Hi,

I have an input in current and need to convert that to voltage. The current is in the order of magnitude of pico. And I need to have the noise reduced down to 5mV or less (in the output). Currently, with the following circuitry (see attachment), I have about 40mV noise and that's no good for my purpose. Please advice on what I can do to make the filter better.

Cheers,
Kevin

 

Can you tell us more about your setup?

 

Well, basically, there is a negative current around that op-amp, so the output is in positive voltage. The current should get amplified over that 5G Ohm resistor and the capacitor is there to slow down the process. The voltage divider with a FET switch is to control addition feedback gain. And after those two low pass filters, there are buffers before going into A/D channel pins of PIC18F4525. I hope that explain more about the project.

Anyhow, like I said, I need to lower the fluctuation of the output voltage of that op-amp down to at least 5~10mV or so. It's still very noisy for my application.

 

Have you done the calculation of the noise produced in a 5G resistor at room temperature?

Your fundamental problem is that you are trying to do the current-to-voltage conversion and have the gain all in one stage. You need to divide the amplification chain into two or three steps; where you put only as much gain as you need to determine the noise figure in the first stage, and then put additional gain into the downstream stages.

 

Well, basically, there is a negative current around that op-amp, so the output is in positive voltage. The current should get amplified over that 5G Ohm resistor and the capacitor is there to slow down the process. The voltage divider with a FET switch is to control addition feedback gain. And after those two low pass filters, there are buffers before going into A/D channel pins of PIC18F4525. I hope that explain more about the project.

Anyhow, like I said, I need to lower the fluctuation of the output voltage of that op-amp down to at least 5~10mV or so. It's still very noisy for my application.

With those filters, it's surprising that you have that much noise. Have you verified that the noise is from the circuit and not from the A/D conversion process.

You haven't provided details of how you implemented this circuit, but the low current and high resistance are challenging issues. Stardard FR4 circuit board may not be suitable, and you may use to use Teflon as an insulator. Also, residual flux can cause a current path. It's very hard to have a reliable 5GOhm resistance without great care. Anyway, this should have nothing to do with the noise problem.

A large 5GOhm resistor at high bandwidth could cause 40 mV noise, but you are filtering at about 0.1 Hz bandwidth, so I would not expect this. These results really point to an issue at the A/D converter. It could be power supply filtering is not adequate to stabilize the A/D converters. Think about how much noise would need to be present before the filter in order for 40 mV to get through a 0.1 Hz filter bandwidth.

 

It is not A/D conversion process. I have scoped the signal before the buffer (cut-off in the image but it's before the A/D converter as well).

I have two plates and one of them is applied with -35V. The other plate which is connected to the op-amp measures the current induced by the opposing plate. I suppose if there is a fluctuation of 40mV in that -35V, that might be causing the noise onto the other plate?

 

It is not A/D conversion process. I have scoped the signal before the buffer (cut-off in the image but it's before the A/D converter as well).

I have two plates and one of them is applied with -35V. The other plate which is connected to the op-amp measures the current induced by the opposing plate. I suppose if there is a fluctuation of 40mV in that -35V, that might be causing the noise onto the other plate?

What I don't understand is how 40 mV is getting through your 0.1 Hz low pass filter. Think about how much noise needs to be on the input of the filter. Have you placed a scope both before and after the output filter? You would need to see many tens of volts of high bandwidth noise at the input of the filter to get 40 mV noise on the output.

Or, are you talking about 40 mV DC offset, and not noise?

 

Another question. - Are you using DC/DC converters for power supplies in your system? You could be dealing with ground loop types of noise.

A simple test would be to ground the inputs of your filters and see if the noise goes away. If it doesn't go away, then your A/D conversion process (which includes all input filters and buffers, as well as power supply and ground quality) is the problem.

 

I have just scoped both the input (inverting) and output of the op-amp, with the plate grounded. I actually separated the apparatus so the -35V wasn't affecting the other plate at all. Anyway, there seems to be only 40mV noise at the input, and the output has about 20~35mV noise. When I say noise, I actually mean pk-pk value, which I think it should be near zero.

 

I have just scoped both the input (inverting) and output of the op-amp, with the plate grounded. I actually separated the apparatus so the -35V wasn't affecting the other plate at all. Anyway, there seems to be only 40mV noise at the input, and the output has about 20~35mV noise. When I say noise, I actually mean pk-pk value, which I think it should be near zero.

OK, I was thinking about the noise after the RC filter you have with the 100K resistor and 1uF capacitor. So my previous comments may be off target.

