SOLVED! Infineon Cutting out Question

[gwhy!]

Hi Jeremy,
The R43/C20 was my first gut reaction to look at until I started metering the circuit, But what you say certainly makes sense. We will have to give it ago and see what happens.

[Jeremy Harris]

I'd be really interested to see what you find out. In my application for these controllers I have no way of loading the motor at low rpm, so the problem doesn't arise (I'm driving a boat prop, whose load is proportional to the cube of rpm).


Jeremy

[matt_in_mtl]

Matt,

Looking at the circuit I'm near-certain that the brutal cut off is the QF1 switch, not anything in the programming.

The fix may be something as simple as increasing the value of C20 by enough to slug down the big spikes, or adding a resistor across it to divide down the voltage.

Here's an experiment to test this idea. Fit a 1k resistor in parallel with C20. This should double the "emergency current limit" point. Should be easy enough to test.

Jeremy

As I think about this more, I think this makes alot of sence. I suspect that if we scope it we will find the RC time constant of C20 and R43 will be quite low. I suspect that the ADC current input is polled by firmware, mabye at around 100hz? This allows average current regulation, but is much too slow to limit pwm frequency current spikes. The output of QF1 possibly triggers an interrupt in the controller causing instant shutdown. A lower C20 value will give faster response to over current-scenarios better protecting the controller. 0.6V drop accross the 4.5mohm shunt would require 133A. if C20 is 1uF, that would give an RC time of 1mS, and 10uF would be 10mS, so I suspect that if the current exceeds 133A for even 1/100 of a second, it would be enough to trigger the over-current shutoff. This actually seems like a fairly smart feature of the controller. I am excited to test this out when I get home. Unless of course you beat me to it gwhy!!

-Matt

[gwhy!]

Well Have a guess what.... 1k across C20 makes the controller WORK, I fitted the 1k to my new virgin controller so its stock settings and stock shunt. I havent tested it under a big load yet ( but that will come ) , without the mod the motor would peak at 10A on the watt meter but cut out and with the mod watt meter still peaked at 10A but no sign of the motor cutting out no matter how you used the throttle. So proof of concept yes! ( maybe ) but more testing to do.

[Jeremy Harris]

Nice one!

I'm glad that this idea seems to have some merit. The next step is to have a think about ways to fully understand this and the impact it has on performance and FET longevity.

It should be possible to fit a small trim pot across C20, so that the peak current cut-off can be fine tuned. Whether this is worth it or not is debatable, though.

Jeremy

[gwhy!]

Nice one!

I'm glad that this idea seems to have some merit. The next step is to have a think about ways to fully understand this and the impact it has on performance and FET longevity.

It should be possible to fit a small trim pot across C20, so that the peak current cut-off can be fine tuned. Whether this is worth it or not is debatable, though.

Jeremy

Yes definitely needs to be looked at as regards understanding the implications on reliability, but that will come, im sure.

[gwhy!]

As I think about this more, I think this makes alot of sence. I suspect that if we scope it we will find the RC time constant of C20 and R43 will be quite low. I suspect that the ADC current input is polled by firmware, mabye at around 100hz? This allows average current regulation, but is much too slow to limit pwm frequency current spikes. The output of QF1 possibly triggers an interrupt in the controller causing instant shutdown. A lower C20 value will give faster response to over current-scenarios better protecting the controller. 0.6V drop accross the 4.5mohm shunt would require 133A. if C20 is 1uF, that would give an RC time of 1mS, and 10uF would be 10mS, so I suspect that if the current exceeds 133A for even 1/100 of a second, it would be enough to trigger the over-current shutoff. This actually seems like a fairly smart feature of the controller. I am excited to test this out when I get home. Unless of course you beat me to it gwhy!!

-Matt

So re-thinking about reprogramming using 218 settings, will this also work

[matt_in_mtl]

Nice one!

I'm glad that this idea seems to have some merit. The next step is to have a think about ways to fully understand this and the impact it has on performance and FET longevity.

