Matt wrote:
Thinking about this more.. If you used an integrator, you need to
set it up so that the integration window is only during the
port-open time.
Otherwise, you might as well use the thing I was talking about
(essentially a mechanical integrator).
Okay, so maybe you need a peak-detect-hold circuit.. The way it
would work is the output from the MAP sensor would be fed to the
input of the peak-detect-hold circuit, and then to the MAP input on
the EI computer.
An output from the EI computer would reset the peak-detect-hold
circuit - preferably when the port is closed on each cycle of
operation. This way, you get to see the full value of the
supercharging effect, and the timing of the MAP sample becomes
a moot issue.. The hold portion of the peak-detect-hold circuit
retains the sampled peak voltage until it gets reset. How's that?
Here's an IC that looks like it might be the right idea. I haven't
priced it..
http://www.amptek.com/ph300.html
Or maybe something like this, though you'd have to rig a switching
power supply to make it work..
http://www.chipcatalog.com/Analog/PKD01EP.htm
I suspect that we are making this problem harder than it really is.
If we make the assumption that the shape of the pressure graph is
linear between IO and IC (which it is in Paul's graph, and likely is
whenever the manifold is "on song" as Paul puts it), then simply take
one measurement at time (Tio + Tic)/2. That is the height of the area
under the curve between IO and IC, and the width of the area is Tic -
Tio. Easy. The less linear the graph shape, the more error this
calculation has, but I suspect that we really only care about the
full-throttle resonant operation anyway.
Russell Kent
We are concerned about the proper mixture over a wide range
of RPMs and that curve will rarely be so simple. Some times it
could even be U shaped.
Paul Lamar
--
The Rotary Engine NewsLetter. Powered by Linux.
ACRE NL web site. http://www.rotaryeng.net
Copyright 1998-2005 All world wide rights reserved.
Okay... How about make it really simple? Just take the samples and sum
them. Do we really care about doing the sum_of_samples/num_samples?
It's not like we are going to get ACTUAL pressure values. Just some
binary representation of a voltage that is proportional to the MAP
pressure. If we want to display it, we'll have to do a bunch of
massaging to make it cubic feet per second. The only thing to worry
about is whether the accumulator register is big enough to hold
num_bits_per_sample *
num_samples.
Using manifold pressure to extrapolate mass-airflow from a sensor that
is located so close to the port means that we are stuck with a bunch of
non-linearity, even if the signal is integrated over port-open time.
This is so because the intake port size isn't constant, even when we say
it's 'open.' It starts to open (low initial flow, regardless of
pressure), is completely open for a while, and then ramps closed
(reducing flow, regardless of pressure). Integrating the values from a
MAF would be better. The MAP would be more useful if the sensor could
be mounted in a plenum with a throttled inlet. It's probably going to
work, but still take a bit of tuning to get the mixture vs MAP vs RPM to
work out.
Regards,
Matt-
No doubt you are totally correct Matt. In fact the flow drops off
as the chamber pressure builds. That is why it is nice to get peak
pressure just before the port closes.
A feedback system based on EGT's so far has been resisted by Tracy for
two reasons.
One; slow response of EGT thermocouples. That can be mitigated
by going down on the size of the thermal couple. Small thermocouples
respond faster.
Two; can't tell what side of the
curve you are one. I suspect that can also be fixed by incremented the
mixture and see if the EGT temp goes up or down. If it goes up you are
on the lean side. If you lean it and it goes down you are also on the
lean side. If you make it leaner and it goes up you are on the rich
side. Simultaneously comparing that technique to an increase or decrease
in engine RPM should also help.
Of course a good wide band mixture monitor would work as well
and I think Tracy needs to try that. Currently the auto tune mode
uses ordinary O2 sensors.
Paul Lamar
Thanks Paul,
Actually, my math was a bit off.. I think you pointed out the correct
method earlier. The register size needs to be num_bits_per_sample +
num_samples - 1 or if num_bits_per_sample = 10 and num_samples is 7, the
register has to be 16bits wide to hold the sum. This is true whether you
div7 or not. Twiddle twiddle.. :)
I think the autotune will be fine with this to get the 95% solution, with
just a bit of remapping to get it 'perfect.' It's not FADEC, afterall,
but it will be better than a marvel schebler..
Matt-
I hope so. My C182 O-470 engine's CHT's and EGT's are all over the map :)
Paul Lamar
The Rotary Engine NewsLetter. Powered by Linux.
ACRE NL web site.
http://www.rotaryeng.net
Copyright 1998-2005 All world wide rights reserved.