Subject: pport tuning peculiarity
From: Rotary Engine
Date: 3/19/2009, 10:53 PM
To: AAA Put this in the To box
CC: Tracy Crook <tcrook@rotaryaviation.com>


  Hello Paul-

  Sent this to Tracy earlier today; thought I would pass it along in case
  someone on the list has seen this before...

  Had some peculiar behaviors in that latest tuning attempts. Hopefully
  you'll have some suggestions.

  Background-

  - 13b, p-ported, with slide throttle design & "tuned" intakes
  - LS1 injectors
  - EC2 with latest p-port firmware
  - EM2
  - 6800 field altitude
  - 55F OAT
  - Known to be under propped (has a 72x72 160hp 2-bladed prop. Needs a
  72x109 230hp 3-bladed prop)

  Symptoms-

  If we run the engine with the manifold pressure lines DISCONNECTED
 from
  the EC2:

  - Engine starts very easily
  - The engine runs quite smoothly across a pretty wide rpm band after
  tweaking the manual mixture control.
  - Reasonably smooth running as low as 1500rpm engine (526rpm prop!)
  - Very smooth running 3000 - 4800 RPM.
  - Above 4800 rpm we seem to "run out of enrichen" on the manual mixture
  control & the engine stops.
  - MAP table view on the EM2 real-time shows the EC2 using map location
  59
  or 91 (MP constant matching atmospheric pressure) depending on high/low
  throttle (as expected)

  If we run the engine with the manifold pressure lines CONNECTED to the
  EC2:

  - Engine noticeably harder to start
  - Engine will not run smoothly at any rpm
  - Significant surging pattern -- rpm increases rapidly, followed by
  engine stumble to low idle rpm, followed by next surge.
  - Engine often stops during the stumble / hard to restart
  - MP pressure varies rapidly based on engine rpm
  - Significant manual mixture changes can help ease the surges somewhat,
  but never enough to get the engine to smooth out enough to let us see
  if MAP
  values could be tweaked.

  Looking at the MAP table with engine off, it seems like all map cells
 are
  at
  0 -- as expected. I did not check all 127 cells, only the top cells (say
  100 - 127).

  Mark (Rocky Mtn High)

  That is good evidence the 555 system will work fine :)

  We noticed this phenomena on the dyno test of your engine.

  I am not privy to Tracy source code but the last time I talked to him
  about it was a couple of years ago and he was taking only one sample
  of the intake manifold pressure per e-shaft revolution.

  Using a computer this takes only about 10 usec.
  http://en.wikipedia.org/wiki/Microsecond
  Light travels only one mile in FIVE microseconds.

  That is 10 millionths of a second. 100 usec is one msec or one thousands
  of a second. This is a very brief period of time so the manifold
 pressure
  is jumping around a lot in a p-port tuned manifold. It could be anything
  or even above atmospheric pressure at some times. That is what p-port
  tuned
  manifolds do. See the enclosed chart. A lot depends on WHEN you sample
  the pressure. It changes with RPM in unpredictable and mysterious ways.

  IMHO this is confusing the heck out of Tracy's computer. IMHO you need
  to feed the manifold pressure from both intake tubes past the slide
  throttle
  using very small hoses into a beer can and then tap out of the beer can
  with
  one normal size hose going to the pressure sensor in the computer.
  Hopefully
  this will damp the pulsations and average the intake manifold pressure.

  Al Gietzen tried this as I recall and it did not seem
  to change anything. I don't know what size hose Al used
  but I think very small hose is essential. On the order of .030
  inside diameter or smaller.

