Looking for a device to characterize JFETs

[MRiddickW]

Hi all,

Can anyone recommend me a curve tracer (or parameter analyzer?) that can handle a JFET with very high input impedance? In particular, I'm trying to characterize some 2N4118A JFETs.

I don't exactly know all the specs I need to test for, but I think Vgs(off) may be the key parameter we need to know. And maybe gain as well? We're getting into "blind man looking for black cat in dark room" territory.

Cheaper is better of course, but I'm expecting to spend around $3000; up to $5000 might not be out of the question, I'd have to check.

I have an Atlas DCA75 which has been helpful for other transistors, but the 2N4118A just registers as a diode junction. Searching around the forum, I found a recommendation for the SysComp CTR-201, but I can't find it for sale anywhere (discontinued?).

Test Equipment Depot has a Tektronix 577D1 that is the right price, but I'm not sure if it will do what we need.

I'm sure I'm forgetting a key detail or misunderstanding something or something else, so please let me know any questions or corrections you have!

Thanks!

Edited to make title more accurate.

[Kleinstein]

For just the VGoff one does not need a full curvetracer:  a simple circuit an the breadboard and a dvm is all it takes. The tricky part is the gate current and off leakage. This needs a way to measure rather small currents (e.g. 1 pA range), which is tricky with an universal instrument that also is made to handle much higher currents.

JFets show quite some scattering, but one usually does not need to measure "gain", which would be the forward transconductance. At a give current to set the working point the transconductance does not scatter that much and most circuits should be build to work with quite some scattering.

Anyway the 2N4118 would likely be used as a switch, not so much as an amplifier.

[Ground_Loop]

Most curve tracers have a voltage stepper and a current stepper for FETs and bipolars respectively.  I have an ancient Heathkit that works with FETs, J and MOS, just fine.

[MRiddickW]

JFets show quite some scattering, but one usually does not need to measure "gain", which would be the forward transconductance. At a give current to set the working point the transconductance does not scatter that much and most circuits should be build to work with quite some scattering.

Anyway the 2N4118 would likely be used as a switch, not so much as an amplifier.

Thanks for the input!

We are rebuilding some boards that use a 2N4118A. You're right it is being used more as a switch, but it's a stupid-high impedance analog input, so I think the transconductance is still a relevant factor. As in, so high that the typical 10 MOhm resistance of a Fluke multimeter is enough to pull the voltage to 0. I know it's more than 40 MOhm because that's what my meter's resistance setting tops out at. So maybe a detour into higher-impedance voltmeters is in order.

For just the VGoff one does not need a full curvetracer:  a simple circuit an the breadboard and a dvm is all it takes. The tricky part is the gate current and off leakage. This needs a way to measure rather small currents (e.g. 1 pA range), which is tricky with an universal instrument that also is made to handle much higher currents.

In this forum post, someone mentions that what you'd want is a dual-channel SMU that can interface with a computer. Would something like a Keithley 2604B do the trick do you think? I suspect we couldn't find even a refurbished one in our budget, but at least for wrapping my head around the problem, it might be a good example.

[RoGeorge]

JFets show quite some scattering, but one usually does not need to measure "gain", which would be the forward transconductance. At a give current to set the working point the transconductance does not scatter that much and most circuits should be build to work with quite some scattering.

Anyway the 2N4118 would likely be used as a switch, not so much as an amplifier.

Thanks for the input!

We are rebuilding some boards that use a 2N4118A. You're right it is being used more as a switch, but it's a stupid-high impedance analog input, so I think the transconductance is still a relevant factor. As in, so high that the typical 10 MOhm resistance of a Fluke multimeter is enough to pull the voltage to 0. I know it's more than 40 MOhm because that's what my meter's resistance setting tops out at. So maybe a detour into higher-impedance voltmeters is in order.

For just the VGoff one does not need a full curvetracer:  a simple circuit an the breadboard and a dvm is all it takes. The tricky part is the gate current and off leakage. This needs a way to measure rather small currents (e.g. 1 pA range), which is tricky with an universal instrument that also is made to handle much higher currents.

In this forum post, someone mentions that what you'd want is a dual-channel SMU that can interface with a computer. Would something like a Keithley 2604B do the trick do you think? I suspect we couldn't find even a refurbished one in our budget, but at least for wrapping my head around the problem, it might be a good example.

Looking for an advice, or for validation?

- If you expect confirmation, sorry, your assumptions about measuring those FETs are not correct.
- If you are looking for advice, Kleinstein already told what's the best thing to do.  Ask for detail if it's not clear how to measure without a curve tracer.
- If you are looking to buy an instrument, that's fine.  Not that would be justified for the job, unless you have to measure a hundred transistors a day, or so.

