Testing N-JFET gate - source breakdown voltage (FAKE jfets!) (Successful finale)

[SilverSolder]

I have a pile of PN4391 N-JFETs on my desk, I'm trying to verify the gate-source breakdown voltage.

The spec says 1uA at 30V.

I am measuring it by grounding Drain and Source, then applying a reverse voltage to the gate (from a 1uA current source), and measuring the current with an electrometer.

The problem is that barely any current flows at all... even at more than 100V compliance voltage, the current is in the low picoamps region - barely measurable...  I maxed out the current source at 120V, but the magic smoke resisted all attempts to get it to budge.

Am I doing something fundamentally wrong, or are the DUTs simply blowing away their specifications?  - made with newer processes that blows away older specs?  ... confused....

[David Hess]

That seems weird to me also.  Try running the same test on a 2N3904 base-collector junction to make sure that your test setup is working.

[duak]

Any idea when the parts were made?  I remember in the late 80s one of my colleagues mentioning that his friend at National Semiconductor said that because the Japanese, and later, most everyone else cleaned up their fabs and processes, the spreads on things like Hfe, breakdown voltages and leakage currents tightened up and really got better.  By then, virtually none of the parts in a lot would have specs in the bottom quartile.  The JEDEC spec for that FET was probably written in the 60's to reflect what could be reasonably delivered and couldn't be changed.  Things have got a lot better.  I remember getting a failure rate  of between 0.1 and 1 %  for US made RAM chips.  When we switched to Japanese, the failure rate essentially went to zero, they were cheaper and used less power.

When you made the leakage current test at 30 V, did you try increasing the temperature to say, 125 C?

Cheers,

[SilverSolder]

To avoid going insane, I stopped messing around with sensitive instruments and instead grabbed a handheld meter with a diode test function...  and found 2 diodes, with both cathodes on the pin that's supposed to be the Source!  ...which is the wrong way round for an N-Fet even if the cathodes were connected to the Gate!

Looks like I've got a batch of fake components - they are clearly marked as PN4391 JFETs (and were sold as such), but they behave more like bipolar transistors...  or something!  eBay...  gotta love some of the enterprising folks you find there.

[Kleinstein]

One diode from gate to source is normal. With an open gate the JFET can behave odd and might fool you in thinking there is a diode.

A good test to see if it is a JFET base current sink circuit:  Gate with a protective resistor to GND, source to GND with a resistor (e.g. 1-10 K) and than measure the drain current from a voltage source of some 12 V and the voltage over the source resistor.

The gate currents are usually maximum values and sometimes older typical currents. Newer parts may be better in the typical values.

[SilverSolder]

These transistors were definitely all 100% fakes...   check this out:

This is what the transistors look like -



And this is what they show up as on one of those cheapy component testers:




Seller refunded instantly when I complained, without even answering.   Caveat emptor!

[OwO]

NOTE: This message has been deleted by the forum moderator Simon for being against the forum rules and/or at the discretion of the moderator as being in the best interests of the forum community and the nature of the thread.
If you believe this to be in error, please contact the moderator involved.
An optional additional explanation is:

[Buriedcode]

I recently was in a similar situation, but with J201's:  https://www.eevblog.com/forum/projects/fake-j201-jfets/

Both the component tester (similar to the one in your pic) and my multimeter (on diode measurement) showed an NPN with a reasonable Vbe and gain.  In fact, they were within 5% of several 2N3904's I had tested.  I was also called out on "crying fake" without proper testing, despite it being rather common for J201s bought on fleabay to be restamped bipolar NPN's mentioned in DIY guitar effects forums. 

However... I did actually bother to do other tests, such as Idss, and Vgs, and concluded they were in fact N channel JFET's, also fell within the extremely wide specs in the datasheet.  Ergo: most likely J201's or N JFETs with very similar specs (the J201 is quite low Idss and Vgs when compared to most other, so stands out).


According o the datasheet for the PN4391, the Idss is between 50 and 150mA, which is huge.  A Vgs cutoff of between -4 and -10V - again quite large.  This would be easy to test - just connect gate and source together (pins 2 & 3) and measure the current through drain to source/gate.  Depending on the pinout, if it is a bipolar, it'll read either nothing (E and B connected), or enough current to destroy it (C and B connected).  If it's a N JFET but perhaps a different part number, it'll show a much lower current.  If you get between 50 and 150mA - it's in spec.

Second test, maybe easier is Vgs.  Gate to ground, Drain to +9V, and put your voltmeter between source and ground.  Negate the voltage to get the Vgs.

[RoGeorge]

Connect a DMM between DS, as you would measure a normal resistor.
Leave the gate floating in the air.

