Noise filtering in Automotive environment

[spamiam]

I want to use a ZX in an automobile. It is actually a somewhat interesting project. I want to measure the motion of a part in a carburetor, so I can determine how to adjust the carburetor mixture. Under certain load/RPM combinations, I will need to adjust a different region of a "needle".

This car has solid copper ignition wires that transmit lots of noise. I may not be able to get resistor plugs for this car.... so...

I may need to measure either analog signal thru a Pot and I'd like to get readings every 1/4 second (4hz) or so.

In an alternate setup, I may need to read a digital signal that will vary from 0-120Hz.

Do you think that a simple RC filter with voltage limiting diodes (what are they called, "clamp diodes", or is it "snubber diodes"?) will do the job? For the digital signal, I would send the filtered signal to a Schmitt trigger inverter.

What would be better to use to get a good (analog or digital) signal without the noise? Other than calling it "ignition noise", I do not know the characteristics of the noise (frequency spectrum, amplitude, etc.). Hopefully it would be as simple as this RC filter too!!!!

-Tony

[Genesis]

Why not use a position encoder?

[spamiam]

[pjc30943]

There are encoders with pretty much arbitrarily high resolutions, you just have to pay for it

Is a few hundred thousand counts per revolution enough, or a few million? They have those, in small packages too. So the ZX could measure whatever movement you like. Since such high-end encoders often have SPI interfaces, for example, data corruption (with appropriate cabling) can be avoided in noisy environments.

I'm pretty sure some of those resolution specs are a bit "promoted", but obviously that shouldn't matter for this application...

[DH*]

Tony,

Do a Google search using ADC anti-alias filter and another using clamping diodes.

[victorf]

[spamiam]

[DH*]

The simple lowpass filter used as the anti-alias filter attenuates frequencies above its cutoff frequency. It should do what you want. Ignition noise frequencies should be RPM * Cylinders / 60 which gives fairly low frequencies - in the audio range. You can find lots of web pages with calculators for low pass filters. For example:

http://www.muzique.com/schem/filter.htm

Try 100K and 0.1 for the RC values.

However, unless you are using your own battery supply, you are likely to have more problems from a noisy DC supply than from ignition noise.

[pdubinsky]

[spamiam]

[DH*]

I think you want to go lower. If I understand it correctly, you want to measure a DC voltage which is 0Hz. The Nyquist limit is to make sure you can faithfully reproduce an AC signal.

[spamiam]

[pdubinsky]

[Genesis]

IF the desired signal is DC of reasonable magnitude AND rise time is not terribly important (that is, you're not terribly interested in catching the "edge") then you can short out ignition noise quite reliably with a reasonably-large value capacitor using an RC network.

HOWEVER, if you need to be able to sense rapid changes in the DC level, then you've got a problem, because the time-constant of the capacitor comes into play in terms of response time to a signal level change.

In any analog situation stability and noise isolation of the DC supply is VERY important. If your reference is not stable and clean then no amount of filtering on the input does you a bit of good.

In those situations I often find that the best option is an optical encoder with the run back to the CPU done over some digital method (e.g. SPI, etc.) This takes the signal up to TTL levels which makes it much more noise-immune, and with reasonable attention to cabling you can usually get reliable results even in quite-noisy electrical environments.

The nightmare scenario is trying to read millivolt-level unbalanced signals in the presence of wideband noise (as ignition noise is) which change relatively quickly OR are inherently moderate to high-impedence sources (the source impedence counts in an RC circuit!) That circumstance bites and when I find myself in that sort of box I look for a way to reduce what I'm trying to measure to a digital signal as close to the acquisition point as is reasonable. Failing that you do what you can to isolate the amount of "antenna" available to pick up the interference, and do a REAL good job of filtering on the power supply side.

[stevech]

an LC rather than RC circuit would really help. If the value of interest is a DC value that does not change during the sampling window, and the sample rate can be very low, this shoud be easy.