DIYEFI.org Forum

I recently upgraded my MS2 to Jean's code that runs sequential injection. I noticed that I needed about 10% more req fuel to achieve similar AFRs at idle. At load, it was also *different* and I am trying to characterize that more effectively. Also, at work I was dealing with a fet driving a brushed DC motor. I relied on my old fet drive circuit and used an appropriate logic level fet VNP35N07 for the job. I was having significant trouble getting appropriate turn on and turn off at the PWM carrier frequency we were using resulting in smoke being released from the fets... Long story short, I was running into the gate charge and/or gate capacitance which had the effect of burning up FETs. The controller could not provide enough drive current to switch on the fet fast enough to keep up. If you go online and look up the vnd5n07 http://www.st.com/stonline/products/lit ... s/4335.pdf scroll to page 8 and look at the total time for switch on and off for a 1k gate resistor and a 10 ohm gate resistor. It is almost an order of magnitude in the time required. Also, the switch off is most affected. And these are both with a 10V gate (input) voltage.

If you go to this link http://www.irf.com/technical-info/appnotes/an-1084.pdf and scroll to page 10 where they talk about gate charge. It is also easy to calculate what resistor would be needed for a carrier freq. We are very low speed, but highly reliant on fast transitions. I know this is what I was fighting with my dc motor drive as I scoped it and was able to get correct function w/o heat when I took my gate resistor from 1k to 100 ohm. I am pretty sure this is what I am fighting in my new sequential install with the need for more fuel scheduling. It is also what we will be fighting when trying to turn on the injectors/ignition at the appropriate time in FreeEMS with merely a fet between the micro and the injector. (this is what I saw as I perused the relevant schematics of freeEMS) For me, I will be installing an ixdn404 http://ixdev.ixys.com/DataSheet/99046.pdf to drive these fets to get solid turn ons and offs to see if there is any difference.

Please lets discuss this as I feel it is relevant.

Gearhead

Interesting data. I hadn't paid much attention on the details of the data sheet and relied on the B&G design for the Microsquirt (which is obviously not a good idea...) when recommending values for the high Z injector drivers. The turn off time should be smaller with 5V but the turn on time will be higher. In any case, a 1K gate resistor doesn't look like a good value to use.

I'll have a look at other logic level FETs to see if the numbers look similar. In any case, a FET driver is probably a good idea but I'd be curious to see if a 5V buffer/driver would be sufficient. However, that might require a heftier 5V power supply.

Jean

jbelanger wrote:I'll have a look at other logic level FETs to see if the numbers look similar. In any case, a FET driver is probably a good idea but I'd be curious to see if a 5V buffer/driver would be sufficient. However, that might require a heftier 5V power supply.

The need for a heftier 5V is what drives me to a 12V fet driver, actually. We actually have a reasonable amount of power available at 12V...

Gearhead

Especially with two regulators... :-)

I agree with the buffering of output devices in general. If we didn't buffer it would be 1.6k not 1k anyway. At that point we will all agree that it's not really good enough.

Fred.

I do not think that burning of the FETs will be an issue for an ECU. Since the switching freq is very low, the FET will hardly heat up. I have an App where I am switching current through a heater, ~1A, at 96khz. The FET gets hot, maybe 80 degrees celcius, if I was switching in the 1khz range or lower I am sure the FET will only be slightly above room temperature.

I generally just put a 10 ohm resistor to the gate to minimize voltage reflections, but it depends on the length and width of the trace.

What is the advantage of running a higher resistance than what is required to minimize voltage reflections? Is it to protect the uC incase the FET gets nuked?

Regards,

Alan To

toalan wrote:I do not think that burning of the FETs will be an issue for an ECU. Since the switching freq is very low, the FET will hardly heat up. I have an App where I am switching current through a heater, ~1A, at 96khz. The FET gets hot, maybe 80 degrees celcius, if I was switching in the 1khz range or lower I am sure the FET will only be slightly above room temperature.

agreed.

I generally just put a 10 ohm resistor to the gate to minimize voltage reflections, but it depends on the length and width of the trace.

What is the advantage of running a higher resistance than what is required to minimize voltage reflections? Is it to protect the uC incase the FET gets nuked?

Regards,

Alan To

yes. to limit the current required of the micro is what I have always understood.

The uC can't drive more than 20mA so there has to be a resistance higher than 10 Ohms. The logical solution is to have some sort of driver.

Jean

The UC can't drive more than 20mA PER 8 PINS :-) yeah, really... not like ms1 which can do that per pin...

I've had mine lock up from excess current. It doesn't reset gracefully, it jams up hard with all pins in fixed state, perfect for frying coils and flooding engines. We need to be fairly cautious with pin current esp with regards CPU heat output if a lot of pins are used. Hence 1.6k is a good number.

Fred.

also note that the vnd5n07 curves and tables are mostly rated with a 10V input. A FET driver with 4A capability (overkill) which operates off a 12V rail will allow us to operate the fet in a region where the data exist for prediction of its operation. I was willing to try a buffer, but the more I think about it, the more I am convinced that this is the *proper way* when you are talking about 1) protecting the micro 2) accurate PW and 3) accurate timing of events. We can make it work, and fudge values to make it work, but I would advocate for precision if it is reasonably possible and it appears that it is, here.

If I understand that IR paper, then:
C= gate charge in Coulombs (usually nano-Coulombs - nC); V= gate drive voltage; I (current) can be reduced to (gate drive voltage)/(gate resistor).

That simplifies to C*V = time*I, so if we know the part we want to use has a gate charge of 18nC (VND5N07 *rated at 10V*) and we use 12V drive with a 100 Ohm gate resistor (.12A max), solving for time we get turn-on in 2.5usec. if we use a 10 ohm, we get .25usec

Gearhead

Interesting stuff. I guess either we need to make it that fast, or reduce the timer precision. I vote for make it fast personally as there is probably no real advantage to increasing the timer period.

Fred.