Has anyone used the p port slide throttle
with Tracy
Cook's ECU?
After changing my engine to the P port
configuration
with the slide
throttle mounted close to the engine the ECU
does
not
work.
There is no
good pick up point for the manifold pressure
sensor
hose. The hose
connected between the engine and slide
throttle gets
the rapidly
fluctuating impulses and also fills with fuel
from the
injectors.
Are there any engines running with this ECU
and slide
throttle
set up ?
John Prendergast
Mark Supinski is trying as we speak. Any day
now. We
had it running
on the dyno
but I ran out of dyno money before we got it
tuned
properly. The
engine would start easily
and run but the mixture was all over the
map..pardon
the pun.
Since then Tracy has changed
the computer a lot. I suspect the problem
may have
been >>when<<<
Tracy was sampling the
intake manifold pressure. He only takes one
sample per
rev so it is
only a few
microseconds long. IMHO many samples should
be taken
and averaged.
The same thing
might be possible by hooking up the two pressure
taps,
one from
each
rotor,
to a vacuum can to smooth out the pulses. Slide
throttles don't
give
good
intake manifold pressure signals when they
are wide
open as
there is
little
pressure loss in the intake system. That is
why they
make more
power.
All of Tracy development has been done with
a plenum
on the end of
the tubes
sampling the pressure just after a single
throttle body.
If your vacuum lines are filling with fuel
something
is wrong
with where
they are mounted. We did not have that problem.
Here is a
picture of
our set up. Can you send some pictures of
yours?
Where
are your
injectors located?
You can also try the apex oil injector ports
in the
rotor housings
for a place
to sample the intake vacuum... if any :)
Paul Lamar ...No rotor no motor.
The injectors are located in a similar position as in
your
pictures
except the elbows go forward and then down to
clear the
cowling. The
vacuum pick ups are located between the slide
throttle and
engine.
I believe the compression wave generated is
causing the
problem. The
vacuum gage shows such a rapid large fluctuation at
the low
rpm range
that it would destroy the gage in short order and
stops
fluctuating
after a short period when the hose fills w/ fuel.
The ECU
T's off the
same hose which of course could not function
properly with
these
conditions.
The Slide throttle plate is so close to the
engine that I
believe the
pressure wave bounces off the almost closed plate
causing
the rapid
fluctuations and the pushing of fuel into the hose.
With the original configuration the throttle body
was
farther away
allowing a larger volume of air to modulate these
vacuum
and compression
wave pulses and the injectors were located at the
engine
not before the
vacuum pick up point as it is now.
John Prendergast
----------------------------------------------------------
I agree with you on the pressure waves.
Try smaller hoses feeding into a large can
located some
distance away.
What you are interested in is the average
pressure and
not the
instantaneous
pressure. I am amazed the gage is responding that
fast. Is
it silicone
fluid filled?
What size hoses are you using now?
We used soft silicone hose which is very flexible. A
restrictor in
the hose will
also work. Drill a small pin sized hole through a
piece of
aluminum
or plastic rod
that fits in the hose. You have got to get rid of
those pulses
as no telling what the pressure sensor is reading as
it is only
sampled for a few millionths of a second. Only
Tracy
knows when.
The first length of the hose could or should be
larger
so capillary action of the fuel will not occur. If
the hoses
are coming out of the top of the intake tubes the
fuel
should drain
back out no problem.
Does the hose come off the top of the tubes? We
did not
have trouble
with
fuel in the hoses as far as I could tell.
Mark Su, Please check and see if you are getting
fuel in
your hoses.
Perhaps we can talk Tracy into a mass airflow sensor
mounted on the
input to the plenum. Mass air flow sensors put out a
zero to
five volt signal just like the pressure sensors.
However
they have
a long time constant as they basically measure
temperature
of a hot
wire
and that does not respond to rapid changes in air
flow. It
automatically
averages out the air flow velocity.
BTW what does your mixture look up table look like?
Perhaps
you can
share yours with Mark Su to get him in the ball park
as your engine is running.
I have also found fuel in the manifold lines. Not a whole
lot, but
definitely present. Sounds like I'll have to mount a
can of
some sort
upstream of the pressure regulator to absorb the pulses.
John -- One thing I would check first is to confirm
exactly
what
fuel
pressure you are running to the injectors. I
discovered that I
appear
to be running 45psi instead of the desired 35psi - so this
could be
causing the flooding problems I am also experiencing.
Note that I have in fact had the engine running pported
with Tracy's
ECU. I got it going last summer. Very tough to get
it firing
the first
time. Once I got it to run the first time, I had a much
easier
time of
it subsequently. Makes me wonder if fuel in the manifold
lines got
'sucked out' after the first run & combined with clean
spark plugs
allowed easier starts after the first. I had quite rough
running below
2850 engine / 1000 prop. I am now further wondering if
some of that
roughness was due to the pressure pulses being
discussed here.
I'm stalled on further startup attempts until I get my
EC2 and
EM2 back
from Tracy - he's updating them to the latest rev so I
have
all the
nifty new EC2 diagnostic modes & the auto-programming
mode for
the EM2.
