2002 power steering shudder fix

Ok, this has gotten my interest:

	pairs	resistance	calculated	ohms	R(name)
1	VT1-VT2	1231	1232.0	612	R(VT1) & R(VT2)
2	VT1-GND	1421	1420.0	1612	R(VT1-GND)
3	VT2-GND	1423	1420.0	1612	R(VT2-GND)
4	VT1-Vcc	615	616.0	4	R(VT1-Vcc)
5	VT2-Vcc	617	616.0	4	R(VT2-Vcc)
6	Vcc-GND	807	808.0		R(Vcc-GND)

This matches the static, resistance measurements. The only schematic that makes sense:


I was able to make a SPICE model and it came up with the same schematic and values.

To get 2.5V out of VT1 and VT2, Vcc would have to be 2.5V. Worse, only one side would change voltage depending upon the torque sign. This is too weird but that is where the static resistance model takes me.

Guess I'll have to go to my car and see what I find.

Bob Wilson

 

My measurements were at the torque sensor connector (at the ECU end), unplugged from the ECU. It could be that the ECU is designed with a crazy low input impedance that would load the voltage down even with Vcc being 5 V (which it definitely is). An ammeter in line with VT1 / VT2 in normal operation would add some information we don't have yet.

But then, the effect of an extra-low input impedance would tend to make the voltage-torque graph nonlinear (the graph in NCFM looks like straight lines). There are still mysteries here.

-Chap

 

I would like to see the resistances with clockwise, none, and counter clockwise torque applied:

• Vcc-VT1
• Vcc-VT2

If the circuit is accurate, we might see a 'limit' in one direction and some sort of torque-proportional set of readings in the other.

Bob Wilson

 

I don't think that's what we do see, at least in the voltages that come out (I don't know if you've looked closely at the chart recorder strip here, but if you click the tiny thumbnail image once it should expand into PriusChat's funky image-viewer-layered-over-the-page thing, and then if you click "Show in original location" you should see the image as a page to itself, and then if you click it again (in Firefox at least) you should see it full size).

-Chap

 

I know which is why I'm puzzled. The static resistance model shows one result but the voltage data shows something else. Sad to say, it was when the contacts were 'noisy.'

What we need are independent metrics and I'm hoping the OP will jump in with his data. After all, he has the 'test article.' But the PS ECU is easy to reach so I'm thinking it may be time to head out to the car with instruments and see what I find. But I want to go prepared to measure 'contact noise'. Assuming there is evidence of a potential problem, I can quantify the effect and then measure the 'burn-out' voltage effects.

Bob Wilson

 

My early strip chart was of the old bad rack, but the later one I posted was of the one-day-old new rack. Still puzzling....

-Chap

 

I figured out the problem with my model:

1. Each side, the pots are separate pads connected by the pickup wires.
2. The pickup wires are electrically connected to VT1 and VT2.
3. Need to include the "lead" resistance to the meter.

I built an improved SPICE model set that cover the key metrics:

1. VT1-Vcc
2. VT1-GND
3. Vcc-GND
4. VT1-VT2

It is a little time consuming but the approach is:

• First find values that give good high-resistance, shortest path, VT1-GND
• Next find values to give good low-resistance, shortest patch, VT1-Vcc
  • Re-tweak the high-resistance, shortest path to minimize error
• Evaluate Vcc-GND and VT1-VT2, find the greatest positive resistance offset
• Divide by 2 and set to the lead resistance
• Repeat all steps until the total error is a minimum

Yes, I know this seems a little confusing. I'm combining a SPICE model with a simplex algorithm to step to a solution. A little 'old school', it is simpler than trying to solve a series of linear equations ... although there may be general solutions that do that.

Regardless, I should have a better schematic and values this evening.

Later,
Bob Wilson

 

So this was my first model:


Look what happens when the signals are rotated 90 degrees:

The sense voltage suddenly become balanced at 2.5V. The offset of the cross bridge network has the same sign and magnitude.

My proposed fix is to add cross-bridge resistors to bring the resistance into the correct range:


I do not know what these parallel resistors will be and suspect we'll have to tweak the values of the four bridge resistors. Note, the "VT1-VT2" and "Vcc-GND" labels in the above uses the original orientation. Rotating 90 degrees swaps them. Regardless, an updated SPICE model will allow finding values that work.

