led's

You could do either, but resonating the PD itself, maybe into a gaasfet gate, sounds good. You could emulate an RF front end, where a resonated phemt amplifier can achieve noise temperatures in the 40K range. And you could tune it with PD bias, nice idea. The sensitivity could be outrageous, and it would reject room light pretty well.

(I had this 'crazy' idea in the past about using a T-coil* as part of a PD front end... only to find that Phil H. had already done it.)

That's better for wideband (time domain) stuff, I think.

ge,

onfrontation.)

spare time while the attendees were doing stuff. So I took along a setup to measure the Johnson noise of a light bulb with a DC current going throug h it. (The measruements were a bit of a pain, I had to abandon the inducto r I was using as a bias element and go with a simple resistor...anyway that 's not important.)

ook that ratio to be the resistance of the bulb. And then I assumed that t he bulb would be making Johnson noise given by v^2 = 4kTR*BW. Where I'd see more noise because of increased temperature of the bulb.

ent

the

noise

Yup, that's what I tried (inductor from voltage source) (Driving a light bu lb with a current source is asking for trouble. IMHO) Problem was I didn't have a big enough toroid and the coil inductor picked up magnetic noise big time. (I already knew this, but sometimes I need my face rubbed into something a few times till I remember.)

So I got some data with a resistor as bias, but at the highest temperatures the bias resistance was much smaller than the DC bulb resistance and it sh unted most of the noise to ground.

ably)

E/dI

Hmm.. OK I'll have to think some more... why dE/dI?

RE the thermal mass: The time constant for a bulb is something like a seco nd maybe 10mS at the fastest. So only at low frequency is that going to be an issue.

ng to

Well I had the thing perched on bubble wrap to try and keep the building sh ake out of it. Better might be to suspend it. I would sometimes see a bun ch of low frequency 'crude'. Which I assumed was adjacent filament loops b umping into each other. But there were long periods of relative quite. (I also could filter out the LF stuff.)

One issue is that the filament loops were also an inductor and they would p ickup local magnetic field interference. (Mostly from the room lights, whi ch I couldn't turn off because those in the workshop would have objected.) The pickup would increase at higher bulb currents.. which still has me a bit confused. So before I try this again I need some big torodial inductor s and some mu metal to shield the bulb.

But this resistance question is great! What does resistance really mean? (Should I be thinking in terms of damping or energy?)

George H.

There's a lot of interesting physics behind Ohm's law. Among other things, it requires that the scattering length in the resistive medium be very short compared with the distance required to acquire enough momentum to cause impact ionization. Really ohmic things, like metals, are pretty remarkable.

Cheers

Phil Hobbs

Johnson noise is pretty small in amplitude, so it's only sensitive to the differential resistance. On time scales long compared with the thermal time constant of the filament, you'd have to worry about the nonlinearity, but at short time scales it would just look like a 2500K resistor, I'd expect.

The filament is obviously not in thermodynamic equilibrium, so the usual math underlying the Johnson noise formula doesn't quite work, but with DC excitation I'd expect it to be very close. It would certainly work if the filament were at equilibrium in an isothermal 2500K oven, and metal conduction is pretty well exactly ohmic, so I wouldn't expect the comparatively small net drift velocity to make much difference at all.

Cheers

Phil Hobbs

There's a theorem of Bode's (*) to the effect that if you have a parallel RC, the integral over all frequencies of the log magnitude of the insertion gain can't exceed pi/RC, regardless of what matching network you pick. T-coils can give you a 2.8x rise time improvement for a given load resistance, which is better than you can usually get with other simple matching networks.

Using the PD as a varactor is a good trick, but one thing you have to watch for is the way the Q varies with bias. It also changes the responsivity dramatically at longer wavelengths, where the light makes it all the way to the bottom of the depletion zone.

Cheers

Phil Hobbs

(*) Pronounced bohd-uh, as in Dutch and German, not bo-dee, as in Hindi and soggy canned spaghetti. (Reference: his obituary pointed this out particularly.)

I did once make a photo-optical link that transmitted usable audio, from a radio. With a flashlight bulb.

The filament is a heated thermistor that makes its own turbulence in the gas. Gotta be noisy.

AC coupled into the filament, it looks just like a resistor. I'm not sure what the small-signal value will be.

