Diode (leakage) current at very low voltage
Kristian Ukkonen
Hello,
What is the current through a 1n4007 or a 1n4148 when
the voltage across it is very low, like +-50mV ?
I'm measuring a leakage current somewhere in my
circuit (about 1 nA) and it MIGHT be the protection
diodes. It has become interesting that no datasheet
actually specifies these "leakage" currents for diodes
at voltage << 0.5V..
Just a magnitude of the current at 25C is enough..
Kristian Ukkonen.
Kristian Ukkonen
Kristian Ukkonen wrote:
> What is the current through a 1n4007 or a 1n4148 when
> the voltage across it is very low, like +-50mV ?
>
> I'm measuring a leakage current somewhere in my
> circuit (about 1 nA) and it MIGHT be the protection
> diodes. It has become interesting that no datasheet
> actually specifies these "leakage" currents for diodes
> at voltage << 0.5V..
To verify that it is the diodes causing the
leakage, I replaced 1n4007 with 1n4148 and surely
enough, the leakage decreased by about a decade.
The leakage is still about 70pA.
Any ideas what diode would be better?
Kristian Ukkonen.
Win Hill
You should have some series input resistance, so that
you feel comfortable to use smaller diodes, like the
1n4148.
Glass diodes can have "leakage" from room light when
in a circuit sitting exposed on the bench.
The gate junction of small-signal JFETs make excellent
low-leakage low-capacitance protection diodes.
Thanks,
- Win
Winfield Hill
Rowland Institute for Science
100 Edwin Land Blvd
Cambridge, MA 02142-1297
Kevin Aylward
"Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
news:3B7FA64A...@pp1.inet.fi...
>
> Hello,
>
> What is the current through a 1n4007 or a 1n4148 when
> the voltage across it is very low, like +-50mV ?
>
> I'm measuring a leakage current somewhere in my
> circuit (about 1 nA) and it MIGHT be the protection
> diodes. It has become interesting that no datasheet
> actually specifies these "leakage" currents for diodes
> at voltage << 0.5V..
>
They ar not leakage currents.
> Just a magnitude of the current at 25C is enough..
>
The equation is
I=Is.exp(q.Vbe/KT)
If you know the current at some particular voltage, i.e. 0.65V, you can work
out Is, that is if you don't have the spice model for it that tells you off
the bat. You can then calculate what it should be at 50mv.
Kevin Aylward , Warden of the Kings Ale
ke...@anasoft.co.uk
http://www.anasoft.co.uk - SuperSpice "Cheap, No Shit!",
GUI xspice, an affordable unlimited component, mixed-mode Windows simulator
with Schematic Capture, waveform display, FFT's and Filter Design.
Opinions of my employer are not necessarily indicative of my own
Oscillators don't, amplifiers do".
Kristian Ukkonen
> > the voltage across it is very low, like +-50mV ?
> >
> > I'm measuring a leakage current somewhere in my
> > circuit (about 1 nA) and it MIGHT be the protection
> > diodes. It has become interesting that no datasheet
> > actually specifies these "leakage" currents for diodes
> > at voltage << 0.5V..
> They ar not leakage currents.
A matter of semantics, eh?
> > Just a magnitude of the current at 25C is enough..
> The equation is
> I=Is.exp(q.Vbe/KT)
> If you know the current at some particular voltage, i.e. 0.65V, you can work
> out Is, that is if you don't have the spice model for it that tells you off
> the bat. You can then calculate what it should be at 50mv.
Yes, I actually did consider that and even run spice for it.
That tells about 1.7nA (reverse) and 5.1nA currents for one
1n4148.
Is(1n4007) = 1.10e-8
Is(1n4148) = 1.07e-9
So 1n4007 should be a decade worse.
Of course, that is two _decades_ more "leakage" current than
I'm really measuring with two 1n4007.
Should the above formula work with so small Vbe?
Amazingly, MURS320 (200V 3A) diode has lowest Is (1.06n vs. 2.52n
of 1n4148) of the diodes included in my spice (Linear SwitcherCAD3).
Kristian Ukkonen.
Kevin Aylward
"Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
news:3B7FD035...@pp1.inet.fi...
>
> Kevin Aylward wrote:
> > "Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
> > > What is the current through a 1n4007 or a 1n4148 when
> > > the voltage across it is very low, like +-50mV ?
> > >
> > > I'm measuring a leakage current somewhere in my
> > > circuit (about 1 nA) and it MIGHT be the protection
> > > diodes. It has become interesting that no datasheet
> > > actually specifies these "leakage" currents for diodes
> > > at voltage << 0.5V..
> > They ar not leakage currents.
>
> A matter of semantics, eh?
>
You can have _real_ leakage, as well the current that's demanded by the
physics of the device.
> > > Just a magnitude of the current at 25C is enough..
> > The equation is
> > I=Is.exp(q.Vbe/KT)
> > If you know the current at some particular voltage, i.e. 0.65V, you can
work
> > out Is, that is if you don't have the spice model for it that tells you
off
> > the bat. You can then calculate what it should be at 50mv.
>
> Yes, I actually did consider that and even run spice for it.
> That tells about 1.7nA (reverse) and 5.1nA currents for one
> 1n4148.
>
> Is(1n4007) = 1.10e-8
> Is(1n4148) = 1.07e-9
> So 1n4007 should be a decade worse.
These are rather high, definitely incorrect. The 1N4148 would be more like
100f or 0.1pa for Is. Typically 1f to 1000f would be usual for most devices.
What spice do you use that has such a bad models?
> Of course, that is two _decades_ more "leakage" current than
> I'm really measuring with two 1n4007.
>
> Should the above formula work with so small Vbe?
I agree, that things might well go amiss at very low Vbe's.
> Amazingly, MURS320 (200V 3A) diode has lowest Is (1.06n vs. 2.52n
> of 1n4148) of the diodes included in my spice (Linear SwitcherCAD3).
>
Ahmmm, oh I see, you mean you don't use a real spice system.
John Popelish
> You should have some series input resistance, so that
> you feel comfortable to use smaller diodes, like the
> 1n4148.