I just noticed that you have ground for the negative rail and the noninverting input is also tied to ground. Is the OPAMP designed to work like this? Even if it is, you may want to use a negative supply rail (with bypass capactors like the positve rail) and then filter the input to the noninverting terminal with an RC low pass filter.

 

So, are you suggesting that I should put 1u and 0.1u on the negative rail to the ground? What would be the importance of the RC filter to the non-inverting input?

The diode to the ground on the inverting input is there to prevent a negative high voltage (>5V) creation at the inverting input when there is -35V on the other plate.

I was also wondering if there would be some error in the oscilloscope. It seems like pk-pk value is 80mV when not connected anything. Does it ever go down near zero when it's supposed to? I don't have a whole lot experience with this.

 

So, are you suggesting that I should put 1u and 0.1u on the negative rail to the ground? What would be the importance of the RC filter to the non-inverting input?.

Not quite. I'm suggesting using a negative supply, and then filtering with the bypass capacitors.

The filter on the non-inverting probably won't help unless you have some type of ground loops generating noise. The power supply bypass, is probably enough.

Basically, the way you have it now can allow noise to be injected into output. You also have not said whether this particular OPAMP is designed to have the noninverting terminal tied to the negative rail. The change I suggested avoids this possible issue.

The diode to the ground on the inverting input is there to prevent a negative high voltage (>5V) creation at the inverting input when there is -35V on the other plate..

Keep in mind that the diode reverse leakage current will modify your current measurement.

I was also wondering if there would be some error in the oscilloscope. It seems like pk-pk value is 80mV when not connected anything. Does it ever go down near zero when it's supposed to? I don't have a whole lot experience with this.

You should ground the scope probe near the OPAMP and then verify that placing the scope probe on ground does not show any noise. What you don't want to do is tie the scope ground far from the OPAMP beause then you are just picking up power supply noise.

 

I scope the ground exactly the way you described - hook up the reference point onto a ground point near the op-amp and scope another ground point nearby (actually several). Its mean voltage is about 3~4mV and pk-pk is 40mV. What would this mean?

 

I scope the ground exactly the way you described - hook up the reference point onto a ground point near the op-amp and scope another ground point nearby (actually several). Its mean voltage is about 3~4mV and pk-pk is 40mV. What would this mean?

Well, it's a red flag that indicates a measurement or environment problem, rather than a circuit problem. However, it's hard say what the cause is. Are there DC/DC converters or other sources of magnetic noise in the system? Are you near any large source of electrical noise? Is the scope or probe defective?

Trying putting your circuit aside and do basic tests on a simple circuit. For example an isolated battery and a resistor. Try the scope in different locations (home and work). Try to get another scope as a comparison. Have someone else visually check your setup.

In situations like this, you need to back up and get to a known point that makes sense to you. Then you can move forward again, changing one thing at a time.

 

There is a DC-DC converter that converts 5V to -35V. And that's located not to far away from the op-amp. Well, I have two versions of boards and one has DC-DC as far as possible from the op-amp but there is still 60mV noise. It seems like anything I measure will have 40~60mV pk-pk no matter what. I'm starting to wonder if the scope can even measure anything below that.

 

Is it possible that you are using a x1 probe setting but reading the x10 voltage scale on the scope? A noise level of 4 mV, although still a little high, might be more reasonable to get.

Perhaps unlikely, but I've done stupid things like this before.

 

The probe and the scope machine are both at 10x. I just created a simple RC low pass filter with 10M Ohm and 23nF. The input is from a DC power supply giving out 2V. Both the input and output pk-pk fluctuate between 40 and 80mV. This is mysterious.

 

.... This is mysterious.

Agreed. I'm sure it will make sense in hindsight. Please let us know what you find.

 

Okay. Here's the verdict. I called the manufacturer of the scope, Tektronix. 40mV seems to be a sampling noise created by the scope itself and it cannot be reduced at all. One way to view a perfect signal is to use the averaging function on the scope, which sort of biases the noise but it represents the signal a bit better. The negative side of this is that I can't really do a single acquisition with the averaging function, only with a higher end scope.

 

Okay. Here's the verdict. I called the manufacturer of the scope, Tektronix. 40mV seems to be a sampling noise created by the scope itself and it cannot be reduced at all. One way to view a perfect signal is to use the averaging function on the scope, which sort of biases the noise but it represents the signal a bit better. The negative side of this is that I can't really do a single acquisition with the averaging function, only with a higher end scope.

Ah, that makes some sense now.

Can't you use a x1 scope probe to improve the resolution to 4 mV?