femtosecond pulse detection

Are you trying to control its amplitude, its timing, or both?

Cheers,

Phil Hobbs

And is it electrical or optical? You don't find a lot of electrical femtosecond pulses.

John

C'mon, you're just a DC guy underneath, admit it. ;)

To the OP: The biggest problem with measuring picosecond and femtosecond pulses is that they tend to saturate the photodiode. If you only care about amplitude, stuff the pulse into a small integrating sphere, which will stretch it out to 10 ns or so--much much easier.

If you care about the timing, you'll have to tell us more about what your actual application is.

Cheers,

Phil Hobbs

Cool, sort of a mushy optical-temporal lowpass filter, with an interesting impulse response.

There are probably ways to make serious optical-temporal filters.

John

this is an optical pulse train with tens of mega herz.

timing is not an issue.

The problem is to detect it's amplitude.

I've built nano second level peak detector works fine.

This kind of short FWHM signal is a problem.

What's integration sphere? never heard of it.

I try to solve from electronics side.

You're going to find it hard to solve in electronics, unless you reduce the pulse amplitude really enormously, and then you have the problem of knowing how much is enough, when you don't have an independent measurement of the attenuation (which you almost certainly won't).

I recommend buying a small integrating sphere (e.g. from Labsphere) and plotting the dc photocurrent you get from your photodiode, with and without the sphere. (If you want to learn about them, Labsphere's web site is a good place to look.)

You'll find that without it, the nonlinearity begins at very low average photocurrents. You should find that when using the sphere, the photocurrent goes _up_ by a large factor due to the enormous duty cycle increase.

Cheers,

Phil Hobbs

Atoms tend to absorb and re-emit photons with half-lives in the area of 10^-8 seconds. I think that can be down to as low as 10^-10 or so and as much as 10^-7 or longer. But roughly speaking, about 10^8 events per second. I think I got that from Hecht and Zajac on optics, many years ago. Phil, if I gather him correctly, is referring to this behavior. The effect of the sphere is to smear out the time period through this absorption and re-emission process. If you integrate the resulting curve, it may accurately represent the short amplitude for you.

Jon

Photodiodes can have really weird behaviour at very short times. One would think that the limiting factor would be the diode capacitance--i.e. that it wouldn't matter how short the optical pulse was, as long as the generated photocarriers didn't shield out the E field in the depletion region, i.e. that the total integrated charge didn't exceed C_diode*(V_bias+0.6V).

This turns out not to be true, for reasons I don't fully understand, but that probably have to do with charge transport. For the shortest pulses, nonlinearity begins significantly sooner than you'd expect from the above simple analysis. One of these times I'd like to spend a day or so with a few photodiodes and a femtosecond laser to try to get to the bottom of this.

Using an integrating sphere gets you down into the tens of megahertz, where everything is sensible.

Cheers,

Phil Hobbs

An integrating sphere works on pure reflection, which as far as I know has zero time delay. The sphere just bounces the light around, which scatters it in time. It's just a metal ball, painted white on the inside.

John

I'll allow Phil to respond to this. He might agree with you and correct me on the subject, or ...?

I'm working from what I recall and interpreted in my own way from Hecht and Zajac's Optics book, earlier on in the book, on transmission and reflection and what I also have read about phosphorescence (stimulation and relaxation of 'illegal' spin-lattice transitions from spin-0 photons.) Those things seem coherent to me and congruent with Phil's comments, but he knows a lot better about what he was talking about.

There is a 'tau', or half-life time, involved in the absorption and emission of photons. It's short, usually -- on the order of nanoseconds, if I recall. In the triplet state case mentioned above for phosphorescence, the improbability of the spin change merely means implies a much longer 'tau', on the order of tens of microseconds and towards milliseconds even, over temperatures that are from say 300C down to -200C, respectively, and with the usual rare earth ceramics I've used. I think this may even appear to be the case when there is no collapse of a wavefunction, as in the integrating sphere. But I could be wrong.

Actually, I'd like him to discuss his thoughts in more detail. I'd learn from it.

One of the key things I've used integrating spheres for is that the concept of emissivity can be neglected when observing the light and estimating temperature from some narrow band. If you try and look at the surface of blocks of aluminum and copper, for example, when both are at exactly the same temerature you will need to know the apparent emissivity (which involves not just the material itself but also the 'shape' from the observing perspective.) But if you drill a deep hole into both of them and look into that hole, you don't need to worry about these differences -- you can take the emissivity value as close to 1, in that case. And if you look, by eye, you will also see the same "color" of emitted light as you look inside the hole but will likely not see the same color from the outside surface.

Jon

Jon,

Sorry, I wasn't intending to be obscure. Integrating spheres, as John points out, are made of metal, and coated inside with a very white dielectric coating, e.g. packed PTFE powder (which is about the best homemade Lambertian high-reflecting coating there is) or paint made from MgO, BaS04, or TiO2. TiO2 is cheap--millions of tons per year go into making ordinary house paint opaque--and also totally insoluble in water, but it has a reflectance dip near 400 nm that makes it less suitable for integrating spheres.

The mechanism is just repeated dielectric reflections, which are broadband and memoryless, and so extremely fast. Light that enters the coating diffuses around inside for a very short time, bouncing repeatedly off the dielectric particles, but rapidly escapes back out the way it came in. (Mathematically this is closely related to the Gambler's Ruin problem, and you know how fast that happens.)

The total reflectance from the surface of the sphere is between 98 and

Because photodiodes really don't respond well to fast pulses, the size of the detected signal may very well go up a lot when the sphere is added, despite the loss.

Your comments about the Lambertian and emissivity advantages of spheres are just right, I think. It's probably possible to use slow fluorescence to stretch pulses, but when you get down to the femtosecond regime I'd worry about coherent effects mucking up the measurement. I might be too nervous about that, but you probably know more atomic spectroscopy than I do.

Cheers,

Phil Hobbs

Thanks, Phil. I had considered the idea of 'c' and bouncing around inside being what you might have meant. I also recalled the rough 'nanosecond per foot' figure but then roughly guessed that the 'ringing' of atoms might actually have a tau on that order or even more and then figured that was probably more what you were saying. Thanks very much for clearing all that up for me!

Jon