>
> Glass diodes can have "leakage" from room light when
> in a circuit sitting exposed on the bench.
I just got caught with this using a 1N4148 to clamp the signal from a
photo multiplier tube into a transconductance amplifier to protect the
input if the amplifier saturated. I had to mod the board by putting a
bit of black heat shrink tubing over the diode.
I found it while exploring around the tube housing with a flashlight,
to look for light leaks while watching the output of the amplifier on
the scope. I happened to sweep the light across the diode, and got a
big reaction out of the amplifier. Oops.
> The gate junction of small-signal JFETs make excellent
> low-leakage low-capacitance protection diodes.
--
John Popelish
Robert
Win Hill wrote:
I recall reading in a not too recent Elektor that a little known fact is
that LEDs have ultra-low reverse leakage when the package is suitably
modified to prevent entry of light. Is there any truth to this?
Jim Thompson
<ke...@anasoft.co.uk> wrote:
|
|
|
|"Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
|news:3B7FD035...@pp1.inet.fi...
|>
|> Kevin Aylward wrote:
|> > "Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
|> > > What is the current through a 1n4007 or a 1n4148 when
|> > > the voltage across it is very low, like +-50mV ?
[snip]
|> Yes, I actually did consider that and even run spice for it.
|> That tells about 1.7nA (reverse) and 5.1nA currents for one
|> 1n4148.
|>
|> Is(1n4007) = 1.10e-8
|> Is(1n4148) = 1.07e-9
|> So 1n4007 should be a decade worse.
|
|These are rather high, definitely incorrect. The 1N4148 would be more like
|100f or 0.1pa for Is. Typically 1f to 1000f would be usual for most devices.
|What spice do you use that has such a bad models?
Neeerp!
1N4007 Is = 14.11nA
1N4148 Is = 2.682nA
|
|> Of course, that is two _decades_ more "leakage" current than
|> I'm really measuring with two 1n4007.
|>
|> Should the above formula work with so small Vbe?
There's a "minus 1" term in there that's ignored at reasonable forward
drop (as a mathematical convenience) but it's important at low
currents/voltages.
Here are the actual equations used by PSpice (hope I got all the
exponents right...cut and pasting from a PDF lost the small raised
fonts).....
Diode equations for DC current
Id = area·(Ifwd - Irev)
Ifwd = forward current = Inrm·Kinj + Irec·Kgen
Inrm = normal current = IS·(e^(Vd/(N·Vt)) -1) <<--- Note "-1"
if: IKF > 0
then: Kinj = high-injection factor = sqrt(IKF/(IKF+Inrm))
else: Kinj = 1
Irec = recombination current = ISR·(e^(Vd/(NR·Vt)) -1) <<--- Note "-1"
Kgen = generation factor = ((1-Vd/VJ)^2 +0.005)^(M/2)
Irev = reverse current = Irevhigh + Irevlow
Irevhigh = IBV·e^(-(Vd+BV)/(NBV·Vt))
Irevlow = IBVL·e^(-(Vd+BV)/(NBVL·Vt))
|
|I agree, that things might well go amiss at very low Vbe's.
|
|> Amazingly, MURS320 (200V 3A) diode has lowest Is (1.06n vs. 2.52n
|> of 1n4148) of the diodes included in my spice (Linear SwitcherCAD3).
|>
|
|Ahmmm, oh I see, you mean you don't use a real spice system.
Certainly looks like the included models suck.
|
|Kevin Aylward , Warden of the Kings Ale
|ke...@anasoft.co.uk
(If replying by E-mail please observe obscure method of anti-spam.)
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| Jim-T@analog_innovations.com Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |
For proper E-mail replies SWAP "-" and "_".
Jim Thompson
[snip]
|
|I recall reading in a not too recent Elektor that a little known fact is
|that LEDs have ultra-low reverse leakage when the package is suitably
|modified to prevent entry of light. Is there any truth to this?
Hi Robert,
I read the same thing somewhere in (I believe) a recent issue of EDN.
Unfortunately I toss them as soon as I scan and rip articles.
Robert
Jim Thompson wrote:
The Elektor application was a standard two amplifier peak detector using the
LED in place of the standard diode. The LED was chosen to minimize the hold
capacitor bleed current to achieve a fairly hefty hold time- many minutes or
fraction of hour -staying within reasonable error range.
Jim Thompson
|
|
|Jim Thompson wrote:
|
|> On Sun, 19 Aug 2001 17:04:37 GMT, Robert <nos...@earthlink.net> wrote:
|>
|> [snip]
|> |
|> |I recall reading in a not too recent Elektor that a little known fact is
|> |that LEDs have ultra-low reverse leakage when the package is suitably
|> |modified to prevent entry of light. Is there any truth to this?
|>
|> Hi Robert,
|>
|> I read the same thing somewhere in (I believe) a recent issue of EDN.
|> Unfortunately I toss them as soon as I scan and rip articles.
|>
|>
|> ...Jim Thompson
|The Elektor application was a standard two amplifier peak detector using the
|LED in place of the standard diode. The LED was chosen to minimize the hold
|capacitor bleed current to achieve a fairly hefty hold time- many minutes or
|fraction of hour -staying within reasonable error range.
What kind of light shield was used? IR will go right through most
plastics. I have a TI IR detector for TV remotes around here
somewhere that's just a molded block of black plastic with leads, no
windows.
Robert
Jim Thompson wrote:
Don't hold me to it- but IIRC they recommended an India ink paint.
Jim Thompson
[snip]
|
|I recall reading in a not too recent Elektor that a little known fact is
|that LEDs have ultra-low reverse leakage when the package is suitably
|modified to prevent entry of light. Is there any truth to this?
I just remembered that low voltage zeners have richly doped junctions,
thus a low Is. I can't necessarily say that low Is, alone, will
guarantee low leakage, but it's worth a try.
William L. Bahn
Kevin Aylward wrote in message ...
>
>
>
>"Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
>news:3B7FA64A...@pp1.inet.fi...
>>
>> Hello,
>>
>> What is the current through a 1n4007 or a 1n4148 when
>> the voltage across it is very low, like +-50mV ?