Mark (Rocky Mtn High)
The Rotary Engine NewsLetter. Powered by Linux.
ACRE NL web site.
http://www.rotaryeng.net
<http://www.rotaryeng.net>
<
http://www.rotaryeng.net <http://www.rotaryeng.net>>
Copyright 1998-2006 All world wide rights reserved.
Making the computer take multiple samples might be harder
than just
doing a little analog signal conditioning on the sensor data.
Multiple samples burns CPU time - maybe not critical,
admittedly.
If the pressure pulses feeding the sensor can be damped
such that
there's no resonant effects in the sensor system (to keep the
sensor
from ringing), doing a little signal conditioning should
be easy.
How are the sensors wired? Are they a variable resistance
like an
oil pressure? It might be as simple as adding an
appropriately
sized capacitor to make a low pass filter. Just need to
filter out
the frequency content that's near the operating RPM of the
engine.
Regards,
Matt-
I am replying to my reply.. Sorry.
Now that I think about it more, in order for digital sampling
systems to
function properly, they must work around Nyquist's
Theorem. That is,
the sampling rate must be 2x the highest frequency content of the
signal
being measured.
http://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem
What this means is that if Tracy's (or anybody else's) system is
sampling the MP (or any other signal) once per cycle, the
samples must
be synchronized with the intake events (probably a tall order given
intake dynamics timing changes with RPM, and as Paul said, intake
configuration). Or, the computer must sample and average often
enough
to guarantee that it sees all frequency/amplitude content of the
pulses. Or (the easiest option), the frequency content of the
analog
signal needs to be filtered such that it's significantly less
than the
sample rate..
It's not just a good idea, it's reality.
Matt-
I agree with this Matt but two samples are better than one.
We are really not interested in the shape of the pulse at this
time. We just want to know what the average manifold pressure
roughly is.
As the software is set up now it could be anything. From zero to 29
inches of
mercury depending on exactly where on the wave it was sampled.
--
Paul Lamar ...No rotor no motor.
Given that Tracy is an electronics guy, I'd be a little surprised if he
doesn't already have some analog signal conditioning built into the
board. If not, I see that as a fundamental design problem...
As far as adding more samples per engine cycle, that might help, but the
results will be unpredictable. Who says that the samples are actually
going to have an averaging effect? That's only going to be true for
some waveform shapes, and for some sample locations. What's the result
if you happen to sample at two points on the waveform when the pressure
happens to be lower than average? That's the problem that Nyquist's
theorem addresses. If the analog bandwidth of the signal coming from
the sensor is anywhere near half the sample rate, we are in essence
grossly under-sampling the signal which will lead to aliasing and other
probably undesirable side effects. Until enough samples are placed on
the signal to guarantee Nyquist's theorem is met, the data can be very
misleading - giving tuning problems.
To know how many samples are required, we'd have to use an o-scope to
look at the signal coming out of the sensor.
Mechanical signal conditioning may be quite a bit easier.. Just
guessing what will solve the problem doesn't sound like a good plan to
me though. I'd definitely want to use a scope to verify the results -
just to make sure that I wouldn't be chasing my tale trying to track
down a tuning problem.
On engines with air filters, I think the WOT vacuum signal is going to
be better, even on engines with slide throttles...
Regards,
Matt-
Well lets say we sample it every 10 degrees of e-shaft rotation.
At 7000 RPM or .0086 seconds. As I recall that is 86 milliseconds
or 860 micro seconds. No problem for the A to D and the cpu.
There are at least 10 instructions per micro second in most CPU's.
So that is about once every 8600 instruction. Surely this is possible.
As far as the pressure wave is concerned that is synced to the rotor
rotation
and that is only 1/3rd the e-shaft rev so we would be sampling the wave
108 times along its length from one end to the other at 7000 RPM
before it
came time to inject the fuel. The wave travels
at the speed of sound. At lower RPM's the strong wave effect peters
out and the
samples occur far more often in a rotor cycle. At 3500 RPM the
second harmonic
occurs. Keeping the sample period constant we would be sampling the
double
wave 216 times before it came time to inject the fuel.
This is one of only two options as I see it. Mechanically damping
the pulses
with a vac can is the other one. Tracy is not going to redesign the
mother board for
electronically damping the signal alone.
Paul Lamar
7000rpm = 116.66 revs/sec : 1rev/0.0086sec = 1rev/8.6ms = 1rev/8600us -
events are even slower than you said.
I agree that it's certainly possible to do all of that sampling. It's
also easy to burn up a lot of CPU instructions, and then start incurring
weird latency which can cause other odd behavior.
Tracy used to talk to us.. Has he given up on the group now?
Regards,
Matt-
I mentioned this to him the last time I spoke to him as I recall. He
did not sound enthusiastic about the idea.
Paul Lamar
The Rotary Engine NewsLetter. Powered by Linux.
ACRE NL web site.
http://www.rotaryeng.net
Copyright 1998-2006 All world wide rights reserved.