How about using several photodiodes with each their own amplifiers

Bloody Google

1. Capacitance is the worst enemy in PD amplifiers. No matter what you do, there's a noise current contribution of 2*pi*f*C_diode*e_N, where C_diode is the total capacitance on the input node (amplifier, pads, and photodiode together). TIAs, bootstraps, cascodes, everything. (Reactive matching networks too, but they're a little more complicated.)
2. Reverse biased PIN diodes are easy to use, and can reduce the capacitance by a factor of about 7 from zero bias. Reverse bias is always a win if you're using silicon or short-wavelength InGaAs--there'll be some voltage that optimizes the capacitance vs dark current tradeoff for your situation, and it won't be zero volts. (I'm running some 32-volt photodiodes at 80 volts, which improves them out of all recognition, and they leak about 10 nA at room temperature.) You do have to make sure that your bias supply is extremely quiet--ideally at least 3x lower noise voltage spectral density than your input amplifier.
3. A good reverse-biased silicon PIN diode will have a capacitance of 40 to 100 pF/cm**2, depending on the thickness. Thinner ones are often faster but don't respond quite as well out beyond 750 nm or so.
4. If you mean collecting more light overall, you're better off collecting it with a lens or non-imaging concentrator and stuffing it into a smaller photodiode, which will have lower capacitance (see #1 and #2).
5. There are some fairly desperate situations where doing as you suggest can make a lot of sense, but they're rare. One was an intracavity darkfield laser particle detector where sample air was drawn through a very small box, whose walls were tiled with CCDs, with the laser light running at right angles to the air flow. It was designed by Bob Knollenberg, and worked very well at the low count rates it was designed for. (You probably aren't doing that, though.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

o

Interesting, thanks

This was originally meant as a post to the thread "Light pulse and photodiode " by George Herold. That's why my opinion about Google in the second post, when it put this into a new thread instead of where it belongs.

If George Herold does not mind, still a (silly?) question. I did not mean those PDs would be paralleled, does total capacitance still add if there are several separate photodiodes and amplifiers each built like you say in #1.

Collecting light is probably easier too if there more than one diode.

To make better measurements, #4 is really the way forward. Lenses are cheap and work amazingly.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

Hmm, I'm going to have to push the numbers around some... But for the above application, it looks to me like you want a Big feedback R. And that is going to dominate the noise. Of course you always want the PD C to be as low as possible!

o

Clearly I'm not 'abusing' my photodiodes enough!

Oh, that's interesting... Say I've got a rod of plastic scintalator (maybe 1 cm in diameter) I was pictureing gluing a PD on each end. (A conical light pipe made out of the same scintalator on each end would be nice...) But would gluing a lens on each end help also? I lose photons but reduce the PD C.

Doesn't the correlation trick help? Half the signal, but two pulses. Or is it a 'wash' in the end. I'll have to do some thinking.

George H.

ou

d ) m f

do

. 0 d t d

What, me mind? Take the thread in any direction you like!

For the application I'm talking about with two PD's you get (about)

1/2 the light in each PD, so if you're just adding the signals then it's clearly better to try and get all your photons into one detector.

George H.

de quoted text -

The R_f noise only dominates at low frequency. The amplifier voltage noise dominates above f_N, where the noise current from e_N equals that from R_f and i_N of the amplifier:

2*pi*f_N*C_d*e_N = sqrt(4kT/R_f + i_Namp**2)

or f_Namp = sqrt(4kT/R_f + i_Namp**2)/(2*pi*C_d*e_N)

Above there, the amplifier voltage noise dominates. Of course, your amp may roll off before that happens, but it gets worse for large R_f and large C_d.

Lenses won't help with scintillator, because you lose NA by a factor of n**2 in going from the rod end into air, so you're much better off cementing a plastic-packaged photodiode onto the end of the scintillator. You have to collect all the available photons.

Multiplying the outputs might be interesting.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

you

and

d.)

d p

Duh, how silly of me. OK please disregard the beginning of this thread.. or it's sire, where I related the needed pulse signal to the Johnson noise... I'll have to amend it. (Adding in the capacitance 'directly' complicates matters)

1000 photons is going to be hard.

I think the best thing to do (for we bears of very little brain) is to do some measurement. And then explain the results. I'll pulse a LED on Monday.

rk

I'm

of

ou do

t

er.

40 n

and

A

OK thanks, I guess the light cone idea only helps when you've got a collimated light beam to begin with.

George H.

est

a t

ned

t -

o

at you

re

, and

ted.)

And

amp

is

dark

=A0(I'm

ut of

=A0You do

at

fier.

of 40

ten

it

1 and

=A0(A

ggest

y

h a

ght

igned

s.

xt -

I feel in need of another "dope slap' for forgetting about TIA /PD noise gain.

So at the end of the day today.

I pulsed a red LED through 50 ohms, And looked at the response with a ~16mm^2 PD, biased to 10V, C =3D 70pF (according to spec sheet) with a OPA134 as TIA, 1 meg ohm feed back and no feedback cap (enough stray C in circuit)

Here's the step response,

channel 1 is the drive AC coupled. And chan. 2 is the PD response. It covers most of the screen. There's this long tail that I'm calling edge effects.

The fast response is about 0.6V or 0.6 uA about 1 uW.

Then here's a 20ns pulse into the led, (about 20E-15 Joules) (with x100 Amp on PD)

That's ~60 -70 k photons.

Except the 20ns pulse was a bit fast for the led... a 200ns pulse gave a pulse response about 30 times larger. So maybe 20 k photons.

