noise in LT Spice

John Larkin schrieb:

Hello,

if 1 amp is flowing 1 second long, 6,241,509,629,152,650,000 electrons are passing. It is hard to detect shot noise under this condition, but this is no proof that it does not exist. If we think about 1 nA flowing

1 µs long, we have only 6,241 electrons passing, but it may be 6,242 electrons also. With current less than 1 nA, shot noise is detectable.

Bye

The shot noise of a 1A current is trivially detectable, if you have sufficiently advanced gear, such as a sense resistor and a blocking capacitor.

The zero-order argument for long resistors having less shot noise is plausible, but its results are wrong in detail, by something like a factor of 3 iirc--it leaves out the fact that the scattering processes that produce the resistance of the rest of the resistor also contribute randomness. Different scattering processes also make the shot noise either better or worse--electron-electron scattering is apparently quite different from scattering from impurity ions or crystal defects, and electron-phonon scattering is different again.

Real solid state physics makes my head hurt. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

h both

to

t noise,

ou

so

in

uit

ways to

rrent

000).

t with

I guess I just have this model in my head that an ionic liquid should look just like a metal resistor, but with slower charge carriers. I could be totally wrong, though. Maybe some sort of screening happens in the liquid?

George H.

Yup. Full shot noise at 1 amp is 5.66e-10 amps RMS.

If we think about 1 nA flowing

It's approaching 0.1% in a KHz bandwidth.

A photodiode pulse of, say, 100 nA has obvious shot noise as seen on an oscilloscope.

We measured a bunch of 50 meg resistors for excess noise. The cermets were very noisy compared to metal films. It's a tricky measurement but not impossible. The spectrum suggested that a lot of the noise was shot noise, but it's hard to keep the bandwidth up at the impedances involved.

I conjecture that some very cruddy materials, like cermets and ionic liquids, don't have the long-term ordering that fixes shot noise in metals.

There's not much available in the books I have, or online, about shot noise, other than that tubes and semiconductors have it, and metallic resistors don't.

John

"John Larkin"

** Only pure metal resistors are free of excess noise - wire wound and " metal foil " types are two examples.

Metal film ( sputtered metal ) resistors are not so blessed having excess noise typically quoted as in the order of 0.1uV/V/decade of frequency - as verified by my tests a couple of years back and posted here.

Even Wiki has a neat summary of the basic facts:

.... Phil

Excess noise (aka flicker noise and 1/f noise) is a different issue from shot noise. Flicker noise is caused by conductance fluctuations, and is concentrated at low frequency, whereas shot noise is caused by decorrelation of the electrons and is white.

(IIRC you pointed that out in one of our recurrent discussions of this topic.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

You forgot to describe one over f.

I still haven't found a reference that says whether nonmetallic resistors and ionic liquids have true (uncorrelated electron) shot noise. Shot and 1/f and whatever are usually lumped as "excess noise."

Electrons do like to spread out evenly. Even a vacuum diode has somewhat less than shot noise because of mutual electron repulsion.

We needed a 50M low-noise resistor, to make a small bias current in about a 2 MHz system, and cermets were terible. We used two 25M Dale axials in series, RN55 types, probably vacuum coated films on ceramic tubes, and they appered to have no excess noise. The rest of the board was all surface mount, so we flew them over other parts, with a mid-air solder joint.

It would be an interesting project to measure various resistors for shot noise. I wish I had the time.

John

But, isn't there recombination noise? The charge carriers can enter and leave the channel by recombination, just as by injection, and that DOES carry some shot noise. It'd depend on the geometry of the resistor, and the (?Debye) distance scale.

Suppose a conventional resistor (metal plates with highly resistive element between) were squashed very thin and flat, so that electrons crossing between plates were far away from others laterally, and crossed quickly enough that they had little likelyhood of interacting with nearby ones. Wouldn't that result in shot noise? Geometrically, that would be like a semiconductor junction, large area but very thin, so electrons cross but don't get much opportunity to correlate.

I found this:

It says that mechanisms like geometry and scattering impurities can break up electron ordering and result in values around 1/3 full shot noise. Some geometries have Fano factors closer to 1.

As a practical matter, probably any resistor cruddy enough to have shot noise would have unacceptable amounts of 1/f noise.

John

"John Larkin"

** Frank Seifert had the time and the inclination:

Nice procedures and results in line with maker's data and my expectations.

Thick film resistors ( Cermet) are the poorest, then carbon comp, carbon film, metal films and the bulk foil and wire wounds come out as top dogs.

The high powered resistors shown in the Fig 9 are all forms of Cermet in TO220, TO246 and similar style packs.

.... Phil

"Phil Allison"

** Just looked up who Frank Siefert is:

Frank has a doctorate in Physics from Leibniz University and is employed at the Max Plank Institute for Gravitational Physics. His special area is control of high powered lasers as used in gravitation wave detectors.

Mention of his 2009 thesis can be seen here.

..... Phil

Does it _really_ have to be high powered lasers? What about "garden" variety normal power?