The initial values for the parallel resistors will be:

• (1067 / 257) * 1067 = 4430
• (1500 / 269) * 1500 = 8364

This will throw the legs off but I'll use the same approach, working the lowest value resistors up to the highest.

The strip-chart reading was a little difficult because I'm used to dual-trace scopes where the X-axis is the same time value. Eventually I was able to read the strip-chart by putting a mental rectangle whose width of the time-delay. Regardless, I think the revised model will solve the problem.

Bob Wilson

 

Hmm, now that I've managed to spark Bob with all my enthusiasm for mathematical swordplay, I'm having second thoughts... I've ordered a cheap salvage rack to see if I can take it apart.

-Chap

 

You are the one with the test article. I'm hoping Chapman gets his test article and we can find out what upper limit on voltage/current flow makes sense.

I appreciate sharing your efforts as you 'tickle the dragon.'

Bob Wilson

 

Interesting but I would probably approach it:

• Increase voltage to steering by "x" times.
  • Using a linear regulator, reduce it to 13.9 - 1.4 ~= 12.5 V to reduce buss noise
  • 12.5 / 5 = 2.5 reduction needed
• Voltage divider by "1/x" times to match original signal levels.
  • Use a pair of 1.5 k Ω trimmers to adjust the output ratio 2.5

This would run all the time the car is in READY yet increase the current on the wipers by a factor of 2.5.

Bob Wilson

 

That long-ago idea of mine was based on a certain assumption about how the torque sensor in the steering gear is wired inside, and a lot of what you've been seeing in this thread is efforts by both me and Bob to figure out whether it's like that or not. The thing is, we don't have any posts yet from anybody who took an actual rack apart to measure the individual resistances inside the torque sensor and how they are wired together. With a part that was selling for $1200 to $2500 and was necessary to steer the car, not many people wanted to start pulling theirs apart.

Without pulling one apart, you can't directly measure how it's wired inside, you can only measure various resistance combinations across the wires that come out, and try to mathematically solve for what's really inside. If the insides were as simple as I thought in 2008, we kind of know what resistances we would see outside at the wires, and we know we don't see that. Bob has been doing the grunt work of trying to find an inside picture that would match what we can measure outside, and so far it not only doesn't match my original simple picture but keeps getting more and more complicated, with extra bridge resistances and connections at the contact springs, etc., that no Toyota book has ever hinted are in there. (That's why I finally broke down and ordered a rack off a parts car for $60 the other day - I've reached the point where I'm not sure our modeling-from-the-outside-in effort might not be missing something basic, and for $60 if it's possible for me to disassemble the thing completely enough, maybe we'll find out.)

But I think for now, you'd be safe to say that everything in my post from 2008 about switching pins around was based on an assumed picture of what's inside there that we can't at all say is correct at the moment. It's rung down the curtain and joined the choir invisible. It is an ex-idea.

Cheers,
-Chap

 

I scanned at 200 dpi for an indication of scale.

This is accessed through the top cap of the steering gear assembly, which comes off with two bolts. The wiring goes through a rubber grommet that you can see in the lower photo (it's dusty on the outside-facing side, and right now is just next to the upper disk detached from the spiral cable housing). Before removing the top cap, the retainer for said grommet has to be removed (two screws and a cable tie) and then the grommet can be popped out of its hole. Then the grommet can be squashed through the hole into the inside of the metal top cap, allowing the wiring to pass through as the cap is removed.

The top cap has a rubber O-ring around the outside where it fits in the case, and contains a cartridge bearing for the steering shaft, a light press fit. After removing the two bolts, I first pried with screwdrivers around the edges to get it started upward ... then I was able to twist it slightly so the holes no longer lined up, thread the bolts back in from behind and use them to push it slightly further up. Eventually I had enough of a gap to grip it with a two-jaw puller bearing against the end of the steering shaft, and pull it the rest of the way off (feeding the wires through the grommet hole as it came). This was not a hard pull - I was actually turning the puller screw bare-handed.