I also measured the noise from the PD circuit. Vrms was about 0.2 mV. With the noise peak at 150kHz.

If I use my previous Q*R*f =3D V equation to predict the number of photons I get about 8k, Which seems at least reasonable.

A faster opamp should help.. that AD fast FET.

George H.

so

op

what you

here

ds, and

cated.)

he

=A0And

ge

at

r amp

s is

s dark

=A0(I'm

out of

=A0You do

ly at

lifier.

e of 40

often

g it

#1 and

=A0(A

ld

se

of

suggest

ity

ugh a

light

b

esigned

ses.

ext -

Yesterday I wrote,

"A faster opamp should help.."

Ahh I think this statement of mine is absolutely wrong! A faster opamp (with everything else equal) will give me more noise at a rate of fc^3/2 and more signal at a rate of fc, so the signal to noise goes down as I increase the bandwidth. (fc is some corner frequency of the circuit.)

This seems a bit counter-intuitive, so maybe someone would check my work. So imagine some PD TIA compensated so the signal bandwidth is fc and noise bandwidth peaks at same fc. Here=92s a Bode plot of noise gain with fc =3D 10 (Opamp GBW =3D 100)

Log Noise gain ^ | 10 | | /\ | / \ 1 |--- \ | \ +---------> frequency 1 10

Then I wave my magic opamp wand and increase the opamp GBW by 100 (GBW =3D 10k). And I can still compensate the thing to have equal 3dB point and gain peaks but now at fc=3D100

Log Noise gain ^ 100 | /\ | / \ | / \ | / \ |--- \ | +------------->

1 100 When you do the integrals for the total noise it goes as fc^3/2!

George H.

Oh, gosh, YES! There are FEW cases where you'd want to run a PD in photovoltaic mode for pulse sensitivity. That would be awful! The non-biased capacitance increases the noise gain of most circuits. Filtering the bias supply is pretty easy, just a couple RC stages and physical separation from interfering traces.

Jon

The problem here is that with two PDs, only half the light reaches each PD. So, it seems a losing proposition in the general case.

Now, I can see a case where you have a long, skinny scintillator, and then putting a PD on each end of the scintillator may capture light better than trying to reflect it all to one detector at one end. Some PETT scanners are built like this. I've seen some other physics detectors that had 10 foot long plastic scintillators with a detector at each end. Time of arrival difference tells the position along the scintillator where the particle hit.

But, in the general case where the scintillator is relatively small, I don;t think the multiple PDs will help.

Jon

Lenses usually suffer from loss at the interfaces. Light guides can be made with silicone optical grease or clear RTV to reduce loss at the interfaces. If it is all done well, you absolutely can't detect the interface between scintillator and light guide. With the super dense scintillators that gets harder, of course, and a glass light guide might work better than Plexiglas, although way harder to fabricate.

Jon

Yes, that's my take, too, unless the scintillator is long and skinny.

Jon

Well, instead of lenses, which generally would require a quadruple interface (scint -> air -> lens -> air -> PD) just a better light guide would be best. Silicone grease for the interfaces, and make the small end of the cone whatever is best for the PD. You can make the guide out of plain Acrylic (trade names Plexiglas or Lucite) as most plastic scintillators are basically doped Acrylic. You can glue the guide to the scint with Acrylic glue.

Jon

George Herold wrote:

The (nuclear instrumentation) literature should have plenty of articles on charge-sensitive preamps that have worked well to this noise level and below. We used all BJTs other than the input JFET device in our charge-sensitive preamps. Don't be surprised by the noise, if the LED is run at low enough level, the photon statistics will start to show up easily. In other words, some of the noise you can observe is REAL optical noise! If you use a PMT and a fast scintillator, then the individual photon counting will fill a scope screen with "fuzz". With a PD and charge-sensitive amp the individual photons will mush together, but the statistical nature will still be quite evident. The key observation is you will see much more noise with the low LED illumination than with the LED off. If you get that, you are ready to see if you get discrete pulses with a scintillator. There are abundant sources of radiation available. Orange glazed dinner plates were once sold with Uranium glaze, you get 5000 counts/minute from a 1" chunk. And, that's just a GLAZE on the surface! Americium sources in smoke detectors, gas lamp mantles have Thorium in them, even a red brick is a bit radioactive. Sources are good as you can move them closer and farther away to prove you are actually detecting something. Cosmic ray showers will hit a scintillator several times a minute, generally creating MONSTER pulses of tens or hundreds of particles in a burst. But, you can't turn these on and off.

Jon

Well, so you see the kind of problems you are running into. The pulse is just a bit bigger than the noise. We usually get these things so that we have about

2-5 mV noise at the output of the TIA. I doubt you are going to get there with any off-the-shelf monolithic op-amp.

Jon

As I think I said already, in your special case, you lose a lot of light from total internal reflection at the scintillator/air boundary, like more than half and probably nearer 2/3.

Ordinary Fresnel reflection from an air/glass interface is a nit by comparison. For that job, you have to glue a plastic packaged photodiode (or even a well-passivated chip) right to the scintillator. There's no way to reduce the phase space volume in that instance, except to use a smaller scintillator, closer to the source. That's why people use PMTs for that job--they're unbeatable for high sensitivity at large area. (See Super Kamiokande, for example.)

But the OP wasn't living in that special world, so the point is sort of irrelevant. Lenses are a huge win in almost all instrumental photodetection situations where they can be applied.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net

e

From my perspective the problem with lenses or light cones is the same, both work great with a collimated beam, but with a rod like scintillator you'd throw away all the 'glancing' rays.

George H.