signal conditioning for eC meter

[Morgoroth]

First all thanks too all who has answered my questions in the pass, has been a great help.

Now, I have another questions, probably basic for someone but not so much for me. I'll try to do this in a very academic style, could be helpful for others too.

I designed a eC (electroconductivity) meter, and for thous who don't know what is that is the measurement of the conductivity of water with something dissolved (kid style explanation), through the measurement of the equivalent resistance of a probe inside of the solution.

My first design was a pure sine wave rectified with same wave through a probe in solution (rectified too), the difference is proportional to the impedance and from there I have my conductivity measurement. The data calculation and simulation are very close, getting a constant difference of 20mV in all range, and even that is not a problem because the calibration of the system should compensate that, better is that difference is constant.

The reason to use a sin wave in my first design is to void degradation on the probe.

The circuit (Image1) is big and expensive to get the resolution I want in my device.

Now, at side in my board I had designed a pH meter who use a AD7793DAC, who have current outputs, each one can give 10uA, 210uA or 1mA. Now I'm taking a AD7794 DAC with more inputs to add the eC meter and make use of the current outputs and void almost all the electronics showed behind.

¿ why a current source given the problems they have (probe degradation in contrast to sine waves) ?, because I can make a 4 point measurement, more precision on the readings, and at uA levels the degradation is very low, anyway I will test it.

The circuit is resumed on Image2.

Now, I'm very tempted to get the signal right from the probe and put it on to the AD7794 differential inputs plus some filters but I'm not so sure if I need more signal conditioning there, I'm aware of the input current bias compensation but even with that I think the input impedance is not high enough and probably worth the effort to get a instrumentation amplifier.

The load is a equivalent resistance between 7ohm up to 10kohm.

So, the question is, worth the price to put a instrumentation amplifier ?, is the only solution to signal conditioning here ?

If I'm missing something important feel free to let me know.

Thanks in advance.

[T3sl4co1l]

Check your dropbox permissions, the images aren't hotlinking.  Or attach to the forum.

Tim

[Morgoroth]

Fixed 

[Kleinstein]

In most cases the conductivity measurement like a PH measurement does not need a really high resolution, as temperature is not that constant. So in principle no really high resolution ADC needed, except maybe for the temperature measurement.

The circuit might need some extra ESD protection to make it more robust.

As the resistance of the probe is usually rather high, there is usually no need for a 4 wire measurement. Form the configuration I would more prefer to have a more or less fixed output voltage and measure the current with a kind of transimpedance amplifier or just a relatively small resistor.

Using a sine ware is not really needed, but it helps to interpret the result, as with electrolytes there can be frequency dependent effects. However the sine ware does not have to be very high quality - just a bout looking right should be enough. For the detection is might be a good Idea to check for phase shifts as well. As coupling to 50/60 Hz might be a problem, it might be a good idea to use a kind of frequency selective detection and not just simple rectifiers.

It really depends on the purpose of the measurements and what should be done with the result. Using a ADC with µC interface suggests the result is somewhat send to a µC.

My suggestion would be using a µC to make a simple DDS type generator and also detect the signal with the internal ADC. This could reduce the hardware quite a lot, but might need quite some software.

[danadak]

I assume you are looking for absolute accuracy.

Using an IA has the advantage its Laser Trimmed for offsets and
G, resulting in very high CMR. Just a thought. The way to approach
the design is set your accuracy, precision, resolution goals then do
an end to end error analysis (including noise) to see what you can
achieve. Matlab makes a good approach to doing analysis, if you do
not want to tackle the algebra, an example attached.

Regards, Dana.

[Morgoroth]

In most cases the conductivity measurement like a PH measurement does not need a really high resolution, as temperature is not that constant. So in principle no really high resolution ADC needed, except maybe for the temperature measurement.

The circuit might need some extra ESD protection to make it more robust.

As the resistance of the probe is usually rather high, there is usually no need for a 4 wire measurement. Form the configuration I would more prefer to have a more or less fixed output voltage and measure the current with a kind of transimpedance amplifier or just a relatively small resistor.

Using a sine ware is not really needed, but it helps to interpret the result, as with electrolytes there can be frequency dependent effects. However the sine ware does not have to be very high quality - just a bout looking right should be enough. For the detection is might be a good Idea to check for phase shifts as well. As coupling to 50/60 Hz might be a problem, it might be a good idea to use a kind of frequency selective detection and not just simple rectifiers.

It really depends on the purpose of the measurements and what should be done with the result. Using a ADC with µC interface suggests the result is somewhat send to a µC.

