jfet as current limiter

[Marco Trapanese]

Hello,

I'm playing with a n-channel jfet (2n4392) as current limiter (resistor
on source, with gate connected on the other end).

Unfortunately, the limit value depends on the resistor value but on the
jfet geometry too. Several transistors bought at the same time, will
have a cutoff voltage quite different. Too much for my application.

I need to limit the current under 30 mA, with no effect under 22-23 mA.

Is there a smart way to set the limit value independently of the
particular jfet?

The circuit must be a bipole: I cannot rely on other voltages or ground
connections.

Thanks
Marco

 

[Rocky]

Hello,

I'm playing with a n-channel jfet (2n4392) as current limiter (resistor
on source, with gate connected on the other end).

Unfortunately, the limit value depends on the resistor value but on the
jfet geometry too. Several transistors bought at the same time, will
have a cutoff voltage quite different. Too much for my application.

I need to limit the current under 30 mA, with no effect under 22-23 mA.

Is there a smart way to set the limit value independently of the
particular jfet?

The circuit must be a bipole: I cannot rely on other voltages or ground
connections.

Thanks
Marco

If you can afford a little voltage drop, use an LM317 in current limit
mode.

 

[bw]

The only solution I know if, amounts to part screening... buy a whole
bunch of JFETs, test them, and bin them. The tighter you need your
current tolerance to be, the more bins you'd need (with each bin using
its own specific resistor value, to achieve the desired current limit).

Or, you can buy "current source diodes", which are essentially
pre-binned JFETs with gate and source shorted together internally.

http://centralsemi.com/product/cld/index.aspx

 

[Jamie]

Hello,

I'm playing with a n-channel jfet (2n4392) as current limiter (resistor
on source, with gate connected on the other end).

Unfortunately, the limit value depends on the resistor value but on the
jfet geometry too. Several transistors bought at the same time, will
have a cutoff voltage quite different. Too much for my application.

I need to limit the current under 30 mA, with no effect under 22-23 mA.

Is there a smart way to set the limit value independently of the
particular jfet?

The circuit must be a bipole: I cannot rely on other voltages or ground
connections.

Thanks
Marco

below is a sim of an inline current limit circuit. I used a loop supply
in the range of 6 .. 24 V as a test voltage in the loop with variable
load currents to show how it stays with in 4..22mA or so, using
different loop supply voltages. I think I can do one with a low voltage
op-amp inline that supplies itself via the loop voltage that will give
you even better response across difference loop supplies. More on that
later..
Jamie

Version 4
SHEET 1 880 680
WIRE 160 32 48 32
WIRE 256 32 160 32
WIRE 48 48 48 32
WIRE 160 48 160 32
WIRE 48 144 48 128
WIRE 160 144 160 128
WIRE 160 144 112 144
WIRE 256 176 256 32
WIRE 112 192 112 144
WIRE 192 224 160 224
WIRE 112 288 112 256
WIRE 256 288 256 272
WIRE 256 288 112 288
WIRE 256 320 256 288
WIRE 256 320 224 320
FLAG 48 144 0
FLAG 224 384 0
SYMBOL voltage 48 32 R0
WINDOW 123 0 0 Left 2
WINDOW 39 0 0 Left 2
SYMATTR InstName V1
SYMATTR Value 12
SYMBOL load 208 320 R0
WINDOW 0 43 41 Left 2
WINDOW 3 107 130 Left 2
WINDOW 123 0 0 Left 2
WINDOW 39 48 84 Left 2
SYMATTR InstName I1
SYMATTR Value PULSE(0.100 0.004 0 .2 .2 .5 1)
SYMATTR SpiceLine load
SYMBOL npn 192 176 R0
SYMATTR InstName Q1
SYMATTR Value 2N2222
SYMBOL res 144 32 R0
SYMATTR InstName R1
SYMATTR Value 15k
SYMBOL References\\LT1634-2.5 112 224 R0
WINDOW 3 -154 3 Left 2
WINDOW 0 -50 33 Left 2
SYMATTR InstName R2
SYMBOL res 144 128 R0
SYMATTR InstName R3
SYMATTR Value 18k
TEXT -40 32 Left 2 !.tran 1

 

[Rocky]

snip<

A little bit of foldback is helpful in sharpening up the knee, which
otherwise tends to be pretty squishy due to the lowish loop gain and to
the current-dependence of Q2's V_BE.

