Fast, efficient IR LEDs?

John Larkin wrote:
On Fri, 05 Feb 2010 11:56:15 -0500, Phil Hobbs
pcdhSpamMeSenseless_at_electrooptical.net> wrote:

I need a fast IR LED (> 20 MHz, < 50 pF) for an optical feedback gizmo.
I have some Stanley DN310s, but they've been discontinued. Other
possibilities are:

Vishay TSFF5410 -- 870 nm, 0.% W/A typ 15 ns rise/fall, 125 pF typ
Vishay VSLB3940 -- 940 nm, 0.4 W/A typ 15 ns rise/fall, 70 pF typ
Panasonic LNA4905L -- 880 nm, 0.3 W/A min 30 MHz typ, no other specs
Osram SFH4550 -- 850 nm, 0.5 W/A typ 12 ns rise/fall, no C spec

It would be really nice to find something with a flat front facet and
(especially) lower capacitance, because it has to work at quite low
currents (5-10 uA).

Any suggestions?

Thanks

Phil Hobbs

PS: Amazing how we're actually talking about electronics at the moment!

Maybe use a visible part? They seem to get the most development effort
lately. I'll measure the capacitance on some of the right-angle
surface-mount Osram parts we use. They are blindingly bright, clearly
on at 1 uA in normal office lighting.

The red response of a silicon detector isn't much below the IR peak.

And now, back to politics...

John

Yes!! Bash them Russkies AKA REDs!

On 2/7/2010 5:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 4:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 12:29 PM, JosephKK wrote:

[...]

You may wish to consider a laser diode operating below critical
current.


Thanks, I know that trick. Thing is, I need a 5000:1 output power
range, or thereabouts--i.e. 3 uW - 15 mW. The bandwidth is going to be
way more than enough at the high end, and the problem is to keep the
feedback poles from crossing at a frequency where there's over-unity
gain.

There are other approaches possible that require different approaches,
but they require more tweaking--e.g. two ranges with two LEDs using
different optical coupling fractions.


Or have an offset in there where the LED (or LD below lasing threshold
as Joseph suggested) runs at a regulated base power level. BTDT, but in
my case that was in order to remain above lasing threshold.


This gizmo is an advanced photoreceiver that maintains
shot-noise-limited performance (2 dB above shot noise) from ~10 nA to
100 uA, with an honest 1 MHz bandwidth over (almost) the whole range.
Doing that down near the minimum photocurrent is a real genuine
parlour trick.


Luckily I never had to do that. BW was always tens of MHz but they gave
me plenty of amplitude to work with. However, up there on that pedestal
it had to be super low noise because we had to extract modulation.


The ones uses two photodiodes wired in series (!) to get a
sub-Poissonian photocurrent to null out the primary photocurrent.
That's a trick I've never seen before, so I might have invented it. It
obviously requires some careful feedback to keep the currents in
balance, but the result is a nice linear photoreceiver with almost no
additional input capacitance.


Neat! But now you've spilled the beans and can't patent it :-(

Patents aren't worth much anyhow these days. Seems like most of what
they do is trigger patent trolls who then bog down whole businesses.


I can patent it for the next year, at least in the USA. I might do
that, we'll see.


Two photodiodes in series have the same photocurrent but *half the
shot noise*, so the cancellation current is actually quieter than the
photocurrent, without needing resistive degeneration. (I also manage
to keep all 300-kelvin resistors out of the signal path, which is key.)

The optical feedback is sort of a poor-man's photomultiplier: most of
the LED light goes to another photodiode, driving an ordinary TIA
which produces the output. It's a really sweet solution overall, with
the one disadvantage that it needs two tweaks.


I assume you mean the balancing of the two PDs in series. Is there no
way to servo that? Maybe by occasionally interrupting the optical path?


There's a bias feedback loop that looks after that. It doesn't have to
be that accurate since the PDs run at 14V of reverse bias--keeping the
junction of the two PDs reasonably still is all that's required.

The tweaks are for making sure that the two photocurrents are reasonably
close to begin with, and to govern the poorly specified efficiency of
the LEDs. (IR LEDs have output power specs that are almost as loose as
the V_T spec of your average JFET.)

