forward leakage of a diode ...

[Leo Bodnar]

Vincent, these curves are 25 years old in case you are using them as selection tables or design references and they were taken from just one random product sample.  Super-bright LEDs were not even invented back then.  I think Bob's intention was not to compile a selection table but to illustrate his thinking process - I was hoping you'd get the book and read it.

This is amazing thread.  As a free bonus, there is some discussion about diodes here.

Leo

P.S. Could we please stop using "leakage" in relation to normal forward conduction?  It confuses the hell out of me and probably a lot of others.

[Kleinstein]

In the very low current range one might have to check / measure the parts used. Except for expensive special low leakage parts, they diodes are not tested for leakage down to the low pA range, as this is expensive (because it takes time).

LEDs might be a good option - though you need to check. The typical current will be low, but one can not be sure. If the high forward voltage is not a problem, especially blue ones would be worth a try.

Extrapolating the diode equation / curve down to very low voltages can be tricky. Quite often there are other contributions than the idea diode current you see at higher currents. Quite often there is at least a contribution with a higher (e.g. 2 ) ideality factor at low currents, that is just not relevant at the more normal currents.

[Alex Nikitin]

Here are the graphs from my measurements for the  1N4148 and BAV199 forward (50mV -500mV) and reverse (50mV -100V) currents at about 26C ambient.

Cheers

Alex

* I've measured a number of samples of each diode, as well as BAV199 diodes from five different manufacturers. The comparative curves are average for the 1N4148 and for the worst make of the BAV199. The best makes of the BAV199 have almost ten times lower forward and reverse currents.

[capt bullshot]

This is an interesting thread ...
IMO using one or more diodes in forward direction to achieve the OPAMP protection isn't the right way.
Most opamp inputs are well protected if one limits the maximum input current to some value in the 5mA ballpark. If you limit the voltage across your sensor input to say +/- 7V using back to back connected transistor BE junctions as zeners, one can achieve OPAMP protection using a simple 4k7 resistor. You might need a zener across the OPAMPs supply pins to catch that injected current, depending on your power supply switching scheme.

I don't know the exact figures of the BE zener leakage, it is believed to be very low. So I'd just consider a architecture change of the protection network.

[tszaboo]

First of all, the correct term, you are looking for is "subthreshold" leakage. It help a lot with the google-fu.
And yes, diodes are not very well defined in this region. But transistors are. You can use 2 transistors of your choice, by connecting the base to the collector, and look up the Collector Cutoff Current. That will be almost your value. It is 0.2nA-15nA for a BC546. Temperature dependent a lot of course.

[Leo Bodnar]

First of all, the correct term, you are looking for is "subthreshold" leakage. It help a lot with the google-fu.

Score!
http://www.google.co.uk/search?q=%22diode+subthreshold+leakage%22

[oldway]

BAV25 is the diode you are looking for.....it has a leakage of 0.1pA at 100 mV forward voltage.

[Alex Nikitin]

BAV25 is the diode you are looking for.....it has a leakage of 0.1pA at 100 mV forward voltage.

Could you provide a link to a data sheet on this diode (Google search draws blank)?

Cheers

Alex

[AndyC_772]

I'm struggling to see the physical mechanism by which a component might be damaged, with a 100K resistor in series with it.

IC damage is caused by local heating, when a non-trivial current flows through somewhere it shouldn't. How much current are you expecting to be able to draw through that 100K resistor under fault conditions?

[Alex Nikitin]

anyway. this is dancing around the subject. all i want to find is a diode that does not leak in forward mode.

BAV199

Cheers

Alex

[oldway]

BAV25 is the diode you are looking for.....it has a leakage of 0.1pA at 100 mV forward voltage.

Could you provide a link to a data sheet on this diode (Google search draws blank)?

Cheers

Alex

[Alex Nikitin]

BAV25 is the diode you are looking for.....it has a leakage of 0.1pA at 100 mV forward voltage.

Could you provide a link to a data sheet on this diode (Google search draws blank)?

