Looking for explanation of erratic FET current

[Trent]

I have built a simple Wheatstone bridge with an MPF102 nJFET in one leg to monitor atmospheric charges and it works, but I'm getting a very erratic current output from the FET. It consists of random jumps in output ranging anywhere from 0.1 mV up to as much as 20 mV.
It is not due to ripple from the power supply. The bridge is insensitive to power supply variations. Also, if the FET is shorted or replaced with a resistor, this "noise" disappears completely.

The part of the circuit that is outdoors is in an enclosure and is under the dome of the antenna so is shielded from weather. I'm sure the antenna has no electrical leakage to ground.

Most circuits of this type which I have seen online show the FET without any biasing. Would this help?
In the attached schematic there is a resistor shown highlighted in yellow connecting the FET gate to source. This is not in place at the moment. Would adding this resistor help?

Also included on the schematic are various measurements and calculations I've made on the circuit.

Due to the low resistance of the FET, I only have a voltage drop on it of about 1.2 V at about 5 mA. I've seen articles that mention using a much higher voltage drop across the FET, but in the past I've noticed that as the voltage and current in the FET rose, so did the extent of the current fluctuations until they practically swamp the signal I want to see.

I've looked up parasitic oscillations and they are said to be in the megahertz range. I don't have an oscilloscope so I don't know the real frequency of these fluctuations. My software samples the data once per second so that's all I see, but it appears to be low frequency.

I'm interested in the daily variation of atmospheric potential so I need to monitor the DC current through the FET as it changes with the potential the FET senses. I also like seeing the spikes from lightning so I don't want to simply connect a filter to remove everything over 1 Hz.

I'm not familiar with working with transistors, certainly not biasing them. I'm way over my head on this so I'm hoping someone can give me a simple explanation for the erratic FET current and hopefully a simple cure.

Thanks

 

[crutschow]

It would seem that, for this circuit to operate properly, it would need a good connection to earth ground to complete the circuit from the antenna. Do you have that?

Also, because of the signal path being between the input and the earth, a bridge circuit may not work quite as you expect. The signal is not between the gate and drain of the FET but between the gate and earth.

And you have no definite bias voltage on the FET. What bias voltage and current do you need? I couldn't find a data sheet on the FET. How did you design the circuit if you don't know anything about biasing the transistor?

 

[Trent]

The circuit is grounded outside the bridge. I found that putting a 10M resistor between the circuit and ground significantly reduced the signal fluctuations.
I could try putting the ground on the source line, inside the bridge to make the "signal" path more directly between gate and earth.

I've seen dozens of similar static sensing circuits online and most are shown without bias resistors. I figured they just weren't very necessary in this type of circuit. I don't know what kind of bias I might need or even if it would be any improvement. There must be a reason so many circuits online have no biasing.
A similar FET which you may find datasheets for is 2N3819.
I have some datasheets for the 2N3819 and MPF102. If you tell me what data you're looking for I'll try to locate it.

Here are a few links to similar circuits which I used to start from and over time gradually modified :
Alan Yates' Laboratory - Simple Electrometer
Static Electricity - Negative Ion Detector electronic circuit diagram
Electronic Idea: Test and Measuring
Build this simple "electronic electroscope," a FET electrometer
Avoid the Shock of Shocks! Build Your Own Super-sensitive Electric Field Detector

 

[crutschow]

You show Rgb from the FET gate to source as not being in the circuit. Have you tried that resistor? That should stabilize the bias.

Is the antenna wire shielded into your house?

What is the nature of the signal you are expecting?

 

[Trent]

The resistor from gate to source, Rgb, has not been tried yet. I plan to try it soon. I've tried everything else I could think of on the indoor part of the circuit.

The cable carying the drain and source wires is shielded and the shiled is grounded. I also have a ferrite core on the cable as a precaution against RF pickup since the cable is 50 feet long.

The signal I'm monitoring is atmospheric charge. It varies from day to night and in thunderstorms. Lightning causes spikes in the data. My graphs match published data so the circuit appears to be doing what it is supposed to do.

The antenna has been through some violent storms without a single malfunction so I'm sure there is no moisture causing unwanted connection between the gate and source or drain.

Depending on atmospheric and operating conditions I usually get a bridge output of 100 - 170 mV between the daily high and low. This corresponds to a change in FET resistance of 6-10 ohms. I don't know how to convert the FET resistance changes to actual voltage sensed by the gate. The FET resistance is about the same as the other resistors in the bridge, 220 ohms. The FET is simply acting as a voltage controlled resistor in the bridge.

I'm at a loss to explain the erratic current through the FET. Having this "noise" hides small details in the signal.
When I read about using FETs in amplifier circuits they talk about having a much larger Vds than the 1.2 volts I have, and, as I said, I've found that increasing Vds or Id makes the "noise" much worse.
The power supply is giving me 14.6 volts and is capable of supplying 1500 mA but I measure a current of only 9.91 mA through the bridge so the circuit is not underpowered.