  Paul Lamar

  Mark,
  You don't need the beer can. If you use 1/8 od tubing, make a coil of
  tubing
  using at least 20 feet of tubing. 30 feet may be better. The length of
  tubing works to damp the intake signal to something that the computer
  makes
  sense with. My experience is the tubing can work better than a reservoir
  as
  it is using the friction of air molecules on the surface of the
 tubing as
  the damper, and will respond faster than a reservoir, but still be
 damped.
  That doesn't mean the maps are correct, as it sounds like the manifold
  pressure side of the map is way off.
  To trouble shoot being rich or lean, have a squirt bottle of fuel handy
  and
  give the runners a quick squirt as the rpms are declining (while
 surging).
  If this makes them instantly recover, the system is going lean. if it
  forces
  it to die, the system is going rich. that will tell you which way to
 go in
  the tuning. My guess is it is very lean with the sensor connected. The
  sensor disconnected simulates wide open throttle with no manifold vaccum
  with high fuel demand, so the ecu increases the pulse width to the
  injectors. When the sensor is connected, and the ecu sees manifold
 pressue
  decreased, it reduces the pulse width.
  make sure you are careful when performing this test...... and of course
  you
  have a fire bottle handy...
  hope this helps,
  Larry

  Hi Larry,

  20 feet of tubing is a fair size expense and bundle.
  My lathe tail stock chuck will not hold a tiny .030 dia. drill or I
 would
  make some orifice up and send them to Mark.
  Perhaps you could make up some orifice that will plug into
  red silicone vacuum plumbing tubing. OD of the orifice would be .125 to
  fit
  the tubing. 0.030, .025 and .020 ought to cover it.

  The rest of the world uses mass air flow sensors for a good reason :)

  Paul Lamar

  Paul,
  If you need a .03 orifice I would suggest going to local motorcycle
  aftermarket
  shop and checking their selection of carburetor pilot jets. If I recall
  correctly I'd check a Kehin size 20 or 25 pilot jet. That should be
 around
  .030 inches.
  Bill Jepson


  Good idea bill.
  Carb jets would be a gold mine of tiny orifices.


  Paul Lamar


 Paul,
 the small needle valves sold at hardware stores for use on small tubing
 would give you an adjustable orfice for developement work.
 Larry

 Another great idea. Are you listening Mark?

 Paul Lamar

 You might also consider using the (unused) oil injector holes - I can't
 remember what size they are.
 That area might provide a slightly less erratic MAP source.
 Plus you may be able to get a true VE-Volumetric Efficiency number.The
 measure value would need to reduced (based
 on calculations) to compensate for compression in the chamber.
 Cheers
 Cary

 Depends on when you do it. Since the computer does it in 10 usec
 it is only a point in time.

 Here is a question for everybody. Assuming you can only sample
 the intake manifold pressure once per revolution of the e-shaft where on
 this curve would you choose? I chose to do it 135 degrees after
 top dead center. The red X marks the spot. Of course the shape of the
 curve changes drastically with RPM because the intake is tuned for
 only a narrow range of RPMs.

 Paul Lamar


 Paul,

 Same place for me. I want to sample the MAP after the intake port opens
 but before the injectors spray fuel.
 I orginally wanted to sample at the oil injection hole like Cary
 mentioned but since my injectors are now directly in housing in that
 area, I run the risk of fuel entering the hoses.I have decided to
 sample upstream of the fuel injection but downstream of the slide
 throttle.

 As an aside, perhaps injecting so far out on the manifold is causing
 erratic readings for Mark's Map sensor.Here is what we know. It runs
 fine at idle and mid rpms without any computer control, besides the
 lean- richen control. This is very carb like and you have the
 identical experience of not being able to run the engine at low rpms
 until you switched to carbs.

 The injection of fuel changes the density and temp. of the air mass
 before it reaches the sensor at which time the computer tries to
 compensate by adding even more fuel?


 Sort of like injecting chlorine into the pool plumbing system before the
 pump sensor.


 Doug in Japan

 Your spot like my spot won't work. Our samples are meaningless. The
 curves moves away at all other RPMs. That specific curve is only good for
  one
 RPMand perhaps only one throttle opening.I get a feeling that people
 are notgetting this.