Cheers

Alex

So it is BAV45 (obsolete from Philips and the replacement from Central Semi has nothing in the data sheet on the forward current values at these levels). About the same performance as the BAV199, which is easily available for a fraction of the price. There are more low leakage diodes available, including FJH1100 from Fairchild (very expensive, photo-sensitive and not as good as the BAV199), PAD series from Vishay (also expensive and not easily available, no specs on the forward current v voltage).

Cheers

Alex

[Alex Nikitin]

Here is a possible solution for the topic starter.  If it is possible to sacrifice a couple of uA to forward bias two diodes, the summary leakage through D1 and D2 diodes in this circuit at 0 to 1V on R2 would be under 0.1pA at 25C (typical figure).

Cheers

Alex

[oldway]

Also the diode FDH300 (very low forward leakage)
http://www.onsemi.com/PowerSolutions/product.do?id=FDH300

In SOT-23, MMBD1501/A

[David Hess]

I'm struggling to see the physical mechanism by which a component might be damaged, with a 100K resistor in series with it.

IC damage is caused by local heating, when a non-trivial current flows through somewhere it shouldn't. How much current are you expecting to be able to draw through that 100K resistor under fault conditions?

The high currents involved can also cause electromigration which is both current and temperature dependent.  CMOS gate structures may suffer from punchthrough of their gate oxide.  The parasitic thyristor in CMOS junction isolated processes may be activated crowbaring the supply if powered.  Base-emitter breakdown damages bipolar transistors when the hot carriers cause dislocations lowering hfe by increasing recombination; this is why bipolar precision operational amplifiers include back to back input diodes to protect against high differential input voltages if their base-emitter junctions are exposed. (1)

Just a series resistor may not be sufficient if its breakdown voltage is exceeded.  For the same reason, creepage and clearance matter.  In the past they used carbon composition resistors for their good short term overload capability but today one of those inexpensive fusible high voltage film resistors would be suitable if you are paranoid enough and cannot rely on additional shunt protection.  The rise time of an ESD pulse is fast so bypass capacitance matters as well.

Things are worse when you consider that it is the machine ESD model which likely applies instead of the human ESD model.

(1) Many PNP input operational amplifiers avoid this problem but I do not know that they are any more ESD resistant.  NPN input cascode operational amplifiers also avoid it but again, I do not know that they are any more ESD resistant.

[Alex Nikitin]

Also the diode FDH300 (very low forward leakage)
http://www.onsemi.com/PowerSolutions/product.do?id=FDH300

In SOT-23, MMBD1501/A

Not really in the same class, less than 400mV for 1uA forward current typical. The BAV199 has around 500-550mV at 1uA .

Cheers

Alex

[free_electron]

   so sorry. sensor output is 1mA .... not 1uA ...
Me : then where does this 1m megaohm requirement come from ?
they: that is input resistance of multimeter so you get no erro ...  (the spelling of erro gives it away right ? )

screw it i'm putting a simple TVs diode across there. and i'm gonna use a 100 ohm sense resistor with an x10 amp.
the only requirement is that the burden voltage of the system should not go above 1.2 volt. as that causes unwanted electromigration in the electrodes.
During normal operation the output is in the order of 50 to 60 microamps. only at full reaction does the sensor deliver 1 mA. but that is not realistic.

something along these lines (image below) will clip enough incoming crap (esd, people with screwdrivers ... ) that it has no impact and protects the input opamp.

Anyway. Interesting puzzle. How to protect a line that has a high impedance ?  Most diodes don't work as they leak to much in the 'subthreshold' zone, and manufacturers don't specify the data for that area. so good luck finding something that works. the transistor as a diode trick would work. That's why keithley uses a matched pair antiparallel as input clamp for their electrometers. ( dave's recent video reminded me of that and i was working on this 1uA pretection thing so one thing lead to another )

And yes , diodes follow that equation but even that is useless information to solve this problem. you don't know half of the parameters used , manufacturers don;t list em. so all you are stuck with is trial and error. which i wanted to avoid in the first place.

One thing learned : when looking for some information that is hard to find : check Bob Pease books FIRST. most likely he's already done it.

[dcarolan]

Here is something I did a while ago and am glad to share.
If youre looking to clamp a hi Z stage. Be very cautious of the diode connected JFet trick. It only works for a couple of milliamps then it looks like a resistor.

