How Should Grounding and Signal Isolation Be Handled in Analog Circuit Design?

[formulajoe]

I designed an analog circuit that transforms a couple of different incoming signals into a digitally compatible signal. I was recently informed there will be a grounding issue at the front of the circuit. The incoming signals will be from a variety of sensors. The first part of my circuit is an op-amp configured as a differentiator to remove dc components of signals. The ground on my op-amp is the same ground as the rest of the analog circuit. I need to factor in the ground from the sensor. Would the easiest way of doing this be to connect the negative of the sensor cable to the grounded opamp input, or use opto-isolation prior to the opamp? My other question is if opto-isolation is the way to go what are the cutoffs as far as the max frequency and min voltage? One of the input signals is a sine wave that increases in voltage as frequency increases, the min voltage I have to consider for this signal is around 50 mV.

 

[Averagesupernova]

Post a schematic please.

 

[formulajoe]

This is a horrible representation, I had to make it in paint because I can't post OrCad schematics on here. Should be enough relevant information though.

 

[berkeman]

This is a horrible representation, I had to make it in paint because I can't post OrCad schematics on here. Should be enough relevant information though.

First of all, there are several ways to post OrCAD schematics here (and in other documentation). In OrCAD, click drag a box to select the part of the schematic that you want, and Cntrl-C to copy it to the clipboard. Then start Word or Paint or Visio or whatever, and do a Cntrl-V to paste it in. Also, you can get a free PDF writer from PrimoPDF.com, which allows you to print from OrCAD directly to PDF files.

As to your sensor question, no, optoisolation doesn't sound like the way you should go. Optoisolation of analog signals is a very involved subject, and you generally won't opto an analog signal directly (especially a low-level analog signal). You need to look more closely at the sensor and its output, as well as the sources of noise, in order to figure out the best strategy for conditioning the signals and getting them into your main monitoring circuit. You also need to be smart about how you handle your sensors and their wiring and grounding -- it's a bad thing, for example, to remote a sensor and run a long wire pair back to the first amplification stage.

What are the sensors exactly, and how are they powered and are they grounded? In general with low-level sensors in a noisy environment, you will do local pre-amplification of the sensor output, then drive that buffered signal back to the main board. If there are grounding or noise issues, you get creative with how you transfer the pertinant info from the sensor signal into your main board. For example, if you are only interested in the frequency of the signal, you might square it up and drive it through an opto. Or if you are only interested in the peak-to-peak amplitude, you might do active rectification and a simple A/D conversion, and then transfer that digital info across some isolated interface...

 

[formulajoe]

Thanks for the tip with OrCad, I'll try that next time.

These signals are coming from wheel speed sensors. The low-amplitude signal is from a passive sensor. The passive sensor varies with the velocity at which the teeth of the gear pass the sensor as well as the air gap. The air gap should be constant for these. There are active sensors that use the hall effect to produce a current. This current is passed through a ohm resistor and the voltage is taken across the resistor. The ground for these sensors is vehicle ground, and the ground for the module I am designing is the ground that is the same as the remainder of the data acquisition box.

 

[berkeman]

I don't understand the implication that the data acquisition box has a different ground from the car ground. Doesn't the data acq box ride in the car? I also don't quite understand the sensor arrangement, with reference to both passive and active sensors to detect wheel velocity. Also, the Hall effect doesn't produce much voltage, so it will give you a pretty noisy reading. Is the passive device just a magnet on the other side of the teeth, and the Hall device detects the delta-B as the ferrous teeth move through the path of the magnetic field?

Do you have the option to use some other sensor? For example, you could drive a high-frequency sine wave into a coil that is next to the passing teeth, and sense the delta-I as the teeth move past (like an inductance measurement). Or if you can swing it, an opto-interruptor and a slotted disk would make a much better speed sensor, depending on the environmentals of where the sensor has to be mounted.

 

[formulajoe]

There are actually more wheel speed sensors besides the ones I discussed my module must work with. I cannot decide what sensors are used, I have to work with what's already existing.
I don't quite understand the difference between vehicle ground and the DAQ ground either. The DAQ does ride in the car, but there must be something about how the rest of the DAQ was designed? I'm going to have to clarify this with somebody. Either I misunderstood something, or somebody on my team misunderstood something.
I was told the signals coming from the sensors will be fairly clean, and from the brief experiment I was shown the signals are clean.

 

[berkeman]

Okay, if the grounds are fairly quiet, and the signals are clean, then just amplify the analog signals, clip them, and turn them into logic signals. You shouldn't run into any big bandwidth issues, I wouldn't think. Have fun!

BTW, I was going to post info about analog opto-isolator modules, but now I won't bother googling them. I did remember since my last post that somebody (Analog Devices maybe?) makes a module that incorporates many of the tricks that I was referring to in order to get actual analog through an opto interface. Anybody interested should be able to find it with a google search and a little refinement.