Large-ish area flat photodiode for NIR (1092 nm)

[David Nadlinger]

Hi all,

TL;DR: Any suggestions for a photodiode which
- can be convinced to perform sensibly at 1092 nm (<1 Âľs rise time;
presumably InGaAs),
- has at least Ø1 mm active area,
- and fits into a 4 mm deep (along direction of incidence) slot
including connections?

The optical powers would be a few to a few tens of ÂľW, so QE is not
critical here. Ge might be problematic in terms of dark current, though.

———

Now, for some context: A while back, I designed these amplified
photodiode modules for intensity stabilisation/monitoring of collimated
laser beams in my lab. The sensor is a Hamamatsu S6775 Si PIN diode,
reverse-biased to ~14 V and bootstrapped with a Hobbsian JFET/PNP
circuit (RIP, BFT92!).

I'm fairly happy with the design and bandwidth and noise are adequate
for the application (~1 MHz at 2 MΩ, ~75 MHz at 3.2 kΩ – the switchable
gain was a bit tricky to get right). While it works beautifully in the
blue, rather annoyingly the response at my two longest wavelengths, 1033
nm and 1092 nm, is absolutely swamped by a slow (~30 Âľs (!)) tail. I
don't think there is anything nefarious going on beyond the textbook
carrier diffusion time constant; still, this is a bit annoying, as the
rest of the system has ~500 kHz bandwidth.

The obvious solution is to use a different sensor material for those NIR
beams, e.g. InGaAs or Ge. I'd really rather re-use the mounts designed
for the S6775 though, so the height of the diode needs to be < 4mm,
which rules out most of the typical TO-… metal-can packages. I also need
a fairly large active area, at least 1 mm^2 or so, as bodging any
focussing optics into my beam-splitter/photodiode assemblies would be a
pain.

My plan is currently to try and find an InGaAs photodiode in a thin SMD
or COB package and mount it on a small carrier PCB to emulate the
S6775's form factor. Hamamatsu's G11193-10R or G13176-010P would be an
option, but there might be better alternatives I'm not aware of.
Roithner Lasertechnik sells a near-ideal part as LAPD-3-09-17-LCC, but
the >$450 price tag is a bit steep even for this low-volume application
(maybe 10 pcs. in total over the next few years).

Thanks!

— David

 

David Nadlinger <david@klickverbot.at> wrote in
news:qdi0ui$ams$1@news.ox.ac.uk:

Hi all,

TL;DR: Any suggestions for a photodiode which
- can be convinced to perform sensibly at 1092 nm (<1 Âľs rise
time;
presumably InGaAs),
- has at least Ø1 mm active area,
- and fits into a 4 mm deep (along direction of incidence) slot
including connections?

The optical powers would be a few to a few tens of ÂľW, so QE is
not critical here. Ge might be problematic in terms of dark
current, though.

———

Now, for some context: A while back, I designed these amplified
photodiode modules for intensity stabilisation/monitoring of
collimated laser beams in my lab. The sensor is a Hamamatsu S6775
Si PIN diode, reverse-biased to ~14 V and bootstrapped with a
Hobbsian JFET/PNP circuit (RIP, BFT92!).

I'm fairly happy with the design and bandwidth and noise are
adequate for the application (~1 MHz at 2 MΊ, ~75 MHz at 3.2 kΊ
– the switchable gain was a bit tricky to get right). While it
works beautifully in the blue, rather annoyingly the response at
my two longest wavelengths, 1033 nm and 1092 nm, is absolutely
swamped by a slow (~30 Âľs (!)) tail. I don't think there is
anything nefarious going on beyond the textbook carrier diffusion
time constant; still, this is a bit annoying, as the rest of the
system has ~500 kHz bandwidth.

The obvious solution is to use a different sensor material for
those NIR beams, e.g. InGaAs or Ge. I'd really rather re-use the
mounts designed for the S6775 though, so the height of the diode
needs to be < 4mm, which rules out most of the typical TO-…
metal-can packages. I also need a fairly large active area, at
least 1 mm^2 or so, as bodging any focussing optics into my
beam-splitter/photodiode assemblies would be a pain.

