amplifying a sub femtoamp of current

I have to do broadband detection -can't go with tuned circuits. John do you know of any commercial cold JFETs which won't have freeze-out at 4 kelvin? One which i can buy?

This can be assured by detecting for along time, approx. 1 sec - the digitizer we have goes at 1 MHz sampling rate - this makes 1 million data points to store, not bad at all.

Also, i am looking use the transimpedance amplifier with a resistor in feedback, described by Bob Pease in "

". This might help me to put a high value resistor in feedback without compromising the bandwidth.

It's my unofficial understanding that regular silicon jfets work cold. I'm sure google has refs, and the Review of Scientific Instruments is rife with stuff like that.

What do you estimate as the spindown time constant of a trapped ion? I guess you'd want minimal resistive losses in the amp to reduce damping of the orbit.

Really low-level charge amps use pure capacitive feedback; see Knoll's book on radiation detectors. Or how about a jfet amp with no feedback at all, just the floating gate on your pickup electrode?

John

I have found some Ge JFETs from NASA, will try to get some. For the time being i am thinking to use the GaAs FETs i have and measure their leekage current. I know it has to decrease when i cool them to 4 Kelvin.

I have to check the numbers for that, if i recall correct the lmiting factor which causes the ion orbit to decay - is collisions with other gas molecules not due to the resistive decay of the preamp. The detection duration is typically in the order of couple of seconds, typically.

How would the FET will be biased in this case? In my system there are 2 detection electrodes for which i am planning some sort of differential configuration.

I'm not well versed in this stuff, so please, guys, be gentle in your flames if I'm completely out of order.

After following all the good advice on your analog signal conditioning, it's my impression from this thread that you'll still be pretty far down in the noise.

If that's not correct, you could simply acquire broadband data at, say,

5Mhz for several seconds and run an FFT on it. This would give you at one shot all the ion species present in the experiment.

If you are indeed buried in noise, you could run correlations on the same acquired data with offsets equal to the period of each expected ion species, which should dig any ion signals out of a lot of noise.

John Perry

As I mentioned, GaAs fets are nasty dudes. The defect densities and trapping states are awful. I'd excpect the Ge's to be leakier than silicon, too.

Cool, pretty high Q once you get up to a MHz. Even so, a resistor will add Johnson noise.

Just use it common-source with floating gate. Gate voltage will be close to zero, so drain current will be close to Idss, gain will be high, and you'll boil a little bit of helium. I did an unbiased jfet infrasonic microphone preamp once, using a GR ceramic mic element, for listening to static tests of the Saturn V main engines, with the mics scattered around southern Mississippi. They accused my amp design of intermittent motorboating but eventually figured out they were picking up the subsonic mating calls of alligators.

Who is that specialty fet company... Interfet? I think they may have some cryo-rated jfets.

John

I am thinking he would be doing dam well to squeek out 100KHz bandwidth...

However, GaAs works down to 4K. Si doesn't. Si JFETs perform beautifully at 77K though.

Make that nitrogen.

Yeah, but not for LHe temperatures. I once did some literature research on the subject, and the upshot is: Depletion-mode Si JFETs don't work at 4K due to carrier freeze-out. Enhancement-mode Si MOSFETs work if you're lucky. GaAs works in general.

In my drawer I have a dual JFET that is rated at 4K. It is actually mounted on a thermally insulating stud inside a TO78 metal can package together with a small heating resistor. It was bought 15 years ago for $500 from a company that doesn't exist any more. Talk about "Not recommended for new designs" ;-)

--Daniel

Thanks for the update. The application would benefit from minimum capacitance, so plumbing the signal out to 77K would be an issue. Maybe a driven guard back from a jfet amp at LN temperature would work.

John

Hmm, i have an option of using a 70K heat shield to mount the JFETs, but then you add the capacitance of a meter long wire carrying the signal from the detector to preamp. I have tried and have still not found JFETs which can be used for liquid He temp. So, am i right in thinking that there are no other commercial devices suited from my applications (low noise and 4 Kelvin) other than GaAs FETs. I have some low noise JFETs from interfet - i can make a simple preamp and cool it to see how far it can go with reasonable performance - though i am sure its not meant for cryogenic operation.