You should be able to buy them in a couple of months, if all goes well.
(No home should be without one, after all.) ;)

Cheers

Phil Hobbs

On Sun, 07 Feb 2010 17:27:11 -0500, Phil Hobbs
pcdhSpamMeSenseless_at_electrooptical.net> wrote:

On 2/7/2010 5:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 4:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 12:29 PM, JosephKK wrote:
[...]

You may wish to consider a laser diode operating below critical
current.

Thanks, I know that trick. Thing is, I need a 5000:1 output power
range, or thereabouts--i.e. 3 uW - 15 mW. The bandwidth is going to be
way more than enough at the high end, and the problem is to keep the
feedback poles from crossing at a frequency where there's over-unity
gain.

There are other approaches possible that require different approaches,
but they require more tweaking--e.g. two ranges with two LEDs using
different optical coupling fractions.

Or have an offset in there where the LED (or LD below lasing threshold
as Joseph suggested) runs at a regulated base power level. BTDT, but in
my case that was in order to remain above lasing threshold.

This gizmo is an advanced photoreceiver that maintains
shot-noise-limited performance (2 dB above shot noise) from ~10 nA to
100 uA, with an honest 1 MHz bandwidth over (almost) the whole range.
Doing that down near the minimum photocurrent is a real genuine
parlour trick.

Luckily I never had to do that. BW was always tens of MHz but they gave
me plenty of amplitude to work with. However, up there on that pedestal
it had to be super low noise because we had to extract modulation.


The ones uses two photodiodes wired in series (!) to get a
sub-Poissonian photocurrent to null out the primary photocurrent.
That's a trick I've never seen before, so I might have invented it. It
obviously requires some careful feedback to keep the currents in
balance, but the result is a nice linear photoreceiver with almost no
additional input capacitance.

Neat! But now you've spilled the beans and can't patent it :-(

Patents aren't worth much anyhow these days. Seems like most of what
they do is trigger patent trolls who then bog down whole businesses.

I can patent it for the next year, at least in the USA. I might do
that, we'll see.

Two photodiodes in series have the same photocurrent but *half the
shot noise*, so the cancellation current is actually quieter than the
photocurrent, without needing resistive degeneration. (I also manage
to keep all 300-kelvin resistors out of the signal path, which is key.)

The optical feedback is sort of a poor-man's photomultiplier: most of
the LED light goes to another photodiode, driving an ordinary TIA
which produces the output. It's a really sweet solution overall, with
the one disadvantage that it needs two tweaks.

I assume you mean the balancing of the two PDs in series. Is there no
way to servo that? Maybe by occasionally interrupting the optical path?

There's a bias feedback loop that looks after that. It doesn't have to
be that accurate since the PDs run at 14V of reverse bias--keeping the
junction of the two PDs reasonably still is all that's required.

The tweaks are for making sure that the two photocurrents are reasonably
close to begin with, and to govern the poorly specified efficiency of
the LEDs. (IR LEDs have output power specs that are almost as loose as
the V_T spec of your average JFET.)

You should be able to buy them in a couple of months, if all goes well.
(No home should be without one, after all.) ;)

Cheers

Phil Hobbs

Just for the heck of it, I asked Jonathan to measure the low-current
linearity of some visible and IR led's. I know that some LEDs can make
visible light at 1 nA, so it will be interesting to see if there is a
linearity knee somewhere. A red LED driving a silicon PIN diode makes
a visually perfect straight-line graph plotted linearly from 0 to 55
mA.

I theory, LED voltage is the log of current, so at some very low
current there won't be enough voltage across the junction to make a
photon of anywhere near the expected wavelength. It could be that
materials defects will kill things before that point.

I did some googling on LED behavior at low currents and found nothing
useful.

I theory, LED voltage is the log of current, so at some very low
current there won't be enough voltage across the junction to make a
photon of anywhere near the expected wavelength. It could be that
materials defects will kill things before that point.

"John Larkin" <jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote in message
news:iet0n5pn6qdm16unmbo0fucgv48k4hossh_at_4ax.com...
I theory, LED voltage is the log of current, so at some very low
current there won't be enough voltage across the junction to make a
photon of anywhere near the expected wavelength. It could be that
materials defects will kill things before that point.

Well, everything is radiating, although at 26meV and no bias, there's damned
little all the way out at 2eV. You're looking at the tail end of two
decades of exponent there.

There's no reason why, for instance, you can't get 2.000eV photons from a
semiconductor with 1.984V across it. ...

