SD vs. GM MAF. Which would be better?

[delta448]

does the sensor actually have a resolution or could it be the resolution of the input on the ecu?

It's mainly a scaling issue in the ECU tables. Every type of ECU has finite table sizes, and a finite interpretation scale for any signal. MAP sensor voltage is scaled on a range the ECU can understand and use in it's calculations. There may or may not be more chance of the sensor's voltage signal being non-linear, but each brand of sensor is different in that respect.

For example, say an 8 bit ADC ECU scales the MAP voltage on a range of 0-255, if you change from a 3 bar to a 5 bar sensor there's a larger real-world psi difference for each step from 0-255. Granted, 255 different values is plenty of resolution for most, for example say your MAP sensor reads from roughly 25inHg to 29psi, so that's just a little less than 1/5 psi per ECU count.

 

[ramsack]

damn, someone ought to make an algorithm that you just need to supply a certain amount of vacuum, then just atmospheric pressure, and then a certain amount of pressure, and the ecu could make a formula based off that and figure out the amount without needing a whole table for every value. it's pretty sad with all the technology we have that an ecu can't estimate a value if it is halfway between 2 knowns, to make a smoother curve. do standalones like aem have more values to set or do they work off more of an algorithm like i was talking about?

 

[Ceddy]

It's mainly a scaling issue in the ECU tables. Every type of ECU has finite table sizes, and a finite interpretation scale for any signal. MAP sensor voltage is scaled on a range the ECU can understand and use in it's calculations. There may or may not be more chance of the sensor's voltage signal being non-linear, but each brand of sensor is different in that respect.

For example, an Evo 8 ECU scales the MAP voltage on a range of 0-255, if you change from a 3 bar to a 5 bar sensor there's a larger real-world psi difference for each step from 0-255. Granted, 255 different values is plenty of resolution for most, a 3 bar sensor reads from roughly 25inHg to 35psi, so that's just a little less than 1/4 psi per ECU count.

The Evo8 ecu scales the MAP to 0 - 1023.


Older Motorola cpu, ecus have a 8-bit ADC, 0 - 255.
(1990 - 1997 DSM ecus)

Newer Hitachi cpu, ecus have a 10-bit ADC, 0 - 1023.
(1998 - 1999 DSM ecus, Evo5 - 10 ecus)

A 10 bit ADC, has four times the resolution of a 8 bit ADC.

ADC = Analog to Digital Converter.


A 3 BAR sensor reads to about 29 psi.

 

[delta448]

Thanks for the clarifications Ceddy, you're much more versed on this topic than I. Post edited.

 

[ramsack]

that would be awesome if there was something with a 24 bit ADC. is this an adc along the lines of say an audio one would output pwm data as say 24bit 192khz? meaning do they have a sample rate?

 

[Ceddy]

that would be awesome if there was something with a 24 bit ADC. is this an adc along the lines of say an audio one would output pwm data as say 24bit 192khz? meaning do they have a sample rate?

The MAP ADC is called once per loop, about 60-100 times a second. Updating quicker wouldn't accomplish anything because the map look ups are done once per loop also.

So sample rate is 60-100 Hz. (About 2000 times slower then an audio ADC)

Really a 10 bit ADC is more then enough.

A 8 bit ADC paired with a 5 BAR MAP you start to get pretty course resolution (~1/3 psi per division), but people have been running that setup for years with no problems.

 

[ramsack]

it's amazing how little computing power is needed to run these cars. i wonder what kind of performance could be had if there were sensors (knock, wideband, egt, map, iat) for each cylinder run by something x64-based. it would be cool to be able to have a computer that would see in slow motion parameters change as the piston moves and variably control things in slow motion.