[Laser] 5 mW lasers & LEDs - Detectors

[Clint Turner]

Having messed a bit with both APDs and PMTs, I can definitely say that 
the PMTs are more sensitive - but as has been inferred, finding one that 
works well at "Red" or longer wavelengths is a significant challenge!  
Shorter wavelengths that fall into the optimal range of most PMTs (e.g. 
wavelengths <green - particularly blue/UV) tend to fall victim to 
atmospheric effects more rapidly - a limiting factor for really long 
distances.

For that, the only real choices are some of the MultiAlkali and GaAs 
units:  The former are more readily available surplus (I've seen them as 
low as $25 on EvilBay, but they are usually closer to $100 or much 
higher).  The GaAs are much better for longer wavelengths, but they are 
very much harder to find and extremely fragile and tolerate no abuse at 
all compared to standard PMTs - which are fragile enough.

As for a PMT supply, that's easy enough if all you need is 1000-1300 
volts:  A modified inverter from a <$5 Harbor Freight electronic 
flyswatter (polarity-reversed, more robust output filter cap, a simple 
feedback regulator) works nicely and is inherently self-limiting in its 
current supply which helps protect the PMT from abuse.

Barry, G8AGN, has done some in-field tests using the '931 PMT and its 
variants and found them to work miserably at "red" - probably worse than 
a good PIN photodiode detector.

* * *

As for APDs, I have an APD-specific receiver that I've tested in the field:

http://www.modulatedlight.com/optical_comms/optical_apd_rx1.html

This uses the Pacific Silicon Sensor AD1100 APD - and I paid full price 
for it:  It is about the same price as the gas used by everyone on an 
expedition where we have several people driving hundreds of miles round 
trip, so if it makes the difference between success and failure, it may 
be worth the cost.

I was able to achieve 6-10dB improvement with it over a conventional 
photodiode-based receiver, but as was the case for the PIN-based 
receiver, the bandwidth is limited and this is a fundamental limitation 
of the physics.  I noted - as was also pointed out on an APD app note 
from Hammamatsu - that best APD S/N occurs in the area where M 
(multiplication gain) is in the area of 3-10 - well below the maximum 
"gain" of the APD, but this it so often the case in an amplifying 
mechanism where the noise floor rises faster than the gain!  In this 
case the "intercept" point (e.g. where further increases in gain caused 
a reduction in signal/noise ratio) occurred at an operating voltage of 
about 35-45 volts.  What this ultimately means is that if it's both the 
ultimate high bandwidth and high gain you want, you are still looking at 
a PMT.

I've not attempted to cool either a PMT or an APD and it's likely that 
some benefit would arise from doing so, but even the plain, old PIN 
Photodiode in the "Version 3" detector is capable of yielding speech 
bandwidth signals from a distant, red (630nm) light source that is too 
dim for the dark-adapted eye to see.

* * *

A while ago I was playing with my MultiAlkalai PMT and various 
high-power LEDs at MHz rate modulation and demodulation and was rather 
disappointed in the PMT's high-frequency response at very low light 
levels - at first.

As I increased the frequency the overall amplitude on the scope stayed 
sort of constant, but the signal disappeared into noise.  At first I was 
puzzled, but then I did some quick, back-of-the-envelope math and 
realized that I simply had too few "photons per cycle" impinging on the 
PMT.  As the waveform became indistinct, the scope was no longer being 
triggered reliably on what had been a sine wave at lower 
frequencies/higher amplitudes.

Once I synced the scope with the original source signal I could see that 
the original signal was still there, but somewhat "undersampled" and 
"fuzzed up" by the uncertainty and the number of photons/electrons and 
still rather noisy.  The upshot is that at some point, sensitivity of 
the detector becomes irrelevant if you don't have enough photons!

* * *

As for the size of the "spot" of light, refer to this page:

http://www.modulatedlight.com/optical_comms/fresnel_lens_comparison.html

Using a good-quality Fresnel lens of reasonable f/D ratio (around unity) 
it should be possible to focus the vast majority of light onto a spot 
that is a few hundred microns diameter - probably much smaller if one 
employs a "secondary" lens.

A "conventional" lens/mirror system should be able to focus to a smaller 
spot, but it's worth noting that even if you accept some losses with a 
"sloppy" Fresnel (e.g. the spot is somewhat bigger than the detector 
itself) better light-gathering power (related to aperture) would be 
easier to achieve with that Fresnel than a much heavier and bulkier - 
but more accurate - conventional lens that was smaller.

One definite advantage of a "conventional" lens - and a 
correspondingly-small active area on the detector - would be related to 
Field-of-View (FOV):  If your primary challenge is discrimination of a 
modulated light amongst a sea of other lights, the comparatively large 
FOV of a Fresnel and its tendency to scatter light may be a detriment.

73,

Clint
KA7OEI


Steve Noll said:
> I'd like to see that discussion too!
>
> I believe one may want to look at it two ways - setups "capable of
> communications at a range of at least 1 km." or setups for balls-out
> record setting distances.
>
> For the latter, really can't beat a photomultiplier tube. Next would be
> an avalanche photodiode.
> I just retired from being senior test engineer at Advanced Photonix
> where we made large area Silicon APDs, 5mm to 16mm dia, the biggest in
> the industry.
> Unfortunately, they're really expensive, even as surplus, as they're not
> easy to make. Also require ~1800V. As the operating point is temperature
> sensitive they're usually TE cooled.
> These were supposed to be a PMT-killer, they haven't been.  They're the
> sensors in the Optos scanners at your better-equipped eye doctor.
>
> Then there's small area APDs. We made those for long-range barcode
> scanners. 100V to 200V.
> I see on eBay right now First Sensor (formerly Pacific Silicon Sensor)
> AD500 small area APDs for $16 from China. How someone can sell these for $16
> when they're $135 from Mouser is a mystery, one hopes they're not
> rejects. Their spectral response curve is not great. About 155V bias
> requirement which is also temperature dependent.
> The biggest disadvantage is the tiny active area (500 um.) Can you get
> most of your receive signal on that 20 mil spot?
> If not then I'd go with a regular silicon photodiode which you can get
> with much larger active areas. And they're cheap.
>
> 73,
> *Steve J. Noll, WA6EJO
> http://www.qrz.com/db/WA6EJO
> http://www.linkedin.com/in/stevejnoll