To Phill Hobbs

It cancels the correlated part of the photocurrent, regardless of where that comes from.

Lasers often have a great deal of amplitude noise, which is referred to as residual intensity noise (RIN). It's commonly 20-50 dB above the shot noise level. The noise canceller gets rid of that pretty well in many cases.

It's much simpler and cheaper than most of the other ways of doing that, and often works a good deal better as well.

Cheers

Phil Hobbs

hat comes from.

Please, "what correlated part" Noises, signal ? the BFW34 is a single packa ge inside which there are two photodiodes ?

s residual intensity noise (RIN). It's commonly 20-50 dB above the shot noi se level.

Never heard about RIN, i'll try googling with ... i'll see. Thx, Habib.

It's a typo for BPW34. (Use the Vishay parts.)

There are two separate photodiodes. CR1 produces the signal photocurrent and CR2 produces the comparison photocurrent. Do you understand how Figure 2 works? Figure 4 is just a cute method of applying negative feedback to automatically balance the subtracter of Figure 2, without screwing up the linearity.

I have a discussion of all that stuff in the introductory section of my longer noise canceller paper, . Have a look especially at the right column on P2.

One of my best gizmos ever.

Cheers

Phil Hobbs

Yep ;)

Paper far more interesting and clearly explained. Fig.1 is essential IMHO. You mention on pp.2 "In addition, since the shot-noise currents of the signal and comparison photocurrents statistically are independent, they obey Eq. ~2! and thus limit the system noise floor to 3 dB above the shot noise of the signal beam alone."That's what i'm expected. The circuitry tri es to cancel the noise induced by the laser beam itself but not the intrins ic Si shot and thermal noises. Smart ! Just a question : what if the two laser trajectoires are not the same lengt h ? As it is implemented in our gas analyser (analyser of monoxyde carbon g as). I think it will defeat the method. BR, Habib.

I go over all that in the second column of P. 16 (see Eq. 16).

Cheers

Phil Hobbs

HO. You mention on pp.2

tries to cancel the noise induced by the laser beam itself but not the int rinsic Si shot and thermal noises. Smart !

ength ? As it is implemented in our gas analyser (analyser of monoxyde carb on gas). I think it will defeat the method.

Strange formula. If Deltaz is a distance in m the the denominator is homoge neous as a distance also to apply correctly the sinus. Strange indeed.

In addition if i take Deltaz = 0.48, fmod=10^6, c=2.10^8 i have not A min = 40

May be i missed something elsewhere ...

The way it came about was sort of funny--one of those cast-your-bread-on-the-waters occasions.

A good friend of mine at IBM was trying to get promoted from technician to engineer. Our boss at the time, who's also a good friend of mine to this day, said that if the tech could design and build a particle detector for finding 2-um bits of crud in the hydrogen belt ovens IBM uses for baking ceramic and glass-ceramic chip carriers, and transfer it to manufacturing successfully, our boss would fully support his promotion. He ran into some trouble, so I just sort of surreptitiously helped out a bit.

The belt ovens look like something out of the throne room of the Great Oz. They're 35 feet long. When the doors open, flames pour out. The whole end of the building is completely covered in blast doors (walls and roof). At the time (1989ish), IBM was the largest industrial user of hydrogen after NASA.

The optical access was a 1-inch window in each end, so he was going to shoot two laser beams next to each other, detect them separately, subtract the photocurrents, and look for extinction events. Of course, a 2-um particle in a 2-mm beam represents only 1 part in 10**6 extinction. (Interestingly it's really 2 parts in 10**6--the part that gets absorbed and the diffraction due to punching a hole in the beam are equal in amplitude [Babinet's principle].)

Anyway, that was way too small to see over the RIN of the He-Ne laser he was using, because he could only get 15-20 dB of noise rejection from his subtraction circuit (he was using separate op amp TIAs and subtracting with an instrumentation amp and way too many trimpots). Phase shifts and amplitude drift kill you when you do that.