The benefit of the LM317L is that it can withstand 40 volts, has
foldback based on power and temperature, it has a very high dynamic
impedance when in current limit and low parts count.

Cct below limits at 23mA
|----------|
--------| LM317L |------/\/\/\---+---
|----------| 51R |
| |
+--------------------+

 

[Marco Trapanese]

Il 10/10/2012 04:03, Jamie ha scritto:

below is a sim of an inline current limit circuit. I used a loop supply
in the range of 6 .. 24 V as a test voltage in the loop with variable
load currents to show how it stays with in 4..22mA or so, using
different loop supply voltages. I think I can do one with a low voltage
op-amp inline that supplies itself via the loop voltage that will give
you even better response across difference loop supplies. More on that
later..

Great!
The actual circuit should be closer to this:

Version 4
SHEET 1 880 680
WIRE 160 -128 -304 -128
WIRE 160 -96 160 -128
WIRE 160 32 160 -16
WIRE 256 32 160 32
WIRE 160 48 160 32
WIRE -304 144 -304 -128
WIRE 160 144 160 128
WIRE 160 144 112 144
WIRE 256 176 256 32
WIRE 112 192 112 144
WIRE 192 224 160 224
WIRE -304 272 -304 224
WIRE 112 288 112 256
WIRE 256 288 256 272
WIRE 256 288 112 288
WIRE 256 320 256 288
WIRE 256 320 224 320
WIRE 224 336 224 320
WIRE 224 448 224 416
FLAG 224 448 0
FLAG -304 272 0
SYMBOL npn 192 176 R0
SYMATTR InstName Q1
SYMATTR Value 2N2222
SYMBOL res 144 32 R0
SYMATTR InstName R1
SYMATTR Value 15k
SYMBOL References\\LT1634-2.5 112 224 R0
WINDOW 3 -154 3 Left 2
WINDOW 0 -50 33 Left 2
SYMATTR InstName R2
SYMBOL res 144 128 R0
SYMATTR InstName R3
SYMATTR Value 18k
SYMBOL res 208 320 R0
SYMATTR InstName R4
SYMATTR Value 200
SYMATTR SpiceLine ""
SYMBOL voltage -304 128 R0
WINDOW 123 0 0 Left 2
WINDOW 39 0 0 Left 2
SYMATTR InstName V1
SYMATTR Value 24V
SYMATTR Value2 ""
SYMATTR SpiceLine ""
SYMBOL current 160 -96 R0
WINDOW 123 0 0 Left 2
WINDOW 39 24 28 Left 2
SYMATTR InstName I1
SYMATTR Value 10m
SYMATTR Value2 ""
SYMATTR SpiceLine load
TEXT 320 -80 Left 2 !.dc I1 0 50m


and the simulation shows the curve I want to see!
I'm sorry for my ignorance, but why it not depends on the physical
characteristic of the bjt?

