque. Rates of change aren't really in play because if you clamp the wheels straight ahead and put a torque on the steering wheel, you'll wind up in a static equilibrium where nothing is moving but the wheel is now at a slight angle from straight ahead in the direction of your force, because the rubber has allowed it to turn that far (just as a heavy spring would). That angle is proportional to your applied force and is what the pots measure.

Now on the road, with the wheels not clamped straight ahead, they'll be able to move and the wheel will be able to continue turning in the direction you are torquing it - but as the whole business turns, the steering wheel and upper shaft will still be leading by that very slight angle that serves to measure how hard you are twisting, and the EMPS will still be using that measurement to determine how much to assist you.

-Chap

 

Ok, here's the first part of the answer (the part that's easy without breaking out the torque wrench). With the torque sensor disconnected from the ECU at the ECU connector, and measuring at that harness connector:

VCC - GND is 745 ohms. This is the parallel resistance of two pots, so assuming the pots are identical, they are 1490 ohms each (or call it 1.5k).

With the wheels straight ahead, steering unlocked, and no applied torque, VT1-GND is 1027 ohms and VT2-GND is 2990 ohms. Obviously these are not identical and also they don't make sense in a simple picture of two voltage dividers in parallel (where each would be about 750 || 2250 or 563):

Code:

           VCC
            |
        +---+---+
        |       |
 750   <       <   750
        >       >
VT1 ---<       <--- VT2
        >       >
 750   <       <   750
        |       |
        +---+---+
            |
           GND

My measurements DON'T confirm this picture!

Also, both VT1-GND and VT2-GND began as larger resistances when I first attached the probes and declined slowly. They never really stopped declining, but seemed to be slowing; 1027 and 2990 were the readings when I got tired of watching. Almost like there's a capacitance involved, but in that case I'd expect a reading to start low and rise, the opposite of what I saw.

I was wondering if it was a fluke of the Fluke, but I went downstairs and measured a 2.2k resistor out of the box o' stuff and got a solid 2342 with no change over time. So the resistance-starts-high-and-falls was somehow a real property of the torque sensor.

This may mean that there is already some sort of mystery signal-conditioning circuit in the steering gear, or it may mean that the picture is more like the below, with some mystery resistances due to wiper contact degradation, and the mystery resistances are not very stable under the applied current of an ohmmeter.

Code:

                VCC
                 |
             +---+---+
             |       |
 750        <       <        750
         ?   >       >  ?'
VT1 ---^v^v-<       <-^v^v--- VT2
             >       >
 750        <       <        750
             |       |
             +---+---+
                 |
                GND

A possible picture?

In that case, ? would be about 464 (1027 - 563) and ?' about 2.4k (yow!). After all, I am making these measurements on a known iffy unit at 131000 miles. Probably more useful to measure a good one.

It just occurred to me now that VT1-to-(VCC and GND shorted), VT2-to-(VCC and GND shorted), VCC-to-(VT1 and VT2 shorted), and GND-to-(VT1 and VT2 shorted) could be a useful set of measurements. I think it can still be assumed that each vdivider portion in the mystery picture is pretty close to centered at zero torque, because under power the ECU really is getting about 2.5v back on each and doesn't act like zero-point calibration is off.

That's what I know for now.

-Chap

 

I wonder about (and haven't measured) the ECU's input impedance on VT1 and VT2 - are they just high-impedance voltage sensing inputs, or low enough to draw a decent wetting current across the wiper faces? Maybe a possible reliability-enhancing measure would simply be a circuit that lowers that impedance (then corrects for the resulting nonlinearity). I have zero knowledge about optimal wetting currents, and obviously it would be limited by what the ECU can supply on VCC and by the heating tolerance of the pots. And maybe Toyota already does it and it's just not effective enough. Just thinking....

-Chap

 

Caution: thinking about wetting current over chips y salsa may lead to the following idea.

Let's assume that there is no fancy circuitry in the steering gear and the simple two-vdividers picture is right, but with the weird readings I got being symptoms of wiper contact resistance.

Wouldn't the idealized picture ...

Code:

           VCC
            |
        +---+---+
        |       |
 750   <       <   750
        >       >
VT1 ---<       <--- VT2
        >       >
 750   <       <   750
        |       |
        +---+---+
            |
           GND

... have just the same electrical behavior in this form, given the way the wipers are mechanically linked?

