GS450h Cuts Out on Heavy Throttle Application SOLVED

[catphish]

I drove in Jackk's vehicle today. The issue he's describing is an extremely violent loss and then return of torque. It only seems to occur at high power levels (150A+ at 600V). Having ruled out the DC current derate, the next most likely problem is a serial communication issue. Since others don't experience the same problem, electrical interference seems a very likely cause. I know the GS450H is very electrically noisy. What cabling are others using for the serial connection between the VCU and the GS450H?
It's a real shame the STM32F103 can't do sync serial with a permanent clock, but it's abundantly clear that this shouldn't matter, since many people are now using this without issue.

[Jackk]

I have set my car as per Damiens video with the 1.11 update and it has fixed my torque drop. My RPM was set at 6000 and
the rev limit was noticable but not violent. Other settings when I drove the car were, as far as I can recall, throtmax 50% and throtramp set at 1.
I have not tried the car with any settings changed as yet.
Hope that helps

Thanks,

Can I ask what Thotomax and Throtramp actually mean?
My throtomax is set to 100% and Throtramp is 100.

[catphish]

Thanks,

Can I ask what Thotomax and Throtramp actually mean?
My throtomax is set to 100% and Throtramp is 100.

Make sure to read https://openinverter.org/wiki/Parameters
throtomax is a way to limit throttle to less than 100%, for example to allow economical driving.
Throtramp is how many % the throttle is allowed to increase every 100ms. So with throtramp=100 you are allowing throttle to go from 0% to 100% in 10ms. This should be fine with the GS450H as the Toyota hardware will limit things as needed.

[catphish]

One thing I'm reluctant to do is change my serial wiring (going to be a bit of a ball ache). The run of the wiring is relatively short id say less than 1m. They are twisted pairs but NOT shielded. Could non shielded serial wiring be the cause? My gut feeling is a code issue as others have experienced the same.

The official answer Damien given in his videos is:

Each serial wire pair needs to be screened and the screen needs to be grounded, and the inverter and box grounded with straps to chassis

[dadiowe]

As per catphish my set up is, each serial wire pair is screened and the screen is grounded, and the inverter and box grounded with straps to chassis.

[crasbe]

Please provide some pictures of your wiring so we can get a better idea of the setup

[dadiowe]

Not sure if you mean me or Jackk, crasbe, but my wiring is as per https://openinverter.org/wiki/File:Zomb-con-et.png.
Also went out today to try the settings that Jackk mentioned but unfortunately the Sun was low and I did not see a traffic island in a narrowing of the road. 2 wheels and tyres gone (as a minimum). Happy Days.

[Bigpie]

Is this actually solved as per the thread title? If not I'll remove SOLVED

[crasbe]

Not sure if you mean me or Jackk, crasbe, but my wiring is as per https://openinverter.org/wiki/File:Zomb-con-et.png.

I would like to see some real pictures of the actual wiring. Four eyes (or even more in a forum) might see more than two

[dadiowe]

As far as my original post is concerned it is. After updating my zombieverter to 1.11. I do not have Torque cutting issue.

It seems Jackk may still have an issue though.

[dadiowe]

I would like to see some real pictures of the actual wiring. Four eyes (or even more in a forum) might see more than two

Why do you need to see my wiring, my car is now fine?

It is Jackk who still has an issue!!!

[crasbe]

As an example of working wiring, so we could be able to compare it. But alright, whatever.

[Jackk]

UPDATE.

I took the time to replace the four twisted serial cables to screened cable. The shields are all soldered to one wire near the inverter plug then this lead approx 20cm long connects straight to the inverter casing. The inverter itself has a seperate ground strap which goes to chassis.

The 4x serial cables are no longer than a metre long. The sheilds are cut off at the VCU end.

And you guessed it! Sadly the car is still violently cutting torque when you boot it. If I am gradual with the throttle it is fine and still goes well.

Can anyone suggest what to do next? Max max Rpm is set to 8000rpm.

Really don't know what to try now. Is there a way of testing for noise on the serial wires?

I kind of feel like this isn't a noise issue. the serial wires are pretty far from any HV cables.

Any suggestions would be great!

Jack

[Pete9008]

Not sure this is the right place to ask this question but it's been bugging me since noticing it.

The serial link for the 450h inverter appears to be using CAN transceivers (based on this thread viewtopic.php?t=205), now this just seems a really odd choice to me. CAN and RS485 may look pretty similar when scoping a board but CAN has lower common mode immunity and lower drive strength in the recessive state. Now this makes sense on a multi-drop protocol like CAN but very little sense on a point to point serial link like on the Toyota inverters. I realise the Toyota chip used is a proprietary Denso se617, and no data sheet is available, but the pinout looks like a standard RS485 transceiver to me. Could anyone point me towards any measurements confirming that Toyota is actually using CAN on the serial bus rather than RS485 or explain what I'm missing?

Sorry for raising this but as mentioned it's been bugging me since noticing it and if the wrong transceiver is being used you would get a system that works but with lower than expected noise immunity.

Edit - take that back, just checked a board I have here and the pinout doesn't match a RS485 transceiver.

Edit2 - Actually the pinout does look like a CAN transceiver. Still interested in the signal levels though if anyone has measured them?

[dadiowe]

UPDATE.

I took the time to replace the four twisted serial cables to screened cable. The shields are all soldered to one wire near the inverter plug then this lead approx 20cm long connects straight to the inverter casing. The inverter itself has a seperate ground strap which goes to chassis.

The 4x serial cables are no longer than a metre long. The sheilds are cut off at the VCU end.