It should be possible to fit a small trim pot across C20, so that the peak current cut-off can be fine tuned. Whether this is worth it or not is debatable, though.

Jeremy

If my numbers were correct, adding a 1k resistor across C20 would double the cutout current to ~260A!! I wonder if your original idea of increasing R43 may be better? That way we can filter out the transient spikes, while still keeping our fets from turning to plasma if the motor phases short?

-Matt

[Jeremy Harris]

It really depends on the duration of these high current spikes. Something like an IRFB4110 can safely take 670 amps for short pulses, limited by junction temperature. Provided that these over-current events are relatively short duration, then big FETs might be able to take it OK.

Jeremy

[Thud]

Let me buy you guys an inernet Beer!

[gwhy!]

Let me buy you guys an inernet Beer!

Internet beer is always good ,

The problem also goes away if you re-program a 206 with the 212 settings so Im guessing it will also be ok with 218 settings, More testing to do . Looks like matts theory about the polling was correct and the firmware indeed plays a part. Now we are cooking....

[matt_in_mtl]

Awesome job guys! I for one am very excited, I think we may have just cracked this egg!! Now who was that testing the mini-monster 6-fet who was having high power cutting out problems, they may be interested in hearing about this?

-Matt

[ZapPat]

The problem also goes away if you re-program a 206 with the 212 settings so Im guessing it will also be ok with 218 settings, More testing to do . Looks like matts theory about the polling was correct and the firmware indeed plays a part. Now we are cooking....

It may just be that by selecting a 212 instead of a 206 you are dividing by half your real soft-current limits (both battery and phase), so it may be this that makes it looks "fixed"? To really be sure, you could select the 212 as controller type but double both current limits as when using the 206 setting - both results should be equivalent.

The transistor-based current overload circuit signal is independant of firmware, so programmed values should not directly affect it... unless using the 212 and 218 settings makes the firmware ignore the overload signal possibly? But then, why would they populate the overload signal parts on their 212's?

[matt_in_mtl]

OK,
time for some more test results (this time with pictures )
I pulled out the usb scope and soldered some test leads to the positive sides of the shunt (after removing my previous shunt mod), C20, and C19. The following were done with my previous settings of 30A.

Here is the first test, Channel A (red) is the voltage across C20 at a current spike. Channel B (blue) is the voltage across the shunt, the green line is at 0.6V
you can see how the voltage across C20 closely follows the shunt voltage with just a little filtering. Also the C20 peak follows the shunt peak by less than 0.5uS, so any spike on the shunt will cause QF1 to switch.

the output of QF1 is further filtered by R42 and C19. Next I measured the voltage across C19 (channel B/blue) and the voltage across C20 (channel A/red).
you can see how right after C20 goes higher than 0.6V (the green line again), the voltage at C19 goes low (causing the controller to quit).

After these tests I replaced R43 with a 10k resistor (1k original). Here is the result (same setup as first test but with much longer time-base).

Note that I had to mess with the timing a little to get the controller to fault. You can see much longer/higher the shunt voltage reached before QF1 switched causing the controller to fault.

I just tested the mod by increasing the phase current to 45A. Result is that it is much better than before, however there is still a remaining problem. I can crack the throttle, and the motor will rev cleanly, but when it reaches about 1/2 full speed, it cuts. I think that even with 10k the change may not be drastic enough. Unfortunately doing this testing I think I just blew my LM317 Oh well, live and learn.

Tomorrow I will grab a replacement and then maybe I will be able to test the setup on the bike.

-Matt

[Jeremy Harris]

Nice work, Matt. It's always nice to see a theory turn into fact.

So, it looks like there is indeed a very big voltage spike on the shunt (which may or may not be wholly representative of the current flowing through it, due to inductive effects) and that this spike triggers the QF1 emergency over-current shutdown circuit. Thinking about it overnight, I suspect that this emergency cut-off circuit is only there to protect the FETs in the event of a motor short circuit. It most probably wasn't intended by the controller designers to be a peak current limiter at all.

