Zombieverter & GS450H Communication Issues Investigation Topic is solved

[nkiernan]

Introduction

• This might seem like a little overkill for something that should be relatively straight forward, but believe me, I’ve been wrestling this one for a long while now!

• I’m going to post a couple of separate posts to break up the back story a little

• Currently having issues getting GS450H inverter starting or communicating with the Zombieverter VCU. To date, have tried three different inverters and two Zombieverters

• Have also been able to get the two Zombieverters and the first inverter tested on Damien’s E39 build and bench testing setup which highlighted important results

• Have an older GS450H VCU that has been able to run the GS450H inverters and transmission to prove components can all work

• Recently have been able to get the Zombieverter running the third inverter and the GS450H transmission but with some issues that need troubleshooting

• The wiring has been checked many times, rewired to simplify, scaled back to basics, and proven to work ruling out simple wiring issues (except for possible noise/interference)

The goal

• I’m hoping to work through troubleshooting these issues in this thread and hopefully finding the source of the issues to get the system up and running

• Previously documented main issues and questions in the Zombie support thread but this issue seems to be unique to me and maybe one or two others, so creating a separate thread not to clutter the support thread

• I am hoping this thread can clarify some tests for the key components and hopefully gather some refences from other users who have working systems for comparison

• Hoping to understand some more about the Toyota sync serial communication data and how to interpret it to help analysing captured comms data

• Noise has been suggested as a potential issue, so hoping to learn/document something on what to check and where, and what equipment may be needed

• If I can work through the issues with some support from others on the forum and get a resolution to these issues, I hope to convert this discussion to a troubleshooting section for the Zombieverter/GS450H Wiki pages

• I’ve been capturing communication data logs, but I wouldn’t be clear on what to look for, what is good or bad results, so hoping the community here can help along the way

[nkiernan]

Results so far:

Older GS450H VCU V3

• Originally the project was going to use the GS450H inverter/transmission setup with the older GS450H VCU from EVBMW

• Bench testing using the GS450H V4 wiring diagrams from the Wiki and forum got the GS450H transmission working first time allowing control of forward and reverse as expected and displaying relevant info on the web interface

• During these bench tests, I had two separate GS450H transmissions/inverter/VCU kits up and running

• The image below shows a Saleae 8 Channel capture of the 8 comms wires while the GS450H VCU was running the GS450H inverter/transmission. The pulses in MTH+ correspond with blipping the throttle and the transmission spinning

Conclusions drawn so far

• GS450H inverters and transmission are functional and can be controlled as needed/expected

• MTH+ data in Saleae capture should not be high so much? (https://openinverter.org/forum/viewtopi ... 835#p48835)

Question to the forum

• Does anyone have similar capture data from a GS450H VCU for comparison (what does good data look like)?

[nkiernan]

Results so far:

Zombie (Firmware V1.01 and older firmwares) & Inverter 1

• Over a period of time, attempted to run the GS450H inverter 1 and transmission with the Zombie VCU (Zombie 1) but web interface would show that InvStat would not change to On and InvUDC would not read the HV correctly. Test HV could be measured at the inverter HV input and 12V power measured at the inverter multi pin connector

• GS450H inverter and transmission were initially ruled out as being the problem as they were proven to run with the older GS450H inverter (understanding was hardware and software to run the GS450H were the same between GS450H and Zombie VCU’s)

• Early diagnosis was that there must be an issue with communication, so wiring was checked and double checked, short twisted pair communication wires were used. Next I moved onto doubting the Zombie board that had parts assembled/soldered by a local phone repair shop

• Had the four comms CAN transceivers changed for known working ones (was a period of suspected dud CAN transceivers mentioned on the forum), but this did not change result

• Finally decided to order a second built and bench tested Zombie VCU (Zombie 2) to ensure the VCU was not the problem (which was the only conclusion that could be drawn after all the testing so far), but this did not change the result either

• Logically the issue was not making sense after all the checks and so Damien offered to take the two Zombies and the inverter to run checks with the E39 project

Damien’s testing:

• Both Zombie’s worked with GS300H bench test setup

• Both Zombie’s worked with E39 GS450H install

• GS450H Inverter 1 would not run in E39 GS450H install with any Zombie (Damian’s original and the two test VCU’s). Spurious data like Invudc giving inaccurate values

• Reconnect Damian’s original inverter and all three VCU’s work again

• Both inverters share same part numbers (G9200-30051)

• Scope clock signal on Damian’s inverter – clean signal

• Scope clock signal on Inverter 1 – has breaks

• Scope clock signal from MCU side of CAN transceiver with Inverter 1 – clean signal

• Scope clock signal on output side – has breaks

• Measure approx. 60R between pairs

• Measure approx. 120k between comms sets on Inverter 1

• Measure infinite between comms sets on Damian’s inverter (TBC?)