So I tried using a JFET variable resistor to servo the subtraction, by looking for the DC output to be zero, which means the photocurrents are the same. Didn't work worth beans. Hmm, why was that? JFET nonlinearity, of course. When I dialled down the photocurrents using a variable optical attenuator, it started to work much better. So then I needed a linear current splitter.

Having just got out of physics grad school, I hadn't done any logging circuits back then, but being an avid reader of app notes and suchlike, I remembered seeing a Pease or Yamatake or some National guy's app note about log amps, and the way you could use matched BJTs to make power law circuits among other things. The main thing for my purposes was that the collector current split in a diff pair is strictly controlled by delta-V_BE, and therefore is linear over some absurd range of I_C.

So I switched to a couple of 2N3904s, and bingo, the thing worked like the bomb. Not monolithic, not matched, just two random BJTs from the drawer, and it gave about 55 dB of cancellation out to a couple of megahertz at all relevant photocurrents, over a usefully wide range of current splitting (4:1 or so).

Of course, at the the first test, we discovered that the temperature gradients and gas turbulence in the oven made the beams dance all over the place, so we couldn't keep them on the photodiodes, so even the noise canceller didn't help.

I suggested a different scheme, which is in Fig 10.8 on P. 377 of my second edition (Go to

That (still with the noise canceller) made the extinction signal AC with a zero background, so it didn't matter much if the beam wandered some.

In the end my technician pal wasn't able to make the system work on his own, so he didn't get his promotion. However, that actually turned out to be a good thing, because it was only a couple of years later that IBM had its first-ever layoff (1992). He'd have been one of the weaker-performing engineers, and might well have lost his job. As it is, he's still there, and I'm still building gizmos. :)

Cheers

Phil Hobbs

The limit comes from the phase shift. If you have two currents, identical except for a phase shift delta_phi, you have

I1 = cos(omega t)

I2 = cos(omega t + delta_phi) = cos (omega t) cos(delta_phi) - sin(omega t) sin(delta phi)

Feedback cancels the first term but not the second. Sine and cosine have the same amplitude, so the maximum cancellation ratio is numerically equal to sin(delta phi).

A 48-cm path difference at 1 MHz limits the cancellation to

Amax = 20 log(sin((2 pi * 1 MHz * 48cm/(3e10cm/s)) = -39.95 dB.

Cheers

Phil Hobbs

le

my

IMHO. You mention on pp.2

ry tries to cancel the noise induced by the laser beam itself but not the i ntrinsic Si shot and thermal noises. Smart !

length ? As it is implemented in our gas analyser (analyser of monoxyde ca rbon gas). I think it will defeat the method.

mogeneous as a distance also to apply correctly the sinus. Strange indeed.

ot Amin = 40

Ok i will think about that. Especially how you turn the initial -Log( (i_co mp/i_signal) -1) into that Log( sin(2.pi.f.Deltaz/c)

I'm tired at the moment. BR, Habib.

Phil, Do you have anything in your book that applies to a single Photo detector? Noise wise that is.

Cheers

Sure. Most of Chapter 18.

Cheers

Phil Hobbs

Thanks, gives me a reason to get your book ;)

Cheers

Thanks, that'll keep my son at university for another 10 minutes. ;)

Cheers

Phil Hobbs

Regardless of the nice optics, If you build instruments you should buy Phil's book. I think the pieces of wisdom are the best. (wisdom = things you learn by making mistakes, or someone telling you their mistakes.)

George H.

I know it's either Learn by example, or Learn by Mistakes ;)

Cheers

I am an expert at the latter >:-} ...Jim Thompson

I've spent an hour on your paper. As far as i could speak on photonics materials, the very topic above is not very clear for me ... i'll speak tomorrow Monday with my guys (one of them is was and remains a very talented "electro-photon-man")

Nevertheless, thank you for the topic.

Best regards, Habib Bouaziz-Viallet.

PS : Strange how this thread was not already haunted by the Decadent and some Bloggers ... not yet ... we'll see.

Habib.

DecadentLoser is not even competent to throw stones at this topic ;-) ...Jim Thompson