My plan is currently to try and find an InGaAs photodiode in a
thin SMD or COB package and mount it on a small carrier PCB to
emulate the S6775's form factor. Hamamatsu's G11193-10R or
G13176-010P would be an option, but there might be better
alternatives I'm not aware of. Roithner Lasertechnik sells a
near-ideal part as LAPD-3-09-17-LCC, but the >$450 price tag is a
bit steep even for this low-volume application (maybe 10 pcs. in
total over the next few years).

Thanks!

— David

Perhaps an IR resistor bolometer? They have windows in front so
only let pass a specific band. Not a diode, but still could be
"read", depending on your need.

 

[Winfield Hill]

David Nadlinger wrote...

Hi all,

TL;DR: Any suggestions for a photodiode which
- can be convinced to perform sensibly at 1092 nm
(<1 Âľs rise time; presumably InGaAs),
- has at least Ø1 mm active area,
- and fits into a 4 mm deep (along direction of
incidence) slot including connections?

How about a small lens, focus to an optical fiber.
There are many fast, affordable fiber detectors.
E.g., QPhotonics QPDI-80, InGaAs, to 2GHz, $109.
I have special RIS-617 versions for these babies.
https://www.dropbox.com/sh/h2xa0m60gs1m5lw/AAB4HgPTwVq-0hV9qZxYxwJ6a?dl=0
Several are installed in our advanced AFM systems.
Free PCBs if you want to experiment.


--
Thanks,
- Win

 

[Arie de Muynck]

On 2019-06-09 06:12, David Nadlinger wrote:

Now, for some context: A while back, I designed these amplified
photodiode modules for intensity stabilisation/monitoring of collimated
laser beams in my lab. The sensor is a Hamamatsu S6775 Si PIN diode,
reverse-biased to ~14 V and bootstrapped with a Hobbsian JFET/PNP
circuit (RIP, BFT92!).

I'm fairly happy with the design and bandwidth and noise are adequate
for the application (~1 MHz at 2 MΩ, ~75 MHz at 3.2 kΩ – the switchable
gain was a bit tricky to get right). While it works beautifully in the
blue, rather annoyingly the response at my two longest wavelengths, 1033
nm and 1092 nm, is absolutely swamped by a slow (~30 Âľs (!)) tail. I
don't think there is anything nefarious going on beyond the textbook
carrier diffusion time constant; still, this is a bit annoying, as the
rest of the system has ~500 kHz bandwidth.

David,

I've had like problems with detecting short laser pulses: a long tail of
many usecs.
After some testing I discovered that the area _outside_ the ring around
the active area was the cause: any electrons generated there would,
after the pulse, slowly drift into the active area causing the tail.
First solution: hacksaw the case, and paint the ring and the area
outside it black. Problem solved.
Second solution for production: a diaphragm covering the inactive area.

Regards,
Arie de Muijnck

 

[Phil Hobbs]

On 6/9/19 12:12 AM, David Nadlinger wrote:

Hi all,

TL;DR: Any suggestions for a photodiode which
 - can be convinced to perform sensibly at 1092 nm (<1 ¾s rise time;
presumably InGaAs),
 - has at least Ø1 mm active area,
 - and fits into a 4 mm deep (along direction of incidence) slot
including connections?

The optical powers would be a few to a few tens of ÂľW, so QE is not
critical here. Ge might be problematic in terms of dark current, though.

———

Now, for some context: A while back, I designed these amplified
photodiode modules for intensity stabilisation/monitoring of collimated
laser beams in my lab. The sensor is a Hamamatsu S6775 Si PIN diode,
reverse-biased to ~14 V and bootstrapped with a Hobbsian JFET/PNP
circuit (RIP, BFT92!).

I'm fairly happy with the design and bandwidth and noise are adequate
for the application (~1 MHz at 2 MΩ, ~75 MHz at 3.2 kΩ – the switchable
gain was a bit tricky to get right). While it works beautifully in the
blue, rather annoyingly the response at my two longest wavelengths, 1033
nm and 1092 nm, is absolutely swamped by a slow (~30 Âľs (!)) tail. I
don't think there is anything nefarious going on beyond the textbook
carrier diffusion time constant; still, this is a bit annoying, as the
rest of the system has ~500 kHz bandwidth.

The obvious solution is to use a different sensor material for those NIR
beams, e.g. InGaAs or Ge. I'd really rather re-use the mounts designed
for the S6775 though, so the height of the diode needs to be < 4mm,
which rules out most of the typical TO-… metal-can packages. I also need
a fairly large active area, at least 1 mm^2 or so, as bodging any
focussing optics into my beam-splitter/photodiode assemblies would be a
pain.