- First, shortly touch the metal of one of the DMM's probe, in order to get your body and the DMM at the same potential.
- Now, shortly touch the gate then leave it floating again.  The parasitic gate capacitance will charge from the surrounding humm voltage induced by the finger.  The DMM will now show a random resistor value, but the value will remain stable for many seconds, if not minutes.
- If you shortly touch the gate again, the DMM will show some other random value, but again, the displayed value will remain stable for a long time while the gate is floating.

In my experience, the gate of a JFET can be so sensitive that it could feel the hand and/or body moving a few steps away from the gate (and show that as a change in the resistance value measured by the DMM).

If the value displayed by the DMM does not hold while the gate is left floating, than it's not a FET.

[RandallMcRee]

As mentioned above you can not trust the readings of this component tester on a component it wasn't designed to test. Don't jump to conclusions.

Hmmm, my cheap component tester reliably identified JFETS versus transistors.
Have tested lots of jfets and never seen it misidentify anything.

[spec]

I have a pile of PN4391 N-JFETs on my desk, I'm trying to verify the gate-source breakdown voltage.

The spec says 1uA at 30V.

I am measuring it by grounding Drain and Source, then applying a reverse voltage to the gate (from a 1uA current source), and measuring the current with an electrometer.

The problem is that barely any current flows at all... even at more than 100V compliance voltage, the current is in the low picoamps region - barely measurable...  I maxed out the current source at 120V, but the magic smoke resisted all attempts to get it to budge.

Am I doing something fundamentally wrong, or are the DUTs simply blowing away their specifications?  - made with newer processes that blows away older specs?  ... confused....

If that were a bipolar transistor the base/emitter junction would have broken down at between 5V and 20V (normally 7V). The base/emitter junction behaves like a low-power zener diode. In fact, there are circuits that use this characteristic.

When datasheet gives a beak-down voltage, it is a minimum break-down voltage and the fact that the gate source did not break down could indicate that you have very good JFETs.

A JFET looking into the gate is exactly like looking into the base of a  bipolar transistor. There is a diode between the gate and source and the gate and drain, just as there is a diode between base and emitter and between base and collector.

The only way to test a JFET, if you don't have a special JFET tester, is to build a simple circuit which you can run off a 9V battery.

I am not saying that your JFETs are not fakes, just that they may not be.

[edavid]

Hmmm, my cheap component tester reliably identified JFETS versus transistors.
Have tested lots of jfets and never seen it misidentify anything.

It depends on what version of the firmware you have.  Markus worked on improving the JFET algorithm a couple of years ago... before that it was hit or miss with low current parts.

I don't think high I_DSS and high V_P JFETs can be tested reliably with the hardware available.

[Kleinstein]

JFETs with very low and rather high threshold could be misidentified with the simple testers. The low voltage ones could behave remotely similar to an NPN, while the high threshold ones could be similar to a diode with D and S connected.

I would be surprised if the PN4391 (4-10 V threshold) is mistaken for an NPN. Due to the missing brake-down - it's also unlikely to be an NPN. The analysis from user spec  is quite conclusive in this point.  So it may sill be a JFET of a different type (e.g. PN4393).  With the low leakage possibly really good ones.

[spec]

+ SilverSolder

Below is a little test circuit for your NJFETS- nothing fancy but it will tell you if the NJFETs are in fact NJFETs and if they are working,

The circle marked A is a multimeter set to the mA range.

Turn the pot to the fully negative line, connect the NJFET, and then connect the batteries, lower first.

The ammeter should indicate no current or a low current.

Turn the pot to make the gate less negative: the ammeter current reading should increase.

And that is it.

https://www.mouser.co.uk/datasheet/2/308/PN4391-1301540.pdf   (surface mount version, which I believe is electrically similar to the PN4391 leaded version that you have)

[SilverSolder]

The component tester only came out to play after all the usual tests had been tried, including the circuit provided by @Spec above.  Subsequent tests with a DMM with a diode test function shows conclusively where the junctions are located on the device:



@RoGeorge:  The (fake) "Drain"-"Source" channel has infinite resistance until you exceed about 8.2 volts, where it acts like a zener diode.  If you reverse the current, it acts like a regular diode with a drop of about 0.65V

@spec:  The 120V was across the (fake) "Gate" - "Source" junction (I was trying to measure gate leakage) which is actually the Collector-Base junction, not the EB junction.  It looks like it was able to withstand 120V.


I buy stuff from China all the time - thousands of products a year.  99.9% of the time I am a happy customer where Chinese engineers and workers have done a good job. This crook isn't going to change my mind about that.

