Constant current diodes

[kalel]

I've only just learned that these exist, so I would like to ask if there's any interesting video that discusses their uses? Maybe there's a thread, or discussion that provides a simplified description of where one might use such diodes.

Below is what I did find:

I have found a datasheet:
https://shop.micross.com/pdf/LSM_J508_TO-92.pdf

It discusses:

J508 Applications:
• Constant-Current Supply
• Current-Limiting
• Timing Circuits

And currents such as 2mA, with an interesting curve (straight line for a large part).

Another model (http://www.datasheetq.com/datasheet-download/882045/1/Unspecified1/2H1002A4?lang=en) mentions "17-40mA 100V" and shows an application for LEDs:

That one is not something I would experiment with, but it's an interestingly simple schematic for LEDs working on such a voltage.

[Keicar]

I've used them to substitute for the load resistor on an emitter follower - where they work well at lowering the distortion. Unless you really need the simplicity of a two-terminal device, though - I'd generally avoid them, as they're hard to come by, and a bit expensive.

I discuss them a bit in my video on JFET current sources (which is what they are internally btw):

https://youtu.be/lll4Ccb-o94

Cheers,

Karl.

[Benta]

You can make them yourself using a JFET with the gate shorted to source. Current spread is quite high, though.

[rdl]

You can make them yourself using a JFET with the gate shorted to source. Current spread is quite high, though.

Yeah, I experimented with doing that a few years back. The "constant current" is not all that constant. I was able to do better with a couple of little npn transistors and some resistors.

[kalel]

The cheapest I could find so far is about $2-3 for 10 TO-252 2H1002A4's.
Supposedly up to 40 mA.

Not through hole, but I guess one can solder some wires or pins or use breakout boards on anything if needed for experimenting.

[digsys]

Long ago, they were available with currents up to 10A. We often used them in bridge format for limiting the inrush currents to huge capacitor banks.
In these cases they were used as current limit, not constant current. Not sure why they stopped making the higher current ones.
There is a application guide on how to convert a say 100mA to 5-10A using a pass element, but that defeated the simplicity.

[mariush]

For led lightning purposes, there are led drives which behave like current limiting diodes, see for example : http://uk.farnell.com/on-semiconductor/nsi45020t1g/led-driver-sod-123/dp/2535067

They work as long as voltage between anode and cathode is less than 45v so with enough leds in series after it, they could be used even on 220v ac

[kalel]

For led lightning purposes, there are led drives which behave like current limiting diodes, see for example : http://uk.farnell.com/on-semiconductor/nsi45020t1g/led-driver-sod-123/dp/2535067

They work as long as voltage between anode and cathode is less than 45v so with enough leds in series after it, they could be used even on 220v ac

Nice, these 2 pin devices are interesting because of simplicity.

[Brutte]

As an alternative I'd recommend a bipolar current mirror. The smallest come in SOT-23 with 4 terminals. Of course you'd need an extra resistor to set current.

[T3sl4co1l]

The JFET version is long obsolete.

The MOSFET version is still current, in fact it's seen a bit of a resurgence in the last decade or two, with depletion mode MOSFETs becoming available with useful ratings (high voltages and high currents).  LND150 is favored among tube amp designers.  DN3545 is a little SMT that can be used for current source/sink, limiting, amplifying and switching.  IXYS has a family of power transistors for similar applications at even higher ratings.

Tim

[Benta]

Main advantage of the JFET version (whether diode or normal JFET) is that it work down to below 1 V. That's not possible with a bipolar combo.

[T3sl4co1l]

Main advantage of the JFET version (whether diode or normal JFET) is that it work down to below 1 V. That's not possible with a bipolar combo.

A BJT current mirror works down to Vce(sat), 10s of mV (though also suffers from full Early effect).  The main downside to BJTs is the need for a separate bias supply.

Tim

[IanMacdonald]

A single BJT with a selected base resistor works quite well for a one-off LED project. Minimum drop is Vce sat.

Provided your LED supply is reasonably constant, no separate bias supply needed.

The reason it's not used commercially is that transistor gains vary quite a bit so the resistor had to be hand-picked.

A small PTC in series with the base resistor will reduce the drive if the LEDs get to more than 90C or so, which is a handy protection feature in case of blocked ventilation, etc. 

[Benta]

A current mirror needs matched and tracking BJTs, way more expensive than a JFET if it's to be done right.

Concerning the single base resistor to bias a BJT, I'll beam that idea back to the 60s where it belongs (and a transistor set you back a week's pay).

[David Hess]

I agree with Keicar; they are interesting until you consider pricing and availability.

A current mirror needs matched and tracking BJTs, way more expensive than a JFET if it's to be done right.

If less compliance is acceptable, then adding emitter resistors can lower the matching requirements so that unmatched bipolar transistors may be used while maintaining high precision.

[Benta]

I agree with Keicar; they are interesting until you consider pricing and availability.

Sadly, I agree as well. Even standard JFETs are very rare these days, they seem to be semiconductor technology that's noone thinks about. Incredibly useful, though.

[David Hess]

I agree with Keicar; they are interesting until you consider pricing and availability.

Sadly, I agree as well. Even standard JFETs are very rare these days, they seem to be semiconductor technology that's noone thinks about. Incredibly useful, though.

As T3sl4co1l points out, depletion mode MOSFETs are available but they are not any less expensive than JFETs and small signal ones are not common or unavailable.  If anything, their absolute tolerances are worse.  I disagree with T3sl4co1l about JFETs being obsolete; they make excellent low bias current and low capacitance cascodes and are less expensive and perform as well if not better than the equivalent MOSFET.

Some circuits do not require absolute Idss but instead matched Vgs or Idss so a random group of JFETs may be matched against each other leaving no waste.  The LM334 and some integrated voltage regulators can be used as absolute 2 terminal current sources and TI inherited some dedicated current source ICs from Burr-Brown.  None of these solutions are as economical as a bipolar transistor current source.

[Brutte]

A current mirror needs matched and tracking BJTs, way more expensive than a JFET if it's to be done right.

Sure, that is why they make a matched and tracking pair of BJTs in a single SOT package, it is $0.1 in low quantities.

[T3sl4co1l]

As T3sl4co1l points out, depletion mode MOSFETs are available but they are not any less expensive than JFETs and small signal ones are not common or unavailable.  If anything, their absolute tolerances are worse.  I disagree with T3sl4co1l about JFETs being obsolete; they make excellent low bias current and low capacitance cascodes and are less expensive and perform as well if not better than the equivalent MOSFET.

Oh, I meant the JFET version of CRDs, just to clarify!

You're more than welcome to use JFETs (while they last; they are slowly disappearing ) or DMOS, but as you say, the tolerances aren't very good, even with source resistance.  If you can afford a trimpot (say for test equipment), that's fine; if not (say for production), you should find a better overall solution.

Tim