My plan is currently to try and find an InGaAs photodiode in a thin SMD
or COB package and mount it on a small carrier PCB to emulate the
S6775's form factor. Hamamatsu's G11193-10R or G13176-010P would be an
option, but there might be better alternatives I'm not aware of.
Roithner Lasertechnik sells a near-ideal part as LAPD-3-09-17-LCC, but
the >$450 price tag is a bit steep even for this low-volume application
(maybe 10 pcs. in total over the next few years).

Thanks!

 — David

I don't know what you mean by 'sensibly'. Silicon will work at some
level at 1092 nm, and is *dramatically* cheaper than InGaAs, so there's
a pretty steep price/performance tradeoff around there. A silicon PD
with a heater nearby might save you some dough.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[George Herold]

On Sunday, June 9, 2019 at 11:15:06 AM UTC-4, Arie de Muynck wrote:

On 2019-06-09 06:12, David Nadlinger wrote:

Now, for some context: A while back, I designed these amplified
photodiode modules for intensity stabilisation/monitoring of collimated
laser beams in my lab. The sensor is a Hamamatsu S6775 Si PIN diode,
reverse-biased to ~14 V and bootstrapped with a Hobbsian JFET/PNP
circuit (RIP, BFT92!).

I'm fairly happy with the design and bandwidth and noise are adequate
for the application (~1 MHz at 2 MΩ, ~75 MHz at 3.2 kΩ – the switchable
gain was a bit tricky to get right). While it works beautifully in the
blue, rather annoyingly the response at my two longest wavelengths, 1033
nm and 1092 nm, is absolutely swamped by a slow (~30 Âľs (!)) tail. I
don't think there is anything nefarious going on beyond the textbook
carrier diffusion time constant; still, this is a bit annoying, as the
rest of the system has ~500 kHz bandwidth.

David,

I've had like problems with detecting short laser pulses: a long tail of
many usecs.
After some testing I discovered that the area _outside_ the ring around
the active area was the cause: any electrons generated there would,
after the pulse, slowly drift into the active area causing the tail.
First solution: hacksaw the case, and paint the ring and the area
outside it black. Problem solved.
Second solution for production: a diaphragm covering the inactive area.

Regards,
Arie de Muijnck

OH good, that was my guess too, an edge effect. Try a mask with some
black electrical tape. (That's what I did.)

Arie, it was someone here(SED) who warned me of this effect, was it you?

George H.

 

[Mikko OH2HVJ]

David Nadlinger <david@klickverbot.at> writes:

My plan is currently to try and find an InGaAs photodiode in a thin
SMD or COB package and mount it on a small carrier PCB to emulate the
S6775's form factor. Hamamatsu's G11193-10R or G13176-010P would be an
option, but there might be better alternatives I'm not aware
of.

Luna SD039-151-001 is one option if you can get them. Of those Hamamatsu
ones I selected G13176, which at least for us is about 40% lower cost
than the ceramic package G11193.

--
mikko

 

[Arie de Muynck]

On 2019-06-10 03:15, George Herold wrote:

On Sunday, June 9, 2019 at 11:15:06 AM UTC-4, Arie de Muynck wrote:
On 2019-06-09 06:12, David Nadlinger wrote:

Now, for some context: A while back, I designed these amplified
photodiode modules for intensity stabilisation/monitoring of collimated
laser beams in my lab. The sensor is a Hamamatsu S6775 Si PIN diode,
reverse-biased to ~14 V and bootstrapped with a Hobbsian JFET/PNP
circuit (RIP, BFT92!).

I'm fairly happy with the design and bandwidth and noise are adequate
for the application (~1 MHz at 2 MΩ, ~75 MHz at 3.2 kΩ – the switchable
gain was a bit tricky to get right). While it works beautifully in the
blue, rather annoyingly the response at my two longest wavelengths, 1033
nm and 1092 nm, is absolutely swamped by a slow (~30 Âľs (!)) tail. I
don't think there is anything nefarious going on beyond the textbook
carrier diffusion time constant; still, this is a bit annoying, as the
rest of the system has ~500 kHz bandwidth.

