poor man's PMT

The Victorian engineering ethos: "Add mass till nothing breaks." ;)

If you use a common-base input stage, you can use solar cells up into the tens of kilohertz. Long ago a friend of mine and I did a gizmo that used diffraction from the latent image in photoresist to implement closed-loop control of the post-exposure bake step (corresponding to the developer for photographic film). It was basically for memory, which at least at the time had a more or less hexagonal layout, so there were six diffraction peaks.

We used seven 1x3 inch solar cells fanned out like a poker hand (except all the way round the circle). With seven cells, we could always get the +-1 orders of at least one of the three directions to sit in the middle of two of the solar cells (as opposed to falling off the edge). The LED light source was in the middle, and was modulated at about 15 kHz iirc. It turned out that by looking at the time evolution of the

(That would be pretty tough to do now, but back in 1993 or whenever that was, the structures were still big enough for visible-light diffraction measurements.)

Cheers

Phil Hobbs

Absolutely. Anything worth doing is worth overdoing.

I put it in a box

With 12 volts and a 2K load, I'm getting about 1.7 us rise/fall from a pulsed laser input, so it shouldn't be hard to push it into the MHz range with better electronics, your cascodes and such.

I wonder if there's a market for a box like this, large-area free-space o/e. A PMT will whup it on speed, but this is small and potentially cheap.

Hey, I bet I could get two more photodiodes on that board! Or four.

The application would be ideal for the Hamamatsu MEMS PMT, if they would ever actually sell it.

Yup, it's all limited by the differentiated voltage noise of the front end.

Maybe. The spatial nonuniformity is a drawback.

Yup. Being a large Japanese company, they're constitutionally chicken. They're just showing the same few development samples for coolness value because nobody wants to take the risk of introducing it as a product.

Eventually people get bored.

Cheers

Phil Hobbs

A medical client of mine was involved with the development of the Hamamatsu MEMS PMT.

I created a Spice model for it in mid 2008. ...Jim Thompson

Interesting. On the off chance that they ever get off their thumbs, can you share the model?

Cheers

Phil Hobbs

Possibly. I'll look into it. BTW, It _wasn't_ the Stheno guys, or it would be easy. ...Jim Thompson

Right. BTW I may get the chance to revisit the Stheno gizmo at some point--there are rumblings in that direction. It was a cute technology.

For everybody else: Stheno was a startup that built polarimetric sensors for verifying chiral purity in medicines.

Left and right hand enantiomers of the same molecule often have very different biological effects. The classical example is thalidomide: L-thalidomide is an excellent and pretty safe drug for morning sickness, whereas R-thalidomide doesn't help nausea and causes horrific birth defects.

Back in the 1950s the technology for making chirally pure drugs basically didn't exist, but nowadays aiui they have to be chirally pure to get approval. (The two enantiomers get interconverted in vivo, so you couldn't get L-thalidomide approved today even if it didn't have such unpleasant associations.)

Unfortunately Stheno didn't survive their last round of venture finance, but their technology is really pretty, so it would be worth another try if they can license it to some company with both dough and patience.

Cheers

Phil Hobbs

It was an interesting (and fun) project, and edifying working with you on it! ...Jim Thompson

Agreed.

Cheers

Phil Hobbs

ut 40

At

m a

nt

pace

en.

an

My example is sotalol, which I take both for its antiarrhythmic activity - which is shared by both enantiomers - and it's action as a beta-blocker whi ch is largerly confined to the R-enantiomer

Chromatography with an optically active stationary phase was perfectly prac tical back then, but I don't think that anybody had thought of using it. On e of my friends pointed this out to a lecturer after some lecture on the su bject in the early 1960's, back when he was still an undergraduate. He's go ne on to invent and make a lot of money out of a better confocal microscope , so it might not have been obvious to everybody.

Sotalol is approved, despite being a cheap racemic mixsture.

Sad. Unfortunately pretty technology still needs customers willing to buy i t.

Geesh, both JT and PH. (along with lotsa others, I assume) And it still didn't fly, that's a tough egg to crack!

George H.

Speaking of a poor man's pmt have you ever reversed biased a photo diode till it avalanched? (maybe not as good as a 'real' APD, but cheaper.)

George H.

The VC's were nasty bastards. I was actually one of the few they decided to pay :-) ...Jim Thompson

It was really JT's show--I was still at IBM, so I was mostly just kibitzing since I couldn't take either their money or their confidential information.

They needed laser noise cancellers to make it work, and Jim came up with an interesting new wrinkle using both PNP and NPN diff pairs, so that you didn't have to run it on one side of balance the way you do with the original model.

It needed a little bit of tweaking to save photocurrent, iirc, and for whatever reason I don't think it performed as well as they hoped, either for optical or electronic reasons.

