Understanding the A/D Converter Function

[KidTurbo]

Been working on problem related to the GM ADC or A/D converter function used by GM ECM's.

Simply wanted to relocate the IAT sensor for my E60 [LLY Duramax] ECM from the combined MAF/IAT unit to the intake runner so it gives a true air temp reading after the turbo and charge cooler. What should be an easy mod has had me racking my brain because the ohms to temperature scale of the different IAT sensors don't match up.

EFIlive gives tuners the ability to modify all sensor scaling. However, GM uses a micro-controller with split "2" resolution scales for each sensor. So we have have a non-linear resistance scale sensor, that is converted to a voltage by a split resolution controller. One scale for hot, other for colder temps, that overlap and the ECM switches them on the fly.... Plus it uses the 10 bit sampling rate scale rather than voltage or resistance levels vs temp.



When I brought this up on the EFIlive forum looking for answers, without a single reply, I knew it was gonna take some research. To a layman, the electrical engineering design or workings of the A/D controller is quite confusing, but I'm slowly understanding it.

Attached is what my "stock" scaling looks like. I have the "ohms of resistance to temp" scale of both IAT sensors. Thanks to EagleMark's explanation on here, I've learned that the GM controller uses a dual 348/4000 ohm Bias resistor to split these scales. I verified this info by using a handy ADC Count Converter in the Megasquirt Manual ..

For the detailed definition on these micro-controllers, see this Berkeley University ADC Explanation PDF

So besides mapping both IAT ohm/voltage conversions manually for each 10 bit step, is there any tools or spreadsheets out there that could be used to convert C or F temp scales directly into that 0-1024 table that GM uses?? I've found the formulas, but am not that good in Excel to start on it from scratch.

[KidTurbo]

Attached is an example of two IAT sensors resistance scales from the GM LBZ engine. The IAT sensor I just installed is close to the the IAT2 listed there, but not exactly the same. However it's a good example of the difference in resolution that needs to be mapped and could also be a good unit for future mods.

I've went with the the ACR-12V-3 - GM#5744430 open sensor, and so far only have "3" points of reference on the ohms scale to build the tables.

40C - 1331.93~1472.13
80C - 362.23~400.35
130C - 102.31~113.07

[1project2many]

Hmm...

I posted a formula that will provide a reasonable approximation of the typical GM CTS curve. IIRC there was a link to a page that would allow one to create formulas based on expected values in the same thread. I would enter them manually however.

Maybe I'll have time to look at this tonight. Is there any overlap in the two calibration tables for the OE sensor?

[EagleMark]

I think this is the link to thread your looking for 1project2many... remember it started at EGR,CTS NS 1K lookup?

http://www.gearhead-efi.com/Fuel-Inj...TS-conversions

[KidTurbo]

Thanks guys. I believe that post is the one that brought me here

About 80% of the temp scales overlap [see EFI screenshots above]. I've mapped out the factory resistance GM used, in ohms to temp 304F to -40F using a 50k POT wired in. So now gotta convert this factory 10bit steps 0-1024 into ohms values, then work backwards with the new IAT sensor scale I found on the GM tech site. Would be nice to write a formula in Excel, then anyone could plug an play in the future. If the sensors used a linear scale would make it easier, but they don't..

Reason I chose the older tune port IAT sensor is the new style IAT GM uses is a 100k ohms scale, and my ECM is a 50k range..

[KidTurbo]

FYI: From what I have verified so far.

Our reference voltage is 5v of course.
ADC Count scale is 0-1024
Scaling1 "Hot" uses 348 Bias Ohms
Scaling2 "Cold" uses 4000 Bias Ohms

I'll post the two resistance to C temp maps I'm using in Excel shortly.. Currently guestimating the the OE Gm values to ohms to get a baseline.

[RobertISaar]

increments of 32 all the way up to 1024? so, a 33 entry table.

it is a lot of calculation, but you can get a pretty reasonable estimate of impedance by treating the sensor as part of a voltage divider. this seems to work well, just not for the 0.00 and 5.00 volt cells, since those involve no voltage and full voltage. you'll just have to enter the minimum and maximum sensor temps for those.
http://www.calculatoredge.com/electr...e%20divide.htm

using the 4K table for example: to bring the 512 A/D count value, you would need a sensor to read 2.5V ((5/1024)X512). with a known resistor A, input voltage and output voltage, resistor B can be solved. 4000 ohms.

32 counts = .15625 volts, is 129.03 ohms.

256 count value is 1.25V, comes out to 1,333.33 ohms.



takes a while, but i've done this once or twice before for situations when sticking in a known/existing sensor scaling table wasn't an option.

or am i looking at the wrong part of this thread?

[KidTurbo]

Yep that works..

I was using the ADC Count calculator form on page http://www.megamanual.com/v22manual/mwire.htm to come up with the counts. Entering the Bias value and then estimating the closest resistance based off the resistance/ temp chart I found on a GM tech site until it gave me the matching count.

While both ways will work, yours is more accurate because it give the exact voltage/resistance value for each count step [besides the ends]. Now I just have to use the new resistance/temp chart to find the matching values. If the sensors were of a linear scale, it would be cake. But I think I can graph this out pretty easy now.

Thanks

-K

[KidTurbo]

OK here it is, resistance map for temp sensor Scaling 1 & Scaling 2 tables. Thanks for the math lesson guys...


Not so hard once ya know the formulas. Now I just have to map the resistance to the temperature in C or F of any sensor. I'm not sure what all GM ECM's use this dual scaling structure, but I'm betting most of the late model units. This is the common structure used when edited with EFIlive, and probably HP tuners also.

Attached is a zipped Excel version for those who need it for editing reference.