This is for experts only. Very hard to set the pwm parameters, very easy to kill the injectors, FETs, etc !

Some time back I was start find out how to drive Low-Z injectors directly from VEMS in peak-hold mode.

There was two reasons for it:

I found that Power Flyback board gives good injector opening but closing time is too long and AFR at low loads is hard to control even with small ~350cc Low-Z injectors. Hihgh voltage flyback gives good control over injector

Second: I had a car with clean wiring harness and I don't want cut it for installing inline resistors.

In fact flyback circuit is simple voltage limiter. It must cut every voltage spike at level safe to FETs which drive injector. Higher cutting voltage gives faster closing of injector.

Using PWM-ing there is two main problems: heat and noise.

Heat. I*U - Higher voltage gives more heat. But at higher voltage current falls too. My tests shows ~15W with 6x LowZ.

Noise is main factor why VEMS developers do not recommend use PWM-ing. It can cause board resets, false sensor readings (especially Rpm sensors), garbage on LCD and so on. Really hard problems!

If High-Z used, VEMS use voltage limiter too. It is 30V TVS diode. It can easily survive with this kind of injectors - current is low.

Solution was simple - external voltage limiter with heatsink. And 4 capacitors for noise reduction.

Here is schematic:


Transistor IRFZ46N or even IRFIZ48N (more robust) both cheap.

need mica insulator kit

Zener is up to 42V (because 18V protection diode together with this device gives 60V - FETs limit)

Capacitors - two is mounted inside ECU between flyback pad and power ground.

Other pair is on the flyback board.

Resistor - 1kOhm

Note: device is connected in same way as common 30V TVS Zener diode.



I try this board on 1600cc for 2L 16V engine. For first was 30V diode and resistors used, but i do not like idle quality. With PWM-ing and this flyback device car pass emissions test! (E85)

Also device survive in race environment where 5 injectors commonly run higher than 60% duty cycle.

Note. Use this on your own risk. device is not officially supported. If you find a way to improve it, please post your experience.

* I've built similar flyback with success. I've got slightly faster closing than 30V zener setup (my idle lambda changed from ~0.9 to ~0.94), both with resistors inline. After resistors removed and changed to pwm mode it was very sensitive to pwm duty %, in fact I couldn't get stable opening below 75% duty. That way low load lambda control results was way better but this high percentage is very dangerous to my 2ohm resistors. Any ideas on this? (MembersPage/GergelyLezsak)

* 75% did not mean 75% of maximal current. From here

Iinj * Rinj = VBatt * dcpercentage/100 - Uflyback * (1-dcpercentage/100)


dcpercentage= 100 * (Iinj * Rinj + Uflyback)/(VBatt+Vflyback)

You will find 75% is LOW setting. Desired Iinj is 1A usualy. Significant higher setting can damage injectors. But lower - your engine! So keep Iinj little more than 1A can be good idea! MembersPage/GintsK

I'm afraid that you simplify this too much, it took a long time for us to find the actual problem with the PWM solution and your powerflyback variant doesn't address the main problem at all. It's pretty much identical to the prototype I built a few years ago (also without knowing what caused the random problems we sometimes saw). As you are obviously willing to experiment and have access to a number of cars to test on you should do some modifications that will address the problem. You need to replace a few diodes in the ECU with faster units. As it is VERY possible that the diode modification can cause other problems we decided to use the safe and in many ways better resistor solution instead.

The current diodes we have are about as fast as a common diode will get, this "ultra fast" diode is however sluggish compared to a Schottky diode. The Schottky type diodes will not survive as much voltage as the standard diodes but they are a lot quicker and they will prevent the very high inrush current that is the cause of most of the noise and that kills the FET when you PWM them. You should start replacing these diodes on all boxes you plan to PWM: D7,D8,D10,D16,D17,D25,D28 and D29. For added safety you should leave the original diodes in the D79,D169,D259,D289,D299,D179 and D89 positions, that way the box will pretty much go back to normal if the schottky diode would fail. You will need to examine the Schottky diodes on the test cars periodically to see how well they survive the beating they get. The behavior of the car will not change if the Schottky fails but the FET's will start to take a severe beating again.

I must say that I'm suprised that you used a FET for this, even if the part you use is fairly rugged I find that most FET's are much more fragile then an IGBT or NPN.

Note that any tests done with the 1600cc Bosch injectors are not likely to test how rugged the box is after it's modified. The 1600cc injector hardly put any load on the electronics and it's always used without a resistor. - Jörgen Karlsson, Gothenburg,Sweden.

Thanks, Jörgen, for this comment. This information must be in wiki long time ago.

Some words for defence: It is not single spike what kills FET. needs spike AND time = energy. Current diodes protect INJFETS from this energy, although those are "slow". If you want I can make high resolution scope shots on injector output!

For statistics: I have two-three fried FETs on ~30 boards. All with clear cause. No single FET fried on 3 boards where described limiter are used. More detailed history you can read here:,189.0.html including why FET in device.

In short: using PWMing injectors have significant better linearity and predictability at idle and low loads than with resistors.

There is some unsolved problem with hardware anyway. Lot of users claims about fried FETs using simple High-Z.

What protects FETs from -20V spikes on gate? MembersPage/GintsK