In my application, there are multiple RTDs being excited with a constant current source and then the voltage across the RTDs are being measured to calculate the temperature.
These RTD signals are being multiplexed and then being fed to the AD8237. (The AD8237 gain is 1 and it is set to work in the low bandwidth.) The output of the AD8237 is connected to the Op-amp LT1997 which works as gain and shift and then signal goes to the ADC.
The RDTs are being switched every 2ms and there are some times that no RTD voltage is connected to the multiplexer. The input voltage to the AD8237 can change from 0 to 2.5V.
There are two identical channels (channel A and B) in this design that they are supposed to function exactly the same way. (see the image below)
So here is the issue:
I have received 25 boards from the board manufacturer and out of these 25 boards, 5 boards have the oscillation issue.
The issue appears to be only on one of the channels shown in the above circuit. I have not had any board that both AD8237 in-amps oscillating.
As the input voltage increases the output overshoots and it begins to oscillate sporadically. Below is the scope trace I have gotten from the output of the AD8237:
This only happens at the beginning of the cycle when the voltage of the first RTD gets applied to the In-amp. All other times it is clean without any problem.
Interestingly, the second channel works perfectly.
I need to understand why this is happening what can be done to prevent this issue to happen again. Please help.
You have two problems. It appears there is no DC return? See fig 5-3 in the Instrumentation Amp Guide, 3rd edition.
A simple diviider, say 10k/10k with 1 meg resistors to one side of the RTDs will solve that.
The AD8237 is one of my favorite InAmps, autozero, micropower, rectangular diamond plot, etc,
but because it's micro power, there is not a lot of current to pull a node back in line if you have an overload.
See fig 37; there is an implied common mode slew rate limit that you have exceeded.
if you have a weak bias on the inputs to 2.5V, you should not have a problem.
Sorry, figure 27.
In addition to what Hooman recommended to do. I created a path to ground by putting 1M ohms resistors on the Pos side (connected from TP77 to GND) and Neg side (Connected from TP78 to GND) of the AD8237. It does not help the oscillation issue.
As I mentioned on Hooman's reply, it appears that the problem is migrating as I put the bad chip on a good board.
Here is a fun fact, when I replaced a bad chip with a good one and the problem went away on the bad board.
Thanks for running the experiment.
I've seen some instability issues resolved by adding a series RC shunt to the output. This will have no effect on your low frequency / DC accuracy.
Also, I wanted to point out to make sure the AD8237 boundaries of G= 1V/V operation as shown in Figure 10 (copied below) are not violated:
I could not tell from your scope photo where the ground is to ensure that Figure 10 is observed?
It seems like with the AD8273 REF pin grounded as you have it, it's going to be very easy for any input voltage shift to hit the limits on output voltage dictated by Figure 10, as you're operating very near the bottom of the "diamond plot".
Any chance of modifying one "bad" board by tying the REF pin to mid-supply (2.5V, decoupled to ground close-by) instead, to see if your instability improves with the input step caused by RTD switching?
Thanks for the prompt response.
The Analog ground is isolated from the digital ground on the PCB. The only connection point between the two grounds is the ferrite FB2.
There will be a massive change if we want to go in that path to change the Vref. I prefer to not do that.
I will check the RC filter to see if helps. in the meantime, can you check this lot number that I took from the back of the chip to see if there is a flag on the manufacturing side?
The lot number is #7124015
I will check that lot number.
Question: Are all of the units that you have from the same lot?