>>
>> I'm measuring a leakage current somewhere in my
>> circuit (about 1 nA) and it MIGHT be the protection
>> diodes. It has become interesting that no datasheet
>> actually specifies these "leakage" currents for diodes
>> at voltage << 0.5V..
>>
>
>They ar not leakage currents.
>
>> Just a magnitude of the current at 25C is enough..
>>
>
>The equation is
>
>I=Is.exp(q.Vbe/KT)
>
>If you know the current at some particular voltage, i.e. 0.65V, you can
work
>out Is, that is if you don't have the spice model for it that tells you off
>the bat. You can then calculate what it should be at 50mv.
>
>
But for such low voltages he needs to use the more complete relationship
(which is still only an approximation):
I=Is*[exp(q.Vbe/nKT) - 1]
Where n, for a discrete diode, is probably close to 2.
Near room temperature, neglecting the '-1' term at Vbe = 50mV would result
in about a 15% error with n=1. With n=2 is would be an error of nearly 60%.
In addition to the temperature dependence embedded explicitly in the
exponential, Is is also temperature dependent and doubles roughly every 5C.
And this doesn't take into account the leakage current itself which can be a
many orders of magnitude larger than Is (and which roughly doubles every
10C).
Kristian Ukkonen
Jim Thompson wrote:
> On Sun, 19 Aug 2001 15:33:00 GMT, "Kevin Aylward"
> |? Kevin Aylward wrote:
> |? ? "Kristian Ukkonen" ?kristian...@pp1.inet.fi? wrote in message
> |? ? ? What is the current through a 1n4007 or a 1n4148 when
> |? ? ? the voltage across it is very low, like +-50mV ?
> |? Is(1n4007) = 1.10e-8
> |? Is(1n4148) = 1.07e-9
> |These are rather high, definitely incorrect. The 1N4148 would be more like
> |100f or 0.1pa for Is. Typically 1f to 1000f would be usual for most devices.
> |What spice do you use that has such a bad models?
> Neeerp!
> 1N4007 Is = 14.11nA
> 1N4148 Is = 2.682nA
1n4148 and google search finds:
http://www.fairchildsemi.com/models/Discretes/diode.html
Is=5.84n
http://www.cadmigos.com/manual/refman/rm0026.htm
Is=2.495e-9
http://www.eletr.ufpr.br/mehl/simulacao.pdf
Is=0.1p
It seems everyone has their own spice Is. :)
There are different manufacturers of these "1n4148" diodes,
so it is propable that no single value is universally correct.
> |? Amazingly, MURS320 (200V 3A) diode has lowest Is (1.06n vs. 2.52n
> |? of 1n4148) of the diodes included in my spice (Linear SwitcherCAD3).
> |Ahmmm, oh I see, you mean you don't use a real spice system.
> Certainly looks like the included models suck.
Well, its a totally free Spice based program with a Windows graphical
user interface etc.. The other free ones don't have a GUI AFAIK.
http://www.linear.com/software/
"LTC SwitcherCAD III is a fully functional Spice III simulator with
enhancements
and models for easing the simulation of switching regulators. This Spice
is a
high performance circuit simulator, integrated waveform viewer, which
also includes schematic
capture. Our enhancements to Spice have made simulating switching
regulators extremely fast
compared to normal Spice simulators. Included in this download are
Spice, Macro Models for
80? of Linear Technology's switching regulators and over 200 op amp
models. Furthermore,
resistors, transistors and MOSFET models are included as part of this
package. With this
Spice simulation, viewing of the waveforms of most switching regulators
can be achieved in a
few minutes on a high performance PC. Also, full circuits using op amps
or transistors can
easily be simulated."
I'm no expert with spice. Just learning it.
How about using some NPN transistor like 2n2222 as diode;
base and collector together? The "spice" tells that it has
"leakage" current <100fA at +-50mV.
Kristian Ukkonen.
Robert
Robert wrote:
This would be a PCB resist pen ink?- available from RS.
Jim Thompson
<kristian...@pp1.inet.fi> wrote:
|
|Jim Thompson wrote:
[snip]
|> Neeerp!
|> 1N4007 Is = 14.11nA
|> 1N4148 Is = 2.682nA
|
|1n4148 and google search finds:
|
|http://www.fairchildsemi.com/models/Discretes/diode.html
|Is=5.84n
|
|http://www.cadmigos.com/manual/refman/rm0026.htm
|Is=2.495e-9
|
|http://www.eletr.ufpr.br/mehl/simulacao.pdf
|Is=0.1p
4:1 would seem reasonable for slightly differing processes, but 0.1p,
give me a break!
|
|It seems everyone has their own spice Is. :)
|
|There are different manufacturers of these "1n4148" diodes,
|so it is propable that no single value is universally correct.
|
|> |? Amazingly, MURS320 (200V 3A) diode has lowest Is (1.06n vs. 2.52n
|> |? of 1n4148) of the diodes included in my spice (Linear SwitcherCAD3).
|> |Ahmmm, oh I see, you mean you don't use a real spice system.
|> Certainly looks like the included models suck.
|
|Well, its a totally free Spice based program with a Windows graphical
|user interface etc.. The other free ones don't have a GUI AFAIK.
|
|http://www.linear.com/software/
[snip]
|I'm no expert with spice. Just learning it.
We all were students at some point in our lives. I first did Berkeley
Spice (2G6, I think) on a VAX with hand-typed netlists ;-)
|
|How about using some NPN transistor like 2n2222 as diode;
|base and collector together? The "spice" tells that it has
|"leakage" current <100fA at +-50mV.
That's probable. Transistors are more richly doped than diodes (see
my other posting about zeners).
|
| Kristian Ukkonen.
John Woodgate
<kristian...@pp1.inet.fi> wrote (in <3B7FD035...@pp1.inet.fi>
) about 'Diode (leakage) current at very low voltage', on Sun, 19 Aug
2001:
>Kevin Aylward wrote:
>> "Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
>> > What is the current through a 1n4007 or a 1n4148 when
>> > the voltage across it is very low, like +-50mV ?