I've tried replacing the FET in case there was some static damage but my results are the same.

I have an idea for making a bridge consisting of 4 FETs connected so that they all change resistance appropriately with the sensed voltage thus giving me a signal up to 4 times stronger. But since I don't know what to do about the noise, I may be getting 4 times as much noise too. I haven't tried this circuit yet. It will be next after I connect the resistor in yellow, Rgb.

 

[Trent]

PS. I'm curious, would there be any advantage to biasing the FET? What would be the difference?
Could it improve the gain enough to offset the drop in sensitivity caused by adding resistance to the bridge? (see example below)

One of the reasons I kept the bridge resistances low is that it seems to improve sensitivity. Say, for example, the resistors are all 100 ohms and I get a certain bridge voltage drop if one resistor (FET) changes to 101 ohms (1% increase). If I make all the resistors 10 times as large, I would need a 10 ohm increase (still 1%) to get the same voltage drop across the bridge.

See the Wheatstone bridge calculator at:
Wheatstone Bridge Calculator

 

[Trent]

I just tried connecting Rgb (1.5M) between the gate and source lines. It did not stabilize the erratic current. What it did do was to reduce the signal to 1/4 its original value.

I then air-wired the gate to the antenna to be sure no PCB surface currents were affecting the gate voltage. This appears to have had no effect.

I have no idea what kind of low frequency signal the antenna could be picking up to cause these fluctuations. I'm lost.

Thanks anyway

 

[ronv]

I think I would try bringing the fet inside to the board and just using a small antenna to see how it works.
The earth ground is what?
It almost sounds like something is charging up and then the fet breaks down.

 

[Trent]

Noise graph

Hello Ronv,

Well, I can bring the antenna inside. I don't recall offhand how it reacted last time it was indoors when I first made it. A major problem with having the FET indoors and just running a gate wire outside to the antenna is that any capacitance between gate and ground severely reduces the sensitivity.

The circuit is connected to the chassis and both are grounded to a water pipe. I get no interference from indoor appliances or fluorescent lights turning on and off. The best thing I've found so far for reducing fluctuations is having a 10M resistor between the circuit and ground.

I have tried more than one FET and I get the exact same results, so I doubt they've all been blown. I handle them carefully and wearing a ground strap. Also, it's not like the voltage rises to some constant value then suddenly decreases, the jumps seem to vary at random.

Here's a sample of what the "noise" looks like:

 

[ronv]

I was just trying to get the 50 feet of wire out of the loop. It's not clear to me that any noise on the source is not like an input to the gate. The whole thing is very sensitive -- are you sure it isn't telling the truth? I have no idea what the output should look like but i guess quiet unless your outside brushing the dog on a dry day :=)

 

[ronv]

Is there any reason the 10 meg to ground can't be smaller?

 

[Trent]

Ronv,

I am assuming the noise is from gate input since I can't find any other source. If I remove the FET the noise disappears.

This is Chicago. I won't have to worry about low humidity for some time yet.

Actually, I began with a direct connection to ground - no resistor. I found (just recently) that I get less noise without a ground, but the whole system acts like a dipole antenna with the antenna and gate being one half and the rest of the circuit being the other half. Therefore, it seems logical that the circuit half should be grounded.

I went with a 10M resistor because it still provides a connection to ground while reducing the current fluctuations about the same as having no ground at all. I'm getting my cleanest signals now, just a fraction of what the current variations used to be while still retaining about 200 mV signal p-p between day and night.

I tried shorting the fluctuating current from the source line to ground with a 4700 uF cap but, surprisingly, it had no effect at all.

Thanks, I appreciate your suggestions.

 

[Sangamo]

Trent, I can't be of much help but I am 150 mi south of you and would like to try this circuit.

 

[Trent]

Hi Sangamo,

I'd be happy to share what I have. I've attached the latest schematic with some notes. Feel free to contact me anytime with comments or questions. This began as an extremely simple, tiny circuit which I've been modifying bit by bit.

I was hoping to be able to detect the approach of charged thunderclouds overhead as a warning that a storm was approaching. At present it is not sensitive enough for that, maybe because it is right next to my building, but it does appear to detect the daily variation in atmospheric charge and my graphs match published graphs of diurnal atmospheric charge.

To monitor the data automatically I bought a Radio Shack digital multimeter, catalog number 22-812, for $50 which comes with a serial cable and software. It can run for up to 5 days sampling once per second and it graphs the data continuously. I save the graphs with a screen-print program.

I'm currently working on what should be a more sensitive detector and have several schematics of those circuits also. They are also simple.

I don't know if we're allowed to give out our email addresses here so we can communicate directly.

Trent