 Paul Lamar


  We are guessing.Sampling may or may not be meaningless.
  Without knowing more about Tracy's system, we have no idea how often it
  samples MAP and if/how it is averaged.
  The same goes for many other factors.
  Cheers
  Cary

  He told me he only samples once per rev. He can not get an average
  by sampling once per rev. It takes about 10 usec or less to read the
  pressure sensor.
  BTW We will never know more about his system as the source code, unlike
  the Megasquirt source code, is proprietary.On the other hand I would
  rather
  be beatenwith a 2 by 4 than learn "C" :) If I were doing a EFI computer I
  would
  do it in assembly language (which I already know) and publish the source
  code.

  http://www.rotaryeng.net/RT1.pdf
  http://www.rotaryeng.net/skid-pad-C&D.pdf

  ---------------------------------------------------

  I have started to enhance the Megasquirt software flow chart to understand
  how it handles manifold pressure measurements.
  Would you like to contribute Cary?

  Paul Lamar



  Paul,

  The time of the sampling may be arbitrary but the sample will only be as
  meaningless as the calibration applied toit:) Yes it is a moving
  target and it won't be entirely accurate any rpm and throttle opening
  but if the sensor is consistent then the error can be accounted for. It
  is a daunting task to tune with EFI as at each rpm and throttle opening
  the look up tables have to be tweaked.And then we fly highwith turbo
  boost into cooler air. Kudos to Dave Leonard who had the patience to
  keep tweaking the tables to get his engine functioning.Tuning with the
  wide band oxygen sensor does have it merits.

  Doug in Japan

 Hi Paul,

 Someone mentioned earlier that the MAP transducer has a response time of 1
 ms.I'm not sure that it's valid to equate the CPU time it takes to read
 the transducer (10 us) with the duration or point in the cycle to which the
 reading applies.Wouldn't the comparitive slowness of the transducer have
 the effect of averaging the variations in pressure that occur between
 readings?This would be similar to using a telephone (with its engineered
 limited frequency response of 300Hz - 3.4kHz) to record an orchestra.
 You're just not going to get the higher frequencies no matter how fast or
 frequently you sample the output from that telephone.

 Regards,

 Tony King
 Queensland

 You have a good point there.
 Lets look at the time it takes to make one revolution of the e-shaft.
 That is a good measure because one rotor goes through all four
 cycles in one rev. At 6000 RPM cruise that is 100 RPM per second or
 one rev in 10 ms.


 Paul Lamar

So at 6000 rpm (10ms per rev) and with a sensor response time of
1ms, it seems reasonable to believe that the reading from the MAP sensor
would represent an average of the pressure over the 36 degrees immediately
prior to the reading.  On this basis, at lower revs it would more
closely approximate an instantaneous reading e.g. at 3000 rpm it would be
the average of the last 18 degrees.  I would have thought it would
be desirable to eliminate this variability and sample a fixed percentage of
the intake cycle (of which I count 3 per rotor per rev - making 120 degrees
for a complete intake cycle) regardless of revs, although for aviation it's
probably not a big deal since it's essentially a constant revs application.

Tony King
Queensland


The intake cycle takes place over a surprising 270 degrees. Call that chamber A.
Another chamber is firing while chamber A is filling so chamber A will not fire
until 270 degrees after the intake port closes. If one takes top dead center
at the end of the compression stroke one would probably want to sample
the MAP near top dead center 22 degrees after ignition. The math CPU overhead
would be considerable if that was how it were done. It would be easier
just to sample it when the plug fires as that is a hardware interrupt
from the crank angle sensor. My guess is that is what Tracy does.

In the Megasquirt it appears to be done over and over again in the
background loop. It is an analog to digital (AD) process.

It would be interesting to check the the speed of the PIC17C752-33/L A to D
Tracy uses. It may take longer than 10 usec.

It is not a big deal for aviation at constant speed but people expect
the engine to run smoothly through out the rev range.

Paul Lamar
-- 
The Rotary Engine NewsLetter. Powered by Linux.
ACRE NL web site. http://www.rotaryeng.net
Youtube key word UTUBPLEASE
Copyright 1998-2008 All world wide rights reserved.