JFET vs. UJT

[Stefan Heinzmann]

JFETs and UJTs seem superficially similar to me:
o They've got a resistive "channel"
o A third terminal is separated from the channel by a pn-junction
o The resistance of the channel is affected by the third terminal

Granted, there are also differences:
o The UJT channel is rather long and has resistance in the kOhm range.
o The UJT pn-junction is operated in the conducting direction
o The UJT channel is too wide to be able to pinch it off with a reverse
voltage on the third terminal

I nevertheless wonder whether there are conditions where the JFET can
exhibit a behaviour similar to UJTs when the gate diode becomes forward
biased, namely whether a region of negative resistance can show up.

 

[Ken Smith]

JFETs and UJTs seem superficially similar to me:
o They've got a resistive "channel"
o A third terminal is separated from the channel by a pn-junction
o The resistance of the channel is affected by the third terminal

Granted, there are also differences:
o The UJT channel is rather long and has resistance in the kOhm range.
o The UJT pn-junction is operated in the conducting direction
o The UJT channel is too wide to be able to pinch it off with a reverse
voltage on the third terminal

I nevertheless wonder whether there are conditions where the JFET can
exhibit a behaviour similar to UJTs when the gate diode becomes forward
biased, namely whether a region of negative resistance can show up.

Yes a JFET works as a UJT ok. The UJT doesn't generally make a JFET thats
any good.

 

[Rene Tschaggelar]

JFETs and UJTs seem superficially similar to me:
o They've got a resistive "channel"
o A third terminal is separated from the channel by a pn-junction
o The resistance of the channel is affected by the third terminal

IMO, the gate of a FET is isolated, there is no diode.
No ?


Rene

 

[Jim Thompson]

IMO, the gate of a FET is isolated, there is no diode.
No ?


Rene

The "J" in JFET means "junction".

...Jim Thompson

 

[John Popelish]

IMO, the gate of a FET is isolated, there is no diode.
No ?

The J in JFET refers to the diode junction that is normally reverse
biased to isolate the gate. If it is forward biased, the isolation
disappears.

 

[John Larkin]

JFETs and UJTs seem superficially similar to me:
o They've got a resistive "channel"
o A third terminal is separated from the channel by a pn-junction
o The resistance of the channel is affected by the third terminal

Granted, there are also differences:
o The UJT channel is rather long and has resistance in the kOhm range.
o The UJT pn-junction is operated in the conducting direction
o The UJT channel is too wide to be able to pinch it off with a reverse
voltage on the third terminal

I nevertheless wonder whether there are conditions where the JFET can
exhibit a behaviour similar to UJTs when the gate diode becomes forward
biased, namely whether a region of negative resistance can show up.

I've seen something like UJT behavior in PHEMT gaasfets, but *very*
fast. I haven't had a chance to explore it much yet; seems it might be
useful for something.

John

 

[Rene Tschaggelar]

The J in JFET refers to the diode junction that is normally reverse
biased to isolate the gate. If it is forward biased, the isolation
disappears.

Thanks Jim and John.
I forgot about those self conducting parts and had mosfets in mind.

Rene

 

[JeffM]

Yes a JFET works as a UJT ok.

Ken Smith

Reliably enough to have been used in one of your designs?

 

[Rick]

Thanks Jim and John.
I forgot about those self conducting parts and had mosfets in mind.

Rene

There are depletion mode mosfets too. Not as common as enhancement mode, but
well...where was I going with this?

Oh, it sounded like you were saying all mosfets were not self conducting.

 

[Ken Smith]

Reliably enough to have been used in one of your designs?

Its not something I would try to put into production. The peak point
voltage isn't specified so the manufacturer can change it all over the
place and still be in spec.

If I made one, I'de expect that one to continue to work forever.

 

[Ron Harrison]

Yes a JFET works as a UJT ok. The UJT doesn't generally make a JFET thats
any good.

--

I don't think so. Can you point to any evidence of this?

 

[Stefan Heinzmann]

Reliably enough to have been used in one of your designs?

Its not something I would try to put into production. The peak point
voltage isn't specified so the manufacturer can change it all over the
place and still be in spec.