Thanks again!
Marco

 

[Jamie]

Il 10/10/2012 04:03, Jamie ha scritto:




Great!
The actual circuit should be closer to this:

Version 4
SHEET 1 880 680
WIRE 160 -128 -304 -128
WIRE 160 -96 160 -128
WIRE 160 32 160 -16
WIRE 256 32 160 32
WIRE 160 48 160 32
WIRE -304 144 -304 -128
WIRE 160 144 160 128
WIRE 160 144 112 144
WIRE 256 176 256 32
WIRE 112 192 112 144
WIRE 192 224 160 224
WIRE -304 272 -304 224
WIRE 112 288 112 256
WIRE 256 288 256 272
WIRE 256 288 112 288
WIRE 256 320 256 288
WIRE 256 320 224 320
WIRE 224 336 224 320
WIRE 224 448 224 416
FLAG 224 448 0
FLAG -304 272 0
SYMBOL npn 192 176 R0
SYMATTR InstName Q1
SYMATTR Value 2N2222
SYMBOL res 144 32 R0
SYMATTR InstName R1
SYMATTR Value 15k
SYMBOL References\\LT1634-2.5 112 224 R0
WINDOW 3 -154 3 Left 2
WINDOW 0 -50 33 Left 2
SYMATTR InstName R2
SYMBOL res 144 128 R0
SYMATTR InstName R3
SYMATTR Value 18k
SYMBOL res 208 320 R0
SYMATTR InstName R4
SYMATTR Value 200
SYMATTR SpiceLine ""
SYMBOL voltage -304 128 R0
WINDOW 123 0 0 Left 2
WINDOW 39 0 0 Left 2
SYMATTR InstName V1
SYMATTR Value 24V
SYMATTR Value2 ""
SYMATTR SpiceLine ""
SYMBOL current 160 -96 R0
WINDOW 123 0 0 Left 2
WINDOW 39 24 28 Left 2
SYMATTR InstName I1
SYMATTR Value 10m
SYMATTR Value2 ""
SYMATTR SpiceLine load
TEXT 320 -80 Left 2 !.dc I1 0 50m


and the simulation shows the curve I want to see!
I'm sorry for my ignorance, but why it not depends on the physical
characteristic of the bjt?

Thanks again!
Marco

Not sure as to what you mean by that last statement? The operation of
that circuit does depend on a bjt characteristics. The Vbe and
Vce(sat) combined will give much less over loss than what you would get
with a jFet or linear regulators

Vce(Sat) should be ~170mV and Vbe being around 0.7v to a sum of 0.870v
loss in this type of circuit.

R1 supplies minimum current in series with R3 to achieve ~ 20mA at
around 6 volts in the loop supply. WHen loop supply is up to 24 volts,
the 2.5V ref will clamp that and attempt to give a steady reference to
the base via R3.

bjt's work in the same manner as fets do with the exception of how
they are biased of course. With a fet, it's all about the Vgs and the
V(Th) with the Vgd that turns it into a constant current device, but the
problem his is, you normally have to get up in the Vg(th) region for it
to start working, that could be higher than the losses you wish to
sacrifice.

With bjt's, it's always current biased instead of voltage and the
voltage here starts ~0.7v and is repeatable within reason between batches.

Jamie

 

[Marco Trapanese]

Il 11/10/2012 02:21, Jamie ha scritto:

Not sure as to what you mean by that last statement? The operation of
that circuit does depend on a bjt characteristics. The Vbe and
Vce(sat) combined will give much less over loss than what you would get
with a jFet or linear regulators

Vce(Sat) should be ~170mV and Vbe being around 0.7v to a sum of 0.870v
loss in this type of circuit.

R1 supplies minimum current in series with R3 to achieve ~ 20mA at
around 6 volts in the loop supply. WHen loop supply is up to 24 volts,
the 2.5V ref will clamp that and attempt to give a steady reference to
the base via R3.

Ok, I've already got it. But...

bjt's work in the same manner as fets do with the exception of how
they are biased of course. With a fet, it's all about the Vgs and the
V(Th) with the Vgd that turns it into a constant current device, but the
problem his is, you normally have to get up in the Vg(th) region for it
to start working, that could be higher than the losses you wish to
sacrifice.

With bjt's, it's always current biased instead of voltage and the
voltage here starts ~0.7v and is repeatable within reason between batches.

....here the good news!
I'm afraid changing the bjt (batches, not models) I run again into the
same issue.
The advantage, as you pointed out, is we rely on a simple junction
threshold.

