If I choose the Wheatstone bridge with 3x 100MOhm and the 4th resistor is my sensor I get voltage varies between 1.633582V to 0.291176V
The minimum acquired diff in voltage I need to be able to detect is 0.00573V (5.7mV)
If I put both V0- and Vo+ into a diff inst-amp (AD620) I can use Rg (Gain resistor) to increase the gain X2.
After the diff inst amp I believe I should put buffer amp and than into the ADC.
Is that sounds ok?

 

Where are you going to get three better than 1% 100megΩ resistors? Have you priced them?

 

Price is not an issue here. I know they are expensive (I already bought 1% )

 

1/2% 3/4 watt. Axial $9

 

Biggest problem I see with your proposal is the the ADC reading is highly non-linear with R{u}. You will have to implement a reverse look-up table to linearize it.

The plot shows V(ad) vs unknown resistance R{u}, which is the independent variable of the simulation. The excitation voltage for the bridge should be the same as the ADC reference voltage to make the readings ratiometric.

Score one for the oscillator proposal, which is intrinsically linear...

 

Here is an alternate proposal. It emulates the way an Ohmmeter works. There is a constant-current source of 58.8nA forcing current through R4 (the unknown). The voltage developed across it is linearly dependent on R{u} as it goes from 1megΩ to 70megΩ. The current-source is derived from the 5V supply that the Arduino uses for its AD reference voltage, making the measurement ratiometric. It requires a single 10.00megΩ precision resistor.

There is a potential problem with the gate-to-body leakage spec for the BSS84. However, the worst case is spec'ed at Vgs=+-20V, and in this circuit, the Vgs is more like -2.7V. You will have to test over some reasonable temperature range to see if this causes a problem...

 

@MikeML
How about a nice jfet. N-ch, reverse the op-amp.
Does this say 500pa gate leakage?

 

Thanks for the interesting solutions.
1. Regarding the non-linearity of my proposal, I know the basic of the W.B is non linearity Vout.
I found an article which may reduce dramatically the non linearization - I am afraid I am not fast in Eagle nor Orcad so I attached the pdf file.
What do you say about that?

2. Regarding your suggestion with the P-ch FET, can you explain more what you are doing there? or point me to some paper on this concept?

 

...
1. Regarding the non-linearity of my proposal, I know the basic of the W.B is non linearity Vout.
I found an article which may reduce dramatically the non linearization - I am afraid I am not fast in Eagle nor Orcad so I attached the pdf file.
What do you say about that?

It looks complicated...

2. Regarding your suggestion with the P-ch FET, can you explain more what you are doing there? or point me to some paper on this concept?

You need a paper to tell you that E=I*R?

Look at Millions of DMMs; it is how they measure Ohms linearly...
The basic idea:

 

It looks complicated...


You need a paper to tell you that E=I*R?

Look at Millions of DMMs; it is how they measure Ohms linearly...
The basic idea:

View attachment 88920

Mike,
I understand the basic idea. What I don't understand is the circuit that you drawed with the BSS84 transistor

 

.......................
2. Regarding your suggestion with the P-ch FET, can you explain more what you are doing there? or point me to some paper on this concept?

If you are referring to how the circuit works, here's an explanation for that. It uses a BJT instead of a MOSFET but the operation is the same.

 

There seems to be a glitch. #12 didn't say anything referred to him in post #29.

 

There seems to be a glitch. #12 didn't say anything referred to him in post #29.

Dunno how that happened. I think the OP got the idea, though...

 

Another questions arised from the circuit described in post #26
1. How did you calculated a constant current 58.8nA from the resistor circuit ?
2. Why do we need the op-amp U2 in the circuit if we can basically run a 5V through a current source directly like in post #29?
3. How fragile or sensitive to noise the circuit build on current source (we are dealing here with nA) ?

 

Another questions arised from the circuit described in post #26
1. How did you calculated a constant current 58.8nA from the resistor circuit ?