Hmm, i have an option of using a 70K heat shield to mount the JFETs, but then you add the capacitance of a meter long wire carrying the signal from the detector to preamp. I have tried and have still not found JFETs which can be used for liquid He temp. So, am i right in thinking that there are no other commercial devices suited from my applications (low noise and 4 Kelvin) other than GaAs FETs. I have some low noise JFETs from interfet - i can make a preamp with floating gate (common source) and cool it to see how far it can go with reasonable performance - though i am sure its not meant for cryogenic operation.

One approach would be to attach the JFET with high thermal resistance leads, such as fine brass wire--the thermal resistance is very high at low temperature, so you might be able to run a normal JFET at 77K in a

4K environment without adding a significantly increased heat load. You can set the dissipation remotely by changing VDD.

The thermal conductivity integral of stainless steel at 77K is about

200W/m, so if you could stand 50 mW dissipation, you could get 77K with three 1-mm stainless steel wires, 1 cm long--with lots of room to reduce these numbers if 50 mW is too much. You can't go too low, of course, since radiation will eventually dominate.

Of course, you'd either have to apply the heat externally, or make sure the JFET was turned on before you transfer the helium.

Cheers,

Phil Hobbs

I remember a very old article talking about a quad CMOS NAND IC (like 74hc00) of the first generation,used at liquid helium temperature as an analog pre-amplifier for a scanning detector in space. I remember they were very suprised about the quality.

People use fine manganin wire for this, too. And somebody makes fine cryo-coax.

How about a silicon diode as a temperature sensor, and a resistor heater, with the control circuit at 77K or room temp? Adds 2 more wires. Below about 20K, a silicon diode's voltage drop gets huge, and gives tons of signal vs temperature around 4K, typically biased at 10 uA.

Or maybe just monitor the jfet current as the temperature indication, and drive the heater resistor as needed. Wow, that's a new jfet bias scheme! Who is it that claims there are no new circuits?

I guess there's still the fundamental problem... is there enough signal? If the thing is to be untuned, I guess it'll need an FFT to find the orbital frequency. How long does it take to do a 2M point FFT these days?

John

In article , John Larkin wrote: [....]

I'd expect they do. If you show up with a bag full of money wanting special testing of an existing part, they are happy to do it for you.

--
--
kensmith@rahul.net   forging knowledge

Ok, so you have a sort of mass spectrometer ? Except the 20pF parasitical capacitance of the plate, which is the parasitical capacitance of the iron-glass pass through assembly from the vacuum chamber to the outside world ? The time constant of any current to voltage converter is about RC (3RC if you want to be more accurate). Converting say 1pA on a precise glass isolated 1 Gohm resistor (and a very accurate electrometric OP AMP, you'll get after the first stage about 1mV ( a reasonable value for the future magnification) if I didn't mistake with the magnitude orders. The parasitic capacitance will give you a time constant of about

1Gohmx20pF = 20mS.

So how you'll be able to get a bandwith of 10Khz to 1Mhz with such a plate capacitance?

greetings, Vasile surducan senior engineer National Institute for Isotopic and Molecular Technology, Cluj-Napoca Romania

Arch wrote:

I like Phil's idea, i have Phosphor bronze wire which i w'ld use.

The Thermal conductivity of this wire is merely 1.6 W/m.K around 4 Kelvin.

I have to heat it externally as the amplifier will be inside the cryostat well before we start to detect. For this as John said we can use a resisitive heater and monitor the JFET current. How about that?

I like Phil's idea, to run the JFETs at around 70 K in 4K environment. i have Phosphor bronze wire which i w'ld use.

The Thermal conductivity of this wire is merely 1.6 W/m.K around 4 Kelvin.

I have to heat it externally as the amplifier will be inside the cryostat well before we start to detect. For this as John said we can use a resisitive heater and monitor the JFET current. How about that?

I like Phil's idea, to run the JFETs at around 70 K in 4K environment. i have Phosphor bronze wire which i w'ld use.

The Thermal conductivity of this wire is merely 1.6 W/m.K around 4 Kelvin.

I have to heat it externally as the amplifier will be inside the cryostat well before we start to detect. For this as John said we can use a resisitive heater and monitor the JFET current. Doing 2M point FFT is not a big deal which yes is done to find the frequency of the ion. Its typically done on the fly, you won't even notice on a good PC.

Make sure you're using the thermal conductivity integral, and not just the value at 4K times the temperature drop--you'll significantly underestimate the thermal loading otherwise. Flip to the second plot in that Lakeshore app note and make sure you understand the distinction.

Cheers,

Phil Hobbs