Log AEM UEGO in evoscan using analog output?
Its all in the formula. I recently just got back to tuning as weather is now permitting. I rebuilt PC in mean time as well. The point is i had to start from scratch in determining the correct formula, like you i got initial formula from forums. I could get lean side to match up but rich side was like .4 points off. the output of digital should be a value of 0-255. My initial formula was:
(18-8.5)/255*x-8.5
Like i said stioch and lean side were good but it was off on rich side
so to figure out WTF i pulled WB sensor and let read atmosphere and it only read leanest value of 248,
Then i dampened a rag with a little gas and richest value was 24. So this altered my initial formula quit a bit:
(18-8.5)/224*x+7.5
now leanest value reads 18 in ECU and richest reads 8.5, hope this helps.
EDIT*****
scratch the above my formula was still off in 14-16 AFR range. Further reading gave me new hope.
i read this post and followed it. my readings in evoscan match my gauge damn near perfectly.
https://www.evolutionm.net/forums/9933717-post119.html
(18-8.5)/255*x-8.5
Like i said stioch and lean side were good but it was off on rich side
so to figure out WTF i pulled WB sensor and let read atmosphere and it only read leanest value of 248,
Then i dampened a rag with a little gas and richest value was 24. So this altered my initial formula quit a bit:
(18-8.5)/224*x+7.5
now leanest value reads 18 in ECU and richest reads 8.5, hope this helps.
EDIT*****
scratch the above my formula was still off in 14-16 AFR range. Further reading gave me new hope.
i read this post and followed it. my readings in evoscan match my gauge damn near perfectly.
https://www.evolutionm.net/forums/9933717-post119.html
I'm seeing 10.x values for my AEM Wideband in EvoScan, this is the post that was mentioned above:
"hey itz,
what i did was measure the voltage at 2 or 3 points, (you don't have to measure voltage, only if you are thorough like me, the afr value on the gauge will suffice for accuracy)
so use 14.7 like discussed before, get the x value from evoscan, and the voltage from the white analog wire. then get the free air value by unscrewing the o2 sensor and reading x value in open air and the voltage. the further away the two points are the better. plus these two values are absolute.
so mine ended up being a parabolic curve, to check that - divide the volts/x value at each of the two measurement points, if they are not equal, then you do not have a linear relationship.
i.e. - mine were 2.422/495= 0.00492 at 14.7AFR and 3.795/770 = 0.00489 at 17AFR
so my line is not linear, the slopeof a line is y = mx +b where m is constant slope... to solve this the easy way...
wolframalpha dot com
enter the following in the box - parabola (x value, AFR),(x value, AFR)
i entered mine as ... parabola (495,14.7),(506,14.8),(770,17)
the solution will be a curve, it won't look right cuz its not zoomed in but just go to the equation box. and click on the equation that looks like this, but obviously yours will be slightly different.
y = -2.75482x10^-6 x^2+0.0118485 x+9.51
you'll get another screen that displays the curve in more detail
scroll down to the equation in alternate forms and click the lower left hand dog ear to get copyable text, choose the one that is similar to this form
x (0.0118485-2.75482x10^-6 x)+9.51
to convert #*10^-6, each -1 equals a decimal shift to the right
i.e
2.75482x10^-6 = 0.00000275482
then just adapt it for use in evo scan like so
x*(0.0118485-0.000002755*x)+9.51
and your AFR values will be exact throughout the entire curve.
***added info to answer any possible questions-
I forgot to mention a few things, first is that if you can find 3 points instead of just 2, the results of the curve will be more accurate, also it will prevent any possible chance of the vertex being placed between the two points, which would severely screw up your resulting equation.
also try to measure these values with the car running or at a voltage of near 14v in the system, like with a charger attached to the battery as the readings for x will vary since it is directly proportional to voltage seen from the analog output,
second is that if you end up having a linear equation, we will use the y=mx+b form, where m = deltay/deltax
so m would equal y2-y1/x2-x1 where y is AFR and x is x value from evoscan. the offset b will have to be found by inspection from logging in evoscan with the new formula. it should be around 10 ish though.