• Appears clock signal is leaking on to comms wires and vice versa

• Removed bottom section of inverter and found corrosion around the pins where data connector feeds into PCB (see image below)

• But Inverter 1 works with older GS450H VCU. Checking code showed the older VCU does not care so much about the return data on MTH, just used for information. However, Zombie uses it to decide if comms have been established between inverter and VCU and won’t enter next steps unless it see’s valid MTH data

Conclusions drawn so far

• Suspected corrosion on PCB leading to crosstalk issue and poor/broken communication, inverter likely source of problem

• From the Saleae capture, CLK, REQ, and HTM data seemed ok in a number of captures. HTM data verified as correct init data for the GS450H inverter so correct inverter selected in web interface. MTH data missing or poor

Zombie (Firmware 1.01) & Inverter 2

• Sourced a second inverter from a breakers but there was evidence the top covers had been removed previously so a little suspicious about how good it would be

• Connected Inverter 2 but same exact results as Inverter 1, InvStat stays off, Invudc not reading HV, no spin, etc

• Chances of getting two inverters with the same issue!?

• Opened Inverter 2 to inspect the PCB that had the corrosion in Inverter 1, but this didn’t look to have any issues or signs of corrosion

• Returned Inverter 2 to breakers

Zombie (Firmware 1.01) & Inverter 1 with replacement PCB from inverter 2

• Not ruling out that Inverter 2 could have another issue, I swapped the clean Inverter comms PCB into the known running Inverter 1 for a test but same exact results as original Inverter 1, InvStat stays off, Invudc not reading HV, no spin, etc

Zombie (Firmware 1.01 and 1.06) & Inverter 3

• Sourced a third inverter from a car that had supposedly driven into the breakers with head gasket issues

• Connected Inverter 3 and with Zombie still on V1.01 firmware tried another test, but same results as previous tests, InvStat stays off, Invudc not reading HV, no spin, etc

• Chances of getting three separate inverters with the same/similar issue!? So I am deducing that there is more to this issue than a bad inverter(s)

• One of the big bug fixes in firmware V1.06 related to Toyota communications and issues starting the inverter, plausible connection to issues I’m seeing

• Updated to V1.06 firmware and repeated test, but same results, InvStat stays off, Invudc not reading HV, no spin, etc

• Decided to get a Saleae capture for reference and to compare to previous inverter captures

• Connected Saleae USB power cable to laptop (running off its own internal battery and not from mains or similar power source to Zombie/Inverter 12V power)

• As per other tests, Ignition on, Start signal applied, HV contactors come on in sequence as expected and for the first time the Inverter 3 starts whining with the Zombie

• At the same time however, the laptop was pinging warnings that the Saleae USB was being continuously connected and disconnected (like what happens when power draw from USB is too much or surging). The Saleae logic software could not keep the connection to get a capture like in previous tests

• Thinking this was not right, the USB cable was unplugged, and immediately the Inverter 3 whine stopped and it shut off!

• Curious to know if the transmission would spin when the inverter whined, reconnected the Saleae USB plug, Ignition On, Start Signal applied, and once again inverter whine and Saleae USB connecting and disconnecting from laptop

• Grabbed screen captures of the web interface showing InvStat was On and Invudc was reading the correct HV value

• Removed brake signal and applied forward drive signal, and when pressing throttle the GS450H transmission spins up as expected. Continued running the transmission at different speeds for a couple of minutes

• Explaining this in the Zombieverter Support Thread, initial feedback was the symptoms could relate to grounding or noise issues

• To try rule out noise issues, I wired up a new VCU connector plug with only the minimum wires needed to run the transmission (originally I had a prewired plug with almost 2.5m of wire for each pin, some of which were coiled up during testing). I also removed the LED indicators I had for HV contactor state

• I tidied up all wiring and fuses and relays to minimise issues

• Repeating the test above with this new wiring didn’t get the same results! While the inverter definitely whinned on a number of occasions, it seemed to stop after a period of time. InvStat seemed to stay off mostly, but Invudc was reading some lower voltages than test HV (as if inverter was on for a while during precharge, read the correct value up to a point, then switched off – possibly at HV+ contactor coming on?)