The circuit consisting of R1, R2, U2, R3, and M1 make up a current-source that drives a grounded load (R4).
I repeat the circuit here with R4=70megΩ. I have labeled the Gate, Source, and Drain of M1, as well as some other nodes, and let LTSpice show us all of the node voltages and branch currents.



R1/R2 just form a voltage divider to create V(ref)=4.412V from the 5.00V Arduino A/D reference voltage. Note that R1 and R2 are standard 1% resistor values. I just needed something close to ~4V.

U2 is wired as a voltage follower such that it forces V(s) to be the same as V(ref). Check it. V(g) is whatever it needs to be to make that happen, because it is inside the feedback loop. V(g) would be different if Vth of M1 changes, but V(s) would not change.

The voltage across R3 is (5-4.412) = 0.588V, so the current through R3 is E/R = 0.588/10e6 = 58.799nA. A somewhat arbitrary choice, however, V(d) must be less than V(s) when R4 is 70megΩ.

Assuming that Ig(M1)=0, then -Id(M1) = Is(M1) = I(R3), so we have created a current source, where V(d) is directly proportional to the resistance of R4.

Note that with R4=1meg, V(d) = 0.059V. With R4=70meg, V(d)=4.12V, well within the full-scale range of the Arduino A/D.

2. Why do we need the op-amp U2 in the circuit if we can basically run a 5V through a current source directly like in post #29?

Explained above. It auto-biases M1 to maintain a constant current through R3, which in turn means a constant current in R4. Note that the circuit is ratiometric with respect to the 5.00V supply V1. If it changes, that changes V(d), but that also changes the reference inside the Arduino's A/D converter, so there is no effect on the value it produces.

3. How fragile or sensitive to noise the circuit build on current source (we are dealing here with nA) ?

You will have to flush all the soldering flux off your circuit board. Opamps must be very low input bias current. The biggest unknown is how Ig(M1) effects the circuit. #12's suggestion might be the answer to that.

 

Hi.
I hope you all OK. I wasn't active much lately and I am back to business.

I tried to re-create the circuit @MikeML designed but I have a few problems - please see picture attached. I don't know what is the problem but the Voutput and Ioutput is not the same as seen in the graphs above. Any suggestions where is my fault?

In addition, I was wondering, what is the difference between running this circuit in compare to a current reference component to establish a constant current? Is it because there is no such current reference component with such low value?
Another question is for whooshing the best op-amp for this task. I tried to filter using the ADI website the best for this including lower Vos, Ib, Iq, noise and high input gain (above 1G). All of them are voltage op-amp.
Here is the list of what I found (There are more but I filtered for SOIC and SOT packages):
AD8554
AD8574
AD8628
AD8630
ADA4077-4
OP4177

Is this correct or should I look for current op-amp?

Thanks,
Sagi

 

Measuring high resistance is the same as measuring low current. I have built a pico-meter from the good old TL0xx opamps. If you used newer ones (LMC662 or 6042), you can easily go to GOhm range. The key is to pick one with very low input bias current, preferrably in the fa range.

The circuit itself is actually very simple.

 

Well, Vbe and temperature effects are the big issues. The I-V converter has a transfer function of Vbias-I*Rf when you connect the + terminal to a voltage source. You don't need a real biasable I-V converter that can work at 100 mA and 0.6 V. That's MUCH harder.

You can search for SMU or (Source Measure Units) and Electrometers. www.keithley.com makes both. Keysight.com, the new separated company from Agilent makes SMU's. Not sure about electrometers.

 

I am still looking for a solution in the board level and planning to build the constant-current circuit.
The DMMs and SMUs are very expensive. I hold some from Keithley and Agilent but they are overpriced for this task.
@dannyf if you have a well established circuit, please publish it. We all want to see it.

I found my problem in the circuit - fixed.
I am still looking for answers for my questions from post #36

Thanks.
Sagi

 

There are electrometer grade OP amps such as this one: http://www.ti.com/product/lmp7721/description