noise in LT Spice

Well, that's just about metallic conductors. Nobody much talks about other stuff.

But 0.1 mm isn't all that small a distance. Resistors inside ICs can be smaller than that.

John

I e s

That's great, Thanks Joerg. John's link was to the theory by Beenakker.

George H.

is

the

nd

Yeah, but what's important (as I read it) is the ratio of scattering length to resistor length. It's not easy to see the shot noise signals in normal resistors.

I tried to measure excess noise in 100 meg resistors at the end of the day today. I failed. It might be a circuit mistake. I had a low noise power supply feeding a 100meg thick film with the other end into a TIA opamp. (Fet, opa134). The feed back resistor was another 100meg resistor. The circuit had an RC corner at about 600Hz. I set the output time constant at 1 second. This gave me a pretty noisy noise signal. (I didn't want to wait around for a longer TC. I was the only one left 'at work' 5PM before the holiday.) So I was only sensitive to excess noise at say the 10% -20% of thermal noise level. I changed the voltage from zero to 100mV to 1 volt and 8 volts. And saw no increase in the noise. When I disconected the input resistor the noise dropped by 1/2. I should try the same circiut with some carbon composite resistors! Do they make those at 100 meg?

George H.

I don't understand that part, the 1 second time constant.

John

This gave me a pretty noisy noise

is

the

f

the

nd

se

hot

ow.

...

w

Oh, sorry. The nosie signals get amplified, filtered and then sent through a multiplier and low pass on the output. The final low pass was set at 1 second. With a noise signals at only ~600Hz, it's like doing 600 separtae measurments in one second... the noise in the noise is then sqrt(600)/600. 10-20% or so.

George H.

You could just dump the TIA output into a scope. Most scopes will calculate RMS these days.

We did a similar thing a couple of months ago, with everything in a cookie can for shielding. We found a huge difference between metal film and cermet resistors, in the 50Mohm ballpark with 10 volts across the resistor. Our measurement bandwidth was also low, 1 KHz roughly, so we were probably seeing 1/f and other junk noise in the cermet. We didn't do spectral analysis, so I didn't know if there was shot noise, or if we could have resolved it in that bandwidth. I wish we'd have taken more data, because I need it now!

Hey, I just had one of those duh! insights. All sorts of transducers could be connected to one another: accelerometers, microphones, antennas*, photodetectors, whatever. If each is at some temperature, say 20C, it generates noise from various mechanisms: brownian motion, local IR radiation, whatever. Any of them can be connected to a resistor, or to one another, and exchange power. If they are all at the same temperature, they must exchange equal amounts of power, or else conservation of energy is violated. So all transducers, and all resistors, and all antennas, must generate the same power at 20C. I suppose everybody but me knew this already.

I've been seeing a lot of very silly "energy harvesting" things lately, like vibrating ribbons in moving air wiggling magnets near coils, 12" diameter windmills, peltiers in a car muffler, or piezo things in your running shoes that charge your iPod. Popular Mechanics has crazy ones on a regular basis, as does the press release mill at MIT. My invention, which I hereby donate to the world, is a wind chime surrounded by microphones.

John

• the antenna must be in a box at 20C, of course. An antenna pointed at the sky can act like a cold resistor, so you can derive free power from the difference in Johnson noise between a resistor at ambient and an antenna.

Well, a contemporary resistor _is_ a metallic conductor. Very flat and thin metal :-)

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Unless it's cermet or silicon or polysilicon or carbon film or tantalum nitride. I have no idea whether the anti-shot mechanisms work in them.

John

er is

in the

1/f a

ut the

(Mind

noise

e shot

flow.

rs

/no...

t n e o

Yeah I do that as a check sometimes. The resolution is poor on the scope.

I think it might be easier (faster) to look at the voltage dependence rather than the spectrum. The cermets are the noisy ones? The only really noiser resistors I've found are the carbon composites. When ever I get back to work I could look a little closer.

Yeah there's kT/2 of energy always floating around. I guess there's some thermal limit in all detectors, lots of times something else bites you first. Photodiodes don't seem to get that much better when cooled. I suppose there might be some dark current change with temperature.

It is strange, but people want to feel they are doing something to conserve energy. The big savings are in having smaller cars and houses, or at least more efficient houses. I cry a little when I see big picture windows facing north.

With all crazy energy harvesting devices you can ask of them one simple question. "Will they generate more energy during their lifetime than they cost in energy to make?"

George H.

No wonder if even the big brass doesn't get it. Crystal-clear proof: This year we installed an evaporative cooler. It uses 0.4kW versus the

7kW the big central A/C guzzles. So I called the utility because I didn't see those coolers on the list of rebate-elegible upgrades. "No, there is no rebate for those" I was told.

