Steve has a good handle on the factors involved here.
His RC snubber definitely addresses the root cause of the problem
and works. Note that this requires removing or
disabling the diode clamps.
My job now is to come up with a fix that accomplishes the same thing
and is easy as possible to implement AND to verify that there are no other
'gotchas'. The first part was relatively easy. It involves
cutting one trace on the circuit board and adding a single component
(resistor [or a MOV as Marv suggested]) in a position that is easy to
retrofit. Completed the bench and flight testing yesterday
with excellent results.
The second part (checking for 'gotchas') was the hard part.
A detailed report would take more time than I've got but here are the
basics.
Compensating for the shorter (actual) injector open times requires
that the EC2 increase the injector pulse widths. The means to do this are
already there in Mode 3 (injector flow rate adjustment). It does require a
big change (about 53% in my case with the Renesis saturated injectors) in the
Mode 3 constant. It was at 150 and had to be increased to 230.
This alone was enough to get the engine mixture back in range and safe to
fly. I spent about 3 minutes on the ground coarse tuning
the MAP table, then took off and let the EM2 autotune do the fine
tuning. All went well until I got to the full throttle / high speed
part of the test. When the engine reached about 6000 rpm the engine went
into rev-limit (ignition & fuel cut until revs dropped below limit then
switches back on until rev limit reached again). Very disconcerting when
in-flight. Engine ran fine when throttled back. This was NOT due to
the normal rev limiter function which is programmed at 7500 rpm.
Much geeky software debugging later I found that it was caused as a side
effect of altering the IPW (injector pulse width) clock beyond it's normal
range. This clock is also used in another function which didn't like
this. This particular bug only affects the Renesis and other single sensor
engines. Very easy software fix once it was found.
2nd & 3rd gen 13Bs and 20Bs use 2 sensors so are not affected by
this. Still checking the dozen or so other engine types that the EC2
is made for.
The final result on my engine (Renesis) did not show any dramatic
change in engine performance or difference in idle smoothness. As Ed
noted, the small primary injectors (290 cc) make it more tolerant of the pulse
stretching effect of the diode clamps. I also got reasonably good idle on
my 2nd gen 13B using 440 cc injectors. The bigger they are, the more
difficult they will be to tune and the more susceptible they will be to idle
problems at low speed.
On the plus side, there were fewer and smaller changes (from nominal)
needed in the map table. This made tuning easier and faster. I
will probably be able to idle slower but I will have to get under the cowl and
adjust the stop on throttle body to try it. It idles at 1600 rpm very
smoothly (before & after the change).
Sudden power chops to idle when entering the pattern go
noticeably smoother now. I was used to the occasional burble & pop
similar to what you've probably heard from P-51s and other warbirds in the
pattern at fly-ins.
More details to follow.
Tracy
----- Original Message -----
Sent: Sunday, March 11, 2007 2:21
PM
Subject: [FlyRotary] Re: A solution? was
: The truth??? / Injector flow rate mystery solved
Ed and
Joe,
The diode in the EC2
allows the current from the collapse of the injector magnetic field to flow to
the positive supply rail (~14V); it doesn’t oppose this. A resistor allowing this current to
flow would also result in a close delay since the current flowing is what
maintains the magnetic field during this delay. What is needed is a way to
decrease the rate of voltage rise just after the EC2 pulse ends so arcing in
the A/B selection relay is suppressed.
After the relay contacts open enough that an arc is no longer possible
(which shouldn’t take long) an open circuit condition now would allow the
injector to close quickly. The
arcing may or may not be a problem any given time the relay is opened since
the timing of the end of the EC2 pulse and the opening of the relay are
independent and arbitrary.
One possible solution is a RC snubber rather
than the resistor that Joe proposed. I think
Tracy is working on
checking this out. I have
installed this in my plane and it works.
Tracy, however can do a
much more thorough job of evaluating this and be sure the change is
reliable. Let’s give him a chance
to do this before we do something we wish we
hadn’t.
Steve
-----Original
Message-----
From: Rotary
motors in aircraft [mailto:flyrotary@lancaironline.net] On Behalf Of Ed Anderson
Sent: Sunday, March 11,
2007 9:16
AM
To: Rotary motors in aircraft
Subject: [FlyRotary] A solution? was :
The truth??? / Injector flow rate mystery solved
Sounds like a
reasonable approach to me Joe. A pull-down resistor
would be relatively easy for me to install - I have the
resistor pack required for the peak-hold type injectors. So I could
easily place four additional resistors in that
box.
If I understand you (please
correct me if I don't), the pull down resistor should go between the injector
and the EC2 sinking terminals. That way the current induced when the
intended pulse terminates and the magnetic field collapses will have a path to
ground rather than being opposed by the diode in the
Ec2.
The value of said resistor could
be around 100 ohms. Since the induced voltage could reach from 50 - 100+
volts an 100 ohm resistor could flow from
0.5 - 1 Amp (for a very short
duration). As far as affecting the 12 Volt signal it would only draw
12/100 = 0.12 amp or 120 milliamps. That would be pulled through the
injectors at all times. The injector resistance is probably (peak
and hold case) around 3 ohms. So the injector would draw 12/3 = 4 amps
(DC case - its undoubtedly less due to the A/C impedance of the coil).
It make take some experimenting - but 100 ohm looks like a good place to
start.
The wattage should probably be
around 5 - 10 watts just to be on the safe side.
So certainly looks like a
suggestion that would work, Joe.