We have a reasonably older version truck (2004 International 9400i with cummins isx500i) for a prototype hybrid drive where we had to use the PV380 plus the PDM as the controller via the analog foot pedal. ( we had tried the PV450 and that could not deal with the existing pedal as well as ECU coding issues, so this was the only method we could get to work without re-coding the ECU)

We use the foot pedal as a measure of "demand" and then control both the cummins throttle and a novel hybrid drive.

The problem we have is that the signal is very "noisy" and i suspect it is electrical interference as the signal signal can jump by over 10 percent OR the A2D converter built into the 380.

I cannot find the spec sheet for the 380 to check, but another post indicates it is a 12 bit device. It is possible the existing vehicle pedal simply does not have enough resistance, but that is a guess by me.

It is difficult to put a oscilloscope on this line to see the actual wave form but the problem is we use this demand to very accurately jump between modes, so seeing the wave form does not solve much either I think(?).

we have tried the built in averaging and that is not really useful as the pedal then has far too much lag and the signal just jumps less, but still far far too much

without building a lowpass filter or doing a numeric filter, I cant see how to solve this?

to summarise, the issue is that we need a accurate measure of the pedal position that does not "jump" for no apparent reason

Any suggestions would be very welcome

Norman

I don't see another way other than doing a filter on the input readings. Like your comment you will have the noise spikes that need to be filtered out without causing the slow motion and delay in averaging the position. The color displays have the Data Smoothing variable but you have the same problem of choosing a value between 0 and 1, and they chose not to do smoothing on the monochrome displays.

One of the software engineers sent a comment on how to do a filter:

Current Pedal Position: this is the A2D reading that will fire the calculation event below on change
Filter Value: a value between 0 and 1 with 0 being no filter and 0.99 being heavy filtering
Pedal Position: has the calculation event and is the final filtered output
Calculation event:
(Pedal_Position * Filter_Value) + (Current_Pedal_Position * (1 - Filter_Value))

Thank you, will try

slight correction, the device is is purpose made unit, the 480 ( not 380) made specifically for the local supplier, but the issue is the same

This is to complete this post for anyone else who picks up the same kind of problem

1) we put an oscilloscope on the line and saw the spikes were quite random and up to 2 volts!!!

2) a simple small capacitor 47 microfarad across the line reduced the noise/error to less than 1% of the original signal

3) we did play with a few bigger and smaller sizes ( bigger caps resulted in obviously slower pedal response and smaller resulted in SLIGHTLY more noise. It was pure trial and error)

so, to summarise, we did NOT need a low or high pass electrical filter, NOR did we need to do any other maths smoothing nor running average etc!!

the solution is maybe trivial but it certainly was a big problem!

normanbutchgrant