You'll see at the gray wiring connector, the pins are partly backed out of the shell. They have to be popped from the shell to feed through the grommet hole to get the metal cap completely off ... now I've just pushed them back into the shell but not far enough to click, just so I don't forget their positions (though I could look them up in the wiring diagram if I had to).

Once the metal top is off, the top cover disk of the spiral cable housing can be removed by releasing its three plastic tabs with a thin screwdriver down through the square openings. Then you see the spiral-wound ribbon cable.

The square black plastic piece in the photos is a wedge. In the upper photo you can see the little plastic "guide" structure built into the spiral cable donut housing at the center where the ribbon passes down through to the resistors. Assembling the spiral cable finished by folding the ribbon on a nice 45° crease so it turns the 90° corner within that little guide area, and comes out winding around the donut. It's retained in that folded position by mashing the wedge in there next to it. You can see the other wedge (undisturbed) out at the outer end of the ribbon where it folds again to pass down to the small PC board that terminates the harness wiring,

In order to lift the spiral cable assembly out of the gear housing, the wedge at the center has to be pried back out (without scarring the ribbon cable!), and the cable gently unfolded at the crease and allowed to feed down through the hole.

Once the spiral cable assembly is lifted out, you can see the three screws attaching the resistor circuit board to the steering shaft.

-Chap

 

How tall are the two sensor modules?

It looks like there are "fingers", not wipers. Mechanically is there something they fit in?

Would you be willing to unsolder the white cable and ohm it out?

Is there continuity between the six brushes and four white cable wires?

If monitoring the resistance, does it change if the "fingers" are pushed or pulled?

Bob Wilson

 

They stand 5.5 mm proud of the circuit board.

This circuit board, which carries the resistance elements, is attached by 3 screws to a carrier on the upper, driver-facing portion of the torsion element. The lower portion, locked in rotation to the pinion, carries the two spring-metal contact assemblies. I won't be posting closeups of those any time soon and probably not without mechanical disassembly of the gearbox to extract the pinion shaft. I may later add more photos along those lines but it'll have to wait. But the springy contacts do look pretty much as they do in the NCFM illustration I posted earlier.

When I have more time I do plan to get as many useful resistance measurements as I can, isolating things as needed.

By brushes, do you mean the metal clips (3 each) at the ends of RR1 and RR2?

You can kind of see the PC traces on the board in the photos, The fat, ground-plane-shaped traces are in fact GND. The several soldered thru-holes join ground plane areas on the two sides of the board. Vcc follows the skinny traces at the outer PC board edge up to RR1 and RR2 - the traces parallel to the outer edges of the ground planes. VT1 and VT2 come back on the even skinnier traces that parallel the inner ground plane edges.

The three terminals of the RR devices are silkscreened (on the other side of the board) D, E, and F for RR1 and A, B, and C for RR2. VT1 is E, VT2 is B. Vcc is A and D, GND is C and F. That does mean the two RRs, which look otherwise identical, have their connections to Vcc and GND reversed, as I inferred they'd have to in post #38. But I also speculated there that the two RRs would have to be different in more ways than that, and maybe they don't; that will have to wait until I have more time for ohming.

Because this much of the sensor can be lifted out of the gearbox but the fingers stay where they are (without disassembly requiring much larger wrenches), it won't be easy for me to do a lot of measuring of that kind. I can explore this much of the sensor to my heart's content, except for interactions with the contact fingers, and for those I have to put it all back together again and then not be able to see the underside.

To make the cleanest demo unit would call for removing the pinion shaft, which probably isn't that hard once I can get to the pinion nut (the infamous pinion nut, subject of a familiar recall). It lives behind the 30 mm hex head plug at the bottom of the gear body, and I do have a 30 mm socket but it's not impact rated, and my progress to date ceases at that point.

There is also a pinion yoke adjustment with a 24 mm internal hex, locked with a 42 mm square(ish) lock nut. Where the electric motor attaches to the body there is a thick metal shim, and I am guessing the shim thickness is selected to adjust the driving gear backlash.

-Chap

 

If you layout the parts and add labels to the text, there might be a publishing oppotunity. GRINS

Bob Wilson

 

Oh, wait, that's the same size as the front axle nuts, so it turns out I do have an impact-rated socket that size.

-Chap