Steering issue under load with track tires

[RedReplicant]

Originally Posted by geraldjust

might have a go on reading the eeprom then.
well the voltage does not dip but ive learned that for engine swaps that VSC and PS are sensitive to it. On the stock system, the current draw sensor on the battery knows when too much current is bring pulled. ie, PS or VSC pump. Then the ECU request the alternator for higher current.

But il be looking into this further then and so a full diagnosic read on the system then since it peaked my interest now.

FWIW, the K swap cars run high output mode at all times.

Anyway, curious to see what you find here.

[geraldjust]

Originally Posted by RedReplicant

FWIW, the K swap cars run high output mode at all times.

Anyway, curious to see what you find here.

you peaked my interest so ima give it a look forsure!

[TRS]

@geraldjust Yes, my car is swapped. But the test decribed in my last post was done on a factory 2018 BRZ. The behavior of this car and my swapped one is fully identical for the test in stopped condition. Ofcourse driving wise it is easier to provocate the cut off in my LS swap due to the higher weight.

I monitored the thermistor temperature during repeating that test on the standing vehicle. I kept the steering torque all the time pretty high to stress the system and also changed direction multiple times. The thermistor temp rised constantly over several minutes, but behavior was same. The cut off jumped in, and as soon as the steering was released slightly, support came back.

It was same when starting the test at 30°C thermistor at the beginning as it was at 40°C in the end. I was provocating cut offs against lock for about 5min. So I highly doubt a mechanical or thermal limitation, since during the complete test run nothing in the basic behavior changed. (Again note: The change request would not be to further increase support, everything needed is to keep that thing running...)

I also was able to define the limits further. The cut off jumps in as soon as the measured steering wheel input torque reached 6Nm. The requested current is reduced to zero in that second.

Then the value just instantly jumps to 7Nm, what seems where the sensor itself tops out, since one still is holding the steering wheel while the motor quits, what inturn results in a instantly higher input torque.

Then, once the steering input torque is slowly reduced, the support kicks in as soon as 5.5Nm is reached. The requested current is then set to the value (or at least ball park) which could be seen during the way towards lock.

This game can be played quick or slow. Means exceeding 6Nm and holding the input torque above 5.5Nm for 10s results in a 9s cut off (about 1s system reaction time...). If you stay above that 5.5Nm for 1.5s, the result is 0.5s cut off. The whole procedure can be repeted in close to 1Hz frequency or any frequency lower then that.

One more observation: The highest current values of above 40+ amps was seen during fast turning of the wheel. Highest current when hitting the lock was about 30-35. So my conclusion: the motor driver is also able to provide even higher amps.
Also the temp is not instantly freaking out, since internal temp sensing keeps in acceptabel range even over longer stress periods and, most important, behavior and values dont change all over the stress test.

The behavior is a simple reaction towards input torque and pretty much doesnt care about any other conditions. Not temp, not current requested, not current controlled, not steering angle nor steering speed. It's a simple and clasic one-variable-condition with hysteressis.

I cannot make any other conclusion but the intension was to create a steering lock overload protection (what absolutely makes sense, no doubt about that), but nobody cared about giving this function a little more then absolute basic logic. This results in a unneccessarily low marging of the EPS system support and the possibility of suddenly losing EPS support during harsh turns and as well the possibility of sudden return of the support, what from my personal expierience, is the far bigger issue towards vehicle stability/control and safety.

The driver is still turning the wheel to stay with the turn, then suddenly support returns and the steering wheel gets "light", what results to a sudden, short inwards turn of the steering wheel due to driver reaction time. This is pretty much like a forced mid-turn-scandinavian-flik.

What actually might be a limitation by physical Design ist the fact, that I was not able to read a steering wheel angle. This somehow triggers me, since the zero position can be calibrated, what draw me to the conclusion that the system must have some kind of knowledge about the steering angle. Though, there is no readable value for that. Only available angle it motor angle. But this can not be easily transfered to a steering angle, since the motor does multiple turns and there is no turn counter. So the absolute position is unknown.

If there actually is no steering angle measured, what somehow makes zero Sense in my head, it's clear that a lock protection has no other option then to relie on input torque only...

[geraldjust]

Originally Posted by TRS

@geraldjust Yes, my car is swapped. But the test decribed in my last post was done on a factory 2018 BRZ. The behavior of this car and my swapped one is fully identical for the test in stopped condition. Ofcourse driving wise it is easier to provocate the cut off in my LS swap due to the higher weight.

