|
Message
Hi Joe,
First, thanks for the offer. You've been
there, done that, so you know how significant accounting for any and all time
delays in an industrial process is.
I've also used simple curve fitting
algorithms (mainly like those that in Excel spreadsheet - of course, they don't
provide any recommendations) to get coefficients for some equations (not related
to EFI)..
However, one of the reasons I've taken the approach
I have on the fuel meter is to avoid the cost of an aircraft quality fuel flow
transducer - like Tracy said they run on the order of $250 (for one and you need
two if returning fuel to a tank).
The traducers I am somewhat familiar with also
require calibration to the installation as well (if you want the most accurate
results). While it would be nice to compare the "actual" flow rate to my
imputed (computed?) flow rate base on the injector timing - I am just not
inclined to spring $250 for one or $500 for two {:>).
Actually, my fuel flow readings are less than 0.1%
in error - far better than the flow transducer method - once I calibrated
it. However, I always had to add a "fudge" factor to my equation - really
did not like that at all. The problem was the injectors
appeared to flow more than my calculations - implying that if my calculations
(and equation) was correct then the injectors "must" flow more fuel than
stated.
BUT, now we know the reason - a diode in the ECU
apparently keeps the injector magnetic field coil energized for some time after
the pulse has terminated - apparently according to Tracy up to 5 msec under some
conditions - well, that really is keeping the injector open longer, therefore
the additional fuel flow over what I would have calculated based on the pulse
duration and the rated flow spec for the injector.
So very much appreciate you offer, Joe.
However, at this time, I think I will wait and see what Tracy comes up with for
the moment.
Ed
----- Original Message -----
Sent: Saturday, March 10, 2007 11:37
AM
Subject: [FlyRotary] Re: The truth??? /
Injector flow rate mystery solved
Ed,
It has been my experience, in industrial control,
environment that there are several sources of latency (delay time) on and off
delay time of the IO circuitry, clock speed of the IO controller, on and off
delay times for the controlled device, mechanical delays etc. When a
machine operator presses a control button, expecting a action to occur such as
starting a motor conveyor, a stacked latency of 5 or even 100 mS is not
perceivable. But trying to put a patterned glue line on a web
moving 5 to 10 feet per second, requires accounting for the IO delay time,
pneumatic solenoid delay, air line charging time, glue head mechanical delay,
glue reaction (like a dripping faucet, the glue will form a droplet before
becoming a stream.)
When I have been up against applications like the
latter, I have found a curve fitting program that has worked very
well. In the example cited, I have run a rudimentary control algorithm
and collected the resultant data. Dump the data into the curve fitting
program, and the program will crunch the data and make recommendations for the
closest formula as well as the required formula constants. This has
worked very well for some difficult applications.
IMO a great way to work your fuel flow situation might
be to install a flow sensor. Run your engine at various
RPMs recording EC2 indications and actual fuel flow sensor
readings. Dump the data into a curve fitting program and viola.
One of the best fitting programs I have found is called XLFit (VBA add
on to Excel). The program cost is $725, but it has a 30 day fully
functioning trial.
If you would like to try an approach such as this, I
could provide IO for capturing fuel flow sensor data, curve fitting program,
as well as coming up to give you a hand (since I am less than an hour's drive
from your location.)
Joe (If I stop rambling, I might actually be able
to turn on the battery masters and fire up the CPU this week
end.)
----- Original Message -----
Sent: Friday, March 09, 2007 8:30
PM
Subject: [FlyRotary] Re: The truth??? /
Injector flow rate mystery solved
Ah, the old induced current caused by the
collapsing magnetic field of the injector has no place to go (quickly at
least) due to the diode. That would indeed tend to prolong the closing
- hummm up to 5 msec - now the duration - that
is surprising.
That's at least part (if not all) of the reason
why I always had to add a "fudge factor" increase over the number I
calculated for the specified flow rate. The injectors were staying on
a bit longer and therefore flowing more fuel than the specified flow rate
would indicate. OK! so now I know why I had to have a fudge factor - I
feel better and will finally be able to sleep {:>)
Thanks for the additional information,
Tracy. It will keep the rumors to a minimum {:>)
Ed
----- Original Message -----
Sent: Friday, March 09, 2007 5:20
PM
Subject: [FlyRotary] Re: The truth???
/ Injector flow rate mystery solved
Still refining the exact numbers but here is the basic cause.
The diode clamp on the injector drivers (in EC2) serves 3
purposes. It limits the inductive flyback voltage to protect the
driver transistor. It recovers some of the energy used to open the
injector thus reducing the current consumption of the system. It
eliminates arcing at the A/B injector switching relay that can cause noise
problems with the processor chips.
Unknown to me until recently, it also significantly increases the
closing time of the injector after the driver pulse ends. I thought
the delay would be negligible. The actual delay depends on several
factors but may be as much as 5 ms (!) Ed will know what a big deal
this is. The factors that go into this and how much it is magnified
by the diode clamp is still under investigation.
The good news is that there is a very simple fix. (Note, Do NOT
go out and remove the diode clamps! That isn't the fix!
Tracy
----- Original Message -----
Sent: Friday, March 09, 2007 7:13
AM
Subject: [FlyRotary] Re: The
truth??? / Injector flow rate mystery solved
I certainly don't know, Rusty. Like I
said the only thing that Tracy and I could conclude was that Mazda was
under-rating their injectors - I wondered whether it might have
something to do with the way an engine is taxed in Japan.
In any case, Tracy is clearly ignoring our
pleas for more information - poor guy is undoubtedly working furiously
getting orders completed before Sun & Fun. But, he really
shouldn't tease like that {:>)
Yes, I understand what you mean about
measured flow rates - I suspect there is something subtle about this
matter. Perhaps the current profile used to open the injectors
plays some role in this as Tracy indicated he was working on a
Fix. Perhaps a two stage or progressive flow rate profile - opens
and provides normal flow for typical automobile usage but, when wide
open (like ours are at cruise) perhaps the flow rate is more. Who
knows - Tracy Crook that's who.
Ed
----- Original Message -----
Sent: Friday, March 09, 2007
12:19 AM
Subject: [FlyRotary] Re: The
truth??? / Injector flow rate mystery solved
Tracy and I have been discussing
this injector flow rate anomaly for years -
For some reason, I was certain that the Mazda
injectors flowed more than they were rated for, but after sifting
through some old logs, that's not true for all Mazda
injectors. I sent 4 slightly used 550's to
RC Engineering, and they tested them at 547, 549, 551, and
551. If that ain't 550, I don't know what is
:-)
I guess I'm baffled now. Does Mazda
operate them at 5 volts or something, rather than 12V? That
would slow down the response
time.
Hey the HKS is a good little engine
in my opinion - not as good as a light weight rotary (which we don't
have yet - Richard! get a move on {:>)), but it'll get you
flying again.
Thanks for the
comments. It can't hurt much to try it, because I've
already got people who want to buy it if it doesn't work out.
Still, it's as much trouble to install as the rotary, so it really
comes down to a weight issue, and perhaps whether I'll ever get my
redrive from
Autoflight...
Rusty
|