... You can electrolyze water the same
way -- it does proceed below the reaction voltage, it's just endothermic and
slow as hell. You don't get something for nothing, so there's still current
draw, with an electron plopping through the bandgap for every photon
radiated, it's just falling through a slightly lower voltage.

Question for Phil: does the process of photon production cause a blip in the
diode voltage? This should be detectable as shot noise on the diode's
terminal voltage -and- on a photodiode directly in front of the LED, and
there should be perfect correlation between the two effects (minus quantum
efficiency, so maybe you'll only detect 1 in 5 events at the photodiode).
Is this measurable? I think it should be.

"John Larkin" <jjlarkin_at_highNOTlandTHIStechnologyPART.com> wrote in message
news:iet0n5pn6qdm16unmbo0fucgv48k4hossh_at_4ax.com...
I theory, LED voltage is the log of current, so at some very low
current there won't be enough voltage across the junction to make a
photon of anywhere near the expected wavelength. It could be that
materials defects will kill things before that point.

Well, everything is radiating, although at 26meV and no bias, there's damned
little all the way out at 2eV. You're looking at the tail end of two
decades of exponent there.

There's no reason why, for instance, you can't get 2.000eV photons from a
semiconductor with 1.984V across it. You can electrolyze water the same
way -- it does proceed below the reaction voltage, it's just endothermic and
slow as hell. You don't get something for nothing, so there's still current
draw, with an electron plopping through the bandgap for every photon
radiated, it's just falling through a slightly lower voltage.

Question for Phil: does the process of photon production cause a blip in the
diode voltage? This should be detectable as shot noise on the diode's
terminal voltage -and- on a photodiode directly in front of the LED, and
there should be perfect correlation between the two effects (minus quantum
efficiency, so maybe you'll only detect 1 in 5 events at the photodiode).
Is this measurable? I think it should be.

Tim

On Sun, 07 Feb 2010 14:10:48 -0800, Joerg<invalid_at_invalid.invalid
wrote:

Phil Hobbs wrote:
On 2/7/2010 4:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 12:29 PM, JosephKK wrote:

[...]

You may wish to consider a laser diode operating below critical
current.


Thanks, I know that trick. Thing is, I need a 5000:1 output power
range, or thereabouts--i.e. 3 uW - 15 mW. The bandwidth is going to be
way more than enough at the high end, and the problem is to keep the
feedback poles from crossing at a frequency where there's over-unity
gain.

There are other approaches possible that require different approaches,
but they require more tweaking--e.g. two ranges with two LEDs using
different optical coupling fractions.


Or have an offset in there where the LED (or LD below lasing threshold
as Joseph suggested) runs at a regulated base power level. BTDT, but in
my case that was in order to remain above lasing threshold.


This gizmo is an advanced photoreceiver that maintains
shot-noise-limited performance (2 dB above shot noise) from ~10 nA to
100 uA, with an honest 1 MHz bandwidth over (almost) the whole range.
Doing that down near the minimum photocurrent is a real genuine parlour
trick.


Luckily I never had to do that. BW was always tens of MHz but they gave
me plenty of amplitude to work with. However, up there on that pedestal
it had to be super low noise because we had to extract modulation.


The ones uses two photodiodes wired in series (!) to get a
sub-Poissonian photocurrent to null out the primary photocurrent. That's
a trick I've never seen before, so I might have invented it. It
obviously requires some careful feedback to keep the currents in
balance, but the result is a nice linear photoreceiver with almost no
additional input capacitance.


Neat! But now you've spilled the beans and can't patent it :-(

Patents aren't worth much anyhow these days. Seems like most of what
they do is trigger patent trolls who then bog down whole businesses.


Two photodiodes in series have the same photocurrent but *half the shot
noise*, so the cancellation current is actually quieter than the
photocurrent, without needing resistive degeneration. (I also manage to
keep all 300-kelvin resistors out of the signal path, which is key.)

The optical feedback is sort of a poor-man's photomultiplier: most of
the LED light goes to another photodiode, driving an ordinary TIA which
produces the output. It's a really sweet solution overall, with the one
disadvantage that it needs two tweaks.


I assume you mean the balancing of the two PDs in series. Is there no
way to servo that? Maybe by occasionally interrupting the optical path?