[egonotto]

Hello,

perhaps i don't understand your problem.
But I think with a two channel Arbiträr Waveform Generator you can build a curve tracer.

Example: https://www.instructables.com/Semiconductor-Curve-Tracer-With-the-Analog-Discove/

Best regards
egonotto

[KE5FX]

Can you replace it with a MOSFET?  Based on what you're saying, that sounds like the right part for the job.

The trouble with measuring a JFET is that you've only measured one JFET.  They are all over the place.  JFET specs should be taken as vague suggestions, never mind what you get when you move from one manufacturer to another.

[MRiddickW]

Can you replace it with a MOSFET?  Based on what you're saying, that sounds like the right part for the job.

The trouble with measuring a JFET is that you've only measured one JFET.  They are all over the place.  JFET specs should be taken as vague suggestions, never mind what you get when you move from one manufacturer to another.

It would be nice to replace it with a MOSFET, but I don't think that's an option for "reasons." 

By one JFET do you mean one type of JFET (4118A), or one physical JFET? We've rebuilt 24 of these units, two of them work, and we bought plenty of components for rebuild/troubleshoot, so we have JFETs out the wazoo.

We *think* the 4118A is the problem child, so we want to characterize the two that work against several that don't and see if there's anything different about them that we might could order custom binning for.

. . .I think with a two channel Arbiträr Waveform Generator you can build a curve tracer.

Example: https://www.instructables.com/Semiconductor-Curve-Tracer-With-the-Analog-Discove/

Thank you for this! That's a good (and cheap) place to start if nothing else.

Looking for an advice, or for validation?

- If you expect confirmation, sorry, your assumptions about measuring those FETs are not correct.
- If you are looking for advice, Kleinstein already told what's the best thing to do.  Ask for detail if it's not clear how to measure without a curve tracer.
- If you are looking to buy an instrument, that's fine.  Not that would be justified for the job, unless you have to measure a hundred transistors a day, or so.

Who doesn't want validation?

- There were three things I implied a request for confirmation on, so specificity on which assumption (or all of them) I'm wrong about is appreciated.
- I got advice on measuring Rdsoff, and appreciate it. If gate current and off leakage are trickier to measure due to the high precision, that's what prompted my question about the SMU with 100 fA to 10 A, 100 nV to 40 V range (And I should have been more specific about why I brought up the Keithley).
- Not 100 per day, but 25-50 for the project depending on how masochistic I'm feeling.

[RoGeorge]

I apologize for being judgemental instead of staying on topic. 



About FET testing with a curve tracer:
- a curve tracer would make sense as didactic instruments, or maybe to pair some components, yet it doesn't add much value in the lab for day to day use, IMO.  Sure, it can be used to sort out bad transistors, too, but a curve tracer will be an overkill.
- a curve tracer doesn't test leakage.  In fact the stairs voltage that will drive the gate will shunt the gate with a much, much less impedance than the gate of any FET.
- the traced curves for FET varies a lot, even for two transistors from the same batch.  Also varies a lot with temperature, so they certainly won't look like the ones shown in the datasheet https://eu.mouser.com/datasheet/2/676/jfet-2n4117-2n4118-2n4119a-interfet.r00-1649084.pdf the datasheet curves are for reference only.



How to measure a FET in general, for no matter what application, might be a difficult task.  However, deciding if a FET from an old module is good/bad, would be much easier.  Usually semiconductors doesn't wear out, so it would be enough to identify the defective ones.  If not defective, measuring its parameters would help at all, since that board was working once.

It's unusual for a FET to be used as a switch and require very high impedance at the same time.  Could be, but very unlikely.  Very hard to treat all the aspects at the generic level.

To measure pA or fA precisely is far from trivial, but usually the exact value is not needed.  OTOH it's almost trivial to compare the leakage of two FETs, or to detect if a FET is not in specs any more (for example because of some invisible dirt film on the exterior.

Would be even possible to measure those pA/fA gate currents with good resolution, just that the measured value would not be calibrated.  To do that just power the DS circuit, for example with a 9V battery and a 100k series resistor, and leave the gate floating in the air.  Now if you touch the gate, it's internal capacitance will charge from the hand (or better charge the gate by connecting it to a known voltage then disconnecting the gate), then it will start to slowly discharge because of the leakage.  This can be observed as variations in the DS circuit, variations easy to measure with any multimeter.  A good FET would take a very long time to self-discharge the Cgs by leakage.