My suggestion would be using a µC to make a simple DDS type generator and also detect the signal with the internal ADC. This could reduce the hardware quite a lot, but might need quite some software.

ESD Protections : Got it, I'm gonna add it asap.

Sine Waves : I change for constant current to save some $ and because I realize in that configuration I'll add all 50/60 Hz noise without chance of filtering after RMS stage.

uC : Indeed I'm interfacing a uC through ADUM5401, already tested communication my pH meter and the only noise I'm getting is 50/60 Hz from my lab (checked with my almost toy oscilloscope), and the AD7793 have dedicated filters integrated for 50/60 Hz noise, and seems to work decently. The problem is as soon I got some communication with the DAC I start to finish the board with all the other things is must have (relays, others sensors, etc) so not sully tested, I'm still programming.

Goals: Measurement of pH and eC with 0.001 resolution.

uC as DDS : that was the idea with the first circuitry, but the noise issue will be a problem and I have to rethink my strategy there, probably a good idea is to test both solutions and choose for the best.

[Morgoroth]

I assume you are looking for absolute accuracy.

Using an IA has the advantage its Laser Trimmed for offsets and
G, resulting in very high CMR. Just a thought. The way to approach
the design is set your accuracy, precision, resolution goals then do
an end to end error analysis (including noise) to see what you can
achieve. Matlab makes a good approach to doing analysis, if you do
not want to tackle the algebra, an example attached.

Regards, Dana.

Thanks for the tip, I've never seen that approach, fortunately I have my Matlab ready to go and will take a loot on it.

[Kleinstein]

The AD779x will not help you much with 50/60 Hz suppression, if this is behind a rectifier.  So the ADC might be a good way to get rid of the 50 Hz in the PH measurement, but does not help much with the conductivity, unless a very low frequency (e.g. 12.5 7 15 Hz) is used so that the ADC is sampling the AC signal. The way to suppress the 50 Hz here is use of a phase locked rectifier. This could also be done in software, by sampling the AC signal.

If it comes to saving, one could save on the ADC and analog isolation. Usually the µC internal ADC should be good enough for conductivity and the signal could be send digital.

Having better than 0.01 PH resolution does not make much sense, if you don't have a very well controlled / measured temperature: the cell voltage is about proportional to temperature, thus 0.01 PH already needs about 1 K stable / measured temperature at least.

[Henrique]

Hi @Morgoroth, i'm also trying to do the signal conditioning to my EC measuring circuit and i was looking to your circuit and i'm intrigued, from what i can see your probe is the R8 and  you are rectifying both ac signals, the ac signal proportional to the probe resistance  and the other original ac signal and after subtracting them. So at your Vout_DIFF you are getting a Vout_Diff=G*Vout - Vout. I think you are doing this to prevent that your PRMS would be afected by variations of WRMS but i don't understand how. Can you explain me ? Thank you in advance.

[Morgoroth]

Hi Henrique,

The only reason to use the instrumentation amplifier is because I need a high impedance and differential amplifier on one chip. Since you now the value peak to peak of the original wave and RMS you can use the difference to found the value of the probe on R8 (RL), and get an idea of the error induced by the electronics. Always assuming the the signal is slow enough to get stable readings after RMS stage. This was a test circuit, not a finished one for production, so you can change the differential amplifier for whatever you feel is the correct output.

wrms = RMS of original wave
prms = rms of probe wave (proportional to the resistance/impedance)

 

[Henrique]

Thank you for answer, i understood it wrongly, i thought you were trying to minimize the error from possible variations of the peak to peak value of the original wave. So this way you are just removing the error induced by the precision rectifier, right ?

I'm designing to high conductivity solutions, maximum 40 mS/cm ( 25 ohms with a K=1cm^-1 probe) and minimum 0.01 mS/cm (10k ohms) , from your experience you think i'm forced to use a 4 point measurement to achive this kind of range of impedance or i can achieve it circuit like yours in figure 1 ?

Thank you again.

[Morgoroth]

What I'm doing with the rectified signals is compare a first wave with a second wave which is the same with a probe in middle, in this way I can know the value of the probe, that's all.... and in the process get some info about the quality of the design with some test points on the PCB.

You will need more than a array of standar buffers,you will need an array of resistor to change the curve up to lineal space, look at the chart attached and you will understand.



The scale is logarithmic. Changing the parameters you´ll be able to move that curve whatever you need to.

Now, if you have a wide range to measure, take the Analog Device CN0349 solution and do something similar. 

Regards,

 JP