[Alex Nikitin]

Here is something I did a while ago and am glad to share.
If youre looking to clamp a hi Z stage. Be very cautious of the diode connected JFet trick. It only works for a couple of milliamps then it looks like a resistor.

Your data on the BAV199 is somewhat suspect (perhaps it was a counterfeit device). Even the worst makes I've measured are much better than the BAV99 standard diode. Here are my measured results for a number of BAV199 from five different manufacturers. Even the worst one is considerably better than your graph shows , and the better makes are similar to what you have measured for the BAS716 (hardly surprising as it is the same device as the BAV199, but a single in a different package).

Cheers

Alex

P.S. - on a JFET connected as a diode. It will look as a diode+resistor for larger currents, however the value of that resistor would depend on the channel resistance. The MMBT4117 is a very small JFET with a very high channel resistance (over 10K). If you use a larger JFET with the channel resistance in tens of Ohms it will happily work as a diode up to tens and possibly even hundreds of mA, though the leakage current will be higher too.

[Wallace Gasiewicz]

Considering the P N Junction of a FET....   
Would a FET with a very low Reverse Leakage also have a low forward sub threshold leakage?

[T3sl4co1l]

Impressed that it's that low: e.g. for 10 to 30mA, Vf goes from ~1.1 to ~1.4V, or ~15Ω.

Yeah not sure what's up with the BAVs, those have some serious uA-nA leakage at low voltage. Surface contamination? Radiation exposure?  Schottky? Well it ain't schottky, the drop isn't nearly low enough in the mA.

Tim

[coppercone2]

how about a BAS low leakage diode

[David Hess]

Considering the P N Junction of a FET....   
Would a FET with a very low Reverse Leakage also have a low forward sub threshold leakage?

I suspect that is the case, but I do not think there is any relationship between reverse leakage and forward conductance.  If there was, then a diode connected JFET would be recommended for log/antilog circuits, and I have never seen that.

Leaky gold doped switching diodes are also leaky when forward biased.

A base-emitter junction is about the best silicon diode; it has low forward and reverse leakage, and high conductance.

Visible light LED junctions with their high bandgap have low reverse and forward leakage, but in the dark.

[Phil1977]

Very interesting thread, I have thought about lots of these considerations for sensor projects too.

One important thing that was missing yet: If I understood it right, you measure a DC signal. That means you can easily get rid of e.g. ESD spikes by

- catching them in a higher voltage suppressor diode
- smoothing the output with an appropriate RC filter

If you do the maths or if you simulate you´ll see:

- e.g. a 2500V ESD spike will be reduced to around 15V by a 6.8V suppressor
- the remaining 15V will be reduced to around 0.2V by e.g. a 100kOhm/2.2nF RC filter

(values are just exemplary for one system I experimented with long ago - they vary greatly with the immanent stray inductance. But the principle stays the same.)

A 100kOhm filter resistor before your 1MOhm shunt will increase the meters impedance by 10% - but maybe that´s a fair price for the pulse protection.

In the RC filter only the cap could create a leakage current, but taking a decent foil cap you can get into the picoampere range.

For some high voltage sensors that have been prone to arcing we even cascaded the protection: Sensor --> 50kOhm HV resistor --> Spark gap --> 50kOhm resistor --> Suppressor diode --> RC 100kOhm 2.2nF --> LM 662 input with several hundreds of MOhm feedback. Of course all higher frequency components of the signal are blocked that way, but we even wanted some analogue averaging there.

[T3sl4co1l]

Large-value resistors are problematic for analog applications; I would avoid maybe even 10k, almost certainly 100k, with regular diodes (BAV99 etc.), certainly 10k with schottky (BAT54S etc.); but lower is okay.

This will be for middle-of-the-road say 12 ENOB applications, not quite generic and loose stuff, somewhat precise, but nothing wild.

For use with LMC662 (there's no LM662?), a bootstrapped diode would be most suitable: a back-to-back diode from input to ESD clamp node (which gets the supply clamp diodes), then a low resistor from that to OUT, with the op-amp wired as unity-gain follower.  Little voltage drop means little current flow, even with leaky diodes.  Not so effective if the same amp needs voltage gain though.

Tim