1974 3.0 CS Electric

[mane]

Ok, thanks.. Like all the way to straight/center position?

By default it seems to have a lot of caster, which helps returning .. so I think it should return easily.

I'm asking, because I was given explanation "it's an old car, don't expect it be like modern cars". I'm not taking that, something is off

It has a electric steering rack, and this got me wondering that if the ecu thinks that I'm resisting the turn, it might apply force to counter the return to center forces. Might be related to steering column binding. Need to check that.

 

[jefflit]

I had a similar issue w/ the VW rack on the CSE after a professional alignment. It has Ground Control coil overs w/ camber/caster plates and the fix was to dial in max available caster.

 

[mane]

I think I know what’s the issue now. That Citroen C2 electric steering gear needs angle sensor, otherwise it want to keep the tires straight. I have couple different ordered and testing those soon as I got my hands on them. I think this will solve my issue.

 

[mane]

Tested my idea today with Citroen C2.. removed angle sensor information from can bus, and it still returns to center .. and even when the whole ecu is in limp mode, it still returns .. so it’s not about angle sensor or the rack. Must be the caster.

 

[mane]

I had a similar issue w/ the VW rack on the CSE after a professional alignment. It has Ground Control coil overs w/ camber/caster plates and the fix was to dial in max available caster.

View attachment 168365

Thanks for the suggestion. This is the next thing I'm looking into, as my theory of angle sensor did not work. Need to first check that my BC Racing hollow bags fit there, with caster plates.

 

[jefflit]

Long, rambling thread about battery coolant leaks and stuff moved to the correct thread.

 

[Bmachine]

Your work on this project will never cease to amaze me, Jeff! Congratulations!

 

[ScottAndrews]

You should try to look up JB Straubel inthe US. He and I designed a hybrid electric car before he joined Tesla. At Stanford he retrofitted an electric drivetrain to a Porsche 944. Man that thing was quick!!! However, he had a zillion lead acid batteries in the trunk of the car as well as a trailer, made from the back end of another 944, that was chock full of batteries..

http://www.evalbum.com/223

My only reservation on electrifying an E9 using a Tesla battery is that that battery is pretty darn big, and pretty darn heavy (I see in some posts above, that you ave also come to see this). I worry that the reltively frail E9 unibody may not be able to support it. You could look at some smaller battery packs. My son has a Geo Tracker that he plans to electrify sometime. He was thinkng he might use a Nissan Leaf battery pack. Not nearly as much range and charge rate at the Tesla P75, but MUCH snmaller and lighter. You could probably fit a couple in the trunk.

I also worry about handling if you significantlychange the weight distribution, so you shodl consider these issues in your design.

Good luck!!
Scott

 

[ScottAndrews]

Biggest problem right now is the steering. It won’t properly return to center. Yesterday we swapped anther Citroen rack, which moves bit smoother but centering issue is still there. Wheels move freely and normally when on a lift and there is nothing binding. Good amount of caster and alignment has been done two times, and values are good. All ideas welcome, if there is something typical to E9s that can cause this?

If you think about the electric steering system, it needs ot be fairly complicated because otherwise you would have system that steered like an old video game (or a 1950's US sedan!!)..too much boost and zero road feel.

The electric rack is trying to provide some sort of a controlled torque vs steering angle to the steering wheel. With steering input, and no boost (i.e., no motor on the rack), the restoring force from the caster will return the wheels to center, and the steering wheel will come along for the ride, and return to center. I have not studued it in detail, but my instinct tells me that there is some semi-linear restoring force vs steering angle that results from caster. That is, at center, there is little caste induced force on the road wheels, and thus minimal torque required to steer the wheels. As the road wheel angle increases, the amount of restoring force from the caster increases, and thus, in order to steer the car the steering wheel torque must increase.

The whole purpose of power steering is to reduce the torque required on the steering wheel, but to do so in a way that still assures that some torque is required on the wheel, and ideally that that torque follows something like the un-boosted torque vs angle characteristic, just at a lower torque level. So the entire control system is seeking to apply enough supplemental force on the steering rack to overcome some portion of the opposing force onthe rack due to caster. Obviously there are speed related aspects to this as well, so presumably the gain of the control loop goes down as speed goes up.