>> >
>> > I'm measuring a leakage current somewhere in my
>> > circuit (about 1 nA) and it MIGHT be the protection
>> > diodes. It has become interesting that no datasheet
>> > actually specifies these "leakage" currents for diodes
>> > at voltage << 0.5V..
>> They ar not leakage currents.
>
>A matter of semantics, eh?
Are they vectors or phasors? (;-)
--
Regards, John Woodgate, OOO - Own Opinions Only. http://www.jmwa.demon.co.uk
Eat mink and be dreary!
John Larkin
Robert, Win,
I've used the b-c junction of a transistor (a 2N4402, I think) as a
low-leakage diode. I recall single-digits of picoamps, I think.
John
John Larkin
Kevin,
it's ironic that Is, the theoretical reverse saturation current, is
usable to compute the foreward conduction curve of a diode, but is
virtually useless to predict the reverse leakage! That's often
dominated by surface effects and defect leakage.
John
Kevin Aylward
"Jim Thompson" <Jim-T@analog_innovations.com> wrote in message
news:gs50ot0tsbf52bodq...@4ax.com...
> On Sun, 19 Aug 2001 19:28:13 GMT, Kristian Ukkonen
> <kristian...@pp1.inet.fi> wrote:
>
> |
> |Jim Thompson wrote:
> [snip]
> |> Neeerp!
> |> 1N4007 Is = 14.11nA
> |> 1N4148 Is = 2.682nA
> |
> |1n4148 and google search finds:
> |
> |http://www.fairchildsemi.com/models/Discretes/diode.html
> |Is=5.84n
> |
> |http://www.cadmigos.com/manual/refman/rm0026.htm
> |Is=2.495e-9
> |
> |http://www.eletr.ufpr.br/mehl/simulacao.pdf
> |Is=0.1p
>
> 4:1 would seem reasonable for slightly differing processes, but 0.1p,
> give me a break!
>
Actually, I am thinking in terms of 10f/u^2, I keep forgetting that my brain
is still wired for i.c. specs. I can certainly state that typical IS for an
i.c transistor are of the order of 5e-18/4u*4u emitter, which might be good
up to maybe 100ua. (yes, 10 to the minus eighteen)
Kevin Aylward , Warden of the Kings Ale
ke...@anasoft.co.uk
http://www.anasoft.co.uk - SuperSpice "Cheap, No Shit!",
GUI xspice, an affordable unlimited component, mixed-mode Windows simulator
with Schematic Capture, waveform display, FFT's and Filter Design.
Opinions of my employer are not necessarily indicative of my own
Oscillators don't, amplifiers do"> For proper E-mail replies
Jim Thompson
<ke...@anasoft.co.uk> wrote:
|"Jim Thompson" <Jim-T@analog_innovations.com> wrote in message
|news:gs50ot0tsbf52bodq...@4ax.com...
|> On Sun, 19 Aug 2001 19:28:13 GMT, Kristian Ukkonen
|> <kristian...@pp1.inet.fi> wrote:
[snip]
|> |Is=0.1p
|>
|> 4:1 would seem reasonable for slightly differing processes, but 0.1p,
|> give me a break!
|>
|
|Actually, I am thinking in terms of 10f/u^2, I keep forgetting that my brain
|is still wired for i.c. specs. I can certainly state that typical IS for an
|i.c transistor are of the order of 5e-18/4u*4u emitter, which might be good
|up to maybe 100ua. (yes, 10 to the minus eighteen)
|
|Kevin Aylward , Warden of the Kings Ale
|ke...@anasoft.co.uk
|http://www.anasoft.co.uk - SuperSpice "Cheap, No Shit!",
|GUI xspice, an affordable unlimited component, mixed-mode Windows simulator
|with Schematic Capture, waveform display, FFT's and Filter Design.
|Opinions of my employer are not necessarily indicative of my own
|Oscillators don't, amplifiers do"
Atto boy ;-)
(If replying by E-mail please observe obscure method of anti-spam.)
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| Jim-T@analog_innovations.com Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |
For proper E-mail replies SWAP "-" and "_".
Gerry Schneider
> Win Hill wrote:
>
> > Glass diodes can have "leakage" from room light when
> > in a circuit sitting exposed on the bench.
>
> I just got caught with this using a 1N4148 to clamp the signal from a
> photo multiplier tube into a transconductance amplifier to protect the
> input if the amplifier saturated. I had to mod the board by putting a
> bit of black heat shrink tubing over the diode.
>
> I found it while exploring around the tube housing with a flashlight,
> to look for light leaks while watching the output of the amplifier on
> the scope. I happened to sweep the light across the diode, and got a
> big reaction out of the amplifier. Oops.
>
Just tried it out here. 1N4148 glass package - open circuit output voltage
at the anode was 2 mVDC with ambient room light (could have been partly
due to rectified AC hum), went to 120 mVDC next to a 50W 12VDC light bulb
run from a gell cell. Short circuit current went to 70 nADC. Surprisingly
high values considering the package doesn't look obviously transparent.
Gerry
Win Hill
>
> Robert, Win,
>
> I've used the b-c junction of a transistor (a 2N4402, I think) as
> a low-leakage diode. I recall single-digits of picoamps, I think.
I can dig it.
Win Hill
>
> John Popelish wrote:
>
>> Win Hill wrote:
>>
>>> Glass diodes can have "leakage" from room light when
>>> in a circuit sitting exposed on the bench.
>>
>> I just got caught with this using a 1N4148 to clamp the signal from a
>> photo multiplier tube into a transconductance amplifier to protect the
>> input if the amplifier saturated. I had to mod the board by putting a
>> bit of black heat shrink tubing over the diode.
>
> at the anode was 2 mVDC with ambient room light (could have been partly
> due to rectified AC hum), went to 120 mVDC next to a 50W 12VDC light bulb
> run from a gell cell. Short circuit current went to 70 nADC. Surprisingly
> high values considering the package doesn't look obviously transparent.
Good work Gerry.
We're most interested in your light-level vs. current measurements...