Code:

           VT1
            |
        +---+---+
        |       |
 750   <       <   750
        >       >
VCC ---<       <--- GND
        >       >
 750   <       <   750
        |       |
        +---+---+
            |
           VT2

That is, apply the 5 V - gnd reference at the wipers, and tap VT1 and VT2 off the pot ends. This will result in a continuous current arount 7 mA at both wipers, without introducing any nonlinearity into the basic response. If I google about wetting current, I encounter rough rules of thumb in the 10 mA range, so 7 mA might actually be in the ballpark. And obviously it's no more power dissipation than the pots handle now. It might not do any good on contacts already as oxidized as mine, but if applied from the start with a new clean pair of pots it might be enough to help keep them that way.

No need now to use anything but ordinary high-Z inputs to sense the VT1 and VT2 signals, because these are now being taken off at soldered or otherwise gastight connections that won't depend on wetting current to stay clean.

Replace the 5 V reference voltage source with a 7 mA current source, and now we can call the voltage across the pot resistive elements a constant 5 regardless of what happens to the wiper contact resistance (which we don't expect to deteriorate as badly anyway because of the wetting current).

Any badness that does develop on the wipers now appears as common-mode noise on VT1 and VT2, which we can ignore by using VT1 and VT2 as differential inputs. Provide two complementary 0 - 5v outputs VT1 and VT2 relative to the ECU's gnd and vcc references, as the ECU's "VT1" and "VT2" inputs.

If the wiper resistance ever does get too bad to live with, it can be detected by the current source exceeding a voltage limit. A clever circuit might signal that by intentionally glitching the VT1 or VT2 output, causing the ECU to log a torque sensor code.

Long term stability of the differential input would be important, as any drift would throw off the zero-point calibration. The nice thing is, if the conditioner circuit is stable enough and somewhere in the ballpark, calibration could rely on the zero-point setting function already built into the ECU.

Am I missing obvious problems? Making some semblance of sense? Needing to find a different source of salsa?

-Chap

 

Interestingly, the symptoms do not seem to be noticeably progressive; in fact, recently it hasn't been annoying at all. The only time I notice it now is when I have a very light fingertip grip on the wheel. Can get a pretty violent shake that way, but I just don't much anymore because it seems without thinking about it I've learned to keep a solid grip almost all the time. My old driver's training instructor would be so happy.

I'm starting to think this makes a real difference. Could be a mechanical feedback/dynamics kind of thing.

As I'm putting torque on the wheel, the torque sensor pot may pass over a noisy spot. The ECU responds to that spike by sending a spike to the assist motor. That impulse goes down to the steering linkage and also up to the wheel. What then happens may depend on how I've gripped the wheel.

With a good two-handed grip, my arms' mass is part of the system and helps to damp the wheel's response. The wheel lags the impulse to the linkage, which reduces the torque sensor signal in the original direction (or increases it in the other direction--negative feedback, either way you look at it) and whatever noise event is pretty short-lived and barely noticeable.

With a loose grip, the wheel can follow the noise impulse at the motor. Obviously it still lags, but not as much, so there's less negative feedback and the impulse is less damped. The gear motion stops when the steering linkage return torque equals the motor-supplied torque, and now the steering wheel lags again, bounces against the torsion element and rebounds, then creates a torque signal in the other direction. The ECU amplifies that and now we have the ingredients for a sustained oscillation out of a passing noisy torque signal. It can last ~ 1 sec and shake the wheel visibly and noisily around 10 Hz.

I'm only guessing at the details, but the effect--a good grip on the wheel seems to keep everything happy--is very real as far as I can tell.

I'd still be interested in anybody's reactions to the conditioner circuit idea I posted above.

-Chap

 

Well ... I don't know .... I can't easily think of what might cause a woosh-woosh-woosh sound, but I don't think it sounds likely to be related to the known steering torque sensor issue, anyway, so IF you are in for an expensive repair, it might be some other kind.

-Chap

 

steering gear jitters - Christmas in July?!

I just got my pinion-nut letter in the mail today. I was expecting that, and it was nice, but since I already know my torque sensor is flaky I wasn't sure I'd even ask Toyota to change the nuts on a rack that's about ready for replacement anyway.

But hmm, there's a second page in this envelope here, what's this?

No ... freakin' ... way! My rack went out of warranty 7 years and 140,000 miles ago, and Toyota's decided to stand behind it? Man, that's class!

And to get this in the mail just now, when I've been putting up with the occasional jitters for almost exactly 3 years now, but just a couple weeks ago for the first time it reached the point of the assist turning off while I was driving, and logging a C1513. (I was on the way home from a trip through the Appalachians, and happily it didn't cut off while I was whoopeeing through the switchbacks, but it waited for a nice straight stretch 3 miles from home.) Came right back on after a restart and hasn't cut out again since, so it's just right at the point where I'd be ready to replace it. I was literally making plans to spring for a Cardone reman rack, probably before this month was out.