The obvious place to start with this is just measure steering wheel position, and input the motor force to achieve that position. The reason they don't do this (at least as the only control strategy) is that there would be ZERO feel on the steering wheel. So, they measure steering column torque, and rack force (which is motor torque through its gearing plus steering column torque through its gearing), and they adjust the motor current as a function of steering angle to maintain the steering column torque aling some pre-defined characteristic as a function of steering angle.

It could just be that the rack you have either has some internal friction, or perhaps the steering column torque sensor has some sort of offset. If that was the case, then the ECU would always think there was some torque applied to the column, even when there wasn't, and so it would continue to apply enough force on the rack to maintain that torque. I am not sure how Citroen does their torque measurement, but if you remember that the electrical system is single ended, then you cannot have a negative signal in one direction, and a positive signal in the other. So the torque sensor output probably has a sign bit, and a value bit (assuming it is digital). If the value bit never goes to zero, then the motor will think there is still some steering wheel input, and it will not allow the road wheels to return to center.

Alternatively, and more likely, given that you have observed this with two different Citroen racks, perhaps the Citroen ECU is tuned for greater caster restoring force than the E9 generates. In this case if the motor torque sensor has a dead spot in the center, then it stops before the road wheels retrun to center, and the force from the caster is not registered by the system, so it does nothing unless the steering wheel input corrects it. A non-electricl version oof this situation woudl be that the rack has enoigh internal friction that near center, where the resoring force is low, the friction force is higher than the caster force, and so the rack sticks near, but not AT center.

With the road wheels off the ground, you have no resotring force from caster, so there is no torque on the motor at all other than to overcome whatever friction there is in the steering mechanism. So this issue would not be observed in that situation.

Cheers

Scott

 

[jefflit]

@ScottAndrews I think you've mixed up builds. My car has been on the road for over two years now. A bit late to be making comments about weight and distribution and handling. Yes, the car weighs 3600 lbs, about 400 more than stock. But I have no steering issues now that I have additional caster -- that was @mane . I have hydraulic P/S w/ a VW rack. Handling is surprisingly great. Not as nimble for an autocross perhaps but really confidence inspiring in the sweeping bends of real roads. Energy density continues to improve at a rate of roughly 10% per year so it is entirely possible I might switch batteries in 5 years for the same range with half the weight. Easy enough to do. But at the time, and for all practical purposes even now, no easily obtainable batteries had the energy density of the Tesla modules.

 

[ScottAndrews]

@ScottAndrews I think you've mixed up builds. My car has been on the road for over two years now. A bit late to be making comments about weight and distribution and handling. Yes, the car weighs 3600 lbs, about 400 more than stock. But I have no steering issues now that I have additional caster -- that was @mane . I have hydraulic P/S w/ a VW rack. Handling is surprisingly great. Not as nimble for an autocross perhaps but really confidence inspiring in the sweeping bends of real roads. Energy density continues to improve at a rate of roughly 10% per year so it is entirely possible I might switch batteries in 5 years for the same range with half the weight. Easy enough to do. But at the time, and for all practical purposes even now, no easily obtainable batteries had the energy density of the Tesla modules.

I was responding to @mane . I THINK this is his thread, although you did put ina fairly substantial post above. Nice work onthe electronics, BTW...

 

[mane]

Megapost! Interesting stuff @jefflit! Though you might want to copy that to your own build thread too

I've seen multiple ev-conversions having coolant leak problems, that have resulted first as a BMS error and when opening the battery-box, it's coolant leak. One guy added a water sensor on the bottom. I thought it was bit too much, but after seeing lot of these leaks, I think it's pretty good idea. My build does not have coolant going through the batteries, so one problem less for me Battery temps are at summer, after 150km driving, somewhere around 38c (max is considered around 60c). Though I know that temp sensor is in the coolant tube and I don't run the coolant - but I'm pretty sure that air would get hot there too, if something would start heating up. I think cooling is only needed is you do constant accelerations, multiple times a row .. or DC fast charging. In normal drive and charge, they don't seem to heat up at all. But it's nice to hear, that you can't start a fire by shorting those

I also have that Tesla parking brake controller. First had problems with used caliper being stuck and controller just made "click". Got instructions to add resistors to power line, to avoid peak - and that worked. Though I had to change both calipers to new, when car had to be towed and tow truck driver broke them both. I had to install them pointing down, so the tow cars lift got in contact with those and broke the motors. Their insurance covered me new calipers. Those things are expensive! About 1200 euros for two. Interesting thing that can-bus integration. Need another Arduino with can shield.