And you guessed it! Sadly the car is still violently cutting torque when you boot it. If I am gradual with the throttle it is fine and still goes well.

Can anyone suggest what to do next? Max max Rpm is set to 8000rpm.

Really don't know what to try now. Is there a way of testing for noise on the serial wires?

I kind of feel like this isn't a noise issue. the serial wires are pretty far from any HV cables.

Any suggestions would be great!

Jack

Silly question but does the version show 1.11 in the software?

[Jackk]

Unfortunately yes it shows 1.11.

One thing to note which may be unique about my setup from most people is that I'm running 600v nominal.

Would this increase the likelihood of noise?

[dadiowe]

Unfortunately yes it shows 1.11.

One thing to note which may be unique about my setup from most people is that I'm running 600v nominal.

Would this increase the likelihood of noise?

Is it possible to easily split your pack down to say 400v to see if it is too much voltage?

[Jackk]

Unfortunately splitting pack down would be something I really want to avoid as it would be a major setback/ very difficult.

There is a chap from USA called mauswerkz he ran a gs450h successfully at 650v. Does anybody know anyone else who has successfully run 600V?

One thing I will say removing the current limit in the code on the 1.11 software does definitely improve the issue. With the current limit code removed it cuts out at much higher loads.

If I use the 1.11 software with the Current limit code it cuts out very easily from relatively small throttle input.

This leads me to think this is still a code issue. I wish I could look into this but I would not know where to start!

One thing to note my encoder wires for MG1 and MG2 are also not shielded. I seem to remember the original oem cables were not but I could be wrong?

[royhen99]

OEM resolver cables are screened/shielded.

https://youtube.com/clip/UgkxKZPA10Tic3 ... ZayEEM8vpM

[Jack Bauer]

To get to the bottom of this is going to require some slow and steady diagnostics and information. Please do the following in this order and post the information here :

1)Use unmodified code straight from github
2)Post your parameter file here
3)Run a plot of the the "torque" paramater from spot values. Trigger the cutout, post the plot here.
4)If the plot shows "torque" drops to 0 and jumps back up etc during the cutout then its the vcu commanding the cutout
5)If its the vcu commanding the cutout , add paramater "idc" to the plot and run again a above. Post plot here.
6)if idc shows very high spikey values in time with the cutout then the vcu is receiving erroneous data from the isa shunt.
7)If neither torque drops nor idc spikes then its the Lexus inverter either loosing comms or losing rotor position or exceeding some factory presets.
8)if idc is spiking as per item 6 then have a look at how the isa shunt is wired.

In the past few days I have driven my e39 over 200 miles attempting to get it to do something like the problem described here. Booting the throttle from a stop or from x to y speed etc etc. Finding the root cause of a problem like this is 99% boring dumb data logging and 1% inspiration. I have the e39 booked into a dyno on feb 23rd where I will beat the tar out of that 450h. I can't do anything else from this end.

[dadiowe]

Damiens route seems a very sensible approach and it will be interesting to see the results.
Re, beating the tar out of the e39. I am not saying that this is not a harware problem.
However, prior to 1.11 my car showed the torque cut and Damiens did not and mine is fine now, so it is unlikely that the problem will show up with 1.11 on Damiens car as it did not show up using v1.10A either.
Unfortunately I can' test this week due to a lack of wheel rims and tyres but happy to test once the car is mobile again.

[Jackk]

Hello,

Firstly thankyou so much for looking into this!

Please see attached parameter file...

Here is the first torque graph. I booted it only once and you can see the torque cut clearly in the graph... It seems while the pedal is depressed after the cut it applies no torque. You have release throttle completely to get it to move again.

Here is the second plot showing torque and IDC... I'm not sure what you mean exactly by spiking values....

What are your thoughts??

I'm still yet to change my encoder wires to shielded. I can do this if you think its necessary.


Many thanks again!!

[evMacGyver]

I'm still yet to change my encoder wires to shielded. I can do this if you think its necessary.

So your resolver cables are not shielded and shield grounded as wiki connections tells to do:
https://openinverter.org/wiki/images/b/ ... ctions.png
You really really really should do it.

[Jackk]

So your resolver cables are not shielded and shield grounded as wiki connections tells to do:
https://openinverter.org/wiki/images/b/ ... ctions.png
You really really really should do it.

Indeed Ill try and get this done today and report back!!

[evMacGyver]

Here is the first torque graph. I booted it only once and you can see the torque cut clearly in the graph... It seems while the pedal is depressed after the cut it applies no torque. You have release throttle completely to get it to move again.

I use older gs450h board so I do not know zombie software that much.

In your graphs there are two torque request ramps, was throttle pedal pressed all the time? You said it need to be depressed to get torque back, but depress is not included in these graphs?

If depress is not in your graph, then I started to think that isn't Torque value only a request, not what inverter is supplying = not a feedback. If it's torque request, as in older gs450h software, there is something dropping your torque request and inverter is doing what its supposed to do. Torque value being request also should eliminate resolver wiring problem. So it could be current sensor spike as mentioned before, but it could also be noisy TPS sensor value. If your TPS value drops for some reason it will drop torque request.

Someone might know better what all inputs could cause torque request limitations on zombie, for me doing graph torque, tps and idc would be interesting.

Edit: looked the source code. If TPS input value would be over its limits (from noise), tps will be limited to min/max. Your TPS is calibrated correctly and works linearly?
If your brake signal is noisy, it would cause tps value to be zero, but it should not feel that aggressive?
There seems to be also limits for Udc and temperature derates.

Anyway, aggressive torque drop points me to EMC problems somewhere, but all things that might alter torque request are another interest.