Measuring Leakage current of a FET

[[email protected]]

Hi all,
I want to measure the gate leakage current of a JFET. The only
measuring instruments i have is an old Tektronix scope and some DMMs. I
have read about Keithley selling meters to do this job, though i want
to design something of my own.
Please suggest.
Thanks

 

[Fred Bartoli]

[email protected] a écrit :

Hi all,
I want to measure the gate leakage current of a JFET. The only
measuring instruments i have is an old Tektronix scope and some DMMs. I
have read about Keithley selling meters to do this job, though i want
to design something of my own.
Please suggest.
Thanks

No need of fancy meters. Just get some 10M/100M resistors from gate to
ground and a low bias current opamp to buffer/amplify it.
Depending on the current value you want to measure, some JFETs opamps
will do (at room temperature) or use the Nationals LMC60xx series which
will get you into fA.
Serious layout is necessary there.

 

[Phil Allison]

** Groper alert !

I want to measure the gate leakage current of a JFET. The only
measuring instruments i have is an old Tektronix scope and some DMMs. I
have read about Keithley selling meters to do this job, though i want
to design something of my own.
Please suggest.

** The gate leakage of a JFET is a very small current as the effective
resistance is many gigohms. If it were my problem to investigate and if
limited to only basic test equipment I might try using a 10nF * polystyrene
* capacitor charged to a few volts DC to apply a gate bias and operate the
JFET as a source follower.

Then, monitor the voltage drop across the resistor ( say 4.7 kohms)
connected from FET source to 0 volts and see how it slowly falls. Leakage
is then given by:

I = 1 exp-8 x dv/dt

Eg , if the source resistor voltage drops by 1 volt in 100 seconds,

I = 1 exp-8 x 1 /100

= 1 exp -10

= 100 pico amps.



BTW:

Great care with cleanliness and insulation will be needed not to spoil the
low leakage of the JFET - maybe just build the test jig circuit as a "
bird's nest " ie with no PCB and in mid air.




........ Phil

 

[Winfield Hill]

MisFat wrote...

I want to measure the gate leakage current of a JFET. The only
measuring instruments i have is an old Tektronix scope and some DMMs.
I have read about Keithley selling meters to do this job, though i
want to design something of my own.

Fred and Phil have given you good advice. The leakage
current of a JFET's gate can be very low, less than 1pA
under the right circumstances. But be careful in your
measurements, it's a strong function of drain current
and drain-to-gate voltage. We have a detailed family of
curves in AoE, page 137, showing a 2n4868 with under 1pA
for Vds under 10 volts, but rising to 2.5nA for Vdg = 30V
and Id = 0.1mA, jumping by nearly 10x to 20nA for Id = 1mA.
In fact there's a region of operation where the JFET's gate
current is nearly 5uA. So you need to know how you want to
use your JFET in your design and adjust your measurement
scheme accordingly.

You may also want to modify your circuit configuration.
For example, if you use a second JFET wired in cascode,
with its gate to the first JFET's source, so the critical
JFET always has low Vds, you can enjoy a sub-1pA leakage
at any operating voltage and current. In fact, such an
approach can mean you don't even need to take the painful
low-current measurements! But one point, you'll need to
carefully choose the cascode JFET, because it must have a
high enough Vgs to give the lower one a reasonable Vds at
your desired current. While doing this be sure to keep an
eye on the wide part-to-part Vgs variation JFETs suffer.
One working example I have found, IIRC, the low-noise
BF862 can be nicely cascoded with an mmbfJ310. That's
just another factoid you can stuff away for later use.

 

[Winfield Hill]

Winfield Hill wrote...

MisFat wrote...

Fred and Phil have given you good advice. The leakage
current of a JFET's gate can be very low, less than 1pA
under the right circumstances. But be careful in your
measurements, it's a strong function of drain current
and drain-to-gate voltage. We have a detailed family of
curves in AoE, page 137, showing a 2n4868 with under 1pA
for Vds under 10 volts, but rising to 2.5nA for Vdg = 30V
and Id = 0.1mA, jumping by nearly 10x to 20nA for Id = 1mA.

I should point out these currents are at 25C room
temperature, the current goes up dramatically with
increased temperature.

You may also want to modify your circuit configuration.
For example, if you use a second JFET wired in cascode,
with its gate to the first JFET's source, so the critical
JFET always has low Vds, you can enjoy a sub-1pA leakage
at any operating voltage and current. In fact, such an
approach can mean you don't even need to take the painful
low-current measurements! But one point, you'll need to
carefully choose the cascode JFET, because it must have a
high enough Vgs to give the lower one a reasonable Vds at
your desired current. While doing this be sure to keep an
eye on the wide part-to-part Vgs variation JFETs suffer.
One working example I have found, IIRC, the low-noise
BF862 can be nicely cascoded with an mmbfJ310. That's
just another factoid you can stuff away for later use.