Envious of that snowless weather there. We're buried in snow here in Finland right now. Haven't been able to take the car out from garage for over 2 months.

 

[mane]

@ScottAndrews good points related to steering problem! I think you might be right about the friction. There might be something just slightly binding with the steering axle, which causes motor/ecu to think that driver wants to apply force to opposite direction than what wheel / caster is trying to do. I think we need to go through that and make it rotate so freely that there is nothing resisting that movement.

Caster angles in Citroen C2 are way smaller than in E9. It's 3.7 degrees in C2 where in E9 I have 8.14. So it should return better.

I also ordered H&R sway bar to front. We removed the original as we thought it wouldn't fit with the battery box, but there is actually space. That might help with the return to center also, as I read that people have had return to center problems after changing sway bar or removing it. Also going to add one to the back.

I also got Mini R56 electric rack, that might just fit when removing that ECU. I'm in process doing that .. but trying first do do everything I can with the C2 rack, to avoid welding and tuning that again .. with no guarantees that Mini rack would be any better. Mini rack ratio is higher, so it would give probably interesting go-cart like feel to E9 .. so kinda tempted. Also the fact that it's actually BMW part makes me happier, than having Citroen parts in BMW Maybe someday I can swap those Tesla batteries and motor to for example i4 or iX ones.

Related to weight, my car weights only 60kg more than original. Can't remember the exact figure, but it was something like 1490kg where original was 1430kg. I think I have 2 less modules than what jefflit has.

I also had problem with those BC Racing shocks. Can't adjust rears so that it would not be bumpy all the time, whatever the setting or ride height. Ended up buying two KW V3 Inox rear shocks. Let's see how things improve. I'm hoping night and day difference.

 

[ScottAndrews]

@ScottAndrews good points related to steering problem! I think you might be right about the friction. There might be something just slightly binding with the steering axle, which causes motor/ecu to think that driver wants to apply force to opposite direction than what wheel / caster is trying to do. I think we need to go through that and make it rotate so freely that there is nothing resisting that movement.

Caster angles in Citroen C2 are way smaller than in E9. It's 3.7 degrees in C2 where in E9 I have 8.14. So it should return better.

I also ordered H&R sway bar to front. We removed the original as we thought it wouldn't fit with the battery box, but there is actually space. That might help with the return to center also, as I read that people have had return to center problems after changing sway bar or removing it. Also going to add one to the back.


Related to weight, my car weights only 60kg more than original. Can't remember the exact figure, but it was something like 1490kg where original was 1430kg. I think I have 2 less modules than what jefflit has.

After I wrote the post about weight and weight distribution, I read the reat of your thread, as well as jefflit's. I see both of you opted for batteries in place ofthe engine, and the motor in back. Seems like the right strategy given that the batteries pretty much offset the engine weight, and to some degree the motor ofsets the fuel weight (at least for a full tank!)

Interesting observation about the C2 caster. That may be the root casue of the issue.

Based on your experience of not experiencing any binding when the wheels are off the ground, I doubt the issue is caused by something pushing back on the rack.

I love a good engineering puzzle!!

I think the issue maybe that BECAUSE of the larger caster on the E9, the restoring force is greater, and so the motor draws more torque, and thus more current per unit of torque applied to the steering column.

This is illustrated in the graphs shown below. The left side shows measured column torque as a function of applied torque (Probably not quite like the real one, which is calibrated for the C2). The right side shows, in blue, the ideal response in terms of motor current vs applied column torque, and in red, the actual motor current as a function of applied column torque. Because you cannot have a discontinuity the motor current must pass through zero at zero applied torque, but, since overall the E9 will require more current per unit of applied torque (especially as the steering angle increases), the result is this S curve, where at low applied column torque the current is still fairly high. (this is, of course, exaggerated to make the point).

Here is what can occur with some re-calibration of the torque sensor. Here, the torque sensr characteristic has a flat spot around the origin, so the motor current falls off faster at lower steering angles (lower applied column torque). The result is a compensated motor current vs torque.

