[dcb]

I think obd is quite accurate enough for the majority of drivers, but it does have it's problems:

1. OBD isn't much of a standard, it is actually many standards with a convenient name. It takes a lot of circuitry (cost/bugs) to be able to reliably plug into an obd connector and expect to communicate. Even then when different folks made the computers for the different cars, they made different interpretations and assumptions, and you sometimes see bogus data on the port. And also it is limited to 1996 or newer.

2. All the mpg parameters have to be inferred and polled and be based on assumption, introducing room for error.

2.a. fuel used, this is based on maf (if available) or map * rpm. But it assumes that the engine is at stoich (14.7:1) so that if it is cold, or enriching, or lean burn, or diesel or ??? then the fuel used can be off .

2b. for distance, it has to periodically ask the car computer "how fast are we going", and remember when and what the value was the last time it asked, and guess at the distance traveled from there.

2c. polling is less accurate also as there is a window for peaks or valleys to get in between polls, the guino gets an interrupt when an injector opens and closes, and when the axle makes a vss tick. It doesn't have to rely on the computers vehicle specific interpretations of what the injectors and vss are doing, it goes straight to the source of the info.

There are some chips (i.e. elm and other projects) that help with obd connections, but they still take a lot of support circuitry to be able to connect to any obd car, far more than the guino has. But you do get "interpreted" access to pretty much all the sensors on the engine once you sort out the obd connection issues.

Having all those sensor readings handy is fun, but the key ones for accurate mpg are how long have the injectors been open and how many vss tics have happened, and those can get a bit muddled in an obd system.