• Also noticed the HV contactors doing a bit of a dance when reapplying ignition on and start where it used to be three clicks?

• Try clean ground lines/sizes, remove LED indicators, check contactor operation, test for noise?, oscilloscope? (need new tests to prep for this)

Initial thoughts

• So the GS450H transmission and Inverter 3 can be controlled by Zombie VCU running V1.06 firmware, but not repeatable?

• Is this latest behaviour and being able to spin the motor (albeit with Saleae help!) due to the changes made in V1.06 firmware?

• If GS450H VCU firmware can run the system ok, would there be a downside to changing Zombie firmware to match what works on GS450H VCU, or is that masking a deeper problem?

• Johannes commented on a slight hardware difference in the VCU’s (https://openinverter.org/forum/viewtopi ... 837#p48837)

• Would this Inverter 3 now work in Damien’s E39?

[nkiernan]

Component checks/testing

• This is where I hope others with running Zombie/GS450H systems might be able to help share info for comparison

• Image below shows a table of resistance measurements to confirm comms connections:

• Some Pulseview/Sigrok captures using a clone 8 Channel analyser:

Pulseview_8Channels_Zombie_IgnOnOnlyInv3_271122.zip

(4.47 KiB) Downloaded 54 times

Pulseview_8Channels_Zombie_IgnOnStartInv3_271122.zip

(70.72 KiB) Downloaded 59 times

Questions to forum

• Would anyone have working GS450H setups with either the GS450H VCU or Zombie to get some comparison measurements. Interested in the approx. 120k readings!

• Open to any comments or suggestions on things to check

• Is there anything in the resistance measurements that stand out. All looks ok to me, 60R between pairs. Are the ~120k readings meant to be infinity?

• Looking to get an oscilloscope to help investigate if noise is an issue

• The file names on the captures explains the main test parameters, maybe someone who knows what they are looking at could help me see if there are details here that could point to the problem?

[crasbe]

From your screenshots it's obvious that it can't possibly be a software issue:

All of the signals are differential pairs. That means if CLK+ for example is high, CLK- must be low. In fact we can see that that's not the case for any of the signals. Even worse, sometimes there are glitches that pull all signals down:

The differential signalling however is done by the CAN transceivers. The software does not have any influence on that.


I don't know what is wrong, but there is something really wrong with your setup.
What are you using to supply the inverter and the ZombieVerter with +12V? What kind of batteries are you using?
It looks like the +12V lines are really thin going to the inverter and the ZombieVerter? Whats the input voltage on the ZombieVerter and inverter measured directly on the PCB for the Zombie and directly at the connection to the inverter?

Can you make a Logic Analyzer measurement with the inverter disconnected? Maybe we can observe if the signals look better or not.

[nkiernan]

What are you using to supply the inverter and the ZombieVerter with +12V? What kind of batteries are you using?
It looks like the +12V lines are really thin going to the inverter and the ZombieVerter? Whats the input voltage on the ZombieVerter and inverter measured directly on the PCB for the Zombie and directly at the connection to the inverter?

Can you make a Logic Analyzer measurement with the inverter disconnected? Maybe we can observe if the signals look better or not.

Thank you Crasbe for taking a look at this. Do the later sections of the captures look any better? The captures start with just 12V power applied and then ignition on signal applied and then start signal applied. I am not sure however if the clone Saleae logic analyser I am using could be adding unwanted artifacts.

I am using a 60V 3A PSU for the 12V power (set at 13.5V). For the HV I am using four 12V lead acid batteries in series giving approx 49.5V for HV testing.