Looks like seller has other unhappy customers and should really be banned from selling on eBay with this track record:

[SilverSolder]

I did manage to source some genuine PN4391 N-Jfets.   These behave as expected in the test setup, blocking the drain-source current when a negative gate-source voltage is applied.  I haven't verified the breakdown voltage yet, that's the next step now that reality and expectations are in perfect harmony!

Interestingly, these JFETs show up as two diodes on the cheapy component testers.   I guess they "don't dare" to apply enough negative gate voltage to detect these devices (the cut-off is almost -15V).

The newer model does at least correctly indicate the Drain-Source resistance, which identifies the device adequately as a Jfet.  The older model is not able to tell it apart from a double diode.

Older model first:



Newer model:

[edavid]

Interestingly, these JFETs show up as two diodes on the cheapy component testers.   I guess they "don't dare" to apply enough negative gate voltage to detect these devices (the cut-off is almost -15V).

The V_P spec for the PN4391 is -4 to -10V.

The component tester has a 5V supply voltage, so it's not a matter of "don't dare", it's just impossible.

[Zero999]

Get an expensive component tester. Mine has a small 12V battery and will no doubt test higher pinch-off voltage J-FETs with no problem, although I've not tried it, as I don't have any.

[SilverSolder]

The V_P spec for the PN4391 is -4 to -10V.

The component tester has a 5V supply voltage, so it's not a matter of "don't dare", it's just impossible.

You might be able to do something with a capacitive voltage doubler?  Doesn't need to make the device much more complicated or expensive, or so it seems to me.

Get an expensive component tester. Mine has a small 12V battery and will no doubt test higher pinch-off voltage J-FETs with no problem, although I've not tried it, as I don't have any.

That sounds interesting - these things certainly are very convenient, for sure.  What component tester do you have there?

[Zero999]

That sounds interesting - these things certainly are very convenient, for sure.  What component tester do you have there?

It's the Peak DCA55. It uses an MN21 12V battery, but I don't know whether it actually puts out that higher voltage. For all I know, it could all run off an internal 5V regulator.

[David Hess]

The common practice is to limit the test voltage to 5 volts or lower until the device is known to prevent damage from things like base-emitter breakdown which is a destructive test.

[SilverSolder]

I downloaded the user manual for the PEAK DCA55, it seems to also limit the test voltage to 5V, so it would probably struggle with this Jfet as well. 

For the first time, I also downloaded and read (some of) the comprehensive documentation on the "cheapy" tester.

On paper, they are quite similar.   A surprisingly massive amount of work has gone into the "cheapy".  That said, the PEAK unit is cleaner and more professionally presented, wont short out against random pieces of metal on your bench (ask me how I know...).   Carrying the cheapy testers around surplus electronics stores in a ziploc bag does also get a little old after a while.  Still, they do a good job for what they are, even if they can't cope with PN4391 and similar devices.

[edavid]

On paper, they are quite similar.   A surprisingly massive amount of work has gone into the "cheapy".  That said, the PEAK unit is cleaner and more professionally presented, wont short out against random pieces of metal on your bench (ask me how I know...).   Carrying the cheapy testers around surplus electronics stores in a ziploc bag does also get a little old after a while.  Still, they do a good job for what they are, even if they can't cope with PN4391 and similar devices.

You can get a "Component Tester" in your choice of case, still for much less than a Peak tester:

https://www.aliexpress.com/item/Handheld-LCD-Digital-multimeter-Transistor-Tester-SMD-Resistance-Diode-Inductance-Capacitance-ESR-Meter/32962735947.html

https://www.aliexpress.com/item/Multi-purpose-Transistor-Tester-128-160-Diode-Thyristor-Capacitance-Resistor-Inductance-MOSFET-ESR-LCR-Meter-TFT/32812264664.html

https://www.aliexpress.com/item/Clear-Acrylic-Case-Shell-Housing-For-GM328-Transistor-Tester-Capacitance-ESR-Mega328/32779436939.html

You can also find cases for your existing testers, for a few dollars.

[SilverSolder]

I was able to measure the gate-source breakdown voltage by using a current source to drive 1uA in reverse through the gate into the N channel and measuring the compliance voltage required to do that.

The pn4391 (manufacturer: Fairchild) junctions break down at 52V at room temperature. 

The Fairchild data sheet says 30V, whereas the Central Semiconductor spec says 40V.  Looks like the 40V number is safe for the Fairchild part too, based on the evidence so far.

For sure you are not stressing it by using it at the max rated voltage of 30V, which is what I really wanted to know in the first place.

Winner, winner, chicken dinner!