S105 OEM ECU

[AndreyB]

SOATE S105 396200-3763010-00 is an OEM ECU running stm32 or Geehy apm32F407VGT7 I think it would be cool to reverse engineer it just to see if we can learn anything.

https://github.com/rusefi/S105/

[AndreyB]

2A low side output ready at https://github.com/rusefi/S105/tree/main/hardware

image.png (22.63 KiB) Viewed 2555 times

[MHTSOS]

This has analog current limit. Nice approach but needs several small parts.

Στάλθηκε από το CLT-L09 μου χρησιμοποιώντας Tapatalk

[kb1gtt]

0.3 ohm sense resistance @ 1 amp = 0.3 watts. Plus your normal switching loses from the transistor. I'd expect it needs good heat sinking, or good injector design. AKA injector would want to be large enough to keep the WOT duty low and lower ECU dissipated watts, but the injector would want to be small enough to idle. It would want to be a tight balancing act.

Of course you can play the odds that most cars don't hit the max ambient temperature and therefore it doesn't overheat and break.

[chaoschris]

@ kb1gtt : The resistors are in parallel, so the resistance is 0,33 Ohm...

[stwillocks]

sweet! knowledge comes from everywhere. I should visit junkyard..

[kb1gtt]

@ kb1gtt : The resistors are in parallel, so the resistance is 0,33 Ohm...

Oops. I just updated the math. Thanks for the clarification. It's still significant heat to deal with.

[mck1117]

what is the idea here? these are reverse engineered schematics as an academic pursuit, or these are intended to be built somewhere?

It's still significant heat to deal with.

0.3 watt is sort of no big deal - a Proteus is about a watt right now, and it barely gets above ambient.

[AndreyB]

what is the idea here? these are reverse engineered schematics as an academic pursuit, or these are intended to be built somewhere?

At the moment I am curious what can we learn from these. Those two ARM OEM ECUs seem to be build against a $50ish price point and overall simpler BOM.

[MHTSOS]

This kind of driver will also need the "fb" voltage to be monitored by the CPU in case the load is shorted. This circuit will limit the current to 2A but in case of a shorted load the mosfet will have to dissipate over 28 Watts because it will go into its linear region.

Στάλθηκε από το CLT-L09 μου χρησιμοποιώντας Tapatalk

[puff]

So, that's a tight balancing act not in terms of the right injector choice, but rather in the number of available pins on the mcu for FB, and its computational resources (vs using a standalone specialized driver with buil-in bells and whisltes)

[mck1117]

So, that's a tight balancing act not in terms of the right injector choice, but rather in the number of available pins on the mcu for FB, and its computational resources (vs using a standalone specialized driver with buil-in bells and whisltes)

This circuit also has no clamping mechanism whatsoever.

[mck1117]

go into its linear region

If there's any inductance in the load (which there is even from a shorted wire), it'll oscillate instead of ending up linear. But still, yes, the dissipation will be significant unless the CPU responds in short order.

[mck1117]

Not to mention that this would set off an EMI bomb if you hit the current limit.

[Dron_Gus]

This circuit also has no clamping mechanism whatsoever.

Only injector outputs have no clamping circuit. To speed up injector closing. MOSFET reversible breakdown is used to clamp high voltage spikes.
All other outputs have diodes to +12V.

[MHTSOS]

The breakdown voltage for this mosfet is 100V minimum. The datasheet doesn't specify nominal and max voltage so this could might as well reach 200V under some operating conditions. I wouldn't want voltage spikes of this magnitude coming into my ECU. Also I don't think it's wise to use a mosfet in a way that on every pulse its body diode goes into avalanche mode. Not every mosfet is built to withstand this kind of abuse. Much safer to clamp the voltage to a safe limit. How fast the injector closes is not that big of a deal since you account for this when trimming dead times.
Also the closing dead times will rely on the breakdown voltage of the mosfet and on each part it would be different and also change with temperature.

Στάλθηκε από το CLT-L09 μου χρησιμοποιώντας Tapatalk

[Dron_Gus]

The breakdown voltage for this mosfet is 100V minimum. The datasheet doesn't specify nominal and max voltage so this could might as well reach 200V under some operating conditions. I wouldn't want voltage spikes of this magnitude coming into my ECU. Also I don't think it's wise to use a mosfet in a way that on every pulse its body diode goes into avalanche mode. Not every mosfet is built to withstand this kind of abuse. Much safer to clamp the voltage to a safe limit. How fast the injector closes is not that big of a deal since you account for this when trimming dead times.
Also the closing dead times will rely on the breakdown voltage of the mosfet and on each part it would be different and also change with temperature.

Yes. You can correct this with dead time calibrations. But I'm not sure about light loads/short injection.

I'm not expert in analog schematic. This is what I was told by one schematic/PCB developer with many-many years of experience.
1. this is allowed mode if you are not going out of working area of MOSFET. Looks like this (IRLR3110) feels good in this conditions.
2. the higher clamping voltage - the faster you discharge all energy from coil - faster it will close. So 100V is ok.

BTW, this ECU has integrated transistors to drive "dump" coils. So it get much higher-voltage pulses on coil outputs when ignition strikes.

[MHTSOS]

I'm not arguing that this design in this application works. It apparently does since its used by OEMs in actual cars.
It is designed for a specific application and with the parts selected it will work fine, driving the engine it was designed to drive.
But, if someone copies this design to use it in another ECU, he might end up with a big surprise.
If you can't find those specific mosfets, because of parts shortages, that are proven to work under these conditions you can't just substitute them with something that has similar ratings.
You won't know if it will work the same way if you use any random injector besides the ones it was designed to work with.
You also have to take under consideration the layout around the mosfets and the clearances of the tracks because of the high voltage, something you don't have to when you clamp the voltage.
The design works but I think that the drawbacks are more that the advantages with the main one being cost savings.
It seems far better to use a "smart" low side driver and have peace of mind.

Στάλθηκε από το CLT-L09 μου χρησιμοποιώντας Tapatalk

[mck1117]

The design works but I think that the drawbacks are more that the advantages with the main one being cost savings.
It seems far better to use a "smart" low side driver and have peace of mind.

Yeah, I think something like this is a pretty terrible idea for any standalone ECU.

[AndreyB]

image.png (236.93 KiB) Viewed 1695 times

[AndreyB]

A bit of lol

Code: Select all

# WARNING: those are USB pins on discovery or rusEFI brain board so make sure to power those NOT via USB
DDEFS += -DEFI_CONSOLE_TX_BRAIN_PIN=Gpio::A9 -DEFI_CONSOLE_RX_BRAIN_PIN=Gpio::A10

OEM boards come not with fancy LIN transceiver but discrete transceiver which seem to just accept UART somehow not producing UART echo.