Date: Thu, 9 May 2002 02:16:54 -0400
Reply-To: David Beierl <dbeierl@ATTGLOBAL.NET>
Sender: Vanagon Mailing List <vanagon@gerry.vanagon.com>
From: David Beierl <dbeierl@ATTGLOBAL.NET>
Subject: Re: A damn OS question again, yea!!! now i need some help,
it's my turn....
In-Reply-To: <5.1.0.14.2.20020508231357.009f9ab0@pop-server.austin.rr.co m>
Content-Type: text/plain; charset="us-ascii"; format=flowed
At 12:14 AM 5/9/2002, John Clavin wrote:
>At 02:12 PM 5/7/02 -0400, you wrote:
>>I have two problems with this; first, you really cannot tell this with a
>>regular meter that takes 2-3 readings a second -- the ECU changes the
>>mixture on its own schedule that does not correspond to the averaging time
>>of the meter. If you want to know what the actual values are you have to
>>use a scope. Second, the feedback loop between the O2 sensor and ECU is
>>very simple: if the ECU sees > 500 mV it leans the mixture; if it sees <
>>500 mV it richens the mixture. It does not attempt to hold exactly at the
>>500 mV mark but rather is constantly flipping back and forth past it,
>>therefore the mixture is constantly going from *very* slightly lean to
>>*very* slightly rich and back.
>
>David,
>
>I have been trying to puzzle out exactly what is happening in the ECU-O2
>sensor loop for some time now. Do you have any idea as to the frequency at
>which the ECU attempts corrections? Recognizing that it is an electronic
>device, it is inherently capable of updating the fuel injector pulse width
>faster than the engine can respond. So it seems to make sense that the ECU
>not change the fuel pulse width quicker than, say, once every N rpm, where
>B is possibly as low as 1 rpm. Or does it?
The Bosch FI book says might be 1-2 seconds/cycle at idle, several c/s at
cruise. If I can I'll stick the scope on tomorrow and see what the Digijet
does.
Book also says that the range of mixture variation by the ECU is about 0.1%
total. It gives the sensor output for a new sensor as about 850 mv for
lambda of 0.99, about 175 mv for lambda of 1.01. For an aged sensor it
shows about 750 and 125 respectively with a slightly different curve.
Lambda is defined as 1.00 for ratio of 14.7, so .99 would be 14.55 and 1.01
would be 14.85 respectively. Note that your best-power mix is around 12.6,
lambda 0.86; and best-economy mix around 15.4/1.05 -- both well outside the
useful measuring range of the sensor.
>The reason I ask, is that I see some inconsistent readings from my O2
>sensor which I think might be caused by the DVM sample rate and ECU update
>rate synchronizing or beating together for an interval. For example, at
>times I see something pretty close to the expected crossovers, something
>like H-H-L-H-L-L-H-L-H-H-L (where H>.5v and L<.5v). But other times, the
>readings seem to tend to the high side and run more like
>H-H-H-L-H-H-H-H-H-H-L-H-H-H-L. The latter suggests that the ECU really is
>updating to keep swinging back and forth over the crossover point, but that
>the DVM's sample rate is running at, say, half the ECU rate, so only
>catching the readings on the high side.
DVM is averaging readings over a certain time period -- typical meter takes
1-3 readings/second, and the measuring technique is to use the input to
charge a cap for a fixed time, then discharge it at a fixed rate and
accumulate counts until it crosses zero. The count becomes the
reading. They call this "dual-slope integration" and there are meter chips
that do pretty much the whole thing except the cap and the input
circuitry. Sensor is going to stay mostly at either its high- or low-point
voltage, with ramps in between of maybe 10% or less, and cycle rate
determined by rpm.
There's a fundamental law of sampling that says if you don't sample at >=
twice the frequency *present in the input* you will get aliasing
errors. Since this input approximates a square wave it has frequency
components much higher than the fundamental cycle rate, so even at idle
this will be affecting your readings. [Note: I'm not sure how the averaging
effect of the meter interacts with this -- it's really talking about point
samples that are in themselves accurate. If there's someone on the list
who understands Fourier transforms they could probably say what happens
exactly.]
>At first, I attributed this to the sampling rate/clock rate differences.
>But more recently, I have suspected that the latter readings are indicative
>of a problem. First, it seems that I see more readings biased to the high
>side, than the low. I almost never see L-L-L-H-L-L-L-L-L-L-H-L during
>normal cruise or idling. Second, I see the bias to the high side mostly
>when I have been cruising a while at 50+ mph. In some cases it seems to
>have been followed by a bout of hesitation/coughing behavior. I am not
>convinced as yet, however, that it is not just a coincidence or my
>imagination.
This could be Vanagon AFM-flutter syndrome. Try shutting the key off for a
couple seconds then back on while the hesitation is happening -- if it's
suddenly cured then you want to look in that direction.
>I would like to catch this on a scope as you suggest. Unfortunately, I can
>only do that in my garage, where not much exciting ever happens ECU or
>O2-wise.
Gee, if you've got a scope, then inverters are really cheap nowadays...you
don't need a high-frequency scope, but you should be using x10 probes to
keep from loading the signal down. What sort of scope is it (if it's one
of those big automotive ones you're probably not going to run it in the
van, I suppose)?
david
--
David Beierl - Providence, RI
http://pws.prserv.net/synergy/Vanagon/
'84 Westy "Dutiful Passage"
'85 GL "Poor Relation"
| 