I'm building the circuit to test it in the real world

Thanks again for the explanations.

Marco

 

[[email protected]]

There is an other simple way, providing you have some stabilized vlatge around:

V unregulated
|
---
V stable \ / -> LED
| ---
100 kR1 |
| c
----------------- b NPN
e
|
///
Iled = ( (Vstable - .7) / R1) * beta

Beta controlled current? Gack! That's about the worst current source/sink
I've ever seen proposed here in SED.

 

[whit3rd]

On Thursday, October 11, 2012 2:31:03 AM UTC-7, Jan Panteltje wrote:
[ a current limiter with sharp knee in 23-30 mA range]

V unregulated
|
---
V stable \ / -> LED
| ---
100 kR1 |
| c
----------------- b NPN
e
|
///

Iled = ( (Vstable - .7) / R1) * beta

It's OK for a quick hack, but requires beta-sorting of transistors;
the given target range (23 to 30 mA) is a bit narrow for standard beta ranges
of parts. I'd not like the lifetime/aging of this, either: audio amps with
this kind of trimmed-gain have turned black on me.

Better by far, if you have negative power supply (-12V is good) would
be
V unregulated
|
---
\ / -> LED
---
|
c
GND--------- b NPN
e
|
Rlimit
|
(-12V)

Iled = (12V - 0.7V)/Rlimit
about 1% temperature variation in 0-50C range

or the nearly-equivalent current mirror solution. Both
will dissipate Vsupply times Ilimit in quiescent state, though,
and require three terminals

 

[[email protected]]

It works very well,
Bullshit.

I have used it.

I have no doubt.

And this circuit also interfaces to a micro / CMOS output, where the micro is on a stabilized supply.

Even worse.

Beta is a physical entity, and you should actually always try to use / take those those into account in your circuits.

Nonsense. Beta is about as uncontrollable as any parameter imaginable. Why
not use 20% resistors, while you're at it. It wouldn't make the circuit any
worse.

Sending transistor into saturation as an idiot and then putting some resistor in series with the LED
in case of the micro, simply costs more parts, and gains very little.

You're clearly a hacker. No one in their right mind relies on beta.

 

[[email protected]]

I've done a few beta-limited current sources, as much to annoy the
"never count on beta" people as any other reason. BCX70 and BCX71 come
in beta graded bands of about 2:1, so in some cases that's safe.

2:1 is piss-poor. Vbe is a *lot* better than that. Use what's easy.

There's too much superstition in this business: Never depend on beta.

I'll stick with that one, thanks. You really can't count on beta,
particularly across lots.

Don't create ground loops.

It's a problem at audio. ...and can be done with audio. ;-)

Don't change reference planes.

They're hard remove.

Don't cut slits in ground planes.

I'm not tempting fate with that one, either.

Ground shields at one end only.

For *long* audio cables, this is the norm. There's too much chance of ground
currents otherwise. I ran into this at the PPoE. Ground at one end wasn't
same as ground at the other and there is a diesel generator wanting to heat
whatever got inbetween. :-(

Never use the enclosure as part of a ground plane.

Ground plane? I wouldn't do that, but bolting the ground plane to it, early
and often, sure.

BNCs are only good to x hundred MHz.

How much power are you wanting to put through them? A kW? No, it's probably
not a good idea at a couple GHz.

You need a zillion bypass caps of various values.

Zillion is a bit much, but various values works. .1uF sprinkled liberally, and
a few caps tuned so their SRF matches the internal clock works.

Your religion (above) is just as strong as it is with those you mock. There
are reasons for the above but like anything, there are limits and reasons for
them.

 

[[email protected]]

You can count on a part to meet its guaranteed min and max specs. I
suspect that you usually do already. Beta is just another spec.

One that is *rarely* guaranteed. If it is, it's basically guaranteed to be
greater than one, and only under very specific circumstances. There are good
reason behind the dogma of not counting on it.