Wires used are 0.75mm2 for the zombie and inverter power lines. Earlier testing used 1.5mm2 for the inverter 12V but changed them during recent wire tidy up. I can get this changed back. I will measure the true input voltage at the Zombie and inverter and report back and I'll get a capture later today with inverter disconnected. Thank you

[dadiowe]

Just a thought, have you earthed the gearbox?

[nkiernan]

Just a thought, have you earthed the gearbox?

Should be ok, both the gearbox case and the inverter case are connected back to the PSU (13.5V) ground with separate 2.5mm2 wires

[dadiowe]

Another quick thought are your shields only connected at one end?

[nkiernan]

Another quick thought are your shields only connected at one end?

Are you referring to the four comms pairs? Currently they're shorter twisted pair only (0.75mm2). Waiting delivery of two core twiested pair shielded to swap in. Do others use the drain pins on the inverter connector?

[dadiowe]

I connected the drains to the inverter connector and I used shielded twisted pairs to the VCU. Having said that the shields going to the VCU are not joined together or earthed.

[Pete9008]

Not too familiar with this but what voltage do the interface lines run at, is it CAN, RS485, 5V, 12V, etc?

It is quite possible that the thresholds on the Saleae do not match well and that is what is causing the odd levels seen (you can adjust the thresholds on some devices, can you on yours?). Getting a scope on them would be the best way to go for this type of problem.

[nkiernan]

Not too familiar with this but what voltage do the interface lines run at, is it CAN, RS485, 5V, 12V, etc?

It is quite possible that the thresholds on the Saleae do not match well and that is what is causing the odd levels seen (you can adjust the thresholds on some devices, can you on yours?). Getting a scope on them would be the best way to go for this type of problem.

Thanks Pete9008. Comms lines are essentially CAN, CAN level but not actual CAN, and use CAN transceivers. I can connect the analyser to both Saleae Logic and Sigrok Pulseview. I'll admit I've only just started using these so need to get more familiar with the settings but I can have a look at this.

Have been following the advice on other threads about best scope options, so will be getting one very soon. Thank you

[Pete9008]

Just checked, my Kingst logic analyser defaults to a threshold of 1.58V which is probably on the low side for this. You can however set it to up to 4V which would be a lot better (although I think you can only do this from the Kingst software not from pulseview). Would be worth checking to see if the feature is available on yours?

Are you going to go for the Rigol scope?

[crasbe]

Thank you Crasbe for taking a look at this. Do the later sections of the captures look any better? The captures start with just 12V power applied and then ignition on signal applied and then start signal applied. I am not sure however if the clone Saleae logic analyser I am using could be adding unwanted artifacts.

I am using a 60V 3A PSU for the 12V power (set at 13.5V). For the HV I am using four 12V lead acid batteries in series giving approx 49.5V for HV testing.

Wires used are 0.75mm2 for the zombie and inverter power lines. Earlier testing used 1.5mm2 for the inverter 12V but changed them during recent wire tidy up. I can get this changed back. I will measure the true input voltage at the Zombie and inverter and report back and I'll get a capture later today with inverter disconnected. Thank you

The later sections don't look any different IMO than at the beginning, the same issues are present.

3A might be a little bit too weak, maybe you can find another lead acid battery to use as a buffer for the 12V system?
The wiki says the inverter needs around 1.5A in idle, but that can be higher when actually operating.
Is your power supply actually grounded at ground? Many lab power supplies have a separate earth connection because the ground is not connected to earth.


0.75mm² is a bit thick for the differential signals because it has the wrong impedance, but I don't really think that's the issue here. No need to be changed I think.
0.75mm² should be okay for the 12V lines, depending on how much current the inverter needs in operation.


Are you using the oil pump? That draws a significant amount of current from the 12V supply

[nkiernan]

Are you going to go for the Rigol scope?

It's looking like the one for me alright

[Pete9008]

It's looking like the one for me alright

Not used the scopes but have one of their sig gens here and it's pretty good (not quite up to Tektronics standards but then it's a fraction of the price). Tempted by the scope myself as it has a much better screen than mine, bigger memory and the protocol decoders would be very nice, but can't justify it it at the moment. It's worth having a look on eevblog.com for reviews and tips on using them.

[nkiernan]

The later sections don't look any different IMO than at the beginning, the same issues are present.