E.g., how dark does the circuit have to be before the leakage current
drops to acceptable (< 0.01nA) limits. For example, with a standard
bright flourescent bench light shining on the bench, is it sufficient
to simply shade the diode? I've made these measurements, years ago,
but I forget, is the answer NO?
BTW, to measure 0.01nA, use a 10Meg resistor with a 4.5-digit DMM on
the 200mV scale, where an LSB is 10uV. Each lsb will be 0.001nA = 1pA.
Detemine the no-signal offset first. It'll be surprisingly small.
Gerry Schneider
> We're most interested in your light-level vs. current measurements...
> E.g., how dark does the circuit have to be before the leakage current
> drops to acceptable (< 0.01nA) limits. For example, with a standard
> bright flourescent bench light shining on the bench, is it sufficient
> to simply shade the diode? I've made these measurements, years ago,
> but I forget, is the answer NO?
After a bit of fiddling with lead dress, I finally mounted the 1N4148 directly
on banana plugs at the DVM (Metex 4-1/2 digit, 10 Mohm). The measurements were
reasonably similar with and without an earth ground connection to the DVM
common. All measurements were done with the diode anode to Vin and cathode to
COM, using the diode as a photovoltaic source (no external bias connection).
Measurements were averaged over 10 runs - same diode.
Interesting results:
very dark room (equivalent to inside a closed box): -39pA
dim lights (equivalent to casual light leakage into an enclosure): -18pA
normal low level computer desk lighting: +20pA
bright workbench lighting: +80pA
I'm willing to believe that the negative offset readings are due to the DVM
doing its dual slope thingy, but have no hard data - anybody know? The DVM reads
+/- 1 pA with the diode removed and all other conditions the same.
Assuming the offset is constant (??), the equivalent photocurrent would be 20 pA
if the box is shaded but not light-tight. Hmmm.. thanks Win! Another obscure but
important fact to impress the client's young engineers with at some future
project meeting. This one will raise a few eyebrows!
Regards,
Gerry
Mark Fergerson
<snip>
> I recall reading in a not too recent Elektor that a little known fact is
> that LEDs have ultra-low reverse leakage when the package is suitably
> modified to prevent entry of light. Is there any truth to this?
While we're into obscure LED properties, I vaguely recall reading an
offhand comment that all LEDs are nonlinear near the origin of their I
V curves especially when driven with low voltage AC (never shown on
spec sheets), but I can't find any solid numbers. Anybody know?
Mark L. Fergerson
Robert Strand
>
> While we're into obscure LED properties, I vaguely recall reading an
> offhand comment that all LEDs are nonlinear near the origin of their I
> V curves especially when driven with low voltage AC (never shown on
> spec sheets), but I can't find any solid numbers. Anybody know?
It makes a lot of sense since the Is is much smaller for an LED.
Is = Id / exp(Vd/(n Vt))
1N914 diode say 600mV @1mA => Is = 2.7nA (n~1.8)
Red LED say 1.8V @ 10mA => Is = 200e-21A (n~1.8)
Transistor BC junction say 0.75V @ 1mA => Is = 0.5e-18 (n~1.02)
Despite of off hand numbers the LED clearly has a much lower Is. It also clear
the transistor gains it's low Is from the smaller n.
Regards
Rob
Jim Thompson
Rob, "N" is only a slope term. "Is" is determined by junction area
and doping levels.
Robert Strand
> and doping levels.
Yes, but I've started use known device voltages and currents to derive Is, so the
junction area and doping levels are implied in these figures? Its probably more the
case that at low currens N approaches 2 anyway (IIRC due to recombination).
Regards
Rob
John Devereux
You can use leds as photodiodes too - not very good ones.
However you can match the package colour to the light
wavelength you want, and hence possibly avoid the need for a
filter. I gather they have a good long wavelength cutoff as
well.
Bad photodiodes, bad "zeners", is there no end to the uses
these devices can be put to? :)
-- John Devereux
Jim Thompson
"N approaches 2". Yep, that happens at the ISE breakpoint.
John Woodgate
<ger...@sympatico.ca> wrote (in <3B815295...@sympatico.ca>) about
'Diode (leakage) current at very low voltage', on Mon, 20 Aug 2001:
>The measurements were
>reasonably similar with and without an earth ground connection to the DVM
>common. All measurements were done with the diode anode to Vin and cathode to
>COM,
Where is the 10 Mohm resistor Win mentioned?
Gerry Schneider
> I read in sci.electronics.design that Gerry Schneider
> <ger...@sympatico.ca> wrote (in <3B815295...@sympatico.ca>) about
> 'Diode (leakage) current at very low voltage', on Mon, 20 Aug 2001:
> >The measurements were
> >reasonably similar with and without an earth ground connection to the DVM
> >common. All measurements were done with the diode anode to Vin and cathode to
> >COM,
>
> Where is the 10 Mohm resistor Win mentioned?
Inside the DVM - 10Mohm resistor from Vin to COM.
Gerry
Win Hill
>
> John Woodgate wrote:
>
>>>
>>> The measurements were reasonably similar with and without an earth
>>> ground connection to the DVM common. All measurements were done with
>>> the diode anode to Vin and cathode to COM,
>>
>> Where is the 10 Mohm resistor Win mentioned?
>
> Inside the DVM - 10Mohm resistor from Vin to COM.
Thankfully most DVMs don't have an internal resistor on the 200mV
scale, and are ultra-high Zin. This allows not only 10M, but 100M
sense resistors for very low current sensing.
John Woodgate
wrote (in <3B81D7B1...@mediaone.net>) about 'Diode (leakage)
current at very low voltage', on Tue, 21 Aug 2001:
>Gerry Schneider wrote:
>>
>> John Woodgate wrote:
>>
>>> I read in sci.electronics.design that Gerry Schneider wrote
>>>>
>>>> The measurements were reasonably similar with and without an earth
>>>> ground connection to the DVM common. All measurements were done with
>>>> the diode anode to Vin and cathode to COM,
>>>
>>> Where is the 10 Mohm resistor Win mentioned?
>>
>> Inside the DVM - 10Mohm resistor from Vin to COM.