Noise cancellers do a good job of cancelling the correlated part of the noise, but anything uncorrelated goes straight through, which is normally what you want. However, if you have some slight decorrelation from something like an etalon fringe or a bit of vignetting, the correlation drops and some of the noise gets through. That means that the more complicated the optical system gets, the more paranoid you have to be if you want really good cancellation (like 50-70 dB).

Things that look as though they're common-mode, such as vignetting before the beam splitter, sometimes aren't, due to effects like polarization sensitivity and fringes inside cube beam splitters. That confuses some folks. You always wind up learning things about your beams that you didn't necessarily want to know, but once you get it right, they work like magic--up to 70-80 dB of noise cancellation at low frequency, with a bit of tweaking. j

Cancellers also put an enormous premium on transistor parameters that don't matter very much in most other uses, e.g. beta linearity and low R_ee', so getting them really right with new components is always a bit iterative. Jim used THAT340s, which I've never tried. Their R_ee' is high, about 2 ohms typical, their betas are pretty low, and there are no beta-vs-I_C curves in the datasheet, so it's hard to know what the beta curvature is like.

The cancellation is limited by the beta nonlinearity to a value of

A_min = |1/h_FE - 1/h_fe|,

i.e. the proportion of the emitter current fluctuations that comes out of the base when it needs to come out the collector and cancel the other branch's fluctuations. To make that number small, you want beta to be high and constant.

Because of the peculiar sensitivity to small effects, they usually don't simulate very well.

I posted a noise canceller transistor tester a couple of months back that I used for a nanoamp canceller for Notre Dame. They really are a bit subtle for such an apparently simple circuit, as well as being dead useful, which makes them fun.

Cheers

Phil

I think the VC's just ran out of patience. We had an inordinate number of construction issues, with the built circuit mis-wired, multiple times :-(

I was off-site most of the time, fighting my failing hip... when I was there at Georgia Tech, they'd drop me off at the loading dock, so I could directly board the elevator... walking from the parking lot, then thru long corridors was killing me.

As I understand it, the scheme _was_ working when the VC's pulled the plug. Nasty bunch, they had a watch-dog present at all times. ...Jim Thompson

about 40

one. At

.jpg

." ;)

from a pulsed

nge with better

front end.

ee-space o/e. A

eap.

our.

hey would ever

hicken.

oolness

s, can

it

ogy.

ess,

ure

nce,

try

Great, thanks for all that. (I recall your transistor matching circuit.)

Now time for true confessions: I just went back and looked at your laser n oise canceler. (Section 18.5.3 in 1st ed. of Phil's book, if you're follow ing along at home.) I didn't quite get this the first time I saw it, but n ow with 'new' eyes... I like it! I wish I had some 'laser noise limited' ex periemtn to use it with... Well maybe not. Is this still under patent? I noticed a 1997 date for your paper, do I have to wait till ~2017? (When we can expect a flood of noise cancelers on the market :^)

George H.

There are lots of laser noise limited experiments.

The patent expired a couple of years ago. US 5,136,274.

My beloved hunchbacks and I are working on an automatically tweaked 10 MHz version that may or may not see the light of day, depending. It has about 12 automatically tweaked parameters:

DC:

AC:

and a couple more for internal functions. The plan is for Version 1.0 to be all-analog in appearance, except for this magic button that tweaks it up for your current operating conditions. Version 2.0 may have a remote display with more controls and stuff, as well as an internal optical calibration facility. (The alternative is Labview, of which I am Not A Fan.)

The analogue board will go out for fab this month, God willing. There are about five things that are liable to oscillate or otherwise misbehave, but we'll see if we can make Rev 0 work right.

Cheers

Phil Hobbs

Build in a lock-in, and a USB interface, as long as you're in there.

There's a USB interface (FTDI) on the CPU board. Young male hunchback got all the USB, display and encoder stuff working this summer, including running I2C over a 2 metre HDMI cable. This was in between working in a tiki bar on the beach and getting back in shape for rowing. He's now gone off to UBC in honours physics.

Young female hunchback has had a long learning curve to climb, but is now much better at layout than I am. She's quietly amused by some of my various stupidities along that line, especially my sins in creating library parts. (Not that I ever claimed to be a big layout person, but she's becoming a pretty good one.)

The lock-in idea is interesting. I actually have an SA604A on the analogue board for tweaking purposes already, but that's more of a network analyzer back end--it measures amplitude and phase, but isn't frequency-selective.

That board is getting a bit full for a single-side load anyway, so the lock-in would have to go on the CPU board. It would also have to be reasonably quick, so maybe I could use some of my new-found Verilog nous to make a programmable divider/phase shifter/phase-frequency detector/lock detector gizmo in a $2 CPLD. At that point all I'd need would be a PGA, a VCO, a comparator, a couple of dual op amps, and maybe an MDAC for setting the loop gain. There ought to be space for that. The key would be avoiding major-league fanciness.

I just got a real genuine PO for an updated version of another one of my old gizmos, ISICL:

Cheers

Phil Hobbs