CN-0359 testing

Hi ksw280,

Apologies for the delayed response. This sounds like it might be similar to an issue we saw with a physically small conductivity cell. CN0359 works well with certain cells, but after it turns out that the op-amp bias current (and maybe the small but nonzero DC component of the excitation waveform) are enough to cause issues with smaller probes.

What worked for the probe in question was to simply capacitively couple everything, except the cell's return/ virtual ground connection. These are some notes from the engineer that tested this theory:

I used a breadboard to make the following connections shown in the attached circuit.

• A 10 µF tantalum capacitor in series between the conductivity cell excitation and J5P1
• A 1 µF polycarbonate capacitor in series between conductivity cell sense output and and J5P2
• A 1 µF polycarbonate capacitor in series between conductivity cell sense output and J5P3
• A 1 MΩ resistor to ground on J5P2s
• A 1 MΩ resistor to ground on J5P3

Could you give this a try?

-Mark

Hi Mark,

Thank you for answer.
I've seen this in "CN0359: Float of signal from Instrumentation Amplifier."
However, the excitation current leaks through the 1M Ohm resistor connected to J5P3. This causes a measurement error and eliminates the effect of using the U14 buffer.
Before asking, I tested by connecting a 1M Ohm resistor to J5P2 without a capacitor, but the DC drift didn't go away.
What do you think about my opinion?

Thank you again for your kindness.

Kwon

Hi Kwon, I apologize for the delay. You really do need the capacitors - the idea is to completely eliminate any DC current flow into your sensor.

On the 1M resistance: This is somewhat arbitrary, and as long as the time constant between the AC coupling cap and the and this resistance is small with respect to the excitation frequency, the error will be negligible.

The purpose of these resistors are to provide a return path for the op-amp's bias current - 50nA max for the AD8253, so even 10M would work fine (producing a 500mV DC offset, but remember - the circuit is digitizing the AC content of the sensor output.)

Same on the ADA4000-1, 10M would be fine (its 40pA max bias current is 3 orders of magnitude smaller than that of the AD8253.)

Also remember to keep the board very clean (water soluble flux is notorious for causing leakage paths in high-impedance circuits.)

-Mark

Hi Mark,

I understand what you say.
But what I wanted to talk about is that the 1M Ohm resistor connected between J5P3 and ground violates the content of the "Fully Automatic Self-Calibrated Conductivity Measurement System" 3 page.

"The voltage applied to the cell, V1, is measured with the AD8253 instrumentation amplifier (U15). The positive input to U15 is buffered by the ADA4000-1 (U14). The ADA4000-1 is chosen because of its low bias
current of 5 pA to minimize the error in measuring low currents associated with low conductivities. The negative input of the AD8253 does not require buffering."

In other words, at low conductivity (= low current), current leakage through the J5P3's 1M Ohm resistor can cause measurement errors.
Therefore, connecting a 1M Ohm resistor between J5P3 and ground does not make sense to use U14.

This is my idea.

Thank you.

Kwon

Hi Kwon,

Indeed, the 1M (or 10M) resistor will introduce an error, but it's worth trying out in order to understand whether the bias current and it's long term effect on your sensor is the root cause of the drift you're seeing. You could start out by just capacitively coupling the excitation (pin 1) - this would remove DC content from the excitation signal. Also note that since you know the value of this resistor, you could compensate in software (if it solves the electrical problem.)

IF you trace the problem back to bias current from the AD8253, you could buffer it separately with an external ADA4000-1 (unity gain buffer.)

And a side question - have you tested the circuit with fixed resistors to emulate your actual sensor? This is always a good reality check to rule out other problems with the board.

-Mark

Hi Mark,

Now I will understand your will.
"The error caused by the 1M Ohm resistor connected to J5P3 is known and can be compensated in software."
I will summarize it like this.

Thank you in the meantime.

Kwon

We'd be interested in your results, it appears there's some subtle issues with certain probes. On that note, can you provide the part number for the probe you're using, or is it your own proprietary design?

-Mark

It is our own proprietary design, looks like bull's eye. Please refer to the attached photo for the shape.

Kwon

Understood - just speculating, but if your sense electrodes are very small, then you may indeed need to take special precautions and use picoamp bias current op-amps on both.

-Mark