If I made one, I'de expect that one to continue to work forever.[/QUOTE]

I decided to actually try this out. I tried 3 different n-channel JFETs
in a curve tracer and I am unable to find a region of negative
resistance. What are the parts you using and how do you bias them?

I tried the following FETs (all are symmetric as far as I can tell -
i.e. drain and source are interchangeable): J113, J309, BF245A

I set up the curve tracer to display the diode curve of the gate-source
diode while modifying the bias voltage applied to the drain. That would
correspond to the E-B1 diode curve of a UJT while modifying the B2 voltage.

With a UJT the curve is essentially a diode curve with Vb2=0. At
Vb2=500mV the knee is very sharp, but there's no negative resistance
yet. Higher voltages at B2 lead to a curve with negative resistance.

With the JFETs I tried, the curve at Vd=0 is a diode curve, as expected.
Raising Vd to 500mV shifts the diode curve by about 50mV (increases the
forward voltage). Raising Vd even more has little further effect. No
negative resistance zone results.

I figure that in order to get UJT behaviour, I would need a JFET that
shows a more pronounced curve shift when raising Vd. I don't know
whether such a device exists, as this behaviour would probably be rather
undesirable in normal operation of a JFET.

 

[John Popelish]

I decided to actually try this out. I tried 3 different n-channel JFETs
in a curve tracer and I am unable to find a region of negative
resistance. What are the parts you using and how do you bias them?

I tried the following FETs (all are symmetric as far as I can tell -
i.e. drain and source are interchangeable): J113, J309, BF245A

I set up the curve tracer to display the diode curve of the gate-source
diode while modifying the bias voltage applied to the drain. That would
correspond to the E-B1 diode curve of a UJT while modifying the B2 voltage.

With a UJT the curve is essentially a diode curve with Vb2=0. At
Vb2=500mV the knee is very sharp, but there's no negative resistance
yet. Higher voltages at B2 lead to a curve with negative resistance.

With the JFETs I tried, the curve at Vd=0 is a diode curve, as expected.
Raising Vd to 500mV shifts the diode curve by about 50mV (increases the
forward voltage). Raising Vd even more has little further effect. No
negative resistance zone results.

I figure that in order to get UJT behaviour, I would need a JFET that
shows a more pronounced curve shift when raising Vd. I don't know
whether such a device exists, as this behaviour would probably be rather
undesirable in normal operation of a JFET.

Thanks for this data. I have been reading this thread with
considerable skepticism, and your results confirm it. For a UJT to
function, there has to be a significant fraction of the total channel
voltage drop between the nearest point of contact of the junction and
the source connection (described in JFET terms), because it is the
reduction in that drop when minority carriers are injected that causes
the negative resistance. In any JFET die I have seen a picture of,
the channel is wide and short after it exits out from under the gate
junction. Your measurements confirm this sort of construction.

 

[Stefan Heinzmann]

Stefan Heinzmann wrote:




Thanks for this data. I have been reading this thread with
considerable skepticism, and your results confirm it. For a UJT to
function, there has to be a significant fraction of the total channel
voltage drop between the nearest point of contact of the junction and
the source connection (described in JFET terms), because it is the
reduction in that drop when minority carriers are injected that causes
the negative resistance. In any JFET die I have seen a picture of,
the channel is wide and short after it exits out from under the gate
junction. Your measurements confirm this sort of construction.

I still think that there might be JFETs that exhibit a UJT-like
characteristic. I measured 3 examples that didn't (the ones I found in
my box of bits), but that doesn't show it can't happen. I'd still be
interested in hearing from Ken what his setup was.

The curve shift I saw tells me that there is /some/ effect, although it
doesn't go as far as producing a region of negative resistance. This in
itself isn't surprising, as there must be some resistance between the
source contact and the nearest junction point. I would expect the effect
to be most pronounced for FETs with very small geometries, i.e. with
very small gate capacitances, as the source-to-junction distance would
matter more in this case. Hence the J113 should be less susceptible than
the other two chips. I measured again, and the result was a 40mV shift
for the J113, 50mV for the J309 and 60mV for the BF245A. I can't say
whether that's significant enough to confirm my theory.