Damien did say he could confirm some of the data was the correct init data, so some captures might have had decent data. These one's obviously not so. Are there decoders for the analysers to convert this data or is that another mine field!?

I'm not using the oil pump during these tests. In standby, the PSU is showing approx 0.17A (from memory) and aftre ignition on and start signal, it sits around 1.5A-1.7A (from memory). No indication its tripping out on amps but yes, low power. I had a 12V battery connected but wondered if the 11.6V to 12V was an issue so tried the PSU to get 13.5V. I can swap back in the 12V battery no issue.

The PSU is a cheap amazon unit, has the three connections on front panel (+, -, Ground) but ground not connected to anything here. Not sure about the mains plug wiring

[Pete9008]

Are there decoders for the analysers to convert this data or is that another mine field!?

Very much doubt that there is an existing one specifically for this as it's Toyota specific although it might be close enough to another protocol for a standard decoder to still be useful (UART??).

You can add custom ones in pulseview, I think it's done in python? As far as I know on the scopes you are limited to the protocols it comes with.

I've done one for the Kingst logic analyser software in the past, to decode a LCD bus, and it wasn't too bad, around an afternoons work.

[mjc506]

Do you have a physical connection between invert case (gnd) and system ground? Ideally, ground every bit of electrics apart from your HV to the inverter case. Your +, -, Gnd connections on the PSU are ok, just treat '-' as your ground.

Connecting your PSU to a car battery, and then to the inverter will do no harm. Worst case, you trickle charge the battery, best case, it provides enough buffer for spikes in load!

[nkiernan]

Do you have a physical connection between invert case (gnd) and system ground? Ideally, ground every bit of electrics apart from your HV to the inverter case. Your +, -, Gnd connections on the PSU are ok, just treat '-' as your ground.

Yes, inverter case is phsically connected to PSU '-' and gearbox case. I could go further and have a better connection to Zombie ground possibly

[crasbe]

Damien did say he could confirm some of the data was the correct init data, so some captures might have had decent data. These one's obviously not so. Are there decoders for the analysers to convert this data or is that another mine field!?

The sync serial data should be decodable with one of the protocol decoders of Pulseview. Either directly by the CAN decoder or the SPI decoder when using the CLK as clock input.
But at the current stage I doubt it would output anything useful.

I'm not using the oil pump during these tests. In standby, the PSU is showing approx 0.17A (from memory) and aftre ignition on and start signal, it sits around 1.5A-1.7A (from memory). No indication its tripping out on amps but yes, low power. I had a 12V battery connected but wondered if the 11.6V to 12V was an issue so tried the PSU to get 13.5V. I can swap back in the 12V battery no issue.

11.6V is pretty low, but I meant that you could connect the battery AND the power supply. Obviously the battery should be charged before that, so that it doesn't load down the power supply

The PSU is a cheap amazon unit, has the three connections on front panel (+, -, Ground) but ground not connected to anything here. Not sure about the mains plug wiring

That's what I wanted to hear. I would recommend connecting your - terminal to the Ground terminal, then the system is "earthed".
In theory that should have the same effect that you observed when you connect the Laptop.

[nkiernan]

Update:

• Following your comments and suggestions above, got to do a little more testing after some setup changes

• Good news is we’re on the right track as I was able to get cleaner contactor operation, inverter to whine when turned on, and motors spinning. Still some quirks to figure out before its ready to put in the car

• One key conclusion is that the Saleae clone logic analyser is not doing its job and has likely steered some earlier conclusions in the wrong direction!

• Image below shows the updated setup. 12V battery and PSU supplying the 12V power (13.5V), 4 x 12V (49.5V) batteries in series supplying the test HV, PSU ‘-‘ connected to PSU ‘Gnd’, 8-channel analyser connected to Zombie and powered by laptop (running off own battery, no mains power connection)

Testing:

• To start, ran some initial tests to get captures with standby power only, then to ignition on, etc

• For the full system test, system was powered up, Brake signal applied, Ignition signal applied, Start signal applied, and inverter starts to whine

• Checking the web interface, InvStat = On and Invudc = 48V

• Measured 13.23V directly at the Zombieverter header power pins

• Measured 49.3V at the inverter HV input and 13.09V at the inverter multi-way plug