 

[Jamie]

Il 11/10/2012 02:21, Jamie ha scritto:




Ok, I've already got it. But...






...here the good news!
I'm afraid changing the bjt (batches, not models) I run again into the
same issue.
The advantage, as you pointed out, is we rely on a simple junction
threshold.

I'm building the circuit to test it in the real world

Thanks again for the explanations.

Marco

Ok go for it. I can say that I've already built and use that however,
I didn't use that 2.5v ref, I used a 5.1V diode instead, because it
seems to be easier to get. The idea of that is to allow me to use a wide
range of voltage in the loop and have the current limit cap stay with in
a window..


Jamie

 

[[email protected]]

below is a sim of an inline current limit circuit. I used a loop supply

in the range of 6 .. 24 V as a test voltage in the loop with variable

load currents to show how it stays with in 4..22mA or so, using

different loop supply voltages. I think I can do one with a low voltage

op-amp inline that supplies itself via the loop voltage that will give

you even better response across difference loop supplies. More on that

later..

Jamie



Version 4

SHEET 1 880 680

WIRE 160 32 48 32

WIRE 256 32 160 32

WIRE 48 48 48 32

WIRE 160 48 160 32

WIRE 48 144 48 128

WIRE 160 144 160 128

WIRE 160 144 112 144

WIRE 256 176 256 32

WIRE 112 192 112 144

WIRE 192 224 160 224

WIRE 112 288 112 256

WIRE 256 288 256 272

WIRE 256 288 112 288

WIRE 256 320 256 288

WIRE 256 320 224 320

FLAG 48 144 0

FLAG 224 384 0

SYMBOL voltage 48 32 R0

WINDOW 123 0 0 Left 2

WINDOW 39 0 0 Left 2

SYMATTR InstName V1

SYMATTR Value 12

SYMBOL load 208 320 R0

WINDOW 0 43 41 Left 2

WINDOW 3 107 130 Left 2

WINDOW 123 0 0 Left 2

WINDOW 39 48 84 Left 2

SYMATTR InstName I1

SYMATTR Value PULSE(0.100 0.004 0 .2 .2 .5 1)

SYMATTR SpiceLine load

SYMBOL npn 192 176 R0

SYMATTR InstName Q1

SYMATTR Value 2N2222

SYMBOL res 144 32 R0

SYMATTR InstName R1

SYMATTR Value 15k

SYMBOL References\\LT1634-2.5 112 224 R0

WINDOW 3 -154 3 Left 2

WINDOW 0 -50 33 Left 2

SYMATTR InstName R2

SYMBOL res 144 128 R0

SYMATTR InstName R3

SYMATTR Value 18k

TEXT -40 32 Left 2 !.tran 1

I'm just not getting that at all...looks more like a wish than anything else.

 

[whit3rd]

I'm playing with a n-channel jfet (2n4392) as current limiter (resistor

on source, with gate connected on the other end).

I need to limit the current under 30 mA, with no effect under 22-23 mA.

There's a classical answer, using TL431 programmable zener

(warning: bad ASCII art, use monospace font)


(+)
|
+-------+
| /
| /
R1 |
+----|
| |>
| \
| |
| |
| |
/---/ |
/ \ |
/ \-----+
--+-- R2
| |
+-------+
|
(-)

I_limit = 2.5V /R2 + (minor correction for zener bias)

which can be redone with an LM4041 for smaller Vref, using a PNP transistor...
the range of 23... 30 mA should be easy to hit with 2V or so of voltage burden.

 

[Marco Trapanese]

Il 12/10/2012 05:12, Jamie ha scritto:

Ok go for it. I can say that I've already built and use that however,
I didn't use that 2.5v ref, I used a 5.1V diode instead, because it
seems to be easier to get.

Well, one good news and one bad news.

The good one is there is almost no difference among single items, so
this issue is fixed.