David,

I've had like problems with detecting short laser pulses: a long tail of
many usecs.
After some testing I discovered that the area _outside_ the ring around
the active area was the cause: any electrons generated there would,
after the pulse, slowly drift into the active area causing the tail.
First solution: hacksaw the case, and paint the ring and the area
outside it black. Problem solved.
Second solution for production: a diaphragm covering the inactive area.

Regards,
Arie de Muijnck

OH good, that was my guess too, an edge effect. Try a mask with some
black electrical tape. (That's what I did.)

Arie, it was someone here(SED) who warned me of this effect, was it you?

George H.

Some years ago I already described this effect in this group, don't know
to whom.

Arie.

 

[George Herold]

On Monday, June 10, 2019 at 5:50:03 AM UTC-4, Arie de Muynck wrote:

On 2019-06-10 03:15, George Herold wrote:
On Sunday, June 9, 2019 at 11:15:06 AM UTC-4, Arie de Muynck wrote:
On 2019-06-09 06:12, David Nadlinger wrote:

Now, for some context: A while back, I designed these amplified
photodiode modules for intensity stabilisation/monitoring of collimated
laser beams in my lab. The sensor is a Hamamatsu S6775 Si PIN diode,
reverse-biased to ~14 V and bootstrapped with a Hobbsian JFET/PNP
circuit (RIP, BFT92!).

I'm fairly happy with the design and bandwidth and noise are adequate
for the application (~1 MHz at 2 MΩ, ~75 MHz at 3.2 kΩ – the switchable
gain was a bit tricky to get right). While it works beautifully in the
blue, rather annoyingly the response at my two longest wavelengths, 1033
nm and 1092 nm, is absolutely swamped by a slow (~30 Âľs (!)) tail. I
don't think there is anything nefarious going on beyond the textbook
carrier diffusion time constant; still, this is a bit annoying, as the
rest of the system has ~500 kHz bandwidth.

David,

I've had like problems with detecting short laser pulses: a long tail of
many usecs.
After some testing I discovered that the area _outside_ the ring around
the active area was the cause: any electrons generated there would,
after the pulse, slowly drift into the active area causing the tail.
First solution: hacksaw the case, and paint the ring and the area
outside it black. Problem solved.
Second solution for production: a diaphragm covering the inactive area..

Regards,
Arie de Muijnck

OH good, that was my guess too, an edge effect. Try a mask with some
black electrical tape. (That's what I did.)

Arie, it was someone here(SED) who warned me of this effect, was it you?

George H.


Some years ago I already described this effect in this group, don't know
to whom.

Arie.

Well then thanks again for that! I see this effect every time I use
blinking LED's to test photodiode amps. I never looked for any
wavelength dependence, maybe next time I'm testing.

George H.

 

[David Nadlinger]

Hi Mikko,

On 10.06.19 10:11 AM, Mikko OH2HVJ wrote:
> Luna SD039-151-001 is one option if you can get them.

Great suggestion, thanks – I wasn't really aware of Luna Inc.
Unfortunately, it seems they no longer manufacture that particular part,
at least there doesn't seem to be any stock at DigiKey or any of the
usual places. I've contacted them; maybe I'll get lucky.

Of those Hamamatsu ones I selected G13176, which at least for us is about 40% lower cost
than the ceramic package G11193.

Thanks, good to know.

Best,
David

 

[David Nadlinger]

Hi Arie,

On 09.06.19 4:14 PM, Arie de Muynck wrote:

After some testing I discovered that the area _outside_ the ring around
the active area was the cause: any electrons generated there would,
after the pulse, slowly drift into the active area causing the tail.

Thanks for the hint – unfortunately, the spatial distribution doesn't
seem to be the problem in this case.

Masking off everything but a central ~2.5 mm x 2 mm area with electrical
tape didn't change the response at all (apart from a small decrease in
the overall amplitude).

As the ~30 Âľs time constant makes up about 50% of the total signal,
small edge effects are rather unlikely to be the cause, I suppose.

— David

 

[David Nadlinger]

On 12.06.19 3:10 PM, David Nadlinger wrote:
> I've contacted them; maybe I'll get lucky.

It seems I might be; apparently Luna were bought by OSI last year and
are now Advanced Photonix.

— David

 

[Arie de Muynck]

On 2019-06-12 16:21, David Nadlinger wrote:

Hi Arie,

On 09.06.19 4:14 PM, Arie de Muynck wrote:
After some testing I discovered that the area _outside_ the ring
around the active area was the cause: any electrons generated there
would, after the pulse, slowly drift into the active area causing the
tail.

