glasselectrode input amplifier

You'd have to tell us how this signal is generated.

Hint: Try to avoid amplifiers with bipolar inputs or anything that has diodes in it. For example, avoid opamps with limited differential mode range because most likely there is some kind of BJT path that can rectify any RF signal it sees. Cell phones and especially the GSM phones on your side of the ocean can cause a lot of grief.

Common mode rejection in op amps is not always good above audio frequencies. Perhaps congifure your sensors as current sources in a cascode circuit. It will help in preserving your bandwidth, lower the output impedance and give you more power to work with alleviating the effects of external influences.

Hi,

I have used the AD8224 which has JFET input. It contains two instrumentation amplifier which I have connected to generate an output signal with a common mode voltage of half my reference. This output then drives an differential ADC input with the same reference. That is VCM_OUT = VREF/2 VDIFF_OUT = ( VIN+ - VIN- ) x G (Where G is 1 to avoid an external resistor with a temperature coefficient)

One input is provided by a glass electrode (pH type). This can be seen as a very high impedance battery. The voltage developed at the electrode is the voltage across the membrane (voltage proportional to pH) and the voltage developed internally between the electrolyte and the electrode (due to chemistry). The second voltage difference, which is unknown, is handled by using a reference electrode (a few kOhm of resistance) which is built typically from the same electrolyte and electrode. The input amplifier now computes the difference between these two inputs which gives an output proportional to the voltage developed across the membrane. The input current of the JFET are handled by placing a grounding rod in the solution (conductivity always > 100uS so this makes no difference for 100 or 200Megaohm). The output of the INAMP is filtered again and is then converted to a digital signal by using the ADC (very slow, high resolution sigma- delta type)

The big challenge here is:

=> Since the inputs are very high impedance I am very sensitive to noise. I have read that some people bias the shield of the electrode cables to the common mode input voltage. In my opinion this would only reduce leakage because but does not have any other benefits than putting it on the ground. => Make sure that the connection to the INAMP are very short on the PCB and that there are no leakage currents. I have some basic spacing and some guard traces around the two electrode signals. The guard traces are on the same potential as the signal and are generated from the inamp feedback pins. => If there are currents or strong electromagnetic field in the solution (although it is conductive but this still happens due to earth currents, ...) I will pick up this in my amplifier.

What I am now asking myself is if there are any better solutions. What I heard is that some people use something with an integrator at the input and some capacitors but I can not figure out any circuits like this.

Kind regards, Christian Walter

The purpose of the driven shield is not so much to reduce leakage as it is to reduce capacitive coupling between the signal leads and shield; if both sides of this capacitor are at the same voltage there is no current through it, otherwise AC common mode signal capacitively couples to the shield presenting an additional load on your signal which will cause error with a high Z source.

Ralph Morrison "Grounding and Shielding" covers the methods of making low level measurements in noisy environments better than anything else I have read, suggest you take a look at it.

An integrator on such a high impedance input would reduce the bandwidth to near DC if that is what you want ...

Consider digital filtering of your acquired signal too; at least averaging of multiple samples.

Thanks for this valuable input - This sounds logical to me and I will test if this gives any benefit in my application.

The step response of a glass electrode is in the range of seconds. To not only see the step but to see how the value are changeing I am happy with a corner bandwidth of a few dozens of hertz. But placing a normal integrator with an opamp does not work because it has a resistor on its inputs. I would have to add an input buffer in any case since i can not draw any current from my electrode. Then I could add an integrator but I do not see any benefit in doing this. I thought maybe you could charge/discharge a capacitor from a constant current source and compare this with the input voltage and apply a closed loop control to it. Again I do not see how it should be better than my input amplifier with some analog filter after the inamp.

Kind regards, Christian

wolti_At wrote in news:b877a9d2-e1fe-4afb-b19e- snipped-for-privacy@m44g2000hsc.googlegroups.com:

With a glass electrode, you have a large surface area of dielectric between two conductive fluids-- a big capacitor!!

You actually need to use a negative capacitance preamplifier with a glass microelectrode. I haven't built one, but they're not too difficult. It's actually an op-amp in a positive feedback mode, with a capacitor in the loop

Hello,

What I have actually adapted is a circuit from an TI application sheet

wolti_At wrote in news:d0d44b3f-c410-4dca-b40a-1b7a055c16d4 @f63g2000hsf.googlegroups.com:

At first blush the AD8224 looks good, seems there is no diff-mode limiter. Still, unlike MOSFETs JEFT inputs can demodulate stuff when the levels reach high enough to briefly get the gate layer to conduct, then they'll rectify whatever bursts they see.

If you have or suspect noise issues I'd start from the diagnostic side. Hang a spectrum analyzer to the output and see what you get in your band of interest. Maybe go up to several tens of kHz. Also hang a scope onto it and crank it all the way up on AC. Hold your cell phone at the shortest distance that can reasonably be expected, turn it off, then turn it back on. If you see spikes you need to filter some more (ask, and we can all give you hints what to do then).

The spectrum analyzer might show 50Hz, 150Hz and so on. Then you need to filter that out. Watch it especially when big machines start such as cranes or elevators.

Apart from all the good advice already given on input amplifiers: also be very careful with the connection cable. Normally a screened cable is used. Most 'normal' screened cables generate electrical charge when moved or bend. For this very high impedance use special cables are available.

And of course mounting method, connections, etc. - all must be very clean and low leakage.

Arie de Muynck

Good point. Also, many cables have "interesting" effects when pressure or humidity changes.

Plus shielded. Else a guy walking by in a shirt with enough Polyester content will cause a system reaction. I hate those shirts.

Christian, Guard rings work great for preventing leakage and they work best if the guard ring voltage tracks to voltage of the source. Let's use a peak detector as an example, tie the cap to the + of an op amp configured as a unity gain buffer and tie the output of the op amp to the guard ring. The guard ring voltage tracks the cap voltage. Does this help you?

ss

I have use this product as well, It is similar to to Burr Browns design, perhaps TI bought Burr Brown,

Hello,

One more question regarding the guard rings and driving the shield.

What is the better solution? The same applies for the guard rings but for the guard rings I would drive them with the same voltage.

Kind regards, Christian

Indeed, stability needs to be considered with any amplifier. Nothing special about this case as far as I can see.

Since you have 2 separate cables to your 2 separate electrodes it might be best for each shield to track its own signal. But you would be much better off reading the grounding and shielding book I recommended and doing the analysis yourself, considering the details of your specific situation, than accepting my guess.

I haven't seen this 5th edition, but it is probably better than the

Is the source a voltage source or current source?

This is important because if what you're trying to measure is, say, a few nanoamps of current - it's not clear what you mean by "glass electrodes" - then a lot of people use current amplifiers to measure the output. The idea here is basically an op amp where the input goes into the inverting terminal, which is connected to the output by a resistor of value R. (Also a small capacitor for some high freq rolloff to kill noise - say 100pF). Let's say R = 100 kilohms. Then assuming the non inverting input is at some ref voltage (Vref... could be 0V but usually raised slightly above there to allow for drift) then 1 milliamp of input current gives an output of 100,000 x 1mA volts, ie 100V. Obviously it's quite good at detecting small currents. The clever bit is that the OFFSET VOLTAGE amplification is low; many op amps have 1mV input offset (in-amps ain't too great) and if you had a voltage gain of 100k the output would be permanently against a rail.

There are drawbacks to this approach like, poor RF rejection compared to voltage amplification. However there doesn't seem much point saying more unless you give some more details on what you're measuring, otherwise we're just rambling on randomly.

nd

to

te

haven't seen this 5th edition, but it is probably better than the