Sounds fascinating! More info please

...Jim Thompson

On 2/7/2010 5:17 PM, Jim Thompson wrote:
On Sun, 07 Feb 2010 14:10:48 -0800, Joerg<invalid_at_invalid.invalid
wrote:

Phil Hobbs wrote:
On 2/7/2010 4:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 12:29 PM, JosephKK wrote:

[...]

You may wish to consider a laser diode operating below critical
current.


Thanks, I know that trick. Thing is, I need a 5000:1 output power
range, or thereabouts--i.e. 3 uW - 15 mW. The bandwidth is going to be
way more than enough at the high end, and the problem is to keep the
feedback poles from crossing at a frequency where there's over-unity
gain.

There are other approaches possible that require different approaches,
but they require more tweaking--e.g. two ranges with two LEDs using
different optical coupling fractions.


Or have an offset in there where the LED (or LD below lasing threshold
as Joseph suggested) runs at a regulated base power level. BTDT, but in
my case that was in order to remain above lasing threshold.


This gizmo is an advanced photoreceiver that maintains
shot-noise-limited performance (2 dB above shot noise) from ~10 nA to
100 uA, with an honest 1 MHz bandwidth over (almost) the whole range.
Doing that down near the minimum photocurrent is a real genuine parlour
trick.


Luckily I never had to do that. BW was always tens of MHz but they gave
me plenty of amplitude to work with. However, up there on that pedestal
it had to be super low noise because we had to extract modulation.


The ones uses two photodiodes wired in series (!) to get a
sub-Poissonian photocurrent to null out the primary photocurrent. That's
a trick I've never seen before, so I might have invented it. It
obviously requires some careful feedback to keep the currents in
balance, but the result is a nice linear photoreceiver with almost no
additional input capacitance.


Neat! But now you've spilled the beans and can't patent it :-(

Patents aren't worth much anyhow these days. Seems like most of what
they do is trigger patent trolls who then bog down whole businesses.


Two photodiodes in series have the same photocurrent but *half the shot
noise*, so the cancellation current is actually quieter than the
photocurrent, without needing resistive degeneration. (I also manage to
keep all 300-kelvin resistors out of the signal path, which is key.)

The optical feedback is sort of a poor-man's photomultiplier: most of
the LED light goes to another photodiode, driving an ordinary TIA which
produces the output. It's a really sweet solution overall, with the one
disadvantage that it needs two tweaks.


I assume you mean the balancing of the two PDs in series. Is there no
way to servo that? Maybe by occasionally interrupting the optical path?

Sounds fascinating! More info please

...Jim Thompson

Normally of course you can't put current sources in series, but by
applying optical feedback you can make series-connected photodiodes work.

Since the diodes have essentially infinite impedance, the noise of each
one splits in half, with half going through each capacitor.

0 +Vbias
|
|
|
*------*
| |
| |
| |
--- ---
/ \ C
--- ---
| |
| |
| |
*------*
| |
| |
| |
--- ---
/ \ C
--- ---
| |
| |
| |
*------*
|
|
0 To TIA

The shot noise current from each diode divides by the ratio of the
conductances of the two branches. Thus goes through the other diode's
capacitance, and so into the external circuit, but half just circulates
round through its own capacitance and hence doesn't contribute to the
output noise.

The shot noise currents from the two diodes are uncorrelated, and so the
RMS noise current arriving in the external circuit is


/ |2eI_dc| |2eI_dc| \
i_N = sqrt| |------| + |------| | = sqrt(eI_dc)
\ | 2 | | 2 | /

which is 3 dB below the shot noise of a primary photocurrent of I_dc.
That's the same SNR you'd get by parallelling the two, but (crucially)
you don't double the capacitance or the photocurrent by doing so.

That makes it a good trick for low photocurrents, though not one you'd
use every day.

Cheers

Phil Hobbs

Phil Hobbs wrote:
On 2/7/2010 5:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 4:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 12:29 PM, JosephKK wrote:

[...]

You may wish to consider a laser diode operating below critical
current.


Thanks, I know that trick. Thing is, I need a 5000:1 output power
range, or thereabouts--i.e. 3 uW - 15 mW. The bandwidth is going
to be
way more than enough at the high end, and the problem is to keep the
feedback poles from crossing at a frequency where there's over-unity
gain.

There are other approaches possible that require different
approaches,
but they require more tweaking--e.g. two ranges with two LEDs using
different optical coupling fractions.


