A non contact way of messuring Johnson Noise?

"Jamie"

** The resident radio ham f****it is on a real howler this time ........

** The replies to this one will be highly entertaining indeed.

.... Phil

...

Sounds quite challenging. In my field guys do something related, noise thermometry at sub-kelvin temperatures, using a SQUID as the amplifier and inductive non-touching pickup from the Cu block whose temperature is to be measured. My boss here has done a much higher temperature version of effectively the same thing, namely an instrument for non-touching measurement of the temperature of molten steel in blast furnaces. The steel is covered by slag, so that pyrometric measurement is typically not possible. Something about the blast furnace instrument was published in Rev. Mod. Instr. vol. 64, p.1593 (1993).

The key idea is to construct a reactive pickup, whose losses (i.e. real part of the impedance) are predominantly due to eddy currents in the conductive medium whose temperature is to be measured. Fluctuation-dissipation theorem then allows one to deduce the temperature of the medium.

Your case is not exactly the same, however. You are not trying to measure the temperature of some material in thermal equilibrium, but rather to measure some sort of excess noise, at some predefined terminals to the medium, right? (1/f when there is dc bias? White noise from some external device which merely passes through your resistor?)

on

If there are other signals present in addition to the noise, the measurement is more difficult still. You'd need enough dynamic range to detect simultaneously the noise and the largest signal present in your system.

Regards, Mikko

Yes, an effect we try to minimize by eliminating any material that is not either dielectric or superconducting from near the SQUIDs.

Interesting paper, thanks. They claim about 1% agreement with a thermocouple.

Molten steel is easier- they use disposable platinum alloy thermocouples- this seemed to be aimed at intermediate temperature range up to 1000°C in applications such as rolling mills where the material is solid but may have nasty slag on the outside.

on

t

Hi Jamie, I'm not sure of your circuit configuration, but can you just attach a cap to the resistor and measure the noise coupled through the cap? What's the resistance and temperature of the wire? (There are bjt opamps with ~1nV/rtHz of noise and FET's with ~4-8nV..) If it's got long cables attached then sometimes the cable capacitance can limit the bandwidth of the system. A good way to measure a small signal is to look for noise differences. If you can switch between having the input shorted and then looking at the resistor you can see noise levels below that of the opamp. (But you've got to keep the bandwidth the same!)

George H.

You do noise thermometry? Do you use PTB gear or have you rolled your own?

I have only done contacted stuff, like measuring Kapitza resistance of our resistive thin films from a van der Pauw pattern. Heated it by current through one contact pair, and coupled the SQUID to the orthogonal contact pair.

I mean Rev. Sci. Instr. (mixing it with Rev. Mod. Phys. ...)

Right. However I remember they did field tests at blast furnaces, too. It seems the steel mill workers have a habit of throwing in all scrap metal they see lying around - one can cast almost anything into such a furnace and it just vanishes. My boss told they had left one instrument prototype lying around in the vicinity of the furnace after one of their field trips. The next time they came around the instrument had disappeared.

on

t t

Inductive probably is the way to go. Although interpretation may be a bit 'fun'. I make Eddy Current NDE Instrumentation that maps/displays a 3D rendition of the resistivity of aluminum sheeting - based upon the changes caused by work hardening of the material during crack formation. These are fuselage layers and not a cylinder, like your wire.

Essentially better than 0.05% relative and 1% absolute resisitivity.

I think the principle can be applied to non-contact resistive measurement for your wire. Do you need values 'along the wire' or a simple aggregate number?

nulling out signals would not be issue, that we already do else where with no issues.

The inductive idea is intriguing though and I may experiment with that a little.

Thanks.

Jamie

And there is the problem, I really can't make a connection to this circuit although I already have, to test the noise and I got a reasonable reading. But the bigger problem is when the High voltage is turned on, Sometimes we get a little problem near by when the voltage comes up faster than the beam current does. A static field builds, it likes to drain itself into this area. The effect does not hurt the R but i am sure it would kill anything sensitive attached to it. The R sits in side a pressured vessel of SF6 that seems to blanket other methods of detection.

This is just a pet project idea I am trying and if works, it'll replace the method we currently use, which is a calculation of 3 other factors to come up with an assumption of what is happening inside.

Jamie

And, I think that may be the only way to make this work. Measuring a fairly small noise in a system where huge "signal" voltages are present is always going to be REALLY tough. Measuring the noise of the resistor under a variety of lab conditions, and then knowing the resistor noise by measuring these equivalent operating conditions may be the only logical way to get what you want. I'm guessing the only parameters that really matter are an intrinsic cal factor for the resistor, its temperature and applied voltage. I'd think that if you start with the right measurements in the lab, prediction would be a lot more accurate than trying to measure microvolts of coupled noise where there are KiloVolts of applied potential on the resistor. Not to mention the problem of HV sparking to the wrong place and smoking everything now and then.

Jon

I think I have come up with a way to commutate the signal down the throat of the 3Mvolt rectifier stack to get statistics on the R.

Problem is, you can't directly acquire values of pretty much anything at that location. Even the heater (gun) on the beam tube uses the ripple voltage into an RF transformer to create the 3 volts or so needed, instead of just feeding some wires down the vessel to it. This means of course you need the stats driven with delwin rods from servo motors, one as a minimum set point and the other as the beam current gain.

I think I can use the same source of voltage for the heater to operate a local circuit and convey the information via a fiber option thread, one that has a blocking jacket on it, because we also have optical monitors in the vessel. This thread would pass down the center of the rectifier stack into a receiver via feed through's that makes connections at a much lower penitential.

By connecting only to the electrical that sits at the end for power and optically coupled, I think I can be safe with a direct connection.

If I spiral the fiber on its wall down the stack I should be fine, famous last words! :)

Jamie

voltage

replace

microvolts

Noise wise corona is far worse.

?-)