The bad one is it depends too much on temperature. At room temperature
with 12k + 12k the current is limited at 25 mA. But after a couple of
seconds it rises up to 30 mA. Heating the bjt about 50 °C the current is
over 35 mA.

I need to guarantee the current never goes above 30 mA over the whole
commercial temperature range (0-70 °C).

I'm going to add a 1n4148 in series of ("of"? is it right?) the zener.

Two questions:

- do I need to add also a resistor on the emitter?
- in order to work fine I'm afraid the diode should be thermally
connected to the bjt. Otherwise they share only the ambient temperature,
but when the bjt heats due to his own current the diode doesn't
compensate at all.

Any thoughts?

Marco

 

[Rocky]

Any thoughts?

What is the maximum voltage drop you can tolerate when not in current
limit?

 

[John S]

Il 12/10/2012 05:12, Jamie ha scritto:



Well, one good news and one bad news.

The good one is there is almost no difference among single items, so
this issue is fixed.

The bad one is it depends too much on temperature. At room temperature
with 12k + 12k the current is limited at 25 mA. But after a couple of
seconds it rises up to 30 mA. Heating the bjt about 50 °C the current is
over 35 mA.

I need to guarantee the current never goes above 30 mA over the whole
commercial temperature range (0-70 °C).

I'm going to add a 1n4148 in series of ("of"? is it right?) the zener.

Two questions:

- do I need to add also a resistor on the emitter?
- in order to work fine I'm afraid the diode should be thermally
connected to the bjt. Otherwise they share only the ambient temperature,
but when the bjt heats due to his own current the diode doesn't
compensate at all.

Any thoughts?

Marco

Try it with a slight modification. Still not really good, but better.

Version 4
SHEET 1 880 680
WIRE 160 -128 -304 -128
WIRE 160 -96 160 -128
WIRE 160 32 160 -16
WIRE 256 32 160 32
WIRE 160 48 160 32
WIRE -304 144 -304 -128
WIRE 160 144 160 128
WIRE 160 144 112 144
WIRE 256 176 256 32
WIRE 112 192 112 144
WIRE 192 224 160 224
WIRE -304 272 -304 224
WIRE 112 288 112 256
WIRE 256 320 256 272
WIRE 256 320 224 320
WIRE 224 336 224 320
WIRE 224 448 224 416
FLAG 224 448 0
FLAG -304 272 0
FLAG 112 288 0
SYMBOL npn 192 176 R0
SYMATTR InstName Q1
SYMATTR Value 2N2222
SYMBOL res 144 32 R0
SYMATTR InstName R1
SYMATTR Value 12k
SYMBOL References\\LT1634-2.5 112 224 R0
WINDOW 3 -154 3 Left 2
WINDOW 0 -50 33 Left 2
SYMATTR InstName R2
SYMBOL res 144 128 R0
SYMATTR InstName R3
SYMATTR Value 1
SYMBOL res 208 320 R0
SYMATTR InstName R4
SYMATTR Value 75
SYMBOL voltage -304 128 R0
WINDOW 123 0 0 Left 2
WINDOW 39 0 0 Left 2
SYMATTR InstName V1
SYMATTR Value 24V
SYMBOL current 160 -96 R0
WINDOW 123 0 0 Left 2
WINDOW 39 24 28 Left 2
SYMATTR SpiceLine load
SYMATTR InstName I1
SYMATTR Value 10m
TEXT 320 -80 Left 2 !.dc I1 0 50m
TEXT 480 40 Left 2 !.step temp list 0 80

 

[Marco Trapanese]

Il 15/10/2012 14:49, Rocky ha scritto:

What is the maximum voltage drop you can tolerate when not in current
limit?

I don't know exactly, but I thinks a couple of V should be tolerated
without any problems. The supply voltage of the current generator is
+24V and it seems it will work until its output reaches 16-17V.

Thanks
Marco