Yet there is a rebate for a higher SEER-number A/C. Which then probably consumes "only" 6kW. Great.

Or facing south in the summer. Or people setting the thermostat for the central heat/cooling unit at 70F and then leaving the sliding door open.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Provided they're in thermal equilibrium. Once you turn on the power, all bets are off. That's why you can get 25 kelvin T_noise amplifiers that work at room temperature.

Sorry, General Dynamics patented it already. ;)

There are genuinely useful applications for energy harvesting...a customer of mine wants to generate power from the temperature gradient across the thickness of a soldier's T-shirt. The idea is to emit a TX pulse periodically as long as the temperature gradient is still there...if you get my drift. :(

Cheers

Phil Hobbs

Brings new meaning to "stiff voltage source."

John

tter is

s in the

he 1/f

tra

but the

. (Mind

e noise

ave shot

c

ot

on flow.

rers

se/no...

out

can

g

the

nto

eg

l.

I was looking at the escess noise of high meg resistors today. I could just barely see some at the highest bias voltage levels. (10 volts.) There was no way I could look at the voltage dependence of the excess noise. So I put a 10k ohm carbon comp resistor into the circuit. Here's the noise as a function of the bias voltage.

The zero bias Johnson noise is 0.67 on the vertical. The bandwidth of the measurment was from 10 Hz to 1k Hz.

It was hard to see if this was linear or quadratic in the bias voltage so I plotted the noise density divided by the bias voltage. Shown here,

So at low bias it looks quadratic, but then rolls-off at higher voltages. Strange.

George H.

better is

ors in the

the 1/f

extra

.. but the

se. (Mind

the noise

have shot

nic

shot

tron flow.

us

turers

oise/no...

about

s can

ing

f the

into

0meg

vel.

nd

or

t

ss

ke

se

s , y

Hey doesn't anyone like my data?

I should add that it's 'noisier' than it could be. At some point during the data taking the resistor started randomly 'switching' between two 'noise states'. The average noise would all of sudden more than double, and then after a while, (0.1 to ~100 seconds) switch back. It took me a while to notice this.... at the end of the day I set my DSO at 50sec/div and the output TC at 0.1sec, let it run at a constant 4 V bias, and looked over my shoulder every once in a while.... clear steps between two 'states'.

Resistors don't have shot noise (whatever they may be made of)

Try the following thought experiment: take a hypothetial resistor with both shot noise and Johnson noise. The current noise density will be: In^2 = 4.k.T/R + 2.q.I Now combine four of these resistors in a series-parallel arrangement to give a combiner resistance R. The current noise density of the combination will be: In^2 = 4.k.T/R + q.I Repeat this N times, and you have: In^2 = 4.k.T/R + 2.q.I/N

So... even if a miniscule element of the resistor does exhibit shot noise, the resistor as a whole does not.

I sometimes do that, but you need to add temperature parameters if you want the noise to stay constant as you change temperature. It's also inconvenient if you want the noise to relate to a circuit current as in shot noise. It's rare that I would want to include external shot noise in a circuit simulation, but it could happen (e.g. if you have a sensor that isn't supported in your version of SPICE). Here is one of many possible ways to introduce the shot noise: You can use an idealised bipolar transistor, drive the circuit current into the base, and set a high current gain (e.g 1000), and drive the collector output into a (biased) CCCS with the inverse gain (e.g. 1/1000). The output current of the CCCS is the same as the circuit current, but with shot noise (plus a fraction) added.

--------------------------------------- Posted through

th

That's pretty neat George! (Though I have to work through the details.) I think you might have to be a little careful how you split the current in each branch though. You can't have 1/2 an electron going one way and the other half going the other.

Have you read Rolf Landauer on shot noise? He shows the shot noise in a resistor is reduced by the ratio of the scattering length to the resistor length...

George H.

ise,

t

to

t .

th

Photodiodes have shot noise. Liquid (ionic conduction) resistors have shot noise. Very small resistors have shot noise. Semiconductor resistors, ditto.

The reason metallic resistors don't have shot noise is a macro-scale interaction of electrons which smooths out the electron flow. Small resistors, where the e-e interactions don't have time to work, have shot noise.

High-ohm cermet resistors have shot noise. We've measured it. And it's hard to find 50M metal-film resistors.

John

oth

oise,

s to

nt

).

ith

Have you looked at shot noise in liquids? (You'd need some 'squishy' metal box.... or a shielded room.) It should go as the scattering length, what's the scattering length of ions in liquids?

George H.

Very small resistors have shot noise. Semiconductor

"George Storm"

** Wot bollocks.

** What a pile of mind numbing sophistry.

Bet this fool has made similar estimates of how many angels can dance on the head of a pin.

.... Phil