I monitored the thermistor temperature during repeating that test on the standing vehicle. I kept the steering torque all the time pretty high to stress the system and also changed direction multiple times. The thermistor temp rised constantly over several minutes, but behavior was same. The cut off jumped in, and as soon as the steering was released slightly, support came back.

It was same when starting the test at 30°C thermistor at the beginning as it was at 40°C in the end. I was provocating cut offs against lock for about 5min. So I highly doubt a mechanical or thermal limitation, since during the complete test run nothing in the basic behavior changed. (Again note: The change request would not be to further increase support, everything needed is to keep that thing running...)

I also was able to define the limits further. The cut off jumps in as soon as the measured steering wheel input torque reached 6Nm. The requested current is reduced to zero in that second.

Then the value just instantly jumps to 7Nm, what seems where the sensor itself tops out, since one still is holding the steering wheel while the motor quits, what inturn results in a instantly higher input torque.

Then, once the steering input torque is slowly reduced, the support kicks in as soon as 5.5Nm is reached. The requested current is then set to the value (or at least ball park) which could be seen during the way towards lock.

This game can be played quick or slow. Means exceeding 6Nm and holding the input torque above 5.5Nm for 10s results in a 9s cut off (about 1s system reaction time...). If you stay above that 5.5Nm for 1.5s, the result is 0.5s cut off. The whole procedure can be repeted in close to 1Hz frequency or any frequency lower then that.

One more observation: The highest current values of above 40+ amps was seen during fast turning of the wheel. Highest current when hitting the lock was about 30-35. So my conclusion: the motor driver is also able to provide even higher amps.
Also the temp is not instantly freaking out, since internal temp sensing keeps in acceptabel range even over longer stress periods and, most important, behavior and values dont change all over the stress test.

The behavior is a simple reaction towards input torque and pretty much doesnt care about any other conditions. Not temp, not current requested, not current controlled, not steering angle nor steering speed. It's a simple and clasic one-variable-condition with hysteressis.

I cannot make any other conclusion but the intension was to create a steering lock overload protection (what absolutely makes sense, no doubt about that), but nobody cared about giving this function a little more then absolute basic logic. This results in a unneccessarily low marging of the EPS system support and the possibility of suddenly losing EPS support during harsh turns and as well the possibility of sudden return of the support, what from my personal expierience, is the far bigger issue towards vehicle stability/control and safety.

The driver is still turning the wheel to stay with the turn, then suddenly support returns and the steering wheel gets "light", what results to a sudden, short inwards turn of the steering wheel due to driver reaction time. This is pretty much like a forced mid-turn-scandinavian-flik.

What actually might be a limitation by physical Design ist the fact, that I was not able to read a steering wheel angle. This somehow triggers me, since the zero position can be calibrated, what draw me to the conclusion that the system must have some kind of knowledge about the steering angle. Though, there is no readable value for that. Only available angle it motor angle. But this can not be easily transfered to a steering angle, since the motor does multiple turns and there is no turn counter. So the absolute position is unknown.

If there actually is no steering angle measured, what somehow makes zero Sense in my head, it's clear that a lock protection has no other option then to relie on input torque only...

ah ok now im seeing the bigger picture. seems like its a limitaiton on the actual sensor used for torque. The thermal issue is still a possibility in my mind because the actual module isnt measuring the important part in my opinion. but thats more of a side note. But i get your point about those numbers. i will try and see what can be done. i retired the eeprom data and i have a feeling it could be calibration data, if so, their a chance to atleast change the upper limit if im correct.

As for the angle sensor, thats actually done on a clock spring. And sent via canbus to the EPS. If anything ima start messing with my car. And start logging and doing some gateway to mess with the data going in and out.

[TRS]

That Info about the steering angle measured at the clock spring leads me to a possible idea for a solution: Of IT would be possible to intersect the Connection between the torque sensor and the actual EPS module/ECU, why not crating a little addon-box, which takes the steering angle via CAN and is between the actual ECU and the torque sensor. It should be a easy logic to simply limit the maximum torque (Sensor voltage) to the equal of 5.9Nm as long as the steering angle is between +/-400° (Just as a example under the assumption that Lock is at +/-445°). This should erase the actual issue and just limit the support to a maximum instead of sending the unit into cut off mode.

[RedReplicant]

I don't know how it works, but this is what the torque sensor looks like out of my spare BRZ column. The column has to be dropped and the front half of the case has to come off to remove the sensor.