>
> Thankfully most DVMs don't have an internal resistor on the 200mV
> scale, and are ultra-high Zin. This allows not only 10M, but 100M
> sense resistors for very low current sensing.
Indeed. I didn't think GS had done the measurement in the way you
proposed. But further thought indicates to me that I don't know exactly
what you were proposing. Was it d.c. supply, diode and 10 Mohm in
series, with the DVM across the 10 Mohm?
Win Hill
>
>> Thankfully most DVMs don't have an internal resistor on the 200mV
>> scale, and are ultra-high Zin. This allows not only 10M, but 100M
>> sense resistors for very low current sensing.
>
> Indeed. I didn't think GS had done the measurement in the way you
> proposed. But further thought indicates to me that I don't know
> exactly what you were proposing. Was it d.c. supply, diode and
> 10 Mohm in series, with the DVM across the 10 Mohm?
Exactly, I use a precision DC current source, but a precision voltage
source could be used if it's setting was increased by the DVM drop.
Thanks,
Mark Fergerson
ISTR they make crappy varicaps as well.
Mark L. Fergerson
D. A.(Tony) Stewart
negative DC leakage? Add a small ceramic cap across the diode or in a
suitable location.
Tony Stewart EE'75
Kristian Ukkonen
Kevin Aylward wrote:
> "Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
> > Kevin Aylward wrote:
> > > "Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote in message
> > > > What is the current through a 1n4007 or a 1n4148 when
> > > > the voltage across it is very low, like +-50mV ?
> > That tells about 1.7nA (reverse) and 5.1nA currents for one
> > 1n4148.
> These are rather high, definitely incorrect. The 1N4148 would be more like
> 100f or 0.1pa for Is. Typically 1f to 1000f would be usual for most devices.
> What spice do you use that has such a bad models?
To get things interesting, I received data on 1n4148 measured
with HP4155A parameter analyzator and it measured:
mV pA
-100,00 -1549,00
-90,00 -1512,80
-80,00 -1468,80
-70,00 -1413,00
-60,00 -1340,10
-50,00 -1245,00
-40,00 -1117,70
-30,00 -949,14
-20,00 -718,21
-10,00 -410,31
0,00 2,00
10,00 549,44
20,00 1277,90
30,00 2223,40
40,00 3441,20
50,00 5015,00
60,00 6991,00
70,00 9475,10
80,00 12670,00
90,00 16544,00
100,00 21323,00
So the Spice results above _are_ reasonable.
> Kevin Aylward , Warden of the Kings Ale
> http://www.anasoft.co.uk - SuperSpice "Cheap, No Shit!",
What does YOUR spice give as answer, eh? :)
Kristian Ukkonen.
Winfield Hill
>
> Kevin Aylward wrote:
>> "Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote
>>> Kevin Aylward wrote:
>>>> "Kristian Ukkonen" <kristian...@pp1.inet.fi> wrote
>>>>>
>>>>> the voltage across it is very low, like +-50mV ?
>>> Yes, I actually did consider that and even run spice for it.
>>> That tells about 1.7nA (reverse) and 5.1nA currents for one
>>> 1n4148.
>>> Is(1n4148) = 1.07e-9
>> These are rather high, definitely incorrect. The 1N4148 would
>> be more like 100f or 0.1pa for Is. Typically 1f to 1000f would
>> be usual for most devices.
>
You had the diode in the lab, exposed to room light
while you took the measurements?
Kristian Ukkonen
Winfield Hill wrote:
> Kristian Ukkonen wrote:
> > with HP4155A parameter analyzator and it measured:
> while you took the measurements?
I didn't perform the measurement, so I don't know (yet).
I did find an interesting article: National AppNote 240 by
Robert A. Pease. It does mention:
"A 1n914 may leak 100pA even with only 1 millivolt
across it"
This article mentiones 1n457 and 1n484 as suitable
low-leakage diodes. Of course, these seem rather
difficult to find. :(
Kristian Ukkonen.
Winfield Hill
>
> Winfield Hill wrote:
>> Kristian Ukkonen wrote:
>>>
>>> with HP4155A parameter analyzer ...
>>
>> You had the diode in the lab, exposed to room light
>> while you took the measurements?
>
> I didn't perform the measurement, so I don't know (yet).
>
> I did find an interesting article: National AppNote 240
> by Robert A. Pease. It does mention:
>
> "A 1n914 may leak 100pA even with only 1 millivolt
> across it"
>
> This article mentiones 1n457 and 1n484 as suitable
> low-leakage diodes. Of course, these seem rather
> difficult to find. :(
OK, I've taken a series of bench measurements and now
do not believe that light pickup is responsible for the
truly horrible low-voltage performance of 1n4148, etc
glass diodes. These diodes can generatea few nA of light
current on the bench, but they get into trouble at much
higher forward currents, deviating strongly from the
expected curves in the 10 to 100nA region, as your data
showed. Ditto for 1n914 1n4153 1n4454 families etc.
As for Bob Pease's 100pA claim, I agree, measuring 75pA
of leakage at 1mV drop, and 1nA at +12mV and -20mV.
Basically at voltages under +/-30mV, 1n4148 etc. glass
diodes look like 10 to 20 Meg-ohm non-linear resistors!
They are indeed _VERY_ GROSS parts in that respect!
Winfield Hill
>
> Kristian Ukkonen wrote:
>>
>> I did find an interesting article: National AppNote 240
>> by Robert A. Pease. It does mention: "A 1n914 may leak
>> 100pA even with only 1 millivolt across it" This
>> diodes. Of course, these seem rather difficult to find.
>
> of leakage at 1mV drop, and 1nA at +12mV and -20mV.
> Basically at voltages under +/-30mV, 1n4148 etc. glass
> diodes look like 10 to 20 Meg-ohm non-linear resistors!
> They are indeed _VERY_ GROSS parts in that respect!
Here are a set of I-V measurements for your edification.