I have no idea how to do this calibration, BUT, since all cars are somewhat different, and Hella (I believe the maker of that steering gear) presumably sells the system for use in many different cars, there is probably a torque sensor calibration table in the ECU. If not, then you might be ble to insert such a calibration between the torque sensor and the ECU.

Here is an interesting video of how the torque sensor works.

 

[jefflit]

My bad. I thought I was in my build thread. Whoops.

 

[ScottAndrews]

My bad. I thought I was in my build thread. Whoops.

No problemo. Love the attention to detail and the Tesla battery tech in your build! I don't know if you saw my first post that included JB's electric 944 Porsche link. I visited him back when Tesla was run by Eberhard and Tarpenning. JB walked me around Roadster #1, still being assembled. Fun and interesting stuff!

 

[mane]

@ScottAndrews that is interesting theory. Are you saying that even when turned front wheels try to return steering to center, electric motor in steering gear needs to react to this? And it cannot just be on "neutral"?

If that's the case, then the caster can probably play a role. So you think that when wheels try to go straight by themselves, ECU thinks it's going faster than "normal" (for Citroen) and starts to put some force against?

 

[ScottAndrews]

@ScottAndrews that is interesting theory. Are you saying that even when turned front wheels try to return steering to center, electric motor in steering gear needs to react to this? And it cannot just be on "neutral"?

If that's the case, then the caster can probably play a role. So you think that when wheels try to go straight by themselves, ECU thinks it's going faster than "normal" (for Citroen) and starts to put some force against?

Not quite. I think part of the problem is that the torque sensor is calibrated for the torque required to steer the C2, and the torque vs steering angle characteristic of the E9 is different.

Because of the larger caster angle, the E9 requires more rack force to turn the wheels. This means that you need to apply more torque to achieve a given steering angle. One possibility then is that areound zero angle, there is still some residual motor current that is sufficient to overcome the restoring force of the caster angle. So the motor torque just doesnt go down as fast as the camber retoring force, and the rack hangs up before it reaches center.

Re-calibrating the torque sensor to reduce the torque signal near zero might allow you to eliminate that effect.

I also did a bit more research on this, and found the attached paper. It is a bit hard to read because it is transated from Korean, but it deals directly with the issue you are experiencing. Notably, they did a sensitivity analysis that indicates that one of the dominant factors in returnability (the technical term for this issue) is wheel CAMBER. So, that might be another thing to explore, and it may be easier than trying to re-calibrate the sensor.

On the other hand, the sensor calibration may not be all that difficult, and it may avoid a situation where you have to adjust the camber to get returnability, but that messes up the handling of the car.

The manufacturer of the rack presumably sells these for use in many different cars. And, since all cars have different inherent steering force characteristics, it is likely that they have provision for calibrating the system to any particular car. This maybe embedded inthe firmware, which may be challenging to get to. If you can extractt he firmware, then you can probably de-compile it and identify the calibration table, adjust that, and recompile. Save a copy of the orignal code though!!

Another alternative would be to impose a calibration circuit between the torque sensor and the ECU. Deep inside the torque sensor is a mechanism that uses the differential rotation of the rack side of the column and the steering wheel side (this differential rotation being proportional to the torque applied to the column) (See the video I posted above). This is usually sensed by the motion of coils inside the sensor, which then couple more or less energy. High torque presumably resulting in higher coupling. This signal (or more likely a pair of signals, one for right and one for left) start life as analog voltages or currents, so you can presumably pre-distort those to effect a calibration that will offset whatever is in the calibration table inthe ECU, and effect your own calibration. You can create compression or expansion of the torque vs angle curve using diodes in the gain path of an op-amp circuit.

Here is what may be a more understadable representation of the problem.

As can be seen in the right hand graph, there are THREE equilibrium points where the steering will come to rest (boost equals restoring force). One on right, one on the left, and one in the center. I'd bet if this is the cause, that if you are turning right, and let the car straighten out, it will stop straigtening a little to the right. If you steer it past that spot, it will settle in the middle, and if you steer it to the left past the left side equilibrium point, it will do the same thing on the left.

