another new beta, new wideband stuff needs testing, a few other misc fixes..
another new beta, new wideband stuff needs testing, a few other misc fixes..
I have missed some great progress with the programm.
I will make some test with the new version and see how it behaves.
If you need free ad channel for true dual wideband you can use pin D31 and pin D27 they have 22.3kohm resistor and will have identical output.
PIN D27 is logged on tside at byte_103
PIN D31 is logged on tside at byte_11E
Also can you check if the actual voltage on the wideband output match the voltage on aldl log.
awesome, please test my patch to see if speed logging works, i've written all this stuff but i have no idea if it works yet!
kur4o can you please test and tell me how TCC lockup in mode4 works? the mode4 document we have seems unclear. i want to add a switch for it. i've already done gear selection.
is there anything else for automatic trans that would be useful?
i wish my bench testing setup worked right now so i could confirm in controlled conditions.Also can you check if the actual voltage on the wideband output match the voltage on aldl log.
im using x/51= to get the voltage, assuming it reads exactly 5v @ peak (linear 0-255=0.00-5.00v). i hope to check its accuracy properly later. i remember with an LM-1 i could never get the linear equation to work properly with the a/c pressure input, and was suspicious that it was actually reading 4.8v@255 or something.
as has been simulated and then tested real-world, the pulldown resistors in the A/D circuits can have a rather large effect on the linearity of a sensor read. either need more drive current(op-amp?) or more pulldown impedance, otherwise the linearity suffers and you need to build a table around known points.
you're right. i'll do that for sure, probably as optional so people have a choice..
do you know what a typical curve would look like, or any data on tests other people have done? are we talking a curve down to a 0.25v dip in the middle or something? i could live with a percent but two or three might be bad news. other ecms must have similar circuits..
also wondering if the circuit for the a/c pressure input is of the same configuration as the other 5v inputs
i'm very happy to increase input accuracy with lookup tables. i already did that for temperature inputs. table lookups against an 8 bit value are very cheap when you have ram to play with, as long as you generate the table as a 256 byte array, the 'table lookup' is just a single address.
once my bench testing setup is working again, i'll grab a potentiometer and some voltage, and build a table.
curve is going to be dependant on exact pulldown value and the sensor(or wideband controller or whatever is being read) itself.
typically, the a/c pressure sensor circuit is setup the same as a MAP sensor A/D channel, using a 51K pulldown. if you attach a 1K pot in place of the pressure sensor, any error is tiny, but if you were to use a 2K pot, you'll start seeing sag, a 5K is very noticeable and it just keeps getting worse. there seems to be good results when the pulldown is at least 40 times the impedance of the pot it is interacting with, or at least I was accepting that amount of error in my testing.
with a wideband controller making an analog voltage via either a digital-analog converter or just a PWM output run through a low-pass filter to approximate it(if there are other methods, I haven't seen them implemented in a wideband controller), hard to say what exactly any individual controller will produce without running it through some known values.
EDIT: a trick you may be able to pull off(not sure how easily) would be to model the differences that would occur with different pulldowns and maximum drive currents from the controller... I get the feeling that would be a bit complex though.
really, instead of a two-point calibration, i should have an optional three-point calibration. for example my wideband controller will output stoich for a particular fuel type when unplugged. that gives me a midpoint, and that should at least get it closer if i apply some curvature to it. i'm not great at math, but i'll fiddle around until i get it working.
edit: this will be next version stuff though, because holy crap i'm way ahead with features and way behind with testing.
Last edited by steveo; 12-13-2015 at 11:34 PM.
Bookmarks