-- Schottky -- -- silicon -- --- JFET ---
1n5711 1n4148 red
1n5819 1n6263 1n4153 1n458 PAD-1 2n4117 LED
----- ------ ------ ----- ----- ----- ----
10mA 270 700 750 800 1330 940 1706
1 200 375 620 670 810 790 1614
100uA 130 280 510 577 710 710 1535
10 60 210 375 490 615 640 1425
1 23 80 255 415 555 580 1330
100nA 0 10 140 330 500 525 1205
10 - 0 65 255 440 460 1080
1 - - 12 185* 370 393 950
100pA - - 1.3* 125* 300 335 840*
10 - - * 70* 230 270 *
1 - - ** * 155 220 *
All diodes were bypassed with 10nF. Room temp was about 20C.
The measurements marked * represent values that were strongly
influenced by light pickup, even though the lab lights were
on low and the parts were shaded.
My 1n458 diode (the type suggested by Bob Pease years ago)
was made by Fairchild in the distant past. It was in a dark
black case, yet it still had trouble with light, picking up
0.25nA in dim light, and 1nA in normal bright bench light.
Bob has new suggestions for present practise, but I'll save
that for a later posting.
Note how good the 2n4117A (Idss = 0.25mA) metal-can JFET
looks when used as a protection diode. Its black-plastic
counterparts are probably also fine. The 2n4117A showed
about 6 ohms of internal resistance at high currents.
The red LED had a nice high voltage drop at low currents,
but it picked up 2nA shaded on a dim bench, and 10nA when I
turned on the light. Furthermore, at high forward currents
its drop is rather high compared to the other choices.
Tony Williams
Winfield Hill <hi...@rowland.org> wrote:
> Here are a set of I-V measurements for your edification.
> -- Schottky -- -- silicon -- --- JFET ---
> 1n5711 1n4148 red
> 1n5819 1n6263 1n4153 1n458 PAD-1 2n4117 LED
> ----- ------ ------ ----- ----- ----- ----
> 10mA 270 700 750 800 1330 940 1706
> 1 200 375 620 670 810 790 1614
> 100uA 130 280 510 577 710 710 1535
> 10 60 210 375 490 615 640 1425
> 1 23 80 255 415 555 580 1330
> 100nA 0 10 140 330 500 525 1205
> 10 - 0 65 255 440 460 1080
> 1 - - 12 185* 370 393 950
> 100pA - - 1.3* 125* 300 335 840*
> 10 - - * 70* 230 270 *
> 1 - - ** * 155 220 *
Checking for log-conformity over just two decades in
the 1uA to 1nA region would be good predictor (greater
than 59mV per decade change in current). Which makes
sense. That 1N458 is still in the groove down to 10pA,
even with the light sensitivity (IR possibly?).
Those Schottky diodes are a bit of a surprising ambush
though, because Schottky are exactly the ones that are
tempting to use for ic input protection applications.
--
Tony Williams.
Bob Wilson
> OK, I've taken a series of bench measurements and now
> do not believe that light pickup is responsible for the
> truly horrible low-voltage performance of 1n4148, etc
> glass diodes. These diodes can generatea few nA of light
> current on the bench, but they get into trouble at much
> higher forward currents, deviating strongly from the
> expected curves in the 10 to 100nA region, as your data
> showed. Ditto for 1n914 1n4153 1n4454 families etc.
>
> As for Bob Pease's 100pA claim, I agree, measuring 75pA
> of leakage at 1mV drop, and 1nA at +12mV and -20mV.
>
> Basically at voltages under +/-30mV, 1n4148 etc. glass
> diodes look like 10 to 20 Meg-ohm non-linear resistors!
>
> They are indeed _VERY_ GROSS parts in that respect!
Even the spec sheet says they are bad. The problem is that most people
use them as general purpose diodes because they are cheap, fast, and
very common and forget to look at the specs. I suspect people forget
that they were designed (back when the earth was cooling) as an
extremely fast SWITCHING diode. Leakage was irrelivant. My understanding
is that the very thing that makes them fast (heavy gold doping) is also
what makes them leak like a sieve. One wouldn't use a 2N3904 as a
microwave transistor, so why use a switching diode in an ultra low
leakage application?
Bob.
Win Hill
>
> Winfield Hill <hi...@rowland.org> wrote:
>
>> Here are a set of I-V measurements for your edification.
>>
>> 1n5819 1n6263 1n4153 1n458 PAD-1 2n4117 LED
>> ----- ------ ------ ----- ----- ----- ----
>> 10mA 270 700 750 800 1330 940 1706
>> 1pA - - ** * 155 220 *
>
> Checking for log-conformity over just two decades in the
> 1uA to 1nA region would be good predictor (greater than
> 59mV per decade change in current). Which makes sense.
Both the small-signal glass silicon diodes and the Schottky
diodes fail to follow a standard 58mV/decade silicon-junction
slope, even for a few decades.
On the other hand the n-channel 2n4117 JFET's gate-to-channel
diode obeys a 58mV slope perfectly over 7 decades of current.
This is clearly seen in some graphs I made from the data,
http://people.ne.mediaone.net/whill/elec/comp/diode/diode-curves.html
Note the multiple slopes that most of the diodes have over
different current regimes. These awkward-looking curves
apply for all the parts in the drawer.
> That 1N458 is still in the groove down to 10pA, even with
> the light sensitivity (IR possibly?).
Although it's not in the 58mV groove above 300nA (it must
be a very small part operating at high current densities),
it does seem to work well down to very low currents, when
the (lightproof) box is closed. :-)
The 2nd logarithmic-presentation graph shows the currents
through the diodes down to just 1mV of forward voltage.
> Those Schottky diodes are a bit of a surprising ambush
> though, because Schottky are exactly the ones that are
> tempting to use for ic input protection applications.
Indeed.
Thanks,
- Win
Winfield Hill
Rowland Institute for Science
100 Edwin Land Blvd
Cambridge, MA 02142-1297
Tony Williams
Win Hill <wh...@mediaone.net> wrote:
> This is clearly seen in some graphs I made from the data,
> http://people.ne.mediaone.net/whill/elec/comp/diode/diode-curves.html
It should be noted that Siliconix spec'd the
PAD-1 for reverse-biassed apps only (although
their little app circuits did use PADs in //
opposition across the inputs of an opamp).
