Measuring 100nA to 1mA and 30V to 110V

[luky315]

I need to measure a weak current signal across a chemical sensor and my problems are that the sensor is grounded on one side so I need to measure it on the high side and that the current and voltage is a bit outside the usual range. We have a maximum 120V voltage source (<3mA, so it is considered "safe" in our application) and I need to control the voltage across the sensor element so that the current throught it is constant and in the (adjustable) range from 100nA to 1mA. The voltage source including a DAC-based control is working fine.
The sensor itself behaves like a nonlinear, temperature dependent and unstable resisitor, so it is also necessary to monitor the voltage across the sensor to shut the power off of in case that the value is outside the tolerances.
At the moment I'm using a high side current monitor with a sense resistor, but the problem is that the resistor needs to be quite large for the lower current range and this creates to much voltage drop in the higher ranges. Unfortunately it is not that easy to use a voltage supply with more than 120V because there are rules and regulations. We like them a lot here :-(
Therefore I cannot have more than 10V drop because I need a maximum of 110V on the sensing element and the next problem with this simple approach is that it is not easy to get the "real" voltage over the sensor element after the current measurement. At the moment I'm manually changing the sensing resistor depending on the desired measurement range and additionally trying to calculate the current through the voltage divider and subtracting that from the current going through the sensing element.
This approach is working somehow, but I'm not really happy with it and would also like to improve the accuracy. The speed is not that critical (20kHz at the moment which is more than enough)
 So what would be a better approach to this?

[guenthert]

I haven't quite understood why current and voltage over the sensor need to vary, but small currents are often measured with a trans-impedance amplifier instead of (absurdly large) shunt resistors.  Then the burden voltage is only the small more-or-less constant offset voltage of the amplifier.

[luky315]

A transimpedance amplifier is not working here because the sensor is grounded - a high side measurement is necessary

[guenthert]

A transimpedance amplifier is not working here because the sensor is grounded - a high side measurement is necessary

What stops you from floating the TIA?  Or, if it for some reason, needs to be all a single device, bootstrap it?

[Alex Nikitin]

A decent electrometer, for example the Keithley 617 I have in my lab, will measure currents from single digit femtoamps up to 20mA floating up to +/- 500V. I use this configuration all the time as this way you can measure the leakage when the other end of the device is grounded - for instance a coaxial cable leakage from the central conductor to the "grounded" screen. The Keithley 617 also has an internal programmable and isolated voltage source which can produce +/- 100V in 50mV increments (2mA max current).

Cheers

Alex

[bdunham7]

This approach is working somehow, but I'm not really happy with it and would also like to improve the accuracy. The speed is not that critical (20kHz at the moment which is more than enough)
 So what would be a better approach to this?

How accurately do you need to measure the current the the low (100nA) end and what do you mean by "20kHz" for what appears to be a DC measurement?  If your sensor has a response time the necessitates a 20kHz BW your problem becomes much more difficult.  Is this a product or a lab experiment?  I was thinking there are decent bench meters that could log both values, but not at 20kHz. 

[Alex Nikitin]

Oh, I've missed the 20kHz requirement  . A floating TIA with 10K gain should give you the range you need and the bandwidth should easily exceed 20kHz. The current noise for 10K in 20kHz BW is about 200pA RMS so measuring from 1nA to 1mA  is OK. You would need to float the TIA and the ADC. The advantage here is that the voltage drop will be close to zero.

Cheers

Alex

[David Hess]

1 volt at 1 milliamp would be 1 kilohm.  100 nanoamps at 1 kilohm would be 100 microvolts.  1 volt with a resolution of 100 microamps is 1 part in 10,000.

With some care, that is all feasible on the analog side.  The largest error will be from self heating of the 1 kilohm shunt resistance.

If only a digital measurement is required, then I would consider placing a delta-sigma converter at the high side to make the measurement across the 1 kilohm shunt resistance directly, and use optocouplers on the digital side for the level shift.  A floating low voltage supply might be needed.

If the problem is dynamic range because high resolution is required at 1 milliamp and at 100 nanoamps, then analog logarithmic conversion could do that but it will require some analog expertise.  This idea here is that error becomes a percentage of the reading rather than the range, so for example 1% error at 1 milliamp and also 1% error at 100 nanoamps, instead of fine resolution at 1 milliamp and coarse resolution at 100 nanoamps.

[Marco]

I need to control the voltage across the sensor element so that the current throught it is constant and in the (adjustable) range from 100nA to 1mA. The voltage source including a DAC-based control is working fine.

Why? Why not just make a high side current mirror to copy the current from a low side current source.

[luky315]

A high side current mirror with a usable range from at least 100nA to 1mA and <1% accuracy would be a solution, yes, but thats not as easy as it sounds and the problem measuring the voltage across the sensor element without additional current drain in the voltage divider remains

[David Hess]

A high side current mirror with a usable range from at least 100nA to 1mA and <1% accuracy would be a solution, yes, but thats not as easy as it sounds and the problem measuring the voltage across the sensor element without additional current drain in the voltage divider remains

The impedance is low enough to support a chopper stabilized amplifier, so I see no reason that the current mirror could not operate over that range with enough accuracy at the 100 nanoamp end.

Bootstrapping a high impedance buffer will allow removing the current drain from the voltage divider, just like an electrometer would do.  This adds complexity but is straightforward.

[luky315]

Adding a FET Opamp as a buffer would be a solution to one part of the problem, but then we need an additional HV power supply.
I found the MAX4007 High-Side Current Monitor, unfortunately only for max. 76V, but the concept looks nice. Maybe I can copy the working principle with a matched transistor.

[Berni]

You can float a high side current shunt amplifier like a INA138 or similar.

Connect the positive supply to the high voltage supply, but do not connect the GND pin to ground. Instead create your own virtual ground for the chip that is always say 10V bellow the high voltage rail. The easy method for this is just putting a 10V zenner across it for regulation and then construct a ~100uA constant current source down to ground using a transistor rated to handle the voltage.

This means the chip is now thinking the high side voltage is always 10V even if it is actually 120V. Next problem is to get the analog output signal down to ground potential. This can luckily be easily done using a  high gain PNP transistor. Connect its base to your virtual floating GND, then pass the chip output signal trough the transistor. This will make the transistor turn on just enugh to let the current trough, keeping itself near virtual ground. This makes the chip see near 0V while the actual ground is way bellow that.

[Marco]

A precision current mirror would be made in the same way you make it for current sensing with an opamp. The offset precision is determined by the opamp, the shunt and high/low side mirroring resistors determine the gain error. You feed the negative supply of the opamp with a bjt at fixed drop from the positive rails.

https://www.analog.com/en/analog-dialogue/raqs/raq-issue-151.html

[luky315]

Is it advisable to use "High Voltage" Opamps (ADHV4702-1, LTC6090 etc.) or would a floating supply be better?

[Marco]

Floating negative supply lets you use better opamps, ie. autozero/zero-drift.