some questions about high precesion constant current source

Have you got a a 10R resistor in series with the gate of the MOSFET, mounted fairly close to it?

MOSFETs have a nasty habit of oscillating at a few hundred MHz - fast enough that a cheap oscilliscope won't see it - and this can invalidate a DC analysis of the circuit. In particular, the inputs of the op amp will tend to follow the peak of the AC envelope of the signal they are looking at at, rather than the average, and this can look like an out-of-spec input offset voltage.

A "gate stopper" somewhere between ten and a few hundred ohms usually kills this sort of oscillation. At Cambridge Instruments, we had enough trouble with oscillating MOSFETs that the 10R gate resistor was compulsory.

It wasn't always necessary, and sometimes 10R wasn't enough, but at least we always had space on the printed circuit board to fit a gate stopper resistor if we needed one.

------------- Bill Sloman, Nijmegen

What exactly do you mean by error of 1%? I assume you have a variable resistor somwhere in your circuit to trim the output current.

the output is shifting from 0.99uA to 1.001uA.

In my circuit,there is a variable resistor,but it is serial with

4990K,the goal is to achieve 5000K resistor.

I don't understant Bill by the 10R resistor.the voltage is 5V,so the

10R is so far.

The 10R resistor has nothing to do with the DC behaviour of the circuit

- it just discourages the MOSFET from using the inductances of the connecting wires and the stray capacitances around the circuit to make a high frequency oscillator.

If the MOSFET is busy turning itself on an off every few nanoseconds. your circuit won't behave the way you'd expect from an analysis based on there being nice stable voltages and currents all around the circuit.

Incidentally, what are you using to see the 0.99uA and 1.001uA numbers? Can you borrow a 5-digit voltmeter, and see if it isn't going from - say - 0.994uA to 1.0006uA? And how fast is moving up and down between these limits?

------------- Bill Sloman, Nijmegen

What are you using for your 5V voltage reference? (Are you sure that it is stable? .01% of 5V is 0.5mV; for comparison, an LM7805 is specced at thermal drift of 0.8mV per degree C.)

Are you seeing thermal effects in your sense resistor? (Easy for air currents to change the temperature of the resistor by .01%)

How well do you trust your meter? (Take a look at its specs - even on a good meter, you shouldn't trust the least significant digit very much.)

There's good information on designing precision devices in Art of Electronics. Getting .01% accuracy is not a matter of getting one thing right, it's a matter of getting very few things wrong.

Thank you,Bill.

The value of the current is sway from abou 0.99uA to 1.01.Such as

0.995uA,my voltmeter is 8.5-digit.

The frequency is low,about 50Hz.of cource this data is read only from the voltmeter.

Thank you Joerg. the voltage reference is AD586,its accuracy is 0.04%,and temp. shift is only 2ppm/c.

I am not familiar with pmosfet,maybe my pmosfet is not right,its type is VP0610l from Vishay.

I have a 8.5-digit voltmeter,so I am sure the value is right.

Whan can I do next?

Add the gate resistor suggested by Bill, and the inverting input- to-output feedback capacitor suggest by me and others many times here on s.e.d. (and in our book). It's imperative that you prevent even a low-level high-frequency oscillation.

--
 Thanks,
    - Win

Thanks all.The circuit is shown as below. Q1:VP0610L; Q2:2N4402; RL:the load(resistor); IC1:OPA602; R1:the resistor(5000K ohm),it is composed of 4990Kohm fixed resistor(0.1%,15ppm/c) and a 20Kohm variable resistor;

+12V,-12V:power; +7V:it is produce by INA105 and AD586(+5V voltage reference); +5.5V:to avoid the voltage of RL over this value;

In fact,my goal is 1mA/100uA/10uA/1uA when R1 is 5K/50K/500K/5000K.

By the way,the VP0610L has been obseleted from last year,would you like recommand one?

result is so bad.I don't know how to modify it.

+12V | .-. | | R1| | '-' | |-----------------------o | +12V | | + | | IC1 | | | |\\| ||-+ o-------------|-\\ ||-> Q1 | >---||-+ +7V--------------|+/ | |/| | | | - | -12V |-------------------o | | |< .-. +5.5V----| Q2 | |RL |\\ | | | '-' | | o------------------ o | === GND

Would by any chance mean 0.99uA to 1.01uA ?

From 1.000uA to 1.001uA it is only 0.1%, not 1%. Could the toggle between 1.000uA and 0.99uA explained by the fact that you meter is not able to display 0.999uA? (It should)

Anyway, a typical 5V reference isn't as precise as you desire. Go to Analog Devices and see what they offer. Be prepared to pay a hefty price since references in the sub 0.1% category are not cheap. Then keep your circuit simple. The less parts the better. If you have to use an opamp check its offset and drift specs since these will play a major role here.

Regards, Joerg

That is not so low a frequency at this current. If the current source has a resistance high enough to keep the current from varying more than 0.01% of 1 uA (0.1 nA) as the load voltage swings over just a 1 volt range, this implies a resistance of more than 10 giga ohms. A capacitance of only a half pico farad passes current peaks of more than .1 nA to a 1 volt peak to peak 50 hertz sinusoidal swing (200 volts per second peak slope). Good luck finding a P-channel fet with less than 1/2 pF drain capacitance and a wiring method that has less stray capacitance than what is left of that 1/2 pF.

+5V is the reference.+7V is produced by +12v and +5v.

On the R1,if there is some noise on +12v,it will be reduced.

I don't understand what ciao Ban mean by -5V.

thanks.

Thank you.Ban.

I don't know whether you have tested this circuit.In my opnion,so many resistors and op-amps,it is almost impossible to reach the goal. You used two op-amps to improve impdance,but the shift will be high.

If the accuracy is low,that is not problem.

During this several months,I found the result is very far from theory.

Thank you.Tony In the circuit shown before,the accuracy is infulenced by

+12V,+5V,resistor and MOSFET.If a pulldown resistor is added between the output of op-amp and -12v,the error of -12V will be included. Are you sure?

best regards.

Thank you.Ban.

I will try again and find the result.

best regards.

Thank you very much .Ban

Your last circuit is suitable for 1mA or more higher constant current source. Normally,the voltage reference is +5V or +2.5V,if the output is 1uA,R1 should be 5000K. Because of R3=R2-R1,the result is R2 is very big. After my simulation,the voltage of load should not more than 60% of Uref.

Tony,thank you. Because the goal is so precesion,something little will be important.I will try your way later.

Editing your drawing, eehinjor, showing our suggestions:

. +12V . | . R1 5k to 5M . | . +-------------------------, . | 470pF | . | ,---||---, | . | | | | . '---+-|-\\ | 220 ||--' S . | >-+-+--/\\/\\--||-> . +7V -----|+/ | ||--, VP0610L . 15k | BS250 . | | VP0106 . -12V +-----------o . | | . |< .-. . +5.5V----| Q2 | |RL . |\\ | | . | '-' . | | . o-----------o . | . === . GND

--
 Thanks,
    - Win

What you do wrong is to put the reference between gnd and +7V, you have to choose +12V as the reference terminal and have -5V from there, so you can use the reference directly on the opamp.

--
ciao Ban
Bordighera, Italy