cap mult (again) at low current.

Bipolars won't work.

How about an RC lowpass filter? I'd expect some ceramic or film caps to work at 4K. Maybe even polymer aluminums?

A c-mult with a fet will have a large, variable voltage drop.

I'd expect a bigger noise problem getting the bolometer signal out, than getting the bias in.

How big do you expect the bolo signal to be?

Yeah maybe getting the signal out is the bigger problem. The signal levels are damn small, you take a spectrum of the lamp, and then lamp plus sample. And then take the difference.

The spectrum is created by a scanning mirror, You can choose the scan rate, which changes the frequency of the spectrum. The signal you receive, is the Fourier transform of the absorption.

It was long ago, but I do remember issues with battery/supply noise.

George H.

Can you put the bolo at the end of a coax (there are cryo coaxes) and do everything at the warm end?

A quickish scan plus signal averaging helps a lot with 1/f noise, but you knew that already.

Really low frequency noise is a bear. A few years ago when I was doing cavity locked lasers for downhole use, I had a brass plaque over my desk that read

DC The Final Frontier

It really is, once you get into the millihertz and microhertz.

Cheers

Phil Hobbs

I was thinking (in the shower this morning) that I could do a driven shield around the signal line. (One aspect that limits the scan rate is the RC product of the bolometer resistance and cable C.) And yeah I have some spendy stainless steel coax that would work. (CC-SS-100 from lakeshore ~$10/ foot.)

George H.

Low-frequency noise can often be simply nailed down as reference-voltage drift. From various sources.

Are you a PhD? Can I call you Dr George?

I get all my best ideas in the shower. I probably generate them in my sleep, and they are delivered in the shower. Water rationing threatens my business.

that I could do a driven shield around

The problem is similar to a photodiode TIA, but the bandwidth is pretty low. The coax thing is at least worth a look, if the project is worth getting compulsive about.

Were you planning to have an amplifier on the cold end? If not, the coax is as good as any other idea, and better than most.

The Lake of Lakeshore is full of money.

On Fri, 15 May 2015 07:30:12 -0700, John Larkin Gave us:

John Larkin and all of his spawn are full of s*it.

Yeah PhD, but the only time I'm addressed as Dr. is when the University comes begging for money.

There is something about the shower, most of your mind is on cruise control, you don't have to think about the washing, there's nothing much to look at, and only the "white" noise of the shower. And so your brain spins around until it finds something interesting to think about.

No cold preamp... that sounds harder than the driven shield.

I don't know about compulsive. I'm asking a favor and some sort of quid pro quo would be nice to offer. I don't think i'll be much help on the physics end... but I can help with the technical aspects. Less noise or more bandwidth, (faster scans).

George H.

And all your s*it is full of your spawn. You are a low-life excuse for a human. Eat your spawn and die, maggot.

On Fri, 15 May 2015 10:53:01 -0500, John S Gave us:

Oh boy, another retard from the total s*****ad bandwagon chimes in!

The problem with DC is that things take forever.

Yup, a lot of it, which is why you have to use nulling techniques down there. The laser locker used a method I came up with around 1992, when I was looking for noise canceller applications.

You use a lowish finesse etalon, and surf on the side of the transmission peak at the point where R = T. That way you can directly subtract the photocurrents from the reflected and transmitted beams, and servo around zero, which helps a lot. I was getting an Allan variance of about 1E-10 at 10,000 seconds, which is pretty good for something that had to fit down a 2-inch cased drill hole (40 mm ID, 38 mm maximum case diameter, 2 concentric zones of temperature control). That was despite using a random chunk of Kapton to attenuate the reflected beam till the curves crossed.

The noise canceller tie-in is that if you attenuate R until the crossing point is where dR/d nu = - dT/d nu, something quite special happens: the sum current R+T becomes independent of laser tuning, and so the amplitude and phase spectra decouple. Thus you can use noise cancellation on R+T to make shot-noise limited intracavity measurements, at least for signals weak enough that the cavity nonlinearity doesn't eat your lunch. (In a lossless etalon, the two points coincide, but in real ones the maximum R is always larger than the maximum T.

It's fun to watch the dip in the noise of R+T as you adjust the optical attenuator.

Cheers

Phil Hobbs

...

Panasonic ECPU works well in LHe, e.g. 1uF in 1210 case. NP0 ceramics work, too, 15nF is the largest I've used.

Regards, Mikko

Guarding the shield helps for bandwidth, but not against amplifier noise. Your SNR will be as if you drove the room- temperature amp without guarding, and then straightened the freq response with a suitable filter.

Some off-the-shelf opamps (like AD8605) keep functioning in LHe if you fiddle with supply voltages a bit, but I guess low-freq noise is more than you can tolerate.

In general reading out cryogenic NTD semiconductor bolometers is a bitch. One of last space missions still using them was Astro-E, the newer ones (like SPICA) are being built around superconducting transition edge bolometers, one of the main reasons being the relative ease of the readout. The NTD bolo readouts typically consist of a J-FET per each pixel, which are located in a specially controlled box at ~ 100 K. J-FETs have non-monotonous dependence of the noise on temperature, and the box temperature is adjusted to a sweet spot. Cumbersome.

Another possibility (depending on the physical size of the bolometer) is to use a resonant transformer a.k.a L-section, to match the bolometer R to the cable impedance, and to bias the bolometer with rf. Because large impedance ratio implies high Q, you'd probably need superconducting resonators, which I guess are out-of question.

My favourite NESG3031 would probably fall a little short for the required Rin ...

Regards, Mikko

But you can measure it instantly!

Oh, just a cap across the bias resistor. That might work wonders! The good news is I can find out. The prof. will be more than happy having me there.

The driven shield idea won't help, the scan mirror in the spectrometer can't go faster than 3kHz. (It's an old piece of kit with lot's of issues.)

The circuit, which I now remember (it has been a while) is a 9V battery up top, feeding a 10M bias resister and ~1Meg ohm photo-conductor (PC) both at 4K. Twisted pair to a preamp. The PC is GaAs doped with shallow impurites... I heard the gap, but it was in wave numbers ... a unit I don't think in any more. (1 cm-1 = 0.124 meV, I had to look it up.)

Thanks,

George H.

(Mikko is there a good fet (jfet?), for a

Oh, The profs have been great, but I'm leaning on several grad students. I was thinking of how to say thank you that involved more than pizza and drink. Electronics help? (I'm can't help much with the physics, w/o a lot of work.)

George H.

Big Grins! Thanks Mikko. Not a bolometer, (my bad)

Still Rin is ~Meg ohms, or worse if there is no light getting to the detector. (Transmission spectroscopy)

George H.

Wow, thanks Phil, I think I got most of that, though I'd have to twiddle the knobs myself to really understand. (When I've tried to sit on the edge of things, it always bounces around a lot... perhaps not enough BW in the loop.)

George H.