Toyota totally rocks!

Has anybody else seen this letter yet, or found the associated documents on techinfo? I was expecting to be easily able to find an official document or campaign number on it somewhere (like the SSC- numbers that have been posted here for other issues) but I really don't see one printed on it. How do we get the details?

-Chap

 

I finally caught the shake-like-a-wet-dog steering wheel jitter on paper!

The chart recorder is one of my most entertaining rummage-sale finds ever, but as soon as I run out of the special roll paper (which costs more than I paid for the recorder) I might just have a conversation piece. So I've been backing the paper up and rezeroing the pens to be able to record three trips per length of paper.


The red pen is VT1, the signal that increases with driver-applied steering torque to the left. VT2 (black pen) increases with driver-applied torque to the right. The traces should be mirror images of each other, but because the recorder's pens must be able to cross without smacking into each other, the red pen is about 6mm ahead of the black (that's 3 of the fine, 2mm grid lines).

Settings were 5 volts full scale, 6 cm/minute chart speed (that's 1 second per mm, or 2 seconds per fine grid line).

You can see the shake-like-a-wet-dog event just as I pull out of my driveway on 20 August (steering at low speed, that's when it happens), where the two pens just colored in solid blocks. The shake lasted 2 full seconds and could have gone on longer; to stop it I grabbed the wheel firmly. The amplitude might well be larger than shown, as the pens can only move so fast.

I notice that the rest of the time (when not noticeably shaking), the VT2 signal does seem always noisier than VT1. The noise looks like frequent sharp spikes toward ground. A tin whisker between VT2 and ground, perhaps?

I am not sure I want to experiment with the whisker-melting techniques, as I do have the letter from Toyota offering to replace the rack assembly, which sounds like a good deal to me (assuming I can trust my dealer to complete such an involved job without breaking something - I've never had to take it in for service before). But anyway, now we know what my steering jitters look like.

-Chap

 

Well, I've finally set the "wheels in motion" by contacting my local dealer with the warranty enhancement letter. With any luck the information we've already got here will be enough to save them the labor expense of trying to do just the pinion nut first, or driving around with a tech long enough for the wet-dog-shake to happen again. Watch this space for updates!

-Chap

 

The effort of documenting the issue well does seem to pay off - the dealer called this morning to say they had discussed the case with Toyota, Toyota is sending a rack under the warranty extension, and I'll be getting a call to come in for replacement when it arrives. They are not asking for a first appointment to try to make the problem appear in front of a tech, so that saves time and trouble for them and for me.

So far so good!

-Chap

 

Rack going in today ... other people are taking my car apart....

And they're not letting me watch....

-Chap

 

... and it was all perfectly painless. They had the needed parts, finished early, and did not seem to run into any surprises. Courtesy of Toyota I now have a new rack (44200-49055) installed with its installation parts kit (04001-31147, I think that's what that was) at 193532 miles, and no hint of steering shake. I notice that I had gotten used to a certain very subtle twitchy feeling that used to be almost always there even when the wheel wasn't visibly shaking (which didn't really happen all that often). All of that is gone. It now feels pretty much as it must have felt when the car was new.

I recorded the new torque signals for two trips this evening and, comparing them to my earlier post, they look much cleaner, especially VT2 (black pen - sorry that pen's a little dry). In answer to Bob's very old question, I can now report that for a brand new 44200-49055 rack, resistances in ohms are:

	VT1-VT2	VT1-GND	VT2-GND	VT1-VCC	VT2-VCC	VCC-GND
1	1231	1421	1423	615	617	807

These resistances did not start at one value and slowly decrease while on the meter, as I had seen measuring the old rack. These numbers were rock steady.

When I picked up the car the first thing I did, before even starting it, was to check for codes. There was a C1213 (ABS and HV ECUs no talkie), but I had come right at closing and no one was around to investigate. I mentioned it to the service advisor, cleared it, and drove home, and nothing seems amiss and the code has not come back. In the book it looks as if the brake ECU is near the steering column and my guess is that a connection may have been jostled during the repair but is now ok again. The advisor told me the tech had checked for codes and seen none at the completion of the job. I am not sure how the code appeared between then and my check, but if it never recurs I'll consider it good. I guess if I can easily see/reach connector B11 without taking my dash apart, maybe I'll check that it's secure.

All in all, a very satisfying resolution to a long thread.

-Chap