For example, to discharge a 2pF at 250fA (typical input parameters of a 2N4118) would take about 8 seconds for each volt to discharge.  Charge Vgs with, say, -5V then disconnect the gate and measure how long until the FET switches off the current through its DS circuit.  Not a precise methode, and sometimes it will take way longer than that to discharge the Cgs, but can detect defective or very dirty FETs.

Sometimes this trick can be achieved by simply connecting a multimeter (disable the autorange) between the DS like would measure resistance of DS, then play around with the floating gate.  Depending on the multimeter, this might or might not work, better use a voltage source and a resistor, and measure the Vd instead of Rds.



Very hard to consider all the aspects at the generic level.  A FET could be as well in a $1 vintage ray-gun toy or in a very sensitive electrometer.  If disclosure is possible, what are those boards to be restored?  Is there a schematic?

[MRiddickW]

Thanks, I know how old it can get to answer the same flavor of question over and over.

We're actually rebuilding smoke detectors (had to check to make sure it was okay to say, otherwise I would have said so earlier), model N12000-3S or NID-58 (different P/Ns for same device). Schematic below.

My understanding: In standby, the A241 element bombards the "collector cup" with positive ions, driving the gate high. When smoke enters the chamber, the ions are blocked, letting the gate voltage fall. As TR2 (our 2N4118A) Vgs falls, so does Rds, turning on TR3 until the voltage at Z2-K is higher than the Zener voltage, tripping a latch made up of TR4 and TR5, pulling the supply rail low, triggering the alarm (in a 2-wire configuration).

The issue we're running into is that the detectors are too sensitive; i.e. they are alarming at too low of a smoke concentration level (0.7%).

Dealing with trying to understand transistors on this level just makes my brain fold in on itself. If I ever found myself teaching an electronics course (God help us all), this would be my number one example of "why should I pay attention in this class devoted to transistors?" 

[RoGeorge]

That's great, having a schematic clarifies a lot of questions.   
Unfortunately I have no hands on experience with ionization smoke detectors.

- Does it look clean inside the sensor?
- Are all of them too sensitive, or only a few?
- Assuming they once worked correctly but now they trigger false alarms too often, did you tried cleaning the inside of the ionisation chamber?  No idea how to do that properly, but please be aware that Americium is dangerous if inhaled or ingested, in case you plan to brush the interior or use compressed air for cleaning.
- Have you tried cleaning the JFET with IPA (onboard, without desoldering it), just in case some invisible conductive film might have covered the FET over time?  AFAIK IPA is not the best chemical for very low leakage cleaning, but it is very common in electronics labs and doesn't damage plastic parts or other components.

(as a side note, the recommendation from one of the best analog engineers was to brush the dirty/leaky IC cases with soap and water to clean grease or fingerprints, and Bob Pease used to use dishwashing machine to clean very high impedance boards, I can not find the original tale, but it's mentioned often https://www.electronicspoint.com/forums/threads/static-is-not-your-friend-vacuuming-pc.13906/#post-85978 - anyways, soap was recommended for experiments in the fA range, not industrial, and the expected current from a ionization chamber is much bigger, in the range of 100 pA according to this guy:  https://youtu.be/oFUUQcpGR3k).

The idea is to provide more current to the gate.  The main radiation coming from Americium 241 is in the form of alpha particles, which are very easy to stop by dust or other kind of dirt that might have deposited over time (it is said they can be entirely stopped by a sheet of paper or by a few centimeters of air).  The lack of enough alpha particles inside the ionization chamber would mean less ions and therefore a current too small for the gate of TR2.  A current too small will trigger the alarm.

Other ways to increase the current available for the JFET's gate would be to increase the amount of radiation (for example getting the Am-241 closer to the chamber - I assume this would not be possible in practice) or to increase the voltage applied to the ionization chamber in the hope that ions would accelerate more and thus arrive in bigger number on the chamber plates (in theory should work, unless all the ions are already captured, in practice IDK).

The last thing to suspect would be the JFET "wearing out", thought it might be some FET damage with time considering that Am-241 also emit some gama radiation that might pierce through the FET and damage its silicon, but this is just a speculation and I think it's highly unlikely to happen.

Also I've read ionization type of detectors are famous for their false alarms when compared to other types.

[1audio]

Here in California you are supposed to discard detectors after 7 years. It doesn't make sense to service them. All new ones are sealed and ditched after 10 years. This must be a special application to justify even opening one up. In any cse I would be replacing the sensor since they do decay.

For the Jfet a simple fixture with a 100M or 1G resistor to the gate and an Ohmmeter source to drain. Slowly bias the gate negative through the resistor with a bench supply until the ohmmeter shows the resistance increasing. Note the voltage. You should be able to screen the Jfets pretty easily.