[geraldjust]

Originally Posted by TRS

That Info about the steering angle measured at the clock spring leads me to a possible idea for a solution: Of IT would be possible to intersect the Connection between the torque sensor and the actual EPS module/ECU, why not crating a little addon-box, which takes the steering angle via CAN and is between the actual ECU and the torque sensor. It should be a easy logic to simply limit the maximum torque (Sensor voltage) to the equal of 5.9Nm as long as the steering angle is between +/-400° (Just as a example under the assumption that Lock is at +/-445°). This should erase the actual issue and just limit the support to a maximum instead of sending the unit into cut off mode.

Seems like the torque sensor is a active type. That on the ring their active circuitry that then puts out to voltages that correspond to the torque. so im thinking some sort of voltage divider but then its a fixed rate and doesnt taper off. You would also lose more "help" that the rack gives.

Second option would be a op-amp and a external chip controlling the curve. i think that more "programmable" but would be more complicated.

Third, maybe some zener diodes to always max out right before the limit. Simple, wont cut the PS, and it will leave it the feeling of its operating range the same.

I dont think the angle data or speed would be needed. if people are calming that this still happens at speeds out at the track i think.

[TRS]

Originally Posted by geraldjust

Seems like the torque sensor is a active type. That on the ring their active circuitry that then puts out to voltages that correspond to the torque. so im thinking some sort of voltage divider but then its a fixed rate and doesnt taper off. You would also lose more "help" that the rack gives.

Second option would be a op-amp and a external chip controlling the curve. i think that more "programmable" but would be more complicated.

Third, maybe some zener diodes to always max out right before the limit. Simple, wont cut the PS, and it will leave it the feeling of its operating range the same.

I dont think the angle data or speed would be needed. if people are calming that this still happens at speeds out at the track i think.

Thanks for those pics. This makes things clearer. Im pretty sure the two rings are not the actual sensor but a rotary union/lead through. I would assume that there is a strain gauge sensor on side of the shaft, which is supplied with the measuring signals/voltage by those two rings.

Nevertheless, the actually important point is, that there is a cable with connector between the sensor and the ECU unit. So it should be absolutely feasible to put an additional unit in between which limits the output voltage and in this way avoids the cut off. Using the correct connectors and pins, this could be a Plug&Play solution.

Angle data would only be needed to still keep the "lock overload protection". Means, If more then eg 90% of max steering angle is achieved, the system behaves as normal/factory. All the add-in unit needs to do is to just hand through the actual signal of the torque sensor without touching it if this angle over 90% condition is met. If the angle is below that (condition is not met), the max torque sensor output is simply limited to the equal of 5.9Nm by that hypothetical add-in box.

This would keep the behavior of the system in Standard driving conditions Like factory. Under Racing conditions, where the problematic higher loads are supplied to the steering rack, the maximum available support is limited to the value which is anyway the maximum allowed by the factory electronics. Only difference is that the cut off is avoided by never exceeding the 5.9 respectively 6Nm.

It could be kind of a "voltage limiter", which allows only voltages below {voltage equal to 5.9Nm} to pass. Everything above that value would be limited to {voltage equal to 5.9Nm}. This unit is pretty much always on and only bypassed (means actual torque sensor voltage is delivered to EPS ECU) if steering angle is above 90% of maximum.

As said, Im no expert in electronics and programming. Im the mechanical guy. But I would guess that such a box/logic should not be that big deal.

CAN hi/lo should be anyway present at the main connector of the EPS ECU and a power supply for the add-in unit is anyway needed, what I guess can also be found in this connector. So there would be two points where the add-in unit is connected inline the factory harness. Once at the main connector for power and can, and once between the torque sensor and the EPS ECU unit.

[Nmstec]

I'm pulling the eep and will see if they have limit set in eeprom, or in flash. Should be eep, if this was a lemon law recall.

[RedReplicant]

After speaking with DCE a little bit more they were able to offer a slightly cheaper kit that changes the connectors from Autosport and uses a more budget (but still extremely nice) enclosure.

They explained that the reason they can't currently build a cheaper BRZ retrofit is because the torque sensor in the BRZ column isn't compatible with their cheaper EPAS controllers.

Price is still not cheap at $4175.

[TRS]

Thanks for doing that request and sharing the information. Im still hoping for a simpler solution that keeps the OEM system, though with no higher EPS support. But as said, Im fine with the amount of support, it Just has to stay and Not cut off. Maybe the idea of that add-in box is adopted by someone capable of doing that kind of stuff...