Or have an offset in there where the LED (or LD below lasing threshold
as Joseph suggested) runs at a regulated base power level. BTDT,
but in
my case that was in order to remain above lasing threshold.


This gizmo is an advanced photoreceiver that maintains
shot-noise-limited performance (2 dB above shot noise) from ~10 nA to
100 uA, with an honest 1 MHz bandwidth over (almost) the whole range.
Doing that down near the minimum photocurrent is a real genuine
parlour trick.


Luckily I never had to do that. BW was always tens of MHz but they gave
me plenty of amplitude to work with. However, up there on that pedestal
it had to be super low noise because we had to extract modulation.


The ones uses two photodiodes wired in series (!) to get a
sub-Poissonian photocurrent to null out the primary photocurrent.
That's a trick I've never seen before, so I might have invented it. It
obviously requires some careful feedback to keep the currents in
balance, but the result is a nice linear photoreceiver with almost no
additional input capacitance.


Neat! But now you've spilled the beans and can't patent it :-(

Patents aren't worth much anyhow these days. Seems like most of what
they do is trigger patent trolls who then bog down whole businesses.


I can patent it for the next year, at least in the USA. I might do
that, we'll see.


Just don't wait until T minus 360 days :-)



Two photodiodes in series have the same photocurrent but *half the
shot noise*, so the cancellation current is actually quieter than the
photocurrent, without needing resistive degeneration. (I also manage
to keep all 300-kelvin resistors out of the signal path, which is key.)

The optical feedback is sort of a poor-man's photomultiplier: most of
the LED light goes to another photodiode, driving an ordinary TIA
which produces the output. It's a really sweet solution overall, with
the one disadvantage that it needs two tweaks.


I assume you mean the balancing of the two PDs in series. Is there no
way to servo that? Maybe by occasionally interrupting the optical path?


There's a bias feedback loop that looks after that. It doesn't have to
be that accurate since the PDs run at 14V of reverse bias--keeping the
junction of the two PDs reasonably still is all that's required.


Good, so it seems automatic. 14V sound like a white-knuckle ride

Not at all. Most Si photodiodes are good to 30-60V. With many (e.g.
the BPW34) the capacitance stops decreasing at 10V or so, but they
continue to speed up with increasing bias, because the series resistance
keeps going down. That's because it's dominated by the (very thin)
diffusion zone, so cranking up the bias until they're really really
fully depleted makes the speed go up amazingly. Silvio Donati's book on
photodetectors is an excellent read for this sort of stuff.

Cheers

Phil Hobbs

On 2/8/2010 12:14 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 5:10 PM, Joerg wrote:
Phil Hobbs wrote:

The optical feedback is sort of a poor-man's photomultiplier: most of
the LED light goes to another photodiode, driving an ordinary TIA
which produces the output. It's a really sweet solution overall, with
the one disadvantage that it needs two tweaks.


I assume you mean the balancing of the two PDs in series. Is there no
way to servo that? Maybe by occasionally interrupting the optical path?


There's a bias feedback loop that looks after that. It doesn't have to
be that accurate since the PDs run at 14V of reverse bias--keeping the
junction of the two PDs reasonably still is all that's required.


Good, so it seems automatic. 14V sound like a white-knuckle ride :-)

Not at all. Most Si photodiodes are good to 30-60V. With many (e.g.
the BPW34) the capacitance stops decreasing at 10V or so, but they
continue to speed up with increasing bias, because the series resistance
keeps going down. That's because it's dominated by the (very thin)
diffusion zone, so cranking up the bias until they're really really
fully depleted makes the speed go up amazingly. Silvio Donati's book on
photodetectors is an excellent read for this sort of stuff.

On Sun, 07 Feb 2010 17:27:11 -0500, Phil Hobbs
pcdhSpamMeSenseless_at_electrooptical.net> wrote:

On 2/7/2010 5:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 4:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 12:29 PM, JosephKK wrote:
[...]

You may wish to consider a laser diode operating below critical
current.

Thanks, I know that trick. Thing is, I need a 5000:1 output power
range, or thereabouts--i.e. 3 uW - 15 mW. The bandwidth is going to be
way more than enough at the high end, and the problem is to keep the
feedback poles from crossing at a frequency where there's over-unity
gain.

There are other approaches possible that require different approaches,
but they require more tweaking--e.g. two ranges with two LEDs using
different optical coupling fractions.

