Problem with Saturating Photodiode

[Paul Hovnanian P.E.]

How can it run out of carriers? Photons make h-e pairs, and more
photons should make more, right? Does high pair density increase
recombination before the charges can be collected? Is it possible that
the photocurrent is limited by ohmic effects? The DC effect does need
pondering.

One of the makers of GaAs photodiodes told us that their devices would
limit above about a milliwatt optical input, so we got access to a
multi-watt pulsed fiber source, and it turns out they were wrong.

John

Perhaps Jim's laser is 'saturating'. Lowering the DC level to get more
AC would seem logical in this case.

 

[Joerg]

Hi Phil,

I have 10V reverse bias, but it's behaving as if the PD is running out
of carriers.

Jim, just make sure they don't change horses later and accidentally put
one in that is only rated at 5V abs max. Luckily the engineer of my
client told me before I designed the bias section that I had to also
accommodate 5V versions.

A larger PD behaves itself.

Modulation is ~2KHz

Yawn... SNCR

Mine has to go 4-1/2 orders of magnitude faster.

 

[Joerg]

I'm pretty much with Phil on this one. I saw this same effect way back when in
grad school at that little joint in Cambridge. The working theory at the time
was that the photocurrent and the diode's bulk resistance reduced the reverse
bias to nothing. It would be interesting if your customer was able to increase
the magnitude of the reverse-bias voltage just at the saturation point to test
this hypothesis. And yes, a bigger diode solved my problem as well, although
we ultimately used a beamsplitter (made of a simple microscope slide, about
4% reflection from each surface) so we could keep the bandwidth that the
smaller diode allowed. In our case the PD was inside a feedback loop so we
couldn't afford the lower BW.

I never thought about local forward-biasing, I'm not sure how this would work
in physics-land, or if it's testable just from the leads.

Did you have that larger PD hooked up to a cascoded TIA to squeeze more BW?

Steve, your post only made it into the CAD newsgroup. Check your
newsreader to make sure it posts to all groups in threads that are
cross-posted.

 

[Steve Goldstein]

Did you have that larger PD hooked up to a cascoded TIA to squeeze more BW?

We were using an op-amp-based TIA with rather low feedback resistance, just a
few kohms. We could get away with this because we had a laser with quite a bit
of intensity (something like a few hundred mW single-frequency at 5145A).

The opamp was the 9826 from Optical Electronics Inc in Arizona. It had huge
GBW, around 1GHz. I learned just a couple of years ago that a now-coworker
actually designed those OEI amplifiers!

BTW, that was quite a laser. Spectra-Physics Argon-Ion laser, the biggest
one they made. It was good for about 8W CW output on all lines (before we put
the modulators and etalon in the cavity for single-frequency locked operation).
It required about 75KVA, delivered from 3-phase 440V. Not very efficient, but
very impressive.

Steve, your post only made it into the CAD newsgroup. Check your
newsreader to make sure it posts to all groups in threads that are
cross-posted.

Right you are. Fixed.

Steve

 

[Joerg]

We were using an op-amp-based TIA with rather low feedback resistance, just a
few kohms. We could get away with this because we had a laser with quite a bit
of intensity (something like a few hundred mW single-frequency at 5145A).

The opamp was the 9826 from Optical Electronics Inc in Arizona. It had huge
GBW, around 1GHz. I learned just a couple of years ago that a now-coworker
actually designed those OEI amplifiers!

Check out TI's OPA847. GBW almost 4GHz and under 1nv/sqrtHz. But I am
lucky since I'll have enough receive level at all times.

BTW, that was quite a laser. Spectra-Physics Argon-Ion laser, the biggest
one they made. It was good for about 8W CW output on all lines (before we put
the modulators and etalon in the cavity for single-frequency locked operation).
It required about 75KVA, delivered from 3-phase 440V. Not very efficient, but
very impressive.

Wow, you could probably burn a hole through a boulder with that one.

 

[LVMarc]

I'm not saturating the amplifier... that was the first thoughts, given
peculiarities in a SwitCap filter output (harmonics) that followed the
TIA.