The later JPAD-series do have a bullet-point
that claims "Wide-ranging logarithmic apps",
so these might be more suitable for fwd use.
In the Farnell cat though, the 2N4117, PAD-5,
and JPAD50 are all fairly expensive, much too
expensive for routine applications. So what
is needed here is the cheap jellybean part.
Since log-conformity could be a good indicator
maybe something like an npn transistor, with
b-c shorted together? What would be needed
there is a transistor that maintains a decent
hfe (>20?) down to very low currents.
So what we need is someone to search the data
sheets for suitable transistors, buy a selection,
and test them............ Win?
--
Tony Williams.
Win Hill
>
> It should be noted that Siliconix spec'd the
Right. An early concern of this thread was with the effect
a protection diode (or a part used as a diode) would have on
an opamp summing junction, if placed across it to ground.
An alternate issue concerns the use of "diodes" for generating
small biasing voltages. Over the years I've made measurements
useful to answer these concerns, these are summarized in two
plots in my correspondence web page,
http://people.ne.mediaone.net/whill/elec/comp/diode/diode-curves.html
The second curve examines the 1mV to 100mV region.
Ideally diodes have no current in their reverse direction,
so this region was not plotted. However, at low voltages
the diodes in the 1-100mV curves act like simple resistors,
passing current equally in both directions, with rather low
resistance values at that!
1n4148 family - 10M
1n6263, 1n5711 - 100k
1n5819 family - 30k (!)
> ... [the Siliconix] app circuits did use PADs in ||
> opposition across the inputs of an opamp. ...
The PAD parts are well suited for this task, but some
JFET types are cheaper, easier-to-get substitutes.
> In the Farnell cat though, the 2N4117, PAD-5, and JPAD50
> are all fairly expensive, much too expensive for routine
> applications. So what is needed here is a cheap jellybean.
I should have pointed out that the 2n4117 family, which was
intended for audio amplifiers, mic preamps, etc., is readily
available in low-cost plastic packages. From Partminer,
Vishay Siliconix PN4117A 0.31 $US
National Semiconductor PN4117A 0.67
Vishay Siliconix 2N4117A 1.02
In general, the gate-drain diode in any small-area JFET
with an Idss under 1mA is probably suited for the task
(the '4117 is 0.25mA).
> Since log-conformity could be a good indicator maybe
> something like an npn transistor, with b-c shorted
> together? What would be needed there is a transistor
> that maintains a decent hfe (>20?) down to very low
> currents.
I'm not sure how important hfe is for this purpose, but
I agree, measurements on a cheap transistor suitable for
the low-current back-to-back diode task is a good idea...
JFET diodes do have the advantage of high reverse voltages.
Ordinary diodes have the advantage of a very small package
with a clear line indicating the cathode. I'd like a low-
cost JFET-style protection diode in a diode package!
Spehro Pefhany
> Right. An early concern of this thread was with the effect
> a protection diode (or a part used as a diode) would have on
> an opamp summing junction, if placed across it to ground.
I don't have low current measurements, but I did these measurements
a few weeks ago for another purpose.. on the protection diodes in a
PIC12C509. These particular measurements are for the internal
protection diode to Vdd.
Current Voltage
(mA) (V, relative to Vdd)
0.020 0.570
0.050 0.601
0.100 0.625
0.200 0.649
0.500 0.677
1.00 0.699
2.00 0.723
5.00 0.762
10.0 0.807
20.0 0.887
Compare a typical 1N4148:
http://www.fairchildsemi.com/ds/1N/1N4148.pdf
> 1n4148 family - 10M
> 1n6263, 1n5711 - 100k
> 1n5819 family - 30k (!)
Interesting. I suppose BAV99's would be similar to the 1N4148, but the
data sheets show some differences at the relatively high currents tested.
(1uA minimum).
It would be nice to see the low current characteristics of jellybean
transistors (eg. C1815 or 2N3904).
Best regards,
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Spehro Pefhany --"it's the network..." "The Journey is the reward"
sp...@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
Contributions invited->The AVR-gcc FAQ is at: http://www.BlueCollarLinux.com
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
John Woodgate
<to...@ledelec.demon.co.uk> wrote (in <4ab0830...@ledelec.demon.co.
uk>) about 'Diode leakage at very low voltage - JFET diodes are great -
graphs', on Mon, 27 Aug 2001:
>So what we need is someone to search the data
> sheets for suitable transistors, buy a selection,
> and test them............ Win?
National Semiconductor AN-30 uses 2N2920 dual transistors for log
converters, but I suppose this application does not depend on quite the
same parameters as that under discussion.
A.Iakovlev
> Here are a set of I-V measurements for your edification.
>
> -- Schottky -- -- silicon -- --- JFET ---
> 1n5711 1n4148 red
> 1n5819 1n6263 1n4153 1n458 PAD-1 2n4117 LED
> ----- ------ ------ ----- ----- ----- ----
> 10mA 270 700 750 800 1330 940 1706
> 1 200 375 620 670 810 790 1614
> 100uA 130 280 510 577 710 710 1535
> 10 60 210 375 490 615 640 1425
> 1 23 80 255 415 555 580 1330
> 100nA 0 10 140 330 500 525 1205
> 10 - 0 65 255 440 460 1080
> 1 - - 12 185* 370 393 950
> 100pA - - 1.3* 125* 300 335 840*
> 10 - - * 70* 230 270 *
> 1 - - ** * 155 220 *
>
> All diodes were bypassed with 10nF. Room temp was about 20C.
>
Is the 1N5712 much different from 1N5711? The datasheets (and the
price) suggest that yes but I have not yet tested. I am looking for a
low forward voltage at up to 5 mA, and low reverse current at up to
3V. The intended circuit is a simple "soft" clipper of the audio
signal (the clipping diodes biased against +U0 (-U0) - about 1v); the
obectives are to have very low distortion for the normal instant
signal values, and rapid clipping above. The source impedance is about
1k.
Thanks!
A.Iakovlev