Flashlight bulbs used to be "filled" with a vacuum, I think they're using krypton in some now.

probably magnetic fields were moving the filament, changing is emission pattern and therefore measured output.

--- The filament of an incandescent lamp is made of pure tungsten, which has a positive temperature coefficient of resistance and is extensively heat-treated in order to make it ductile enough to survive the drawing and coiling process required to form it into the shape required for a particular filament.

Thermistors are an entirely different breed of cat in that they're constructed by pressing and sintering metal oxides, generally have a negative temperature coefficient, and can only be enticed to emit humanly visible photons once, and then only for a very short time.

---

--- Marie and Pierre Curie sifted through a ton of pitchblende, in the winter and in unheated quarters, and discovered Radium.

You couldn't make your measurements after everyone else left your comfy lab and you had it all to yourself?

---

--- Why?

---

--- It's arguably best to think of resistance as collisions between electrons which liberate heat.

----

--- With AC coupled into the filament, its wound nature will cause it to look inductive, so it can't look like a pure resistance regardless of how lightly it's driven.

-- JF

There was that few hundred pieces of the 1.3 um APDs surplused from Terabeam that went for about 1 cent on the dollar three or four years ago, but nothing interesting since.

We should still figure out something useful to do with those.

Cheers

Phil Hobbs

ge,

confrontation.)

e spare time while the attendees were doing stuff. So I took along a setup to measure the Johnson noise of a light bulb with a DC current going throu gh it. (The measruements were a bit of a pain, I had to abandon the induct or I was using as a bias element and go with a simple resistor...anyway tha t's not important.)

took that ratio to be the resistance of the bulb. And then I assumed that the bulb would be making Johnson noise given by v^2 = 4kTR*BW. Where I'd see more noise because of increased temperature of the bulb.

OK, I was thinking about this. (I couldn't wait for the weekend.) So there's some energy dissipated in the bulb given by the DC V*I. But that doesn't come into the Johnson noise. Then I was thinking of the derivation of Johnson noise that starts with a p arallel RC and the equipartition theorem. (the capacitor holds 1/2 kT of energy = 1/2 CV^2) and then you integrate over the RC bandwidth to get th e Johnson noise formula, So 'R' in this case is, as you say, the dynamic R . And that's pretty cool!

So the data for my first crack at this was pretty crappy. (not much signal , since it was all lost in the small bias resistor.. big error bars.) But I got a number for the temperature of about 1500K for a bulb that was a li ttle past orange and getting to yellow. And that number seemed small. But I was using a DC resistance value. The dynamical resistance looks to be ~

Ah OK the drift velocity looks pretty small compared to all the other corre ctions.

George H.

So what is the AC, small-signal resistance of a hot filament?

I'm guessing that's it's about equal to the gross resistance, E/I, at that operating point. The DC slope resistance, dE/dI, at that point will be higher, maybe 2x or thereabouts.

Last time I checked - the only way you can buy a flashlight bulb is buy a flashlight with one in it.

There are still one or two discount stores that carry bulb type flashlights - but increasingly they're dual type with a ring of LEDs round the reflector.

You can buy spare Maglite bulbs.

I bought a Maglite - and quickly formed the conclusion that they're overrated.

A 3W LED flashlight is brighter, uses less batteries and the 'bulb' will probably outlast me.

The Maglites are physically very rugged, sealed, waterproof. Lots of LED lights are junk and *don't* last long. The Maglite bulbs last because such lights are not used for weeks at a time. They usually have a spare inside, too.

I got stuck in the dark on a cliff-side hiking trail on the north coast of Kauai. The sun goes down fast, splat-sizzle, at that latitude, and then the killer frogs attack. The trail was steep, dangerous, and covered by slippery wormy guava fruits. My little Maglite saved my life, and my GFs as an added bonus.

Have you been licking toads again!!!

(If you're playing sniper in the jungle - you can stick a frog on the end of your rifle to quench the muzzle flash).

As I do a fair bit of cycling, and often after dark - I keep a spare 5-LED cycle light in my pocket, it does for almost anything I need a flashlight for. At 1GBP I'm not too worried about losing one ever once in a while.

Some of the single, power LED ones are nearly as pricey as a maglite, so I use them mainly where they can't go far if I put them down without thinking.

The GF was quite pretty, in fact.