Or have an offset in there where the LED (or LD below lasing threshold
as Joseph suggested) runs at a regulated base power level. BTDT, but in
my case that was in order to remain above lasing threshold.

This gizmo is an advanced photoreceiver that maintains
shot-noise-limited performance (2 dB above shot noise) from ~10 nA to
100 uA, with an honest 1 MHz bandwidth over (almost) the whole range.
Doing that down near the minimum photocurrent is a real genuine
parlour trick.

Luckily I never had to do that. BW was always tens of MHz but they gave
me plenty of amplitude to work with. However, up there on that pedestal
it had to be super low noise because we had to extract modulation.


The ones uses two photodiodes wired in series (!) to get a
sub-Poissonian photocurrent to null out the primary photocurrent.
That's a trick I've never seen before, so I might have invented it. It
obviously requires some careful feedback to keep the currents in
balance, but the result is a nice linear photoreceiver with almost no
additional input capacitance.

Neat! But now you've spilled the beans and can't patent it :-(

Patents aren't worth much anyhow these days. Seems like most of what
they do is trigger patent trolls who then bog down whole businesses.

I can patent it for the next year, at least in the USA. I might do
that, we'll see.

Two photodiodes in series have the same photocurrent but *half the
shot noise*, so the cancellation current is actually quieter than the
photocurrent, without needing resistive degeneration. (I also manage
to keep all 300-kelvin resistors out of the signal path, which is key.)

The optical feedback is sort of a poor-man's photomultiplier: most of
the LED light goes to another photodiode, driving an ordinary TIA
which produces the output. It's a really sweet solution overall, with
the one disadvantage that it needs two tweaks.

I assume you mean the balancing of the two PDs in series. Is there no
way to servo that? Maybe by occasionally interrupting the optical path?

There's a bias feedback loop that looks after that. It doesn't have to
be that accurate since the PDs run at 14V of reverse bias--keeping the
junction of the two PDs reasonably still is all that's required.

The tweaks are for making sure that the two photocurrents are reasonably
close to begin with, and to govern the poorly specified efficiency of
the LEDs. (IR LEDs have output power specs that are almost as loose as
the V_T spec of your average JFET.)

You should be able to buy them in a couple of months, if all goes well.
(No home should be without one, after all.) ;)

Cheers

Phil Hobbs

Just for the heck of it, I asked Jonathan to measure the low-current
linearity of some visible and IR led's. I know that some LEDs can make
visible light at 1 nA, so it will be interesting to see if there is a
linearity knee somewhere. A red LED driving a silicon PIN diode makes
a visually perfect straight-line graph plotted linearly from 0 to 55
mA.

I theory, LED voltage is the log of current, so at some very low
current there won't be enough voltage across the junction to make a
photon of anywhere near the expected wavelength. It could be that
materials defects will kill things before that point.

I did some googling on LED behavior at low currents and found nothing
useful.

John

Please be so kind as to give a few makers and part numbers for

John Larkin wrote:
On Sun, 07 Feb 2010 17:27:11 -0500, Phil Hobbs
pcdhSpamMeSenseless_at_electrooptical.net> wrote:

On 2/7/2010 5:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 4:10 PM, Joerg wrote:
Phil Hobbs wrote:
On 2/7/2010 12:29 PM, JosephKK wrote:
[...]

You may wish to consider a laser diode operating below critical
current.

Thanks, I know that trick. Thing is, I need a 5000:1 output power
range, or thereabouts--i.e. 3 uW - 15 mW. The bandwidth is going
to be
way more than enough at the high end, and the problem is to keep the
feedback poles from crossing at a frequency where there's over-unity
gain.

There are other approaches possible that require different
approaches,
but they require more tweaking--e.g. two ranges with two LEDs using
different optical coupling fractions.

Or have an offset in there where the LED (or LD below lasing
threshold
as Joseph suggested) runs at a regulated base power level. BTDT,
but in
my case that was in order to remain above lasing threshold.

This gizmo is an advanced photoreceiver that maintains
shot-noise-limited performance (2 dB above shot noise) from ~10 nA to
100 uA, with an honest 1 MHz bandwidth over (almost) the whole range.
Doing that down near the minimum photocurrent is a real genuine
parlour trick.

Luckily I never had to do that. BW was always tens of MHz but they gave
me plenty of amplitude to work with. However, up there on that pedestal
it had to be super low noise because we had to extract modulation.