If my theory is right I would expect the effect to be more noticeable in
FETs inside IC's. Maybe someone from the IC designer fraction can
comment on this.

 

[John Larkin]

I still think that there might be JFETs that exhibit a UJT-like
characteristic. I measured 3 examples that didn't (the ones I found in
my box of bits), but that doesn't show it can't happen. I'd still be
interested in hearing from Ken what his setup was.

The curve shift I saw tells me that there is /some/ effect, although it
doesn't go as far as producing a region of negative resistance. This in
itself isn't surprising, as there must be some resistance between the
source contact and the nearest junction point. I would expect the effect
to be most pronounced for FETs with very small geometries, i.e. with
very small gate capacitances, as the source-to-junction distance would
matter more in this case. Hence the J113 should be less susceptible than
the other two chips. I measured again, and the result was a 40mV shift
for the J113, 50mV for the J309 and 60mV for the BF245A. I can't say
whether that's significant enough to confirm my theory.

If my theory is right I would expect the effect to be more noticeable in
FETs inside IC's. Maybe someone from the IC designer fraction can
comment on this.

If it did happen, you might expect jfet source followers to oscillate
in relaxation mode.

John

 

[Stefan Heinzmann]

If it did happen, you might expect jfet source followers to oscillate
in relaxation mode.

Why? Only if the gate became biased into forward conduction, I'd expect.

 

[John Larkin]

Why? Only if the gate became biased into forward conduction, I'd expect.

Hmmm, yes, I suppose so. And the jfet depletion-mode conduction
generally prevents that from happening.

Maybe the doping is all wrong for a jfet to do the ujt thing.

John

 

[Ken Smith]

[...]
I decided to actually try this out. I tried 3 different n-channel JFETs
in a curve tracer and I am unable to find a region of negative
resistance. What are the parts you using and how do you bias them?

I think I may have to pretend I never said it would work.

I was told it worked and confirmed it once. Its not like I did this a
lot. The JFET I used to confirm it was, I believe, a 2N4867. I just
tried several fets from the junk box but no 2N4867s. None actually
worked.

[...]

I figure that in order to get UJT behaviour, I would need a JFET that
shows a more pronounced curve shift when raising Vd. I don't know
whether such a device exists, as this behaviour would probably be rather
undesirable in normal operation of a JFET.

Yes it would be a bad property for them to have so this may explain why
none today do it but that I was told and confirmed it many years ago.

Live and learn.

 

[Stefan Heinzmann]

Yes a JFET works as a UJT ok.

[...]

I decided to actually try this out. I tried 3 different n-channel JFETs
in a curve tracer and I am unable to find a region of negative
resistance. What are the parts you using and how do you bias them?

I think I may have to pretend I never said it would work.

I was told it worked and confirmed it once. Its not like I did this a
lot. The JFET I used to confirm it was, I believe, a 2N4867. I just
tried several fets from the junk box but no 2N4867s. None actually
worked.

[...]

I figure that in order to get UJT behaviour, I would need a JFET that
shows a more pronounced curve shift when raising Vd. I don't know
whether such a device exists, as this behaviour would probably be rather
undesirable in normal operation of a JFET.

Yes it would be a bad property for them to have so this may explain why
none today do it but that I was told and confirmed it many years ago.

Live and learn.

Indeed. I shall try to get a 2N4867 when I have the opportunity. The
InterFET databook doesn't show any particular geometry for the part that
would explain why it would work as a UJT, but you never know until you
try...

Thanks for your info, anyway!

 

[Ken Smith]

Indeed. I shall try to get a 2N4867 when I have the opportunity. The
InterFET databook doesn't show any particular geometry for the part that
would explain why it would work as a UJT, but you never know until you
try...

WARNING! WARNING! WARNING!

The Interfet 2N4867 is not really a 2N4867. The 2N4867 was made by
Siliconix many years ago. Somehow in about 3 steps they lost the art of
making them. I think the guy who knew when to kick the machine quit.

After a talk with the nice folks at Interfet they agreed to do a sort on
one of their production JFETs to pick out the ones that performed like the
2N4867 for me. Obviously if "one dam fool" wants 2N4867s there may be
another so they put it in their data book.