That's a pmbfJ310 I've been using (i.e. made by Philips).
The batch I have in stock has Vgs of -2.25V at 3mA, and
-1.55V at 10mA (it has an Idss of about 30mA). This means
a Vds of 1.5 to 2.5 volts for the lower JFET, depending on
amplifier current. These are fine operating voltages for
my bf862 JFETs and insure very low gate-leakage currents.
My amplifier design has a 1.2nV noise level, a 1pA input
current, gives -3dB at 45MHz, with a 280V/us slew rate.
I don't know its input capacitance, but it's very low.

 

[Sven Wilhelmsson]

MisFat wrote...

Fred and Phil have given you good advice. The leakage
current of a JFET's gate can be very low, less than 1pA
under the right circumstances. But be careful in your
measurements, it's a strong function of drain current
and drain-to-gate voltage. We have a detailed family of
curves in AoE, page 137, showing a 2n4868 with under 1pA
for Vds under 10 volts, but rising to 2.5nA for Vdg = 30V
and Id = 0.1mA, jumping by nearly 10x to 20nA for Id = 1mA.
In fact there's a region of operation where the JFET's gate
current is nearly 5uA. So you need to know how you want to
use your JFET in your design and adjust your measurement
scheme accordingly.

You may also want to modify your circuit configuration.
For example, if you use a second JFET wired in cascode,
with its gate to the first JFET's source, so the critical
JFET always has low Vds, you can enjoy a sub-1pA leakage
at any operating voltage and current. In fact, such an
approach can mean you don't even need to take the painful
low-current measurements! But one point, you'll need to
carefully choose the cascode JFET, because it must have a
high enough Vgs to give the lower one a reasonable Vds at
your desired current. While doing this be sure to keep an
eye on the wide part-to-part Vgs variation JFETs suffer.
One working example I have found, IIRC, the low-noise
BF862 can be nicely cascoded with an mmbfJ310. That's
just another factoid you can stuff away for later use.

Thanks for a good piece of information!
I just want to add that it also depends on temperature,
as any diode reverse current. About 8% increase per K at 25C.

 

[[email protected]]

Thank you everyone for the great notes of knowledge.
I am going to try both the methods (by Fred and Phil).
Phil, I have seen AN47 by Jim and i assume thats what you meant by
building circuits in air.
Also, thanks Will i will try these JFETs - i have heard about BF862 as
a great choice for low noise applications.

 

[[email protected]]

Winfield Hill wrote...

I should point out these currents are at 25C room
temperature, the current goes up dramatically with
increased temperature.


That's a pmbfJ310 I've been using (i.e. made by Philips).
The batch I have in stock has Vgs of -2.25V at 3mA, and
-1.55V at 10mA (it has an Idss of about 30mA). This means
a Vds of 1.5 to 2.5 volts for the lower JFET, depending on
amplifier current. These are fine operating voltages for
my bf862 JFETs and insure very low gate-leakage currents.
My amplifier design has a 1.2nV noise level, a 1pA input
current, gives -3dB at 45MHz, with a 280V/us slew rate.
I don't know its input capacitance, but it's very low.

I for one would be interested to have a look at that amplifier circuit.
Is it a standard JFET diffamp + common emitter OTA? (ps: when is the
next AOE coming out? Alot of the transistors in the present edition are
obsolete now, same with opamps etc etc. Will it still be just the two
of you?)

A few cents worth of advice for measuring JFET leakage. The AD549 will
come in handy, its readily available, and has the lowest bias for
buffering or amplifying leakage. This is probabally obvious, but with
the other techniques mentioned above, a really low-pass filter will be
necessary. Some of the ultra-low leakage JFET transistors have a
leakage comparable to their rediculously small noise current. If you
put a capacitor on the gate of a cascoded JFET the voltage will random
walk more than it discharges.

Be interesting to hear from a semico test engineer to hear how they get
those perfectly smooth JFET leakage (and current noise) plots.

/Andrew

 

[Ken Smith]

I just want to add that it also depends on temperature,
as any diode reverse current. About 8% increase per K at 25C.

It also varies with the drain current of the FET. This means that in the
active area, the curve doesn't look like a diode or a resistor.

 

[[email protected]]

No need of fancy meters. Just get some 10M/100M resistors from gate to

ground and a low bias current opamp to buffer/amplify it.
Depending on the current value you want to measure, some JFETs opamps
will do (at room temperature) or use the Nationals LMC60xx series which
will get you into fA.
Serious layout is necessary there.

I am still awaiting my opamps to be shipped to me. But i have the
OPA637, jfet input opamp with input curret of <10pA. So i decided to
check some of the MESFETs which are supposed to be leakier than JFETs.
Here's the circuit.
http://img120.imageshack.us/img120/8167/leakagecurrentfb6.jpg
And as Phil suggested i have made the circuit in air, with no tracks on
a PCB. I measured Ig_leakage of around 1 nA with VDS = 0.6; IDS = 3.5
mA at room temperature.
One thing i have noticed is the annoying 60 Hz (100 mVp-p) at the
output of opamp and making some sort of shield to cover the circuit and
try to get rid of it.
Thanks again guys.
Will, i got the BF862 and will try your idea. thanks