You can also see here that if you increase the restoring force by increasing the caster, you can mitigate some of this negative boost, but you can also do that by changing the torque vs motor current characteristic to keep the motor force below the restoring force over the full range of steering angles.

 

[mane]

Thanks for giving a thought for this! Really appreciate it!

Wow, there is a lot of good information in that PDF!

I feel that the "manual" power to turn the wheels is bit greater in E9 than in C2 (My son also another C2 in our yard - easy to test the differences). But it's nice feel and not too difficult/hard to turn. Low speeds it works well too, as I have it adjusting by speed. I have noticed that after turning, it returns back for a bit, but then just sticks there .. and does not really move anywhere. That really feels like something is just holding it back, and keeping it in same place.

But.. I would think that if I ease the steering return with my hands (making sure that it won't try to return too quickly), it should have time to decrease the torque. This does not happen. So I'm not sure if it's that either. I would think I could emulate lower caster angle manually.. and make it return when resisting a bit. So not sure if that's the case either.

I have opened the cover of one ECU (I have 4 of those), and it has power electronics on top and there is no way to remove those without breaking it, or desoldering the power electronics. If I would do that, and got the actual board in my hands .. I don't know how to read the memory out of some rom. I guess I would need to identify the chip, by some equipment off from Aliexpress and read the binary out. Then try to figure out from hex data, what is the table for this these torques. Change some values, program it back and solder and put it back together and hope for the best. I think the first step already is difficult.. Modifying the torque sensor values on a fly, does not sound any easier. One thing could be, to find another ECU to control the same unit. I know the wirings from ECU to rack: it has primary and secondary torque sensors return info, 12v, ground, temperature and some diagnostics info wires.

I think I could try modify the camber and see what it does. Also after having the missing sway-bar installed, it might change things.

 

[ScottAndrews]

My theory is that this "resistance" you have described is actually the rack motor pushing back on the rack in the wrong direction (still trying to steer it right or left, but with the forces balanced at the place the wheel returns to.

The null points in the diagram (the places where the motor force and the restoring froce are equal) are the places where the steering will stop. It stops there because (looking at the right hand part of the graph) the restoring force and the motor force are equal and opposite. If they are equal and opposite the rack will simply stop there. At that null point, turning the wheel to the right will meet some resistance from the restoring force, and turning it to the left will meet resistance from the motor (because the motor force is still expecting to steer the car to the right (see my graph above. When the dashed lines are above the solid lines (or below onthe left side) the boost is actually fighting you.

I am still not sure I understand the steering wheel feel you are getting. I understand that if you turn it one way, it returns but not all the way back to center.
I assume you can then steer it back to center by turning the wheel manually. The key question relative to my diagram is if there is another null spot right around center where it will drive straight and not wander, and also if there is an other "resistance" spot when turning the other direction. If these "resistance" points are symmetrical (i.e on both sides, right and left), then I doubt there is any mechanical resistance causing the issue. The issue is probably that the torque reported to the ECU is too high at low steering angles, so you will get the boost working against you in either direction (those little humps inthe right hand diagram).

I think modifying the torque sensor is the easiest way to adress this. As an initial step you might just be able to add a resistor in the output. That would raise the steering effort a bit (because it would be reporting less torque to the ECU, but that's not terrible. Reducing the torque reported to the ECU should have more or less the same effect as increasing caster, but without affecting other suspension dynamics.

It is also pretty easy to make a circuit that reduces the lower values, but does not reduce the higher values, thus eliminatng the resistance spot but still providing reasonable boost at higher steering angles.

The boost is lower at higher speeds, so one questions, does this "resistance" effect go down the faster you drive? If it is the motor causing this issue, and the motor torque goes down as speed increases, then then this resistance effect shoudl also decrease as speed increases.

Another experiment is to just disable the power boost, and see if in that case, the wheel returns to center or still encounters resistance.

With the power steering system on, try driving in a safe open area. Steer to the right, and then note where the wheel returns to. Then steer to the left, and note if it stops short on that side as well, and approximately how much angle is left. Nudge the wheel to center. Does it then stay in the center?

If it exhibits this behavior, then the issue is that the torque reported to the ECU results in a higher motor force than the restoring force, so the dashed line crosses over the solid line (as shown in my diagram above), and the boost effectively inverts.