[David Hess]

Any curve tracer which can handle JFETs can measure Vgs(th) and transconductance which should be all that you need if you make separate gate and channel leakage measurements which will require a different instrument, like an electrometer.  I would take a very close look at gate leakage which would indicate that the part was damaged.

Besides making a test circuit for gate leakage, a common 3.5 digit voltmeter with 10 megohm input resistance and 200 milivolt range can measure current with 10 picoamp resolution, so it is not always necessary to have an electrometer or equivalent.

[MRiddickW]

Whoops, forgot to come back to this.

Something I forgot to mention is that this is more of a rebuild project, rather than a troubleshooting project. Essentially once we have a BOM that works reasonably consistently, we will replace all semiconductors and capacitors and hope for the best. Something along the lines of 70% success or better is what we're hoping for.

- Does it look clean inside the sensor?
- Are all of them too sensitive, or only a few?
- Assuming they once worked correctly but now they trigger false alarms too often, did you tried cleaning the inside of the ionisation chamber?  No idea how to do that properly, but please be aware that Americium is dangerous if inhaled or ingested, in case you plan to brush the interior or use compressed air for cleaning.

1) They all seem clean enough; I'm sure they could be better, but looking inside doesn't immediately make you think "this is super dusty."
2) They're all too sensitive.
3) We've been instructed to steer clear of the ionization chamber as much as possible, so not a whole lot we can do in this case.

Have you tried cleaning the JFET with IPA (onboard, without desoldering it), just in case some invisible conductive film might have covered the FET over time?  AFAIK IPA is not the best chemical for very low leakage cleaning, but it is very common in electronics labs and doesn't damage plastic parts or other components.

(as a side note, the recommendation from one of the best analog engineers was to brush the dirty/leaky IC cases with soap and water to clean grease or fingerprints, and Bob Pease used to use dishwashing machine to clean very high impedance boards, I can not find the original tale, but it's mentioned often https://www.electronicspoint.com/forums/threads/static-is-not-your-friend-vacuuming-pc.13906/#post-85978 - anyways, soap was recommended for experiments in the fA range, not industrial, and the expected current from a ionization chamber is much bigger, in the range of 100 pA according to this guy:  https://youtu.be/oFUUQcpGR3k).

VERY interesting. The JFET is mounted in such a way that replacing only that transistor is highly unlikely to cause any leakage (see picture below), but when I replace other parts, I've been cleaning with IPA. The boards have actually been reworked once to a BOM we made when we thought it would be a straightforward project. I know they cleaned the flux off after rework, but I'm not sure with what. I will ask about that.

This must be a special application to justify even opening one up. In any cse I would be replacing the sensor since they do decay.

Assuming the detectors are 50 years old, there would still be 92% of the Am-241 left. But in an application as sensitive as this, maybe that actually is enough to make them not work. I'm hoping we can design around it if that is the case though.

Thanks for the testing recommendations y'all, I will be back in the office tomorrow to try some of these schemes out.

[RoGeorge]

This looks less and less about curve tracers, you may want to change the title into something else related with electrometer leakage cleaning, or alike.  It is possible to edit the title of the whole thread by editing the first post.  Otherwise people experienced in low leakage currents might not be interested to click on a topic about curve tracers.

It might be just a problem of properly cleaning the old devices, and no rebuild.  I do not have proper experience about how to do that, you may want to talk with electrometer enthusiast.

There are a few topics about cleaning vintage electrometers, i.e. here https://www.eevblog.com/forum/metrology/v7-45-electrometer-made-in-belarus-(welcome-in-attoamps-world)/msg3238194/#msg3238194

[MRiddickW]

Yeah, I was starting to think that too.

I meant to post that info more to inform the context and maybe the important things I should be testing the 4118A for, rather than turning it into another discussion about leakage (though I greatly appreciate the pointers you’ve given).

So with all that in mind, if I wanted a single unit that could characterize the JFET to the extents previously discussed, what should I be shopping for? An SMU or something else?

And any particular recommendation, preferably under $5000 (used is fine)?

[mjs]

2-channel SMU is great for measuring and sorting transistors. I've got a boat anchor HP 4142B with 3 channels - it's been great for sorting and analyzing. I'm getting rid of it or changing to a smaller unit (even "high current" versions with pA resolution would be ok for my current work). I could sell it, but it's too heavy to ship.

Keithley 2636 can sometimes be found below 5k, but the lowest 1 nA range seems to have 800 fA 0.1-10 Hz noise and 2N4118 is specced close to it. But you might identify any faulty ones easily. At these currents you'll need also correct cables, connectors and a closed measurement box.