Hello guys,

This is an efficient current limiting
peak and hold injector driver
from simple components. You do not have to chase fancy
components that are hard to buy in small quantities.
Works for low or high impedance injectors.

Having 4 or 8 pieces of this on a PCB is preferred.

http://x-dsl.hu/genboard/injector/injdrive.pdf
(the geda .sch and .ps is also there, this 
helps if you're designing a PCB. Please
drop me a mail if so.) 

====== Circuit operation
The circuit works by masking the master injection signal 
(look at Qfeedback and U1)
when injector current exceeds a threshold
(measured on Rcurrent). It has a histeresis
comparator. The histeresis can be set with Rhisteresis.
(200k is the minimum, 470k seems better)
The frequency will not be very stable, but it's not a 
problem. We want specified current, not specified frequency
after all. 

The limiting current is basically wired in by Rcurrent (so
that voltage drop is ~0.6V), you can change
that to match your desired injector current.

Peak current is determined by R11, R12, C13 and Z14.
It works by charging C13 to 5V (or 7V depending on
S15 which is not a real switch but wired according to user taste)
before master pulse becomes active. Therefore a higher
injector current will be maintained for about a 
millisecond, before voltage of C13 decays (exponentially).
The Z14 zener makes a voltage offset, so that given 
voltage drop on R11's 'input' pin (appr. 5V) is relatively more significant.
Note that dropping injector current to about 70% in hold mode will
result in 49% power dissipation.

Q2, Q3 and Q4 is a simple 2 stage bipolar driving for the
power FET. Nothing fancy. C3 was put in after Alexei Pavlov
pointed to a motorola appnote. It helps when Q3 and Q4 bases
go down, but worsens when they go up. (so don't make its capacitance much higher)

Rlimitn and Rlimitp be small, like 47..100 Ohm.

That's all, simple, isn't it?
====== End of Circuit operation

Some concerns:
This circuit relieves the ECU from high freq switching 
of injector signals (eg. from software) for current-limiting 
which is nice if you drive many sequential
injectors and your particular uC does not have enough hardware PWM outputs.

The ECU can have diagnosis data about proper (or electrically failed)
injector operation by looking at one of the digital signals,
eg. the gate of Q1 or the masking input of U1 which is the output
of Qfeedback. If it's not flipping back and forth within a freq limits
after master injector pulse is applied (and longer than a minimum value)
than something is wrong (too high freq if injector is short-circuited.
No masking if injector is not connected at all. Other problem
if mask freq is way too low). No need to monitor it continuously,
just once in a while..

This driver relies on a Dschottky diode to dissipate the energy of
the injector at switchoff, so switchoff isn't particularly fast.
However if you want less flow, use a shorter master pulse. Simple.
The turnoff is deterministic.

Do you have comments on the circuit? 

thanx:
 Marcell
 cell@x-dsl.hu