Ok, to answer why we measured voltage, it is more accurate, and if you choose to use that instead of AFR to solve for the equation, you will have to develop a relationship between the AFR value and the voltage value to offset the formula, so your + xxxx value on the end will be diffferent to get the proper afr value. This is up to you, but my values are pretty close and .1 difference from rounding up or down does not really bother me when logging.
note: a drawback of computational delay which is in milliseconds may occur depending on the processor power of the laptop you are using.
plugging the laptop into an inverter while logging to maintain full power usage is recommended to minimize any lag in the afr values vs rpm curve. especially during spool up when the values and load cells transition rapidly.
hope this helps
Randy"
what i did was measure the voltage at 2 or 3 points, (you don't have to measure voltage, only if you are thorough like me, the afr value on the gauge will suffice for accuracy)
so use 14.7 like discussed before, get the x value from evoscan, and the voltage from the white analog wire. then get the free air value by unscrewing the o2 sensor and reading x value in open air and the voltage. the further away the two points are the better. plus these two values are absolute.
so mine ended up being a parabolic curve, to check that - divide the volts/x value at each of the two measurement points, if they are not equal, then you do not have a linear relationship.
i.e. - mine were 2.422/495= 0.00492 at 14.7AFR and 3.795/770 = 0.00489 at 17AFR
so my line is not linear, the slopeof a line is y = mx +b where m is constant slope... to solve this the easy way...
wolframalpha dot com
enter the following in the box - parabola (x value, AFR),(x value, AFR)
i entered mine as ... parabola (495,14.7),(506,14.8),(770,17)
the solution will be a curve, it won't look right cuz its not zoomed in but just go to the equation box. and click on the equation that looks like this, but obviously yours will be slightly different.
y = -2.75482x10^-6 x^2+0.0118485 x+9.51
you'll get another screen that displays the curve in more detail
scroll down to the equation in alternate forms and click the lower left hand dog ear to get copyable text, choose the one that is similar to this form
x (0.0118485-2.75482x10^-6 x)+9.51
to convert #*10^-6, each -1 equals a decimal shift to the right
i.e
2.75482x10^-6 = 0.00000275482
then just adapt it for use in evo scan like so
x*(0.0118485-0.000002755*x)+9.51
and your AFR values will be exact throughout the entire curve.
***added info to answer any possible questions-
I forgot to mention a few things, first is that if you can find 3 points instead of just 2, the results of the curve will be more accurate, also it will prevent any possible chance of the vertex being placed between the two points, which would severely screw up your resulting equation.
also try to measure these values with the car running or at a voltage of near 14v in the system, like with a charger attached to the battery as the readings for x will vary since it is directly proportional to voltage seen from the analog output,
second is that if you end up having a linear equation, we will use the y=mx+b form, where m = deltay/deltax
so m would equal y2-y1/x2-x1 where y is AFR and x is x value from evoscan. the offset b will have to be found by inspection from logging in evoscan with the new formula. it should be around 10 ish though.
Ok, to answer why we measured voltage, it is more accurate, and if you choose to use that instead of AFR to solve for the equation, you will have to develop a relationship between the AFR value and the voltage value to offset the formula, so your + xxxx value on the end will be diffferent to get the proper afr value. This is up to you, but my values are pretty close and .1 difference from rounding up or down does not really bother me when logging.
note: a drawback of computational delay which is in milliseconds may occur depending on the processor power of the laptop you are using.
plugging the laptop into an inverter while logging to maintain full power usage is recommended to minimize any lag in the afr values vs rpm curve. especially during spool up when the values and load cells transition rapidly.
hope this helps
Randy"
BUMP- can someone PLEASE help me figure out this formula stuff before I drive my evo off a bridge....
I'm seeing 10.x values for my AEM Wideband in EvoScan, this is the post that was mentioned above:
I am willing to try all of this to get accurate results before giving up and switching to serial but the post isn't exactly 100% clear to me...
I'm seeing 10.x values for my AEM Wideband in EvoScan, this is the post that was mentioned above:
I am willing to try all of this to get accurate results before giving up and switching to serial but the post isn't exactly 100% clear to me...
The idea is to capture voltage data and correlate it to specific AFR readings found on the AEM gauge. You can input that data into a computational website such as wolframalpha to give you a function (whether it be linear, exponential, or parabolic) that you can plug into Evoscan and have super damn accurate results when you are logging.
Thread
Thread Starter
Forum
Replies
Last Post