• Recorded some Pulseview captures

• Removed brake, applied Fwd signal, applied throttle and motors span up as expected. Ran motors for a while, then removed Fwd signal, applied Rev signal, applied throttle and motors span backwards as expected

• To finish, I applied Brake signal, removed Rev signal, removed Ignition On signal and contactors opened and Zombie was in standby power mode

• Trying to repeat the test, I reapplied Ignition on signal, and Start signal, but system would not power up as before (no inverter whine or inverter data on web interface). Seemed I had to completely shut off power to Zombie (power cycle) before I could restart the system
  I need to double check this again though as this doesn’t sound right if Zombie has permanent 12V power in a conversion

Conclusion so far:

• System can be run as expected, so components should all be ok

• Logic analyser captures appear corrupt even though system is running, so Logic analyser must not be working as expected

• Cleaner contactor operation. Had some issues with Udcsw setting and contactor chatter earlier which I had put down to Inverter not powering, but appears it could have been PSU struggling to give enough Amps as the three contractors were expected to come on so close to each other as I kept lowering the switch threshold. So the 12V battery helped provide the amps needed

• Possible knock-on effect was likely spikes in amp demand for contactors interfering with the Zombie/Inverter power and could have interfered with communications at start up request

Laptop influence:

• During the tests above, the laptop was always connected to the system (USB cable to Logic analyser running off laptop’s own battery power)

• Previously, the laptop was giving warnings that the USB connection was being continuously disconnected and reconnected while the system was powered up and gearbox could be run. No captures could be taken then

• In these latest tests, the USB connection was not lost and captures could be taken (better grounding/earth?)

• Previously, when the USB cable was disconnected, the inverter whine would stop immediately and system could not be run

• So with the system powered up and motor able to run, I unplugged the USB cable, and this time the inverter stayed whinning and I could continue spinning the motor with throttle input. However, after a short period of time, the whine faded away and the inverter stopped and the motors could not be run any further. So laptop connection still has an effect

• Without shutting off the system or restarting, I reconnected the USB cable and the inverter started to whine again and I could spin the motors again with throttle intput. System picked up where it left off (safe in real world application?). However, InvStat in the web interface stays off during this inverter restart even though motors can be run and inverter whines

• Next I powered down and restarted the system with laptop Logic analyser connected, inverter whining and motors spinning ok

• This time I disconnected the Logic analyser test lead connected to the Zombie ground pin (to see if it was ground/earth from analyser that was having the effect)

• System continued to whine and run with no issue

• Disconnected USB cable completely and appears system continued to run (I need to double check it did not shut down after a period of time as above)

• Finally, I powered down the system and with USB cable unplugged (no laptop connected), I restarted the system

• Inverter did indeed start to whine as before, but after a number of seconds the whine faded away and the motors could not be run

• Appreciate testing is not complete, was late and ran out of time. Wanted to get a capture with inverter plug disconnected and would need to repeat some of the tests above to check repeatability, but some intereting results and changes to operation. Thank you all for your help

Captures:

I’ve included a number of captures from Pulseview. The file sizes are ridiculously small compared to the equivalent Saleae Logic captures so hope they’re complete. The file name tries to explain the test setup. Key one is the AfterStartInv3GndConnectedGearboxRunning file. This is the capture when inverter was on and I was spinning the motors. What I’ve noticed is the capture data is pretty much garbage! So better test equipment needed. Will be going for the Rigol DS1054Z. GndConnected refers to connecting the PSU ‘-‘ to PSU’Gnd’ as per Crasbe’s comment above. One capture is without this connection

Pulseview_8Channels_Zombie_AfterStartInv3GndConnectedGearboxRunning_281122.zip

(79.9 KiB) Downloaded 60 times

Pulseview_8Channels_Zombie_AfterStartInv3GndConnected_281122.zip

(16.45 KiB) Downloaded 54 times

Pulseview_8Channels_Zombie_IgnOnOnlyInv3GndConnected_281122.zip

(3.79 KiB) Downloaded 54 times

Pulseview_8Channels_Zombie_12VPowerOnlyInv3GndConnected_281122.zip

(4.73 KiB) Downloaded 52 times

Pulseview_8Channels_Zombie_12VPowerOnlyInv3NoGnd_281122.zip

(3.16 KiB) Downloaded 51 times

[dadiowe]

Good news in some ways. You seem to be going in the right direction. Perhaps a few more Volts on the HV side might help

[nkiernan]

Looking like success!!!