It looks like we can either completely disable this function (and live without short circuit protection) by removing QF1 completely, or we can try and find a compromise set of component values that will still protect a set of big FETs in an emergency.

The key thing here is to try and get a set of component values that will keep the high current pulse within the time/current limit for the particular FET being used. This may be challenging, because maximum phase current is almost wholly dependent on the total circuit resistance and the supply voltage, meaning that we might need different values for different motors, wiring configurations, battery types and voltages.

Also, we don't know whether or not the controller firmware allows pin 27 to go low for short periods without triggering a shutdown. My guess is that it doesn't and that pin 27 is effectively an interrupt, but it might be nice to check.

We can work out the peak phase current easily enough, provided we know the motor Rm, the phase wire resistance, the FET Rdson, the shunt resistance, the supply wire resistance and the battery pack internal resistance. If the total circuit resistance was around 200mohms (a complete WAG, but probably not too far off the mark for an average set up using a big HXT and 8 or 10g wires with some lipo) and the supply voltage was 100V, then the peak phase current would be around 500 amps, but the rate of rise of current WRT time would probably be quite modest, due to the total circuit inductance.

It may be that we can set the trigger threshold for QF1 to around 600 amps or so fairly safely, knowing that this peak current might only be achieved with a motor short.

I think my preference would be to try and keep the time constants near enough as they are, but increase the trigger threshold current. This seems logical to me, as we've uprated the FET current capability, but wish to retain the fast emergency cut off for a motor or wiring fault. The simplest way to do this seems to be to add a resistor across C20, to divide the voltage that QF1 sees.

Jeremy

[gwhy!]

It may just be that by selecting a 212 instead of a 206 you are dividing by half your real soft-current limits (both battery and phase), so it may be this that makes it looks "fixed"? To really be sure, you could select the 212 as controller type but double both current limits as when using the 206 setting - both results should be equivalent.

The transistor-based current overload circuit signal is independant of firmware, so programmed values should not directly affect it... unless using the 212 and 218 settings makes the firmware ignore the overload signal possibly? But then, why would they populate the overload signal parts on their 212's?

I quickly re-programmed my modded 6fet (with soldered shunt) to try this last night ( so I havent done any real testing with it ) programmed current was set to 20A and phase current to 60A , Before re-programming the controller would cutout and the watt meter peaked at 10A (as virgin 6fet ) after re-programming watt meter peaked at 12A and no sign of cutting out ( this was with no load on the motor ). My thinking is maybe the board selection in the software is that one of its functions is to delay the hard cutoff for the type of controller selected... If all the controllers have all the same values of components then the only thing that can change it is the sofware because my 12fet works ok .

[gwhy!]

Just connected up the re-programmed controller to a load and with the software current set to 20\50 the controller works 100%. The shunt is about half the original value this gives me a real current limit of 20A ( and the controller shuts down if the motor is stalled for about 1-2 secs - this is about the same as my older 6fet )so it looks like the only real difference between the 6fet and the 12fet is the shunt ( stock shunt on 6fet is about double the value of the stock 12fet ) so it all works out. I need to fit a lot smaller shunt to really give it a full 80 to 90 A load test.

Edit: I just re-programmed my controller to 40/100 but this, every now and again do cut out from a dead stop but nowhere near as bad as original ( and is just about usable ) , it current limits at 38A. I was hoping that if the shunt was halved then the 212 software settings will make the 6fet as the stock 12fet ( it was worth a shot and do improve things but dont fix them )

Edit2: Thinking about it my 12fet is running stock fets so it will not be programmed for so high a current ( from memory its only set for 25/62) and the shunt is halved. I wonder if the 12fet will also cut out if I upped it to 40/100 ( I cant try it cos it will blow up ) .... Burtie had a 12fet fitted to a bike he was having similar problems with cutting out same as the 6fets.. I wonder what current settings he had programmed into the controller.

[12p3phPMDC]

You guys rock!!

Excellent work.

[gwhy!]