Thanks for the hint – unfortunately, the spatial distribution doesn't
seem to be the problem in this case.

Masking off everything but a central ~2.5 mm x 2 mm area with electrical
tape didn't change the response at all (apart from a small decrease in
the overall amplitude).

As the ~30 Âľs time constant makes up about 50% of the total signal,
small edge effects are rather unlikely to be the cause, I suppose.

 — David

No, that rules out the edge effect as major contributor.
In my case it was about 10% of the amplitude.

You do have _stiff_ biasing? Maybe amplifier related causes?
A diagram with some signal level indications I could review.

Arie

 

[David Nadlinger]

On 12.06.19 8:43 PM, Arie de Muynck wrote:

You do have _stiff_ biasing? Maybe amplifier related causes?
A diagram with some signal level indications I could review.

The very same photodiode module shows clean <40 ns [1] rise/fall times
at 674 nm, and even does if I turn up the intensity for an
order-of-magnitude increase in photocurrent (cf. biasing). The behaviour
is also repeatable across different devices. The source of the problem
is definitely wavelength-related.

(I'd post a schematic, but apparently my web server just happened to go
down. It's an ADA4187 TIA with a BF862/BFT92 bootstrap; nothing out of
the ordinary if you are familiar with the work of the gentlemen Hobbs
and Hill.)

— David


[1] Likely another factor of ~10 faster than that, even – I just didn't
test that particular one with a faster signal source.

 

[George Herold]

On Wednesday, June 12, 2019 at 6:33:59 PM UTC-4, David Nadlinger wrote:

On 12.06.19 8:43 PM, Arie de Muynck wrote:
You do have _stiff_ biasing? Maybe amplifier related causes?
A diagram with some signal level indications I could review.

The very same photodiode module shows clean <40 ns [1] rise/fall times
at 674 nm, and even does if I turn up the intensity for an
order-of-magnitude increase in photocurrent (cf. biasing). The behaviour
is also repeatable across different devices. The source of the problem
is definitely wavelength-related.

(I'd post a schematic, but apparently my web server just happened to go
down. It's an ADA4187 TIA with a BF862/BFT92 bootstrap; nothing out of
the ordinary if you are familiar with the work of the gentlemen Hobbs
and Hill.)

— David


[1] Likely another factor of ~10 faster than that, even – I just didn't
test that particular one with a faster signal source.

Huh, Some shallow impurity (doping) in the PD? Do you have access
to a spectrometer? You can turn it monochromator* and look at the response..

Maybe some interference fringes from the coating? If you put a
little squeeze on it (maybe an alligator clip) does it change?

George H.
*weird spelling, my first guess had a meter on the end...
maybe we could all agree to call it a monochrometer,
from now on. :^)

 

[Phil Hobbs]

On 6/12/19 10:21 AM, David Nadlinger wrote:

Hi Arie,

On 09.06.19 4:14 PM, Arie de Muynck wrote:
After some testing I discovered that the area _outside_ the ring
around the active area was the cause: any electrons generated there
would, after the pulse, slowly drift into the active area causing the
tail.

Thanks for the hint – unfortunately, the spatial distribution doesn't
seem to be the problem in this case.

Masking off everything but a central ~2.5 mm x 2 mm area with electrical
tape didn't change the response at all (apart from a small decrease in
the overall amplitude).

As the ~30 Âľs time constant makes up about 50% of the total signal,
small edge effects are rather unlikely to be the cause, I suppose.

 — David

Sounds like you haven't depleted the PD all the way to the back contact.
That causes diffusion tails selectively at longer wavelengths. If you
crank up the bias, it's common to see gradual improvement followed by a
dramatic speed-up when the die is fully depleted.

This isn't usually possible with PN diodes, but it usually is with PINs.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[David Nadlinger]

On 13.06.19 2:35 AM, Phil Hobbs wrote:

Sounds like you haven't depleted the PD all the way to the back contact.
 That causes diffusion tails selectively at longer wavelengths.  If you
crank up the bias, it's common to see gradual improvement followed by a
dramatic speed-up when the die is fully depleted.

Yes, seems like that is what must be happening. I am already biasing the
diode at 14 V, so didn't expect to see this, but then again, the S6775
is probably rather wide (which matches the 35 V abs. max rating).