The ones uses two photodiodes wired in series (!) to get a
sub-Poissonian photocurrent to null out the primary photocurrent.
That's a trick I've never seen before, so I might have invented it. It
obviously requires some careful feedback to keep the currents in
balance, but the result is a nice linear photoreceiver with almost no
additional input capacitance.

Neat! But now you've spilled the beans and can't patent it :-(

Patents aren't worth much anyhow these days. Seems like most of what
they do is trigger patent trolls who then bog down whole businesses.

I can patent it for the next year, at least in the USA. I might do
that, we'll see.

Two photodiodes in series have the same photocurrent but *half the
shot noise*, so the cancellation current is actually quieter than the
photocurrent, without needing resistive degeneration. (I also manage
to keep all 300-kelvin resistors out of the signal path, which is
key.)

The optical feedback is sort of a poor-man's photomultiplier: most of
the LED light goes to another photodiode, driving an ordinary TIA
which produces the output. It's a really sweet solution overall, with
the one disadvantage that it needs two tweaks.

I assume you mean the balancing of the two PDs in series. Is there no
way to servo that? Maybe by occasionally interrupting the optical path?

There's a bias feedback loop that looks after that. It doesn't have
to be that accurate since the PDs run at 14V of reverse bias--keeping
the junction of the two PDs reasonably still is all that's required.

The tweaks are for making sure that the two photocurrents are
reasonably close to begin with, and to govern the poorly specified
efficiency of the LEDs. (IR LEDs have output power specs that are
almost as loose as the V_T spec of your average JFET.)

You should be able to buy them in a couple of months, if all goes
well. (No home should be without one, after all.) ;)

Cheers

Phil Hobbs

Just for the heck of it, I asked Jonathan to measure the low-current
linearity of some visible and IR led's. I know that some LEDs can make
visible light at 1 nA, so it will be interesting to see if there is a
linearity knee somewhere. A red LED driving a silicon PIN diode makes
a visually perfect straight-line graph plotted linearly from 0 to 55
mA.

I theory, LED voltage is the log of current, so at some very low
current there won't be enough voltage across the junction to make a
photon of anywhere near the expected wavelength. It could be that
materials defects will kill things before that point.

I did some googling on LED behavior at low currents and found nothing
useful.


Academics have tried to do single-photon generation with LEDs. Pulsed at
very low current. IIRC one paper was from Syracuse, NY. But I don't
think you get access to this stuff directly on the web, probably needs
some paid access like IEEE Explore. Or good connections to a university.

...and HOW does one determine / detect a single photon? Tap into nerve of

On Mon, 08 Feb 2010 20:27:22 -0500, Phil Hobbs
pcdhSpamMeSenseless_at_electrooptical.net> wrote:

[snip]

Not at all. Most Si photodiodes are good to 30-60V. With many (e.g.
the BPW34) the capacitance stops decreasing at 10V or so, but they
continue to speed up with increasing bias, because the series resistance
keeps going down. That's because it's dominated by the (very thin)
diffusion zone, so cranking up the bias until they're really really
fully depleted makes the speed go up amazingly. Silvio Donati's book on
photodetectors is an excellent read for this sort of stuff.

Cheers

Phil Hobbs

Where do you get that book? Searching on "Silvio Donati"
"photodetector" seems to scramble google's brains

...Jim Thompson

On 2/8/2010 8:52 PM, Jim Thompson wrote:
On Mon, 08 Feb 2010 20:27:22 -0500, Phil Hobbs
pcdhSpamMeSenseless_at_electrooptical.net> wrote:

[snip]

Not at all. Most Si photodiodes are good to 30-60V. With many (e.g.
the BPW34) the capacitance stops decreasing at 10V or so, but they
continue to speed up with increasing bias, because the series resistance
keeps going down. That's because it's dominated by the (very thin)
diffusion zone, so cranking up the bias until they're really really
fully depleted makes the speed go up amazingly. Silvio Donati's book on
photodetectors is an excellent read for this sort of stuff.

Cheers

Phil Hobbs

Where do you get that book? Searching on "Silvio Donati"
"photodetector" seems to scramble google's brains

...Jim Thompson

*Silvano* Donati, my bad.

http://www.amazon.com/exec/obidos/asin/0130203378

Cheers

Phil Hobbs