And so it continues ,
I just re-programmed my modded 6fet using the '218' settings to 20/50 and adjusted the shunt for about 60a limit , The controller works sweet as a nut. I conclude from this that the software indeed to play apart in the hard cutoff ( be nice if someone else can confirm the results as Im getting ). What I think would be nice though is to have some sort of switch for 'on the fly' current limiting settings, and from what matt have shown and with Jeremy's suggestions this very much looks doable, but would it be usefull thats another story.

[Jeremy Harris]

Nice work.

What I think might be happening is that changing the controller type changes the time delay before pin 27 going low has any effect. This makes sense, as the controllers with the larger number of FETs will have more junctions to heat up from the over-current pulses, so will be more tolerant of a longer period at over-load current.

So, it looks like the overload circuit tells the controller that the peak current is too high, and the time between seeing this overload signal and shutting the controller down depends on the number of FETs that the controller thinks it's driving. If the overload signal goes away before the set time, then the controller just ignores it and carries on running.

The other function of the controller type setting is the one we already know; it sets the steady state current limit based on the shunt resistance, which changes with the size of the controller.

Jeremy

[gwhy!]

Nice work.

What I think might be happening is that changing the controller type changes the time delay before pin 27 going low has any effect. This makes sense, as the controllers with the larger number of FETs will have more junctions to heat up from the over-current pulses, so will be more tolerant of a longer period at over-load current.

So, it looks like the overload circuit tells the controller that the peak current is too high, and the time between seeing this overload signal and shutting the controller down depends on the number of FETs that the controller thinks it's driving. If the overload signal goes away before the set time, then the controller just ignores it and carries on running.

The other function of the controller type setting is the one we already know; it sets the steady state current limit based on the shunt resistance, which changes with the size of the controller.

Jeremy

Yes, I think this sums it up nicely . The real acid test will be when I put it on the bike ( hopefully soon ) .

[nieles]

i also added hall sensors to a 6374 outrunner. and it is having the same symptoms you guys discribe.

so what is the easiest thing to try? i think the reprogramming with 212 settings?
also if i reprogram, what should be my settings? i have not modded anything yet to the controller.

Niels

[Burtie]

I tried the 1k resistor mod across C20 and it fixed my cutting out problems

Thanks for all your hard work guys!

Burtie.

[gwhy!]

i also added hall sensors to a 6374 outrunner. and it is having the same symptoms you guys discribe.

so what is the easiest thing to try? i think the reprogramming with 212 settings?
also if i reprogram, what should be my settings? i have not modded anything yet to the controller.

Niels

I think re-programming with 212 setting will be the easiest, If you are running stock fets then program to 20/50 and halve the shunt value this should give you around 20A limit. Once you have confirmed that this works just reprogram with higher values ( with the 212 settings and halve shunt value it started cutting out again at 40a ) this will depend on motor. I think 30/75 with halve shunt value will be what you should be aiming for ( with stock fets ). Hope this helps.

[Thud]

Good work & thanks to all the electronic's minded folks who have looked at this issue.
Jerremy H, Gwhy, Matt,Burtie....you guys all rock!

Now I would like to steer this into the NOOB teritory & ask for a couple points of clarification.
Here is where I am at:

Gwhy- you mention 1/2ing the values...(I am interpreting this as the "programed value" in parameter designer)
this is because of the variance of shunt value from the b206 & b212 (?) essentialy inputing 20 for an expected 40?

I like the idea of adding a resistor accross C20 because, (I assume) the programed values (in the Current/voltage design box of PD) acuratly depict the expectation?

My goal is to write a SOP for the 1 or 2 controllers I intend to use with a couple levels of modifications.
(al'a Will newton)
A step by step guide for those who have a real hard time visualizing these things & don't comprehend the variables.

I have talked to a few mechanicly inclined guys with my level of ellectronics knowlege who read the infineon threads & quit....too much info & theroy.

Ive also laughed at the few questions I got pm'd on controllers, As I am an ellectronics dufass. I can solder like the wind But to change values on components & start modding shunts (like methods & a few others here) to affect re-gen voltages is WAY outside of my comfort zone. At best I am a gifted mimic.
Thanks again guy's.