Got to spend time yesterday on this and now have two separate Zombie VCU's working, two inverters working, and the GS450H transmission spinning...and this is without the laptop or logic analyser connected and just a 12V battery for the LV supply, no PSU!

Looking back at the comments in this thread, I focused more on the ground/earth connections and routing. I didn't expect this to change anything but wanted to give the list of tests I had planned the best chance (new Rigol scope had just landed too)


Setup changes:

• Re-pinned the Zombie ground wire with a larger 1.5mm2 wire, originally this was 0.75mm2 (thinking about the three contactors using this ground path as well as the VCU power)

• Changed the inverter LV power wires to larger 1.5mm2 wires, originally both '+' and '-' were 0.75mm2 (looking at the GS450H V4 wiring diagram and fuse rating)

• Routed inverter LV '+' to 12V battery '+' terminal via 10A fuse (as per GS450H V4 wiring diagram, originally had this coming into a shared fuse block)

• All ground and earth wires were routed back to the 12V battery '-' terminal (previously some were routed to a common fuse block ground)

• Added a switch panel for easier testing (previously I was swapping wires in a quick connector (thought I was in for the long haul on this one)

Test 1:

• I had a number of tests planned out, with/without laptop connected, different VCU firmwares, chasing CLK signal from MCU to CAN transceiver to inverter, etc (Zombie 2 on V1.06A and Inverter 3)

• To kick things off, I fired up the setup without the laptop or logic analyser connected as a benchmark (12V battery and PSU @ 13.5V connected together as per previous tests)

• And there it was, inverter whining, and system appeared to be up and running - a surprise! Checked the web interface and Invstat was reading 'Off', but Invudc was reading approximately correct HV value

• Brake off, Forward signal applied, throttle pressed, and motors spin...nice one

• Ran a number of tests with this set up and noticed that sometimes, InvStat would read 'Off' and sometimes 'On', even though motors could spin each time and inverter was whining

• Decided to see if the PSU LV supply had any effect, so disconnected it leaving just a 12V battery for the LV, and system ran as before

Test 2:

• Swapped in Zombie VCU 1 (on V1.01A firmware) to check if the recent firmware updates had an effect (not expecting it to run)

• Powered up the system and inverter whined and motors could spin!!!

Test 3:

• On a roll, so decided to test the original inverter (that wouldn't work on Damien's E39 due to the corroded comms PCB. I have replaced this PCB)

• Connected Zombie 2 (V1.06A firmware)

• Powered system up, and again, inverter whined and motors could spin (so two working VCU's and two working inverters, and is it really all down to ground wire size and routing!?)

Observations/Queries:

• As mentioned above, I tested the system a number of times and periodically InvStat would read 'Off' even though inverter was whining and motors could spin. Other times it would read 'On'. Is this normal or still pointing to a small communications issue?

• Wondering what resolution the inverter HV measurement has (Invudc reading). Test HV was approx 49+V, Invudc always reading 48V (not an issue in accuracy, but next point leads me to double check its not a value just locked in the system and its not actually reading correct)?

• When system is up and running, Invudc reads 48V which seems spot on, test HV is approx 49V. When system is powered down, HV at the inverter inputs reduces as expected and refreshing the web interface shows this reducing (udc value). However, Invudc just stays at 48V? Is this normal as this was the HV inverter was seeing when it got switched off, or should it update?

• Big question on restarting the system after turning it off. With Permanent 12V applied to Zombie VCU, so LED's remain blinking (some conversions would leave the VCU permanently powered like this?), when I turn ignition off, contactors go off, and HV udc reading starts to drop. I expected that I could reapply ignition signal and start the system up again. If udc had not dropped below udcsw, system should just start. If udc had dropped below udcsw, precharge would start and then system start up. However, contactors click and I measure HV input at the inverter drops to a few volts (even if udc had been higher while still falling after system shut down) and so no precharge or HV contactor and system won't start. I found I had to power the VCU off completely for a few seconds, then power it on and starting would act normally then?

Thank you all for your help and suggestions