Would the amount of charges generated outside the depletion region grow
more or less proportional to the reduction in QE/absorption coefficient
near the cut-off? Is there a way to estimate the punch-through voltage
from looking at typical datasheet capacitance/dark current diagrams? I
really need to have a more principled look at detector physics one of
these days.

I'll try to get some step response vs. reverse bias data next week, at a
few different wavelengths.

— David

 

[David Nadlinger]

On 09.06.19 1:51 PM, Winfield Hill wrote:

How about a small lens, focus to an optical fiber.
There are many fast, affordable fiber detectors.
E.g., QPhotonics QPDI-80, InGaAs, to 2GHz, $109.

In this instance, I have some 35 of these detectors on a single optical
table, so I wanted to avoid the extra complexity that comes with
fibre-coupling. The beam sampler modules use large-area photodiodes and
integrate PCBs as well as beamsplitters/filters to reduce alignment
complexity. (There are already ≈40 fibre collimators on that table to
worry about.)

I have special RIS-617 versions for these babies.
https://www.dropbox.com/sh/h2xa0m60gs1m5lw/AAB4HgPTwVq-0hV9qZxYxwJ6a?dl=0
Several are installed in our advanced AFM systems.
Free PCBs if you want to experiment.

Many thanks for this very kind offer! It was also interesting to flip
through your designs. Mine are not entirely dissimilar, which might have
something to do with the fact that I am using AoE as my monitor stand
(those references to 3.x in the main text sure were tantalizing when
first looking into it – I am definitely looking forward to the
x-chapters release).

Did you ever look at negative output swing on the ADA4817? I am using it
with positive photodiode bias on said boards, and measuring only ≈ -3.6
V swing off Âą5 V supplies. This is in violation of the datasheet specs,
but reproducible in isolation (>> 1 kΊ output load) although all samples
I tried were from the same reel. Had I known this, I would have chosen
different supplies. The joys of iterative design…

— David

 

[Phil Hobbs]

On 7/12/19 11:47 PM, David Nadlinger wrote:

On 13.06.19 2:35 AM, Phil Hobbs wrote:
Sounds like you haven't depleted the PD all the way to the back
contact.   That causes diffusion tails selectively at longer
wavelengths.  If you crank up the bias, it's common to see gradual
improvement followed by a dramatic speed-up when the die is fully
depleted.

Yes, seems like that is what must be happening. I am already biasing the
diode at 14 V, so didn't expect to see this, but then again, the S6775
is probably rather wide (which matches the 35 V abs. max rating).

Would the amount of charges generated outside the depletion region grow
more or less proportional to the reduction in QE/absorption coefficient
near the cut-off?

It depends on the details of the doping profile. One common method for
making ohmic contacts is to crank up the local dopant density till the
band gap goes away, in which case you're never going to get it fully
depleted.

Is there a way to estimate the punch-through voltage
from looking at typical datasheet capacitance/dark current diagrams? I
really need to have a more principled look at detector physics one of
these days.

Not that I know of. Photodiode datasheets are almost as bad as diode
laser datasheets.

They never, ever tell you about the series resistance, which is a
crucial parameter for low-noise photoreceiver design.

I'll try to get some step response vs. reverse bias data next week, at a
few different wavelengths.

I'd be interested to see them.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[David Nadlinger]

On 13.07.19 7:42 AM, Phil Hobbs wrote:

They never, ever tell you about the series resistance, which is a
crucial parameter for low-noise photoreceiver design.

Yes, that's pretty annoying. I have some VNA S11 traces of the S6775 vs.
reverse bias, but they are a bit harder to interpret than I'd hoped. The
real part of the impedance at 10 MHz ranges from 110 Ί at 66 mV bias to
80 Ί at 15 V. Makes sense; the width of the undepleted region gets
smaller. However, there is a resonance around 410 MHz that doesn't move
with bias voltage – its residual (real) resistance at 66 mV bias is 15.5
Ί and reduces slightly to 13 Ί at 15 V.

This very much doesn't look like a current source with shunt C and
series/shunt R. Just soldering a capacitor on legs into the production
version of my frontend PCB would probably be the easiest way of figuring
out the (noise-)equivalent series resistance for later reference.

I wish I had the time to buy a few of every (sort-of-)large-area
photodiode out there and compile a database of parasitics as a function
of reverse bias. This would be seriously useful.

— David