But something is happening in carrier generation... reducing light, or
going to a different PD (larger) solves the issue.

I'm just looking for some direction on how running out of carriers
might be modeled. It's sort of like channel saturation in a MOSFET or
maybe current-crowding in a bipolar device.




Can't do that... proprietary... start-up with VC money ;-)

...Jim Thompson

JT,

Why not measure the "sauration you observe, mathematically and directly,
INstad of terying to fool spice with circuit elements, just insert a
light level dependant A/W, then the saturation will be exactly what you
observe anyway!

as you indicate increasing area or decreasing the power (density?)
displays a decrease in the A/w conversion.

You may be able to get away by adding a a varible impedance to the
source, as this would give a linear explaination of the V va I response
observed.
Best regards,

marc

 

[Jim Thompson]

JT,

Why not measure the "sauration you observe, mathematically and directly,
INstad of terying to fool spice with circuit elements, just insert a
light level dependant A/W, then the saturation will be exactly what you
observe anyway!

as you indicate increasing area or decreasing the power (density?)
displays a decrease in the A/w conversion.

You may be able to get away by adding a a varible impedance to the
source, as this would give a linear explaination of the V va I response
observed.
Best regards,

marc

From what little data I've seen it looks like a MOSFET running out of
hole-electron pairs.

I'm trying to cajole the client into taking careful data but he's
resisting... he's sure he understands what is happening... he's a PhD
;-)

...Jim Thompson

 

[Joerg]

From what little data I've seen it looks like a MOSFET running out of
hole-electron pairs.

I'm trying to cajole the client into taking careful data but he's
resisting... he's sure he understands what is happening... he's a PhD
;-)

Just curious, why do you want to spend so much effort in modeling what
happens if the PD receives excessive optical energy? The docs I have
seen indicate that some irreversable damage might occur. Example on page
9 of this one:

http://www.newfocus.com/Manuals/15XX-B_Manual_RevA.pdf

I had a situation that was probably similar. The lasers for our market
were mostly 5mW and above but the PD couldn't stomach more than 1mW,
some of them 2mW. Called the client about it last week and we quickly
agreed to use optical attenuation.

 

[Jim Thompson]

Just curious, why do you want to spend so much effort in modeling what
happens if the PD receives excessive optical energy? The docs I have
seen indicate that some irreversable damage might occur. Example on page
9 of this one:

http://www.newfocus.com/Manuals/15XX-B_Manual_RevA.pdf

I had a situation that was probably similar. The lasers for our market
were mostly 5mW and above but the PD couldn't stomach more than 1mW,
some of them 2mW. Called the client about it last week and we quickly
agreed to use optical attenuation.

Subtle usage of PD. NOT overpowering, just getting into
non-linearities that cause modulation to generate harmonics.

...Jim Thompson

 

[Joerg]

Subtle usage of PD. NOT overpowering, just getting into
non-linearities that cause modulation to generate harmonics.

Ok, that's different. If you are using a special low noise laser source
that might matter. With normal DFBs from what I could see they seem to
dominate the distortion and noise budget. This matters a lot in our
case. It'll be interesting.

 

[Chris Jones]

I just ran onto a situation where a modulated laser impinging upon a
photodiode had such a high light level that the photodiode was
"saturated"... no more available carriers.

Thus "clipping" of the modulation was occurring in the photodiode
itself, creating harmonics.

And lowering the DC bias on the laser raised the AC signal level...
really strange lab results to ponder when you're tired ;-)

Any ideas how one might *model* the photodiode in Spice?

...Jim Thompson

Just a possibility: do you think maybe the effect is due to series
resistance between the active diode area where the light is being
collected, and the bond pads or terminals of the diode? If that series
resistance multiplied by the current were great enough, then the actual
junction could get forward biased over some part of the area even though it
is reverse biased at the terminals. A lot of this would depend on the
sheet resistivity of the diffusion which would depend on the diode design
e.g. if they wanted it to be sensitive to UV etc.

See if it gets significantly worse when you drop the bias to 8V, or better
if you increase the bias to 12V. This would not rule out other effects but
it would be a good clue.

Chris