Multiplex sensors... how to ground?

You shouldn't be using the 4051, but the 4052, and selecting the local ground at each LM35 as well as the LM35 output pin. The differential output for the mux should then be fed through a differential amplifier

- either a subtractor-with-gain or an instrumentation amplifier - to amplify the signal to fill the A/D converter range, and convert it from a differential to a single-ended signal referred to the ground voltage at the common point of the DAQ analog and digital grounds.

If you could arrange things so that the wasn't any current running through the 8-24 tracks connecting your AD35 grounds to DAQ "star" ground point, you mightn't need to bother with this. The 60uA drawn by the LM35 isn't going to produce a 0.5V offset in any remotely likely length of track, and 0.5V is large - but by no means unknown - voltage drop across a ground track, though it would take a pretty thoughtless layout to generate it.

Since that level of error voltage is probably not being generated by a DC voltage drop, but by some kind of rectification of hgh frequency interference at the LM35, the buffer amplfier or the analog input to the A/D converter input, the differential multiplexer and the differential amplifier is probably a safer route to follow. It has worked well for me in similar situations.

--
Bill Sloman, Nijmegen

Hi Bill, Thanks a lot for the reply. From this and other things I've read, I'm realizing that I am being very sloppy about my grounds. Right now, this circuit is just laid out on a breadboard, and I am just plugging anything that needs grounding to the nearest available bus. I am going to rip the circuit out and start over, keeping things separate and see if this helps. I'll also order up some dual muxes (I think that I will go with some 8x2 like an ADG707 rather than a 4052 just so I don't have to re-write the software. I only need 3 of them, so the cost isn't really important) and a differential amp.

Is there a standard differential instrumentation amp that anyone can recommend? This is another of those devices where a quick search at Digikey turns up 100s of items with similar looking specs and prices from $0.40 - $40.

As for the source of the offset voltage, it seems like it is coming from the digital side of the mux, regardless of whether the LM35s are connected. Don't know if that means anything.

Thanks again, mike

Mike,

think about where the current is flowing through the ground and creating the 0.5 volt diff..also try to keep you low level signal differential., even if one side is actually ground, you can "pretend" it is differential and connect it to ground at a place where it is the "correct" ground...

Mark

Mark and Bill, Thanks again for the replies. I actually found that I had everything I needed in my Misc. Stuff box, so I was able to wire it up last night. Keeping the signal differential has fixed the ground shifting problems, and everything seems to be working... at least, I'm getting reasonable readings out. I'm still puzzled slightly by the behavior of this circuit, and I wonder if anyone can shed light on it.

I'm running both the signal and the LM35 gnd wires through the 4052 and from there to the instrumentation amp (I had a burr brown INA122 around). The data sheet on this recommends using 47K resistors from each input to gnd to dissipate input bias currents. My actual output seems to shift if I vary the size of this resistor, but 47K seems to give me exactly the right value.

The most puzzling thing is that if I measure the voltage across the 2 outputs of the 4052, I get a value exactly 2x what I expect. This turns out to be fortuitous, because the amp has a minimum gain of 5x, so between the two of those, I get 10x gain. Again, the voltage between the two channels is dependent on the size of the input bias resistors, which I guess makes sense because the V- side eventually has another path to ground. So, my question is, is the 47K resistor a magic number that matches the internal circuitry of the amp? At this point, should I just relax and enjoy the fact that it is actually working?

Thanks again for all the help. Rewiring this whole circuit from scratch has helped in many ways, especially in clearing up noise... the new circuit is much cleaner (I remembered to tie the shield from the sensor cables to ground, which helped a lot) so I don't even think I need the low-pass filter.

cheers mike

Without a complete circuit diagram, all I can say is that this is weird. The LM35 is a relatively low-impedance voltage source (at least for small currents), and the "on" resistance of the 4052 ought not to be high enough to do anything interesting, so the odds are that what you are reading doesn't mean what you think it means.

You always need the low pass filter - you just don't know when ...

--
Bill Sloman, Nijmegen

Hi Bill, Yes, that is my impression too. When I get back to the lab on Sunday or Monday, I'll upload a schematic. I realized that it isn't working as well as I thought... the values are great in a narrow range around room temp, but when I bring the probe down to 0, it is still reading 12 C. I notice that there is a big voltage offset, too.

Hopefully this will make more sense in a new week. I'll put up the diagram to help others.

Thanks again. mike

I have put a circuit diagram here:

The ground is connected to the digital and analog ground of the DAQ card after pin 5 on the amp. Thus, the only path for current through the LM35 leads through the amp... is that a poor design?

Thanks, mike

If anyone has suggestions of things I should check, I'd appreciate it.

The schematic helps. Your path to ground isn't "through the amp" (since the input impedance of the INA is up in the gigohms, that wouldn't be much of a path); it's through the 47k resistor.

The LM35 draws nominally 60uA of supply current. That current has to flow to ground (since it came from Vcc). In your case, it has to flow through a

47k resistor. 47k * 60uA = 2.8V. So the ground of the LM35 is 2.8V above actual ground. But, since Vcc is 5V, that means you're powering the LM35 with 2.2V. It requires 4V, minimum, to work.

To fix it: Instead of connecting the - input of the INA to ground through

47k, just connect it directly to ground. And don't bother with the 47k from the + input to ground; that is only necessary when the input is floating with respect to ground.

Walter,

Thanks a bunch, that solved it! The explanation makes perfect enough sense that I might not even make that mistake again, now that I understand it.

Bill and Mark and everyone who looked at this, thanks for your insight. I'm really happy to have this device working and to have made some progress learning about circuit design.

cheers, mike