Zener noise source (freq < 1MHz.)

IIRC, the noise in an avalanche diode is proportional to current.

Could you have been falling prey to the variability in the Zener voltage and the low overhead between the diode and the rail? A constant-current sink would fix that...

ttdesign.com- Hide quoted text -

Hi Tim, I just had 12V zeners +30k ohm across 15 volts. Zero correlation between zener voltage and noise. (I only tested 5.)

George H.

Perhaps you meant inversely proportional to the square root of the current? It's actually a little bit more complicated than that. At low currents, the avalanche self-quenches from time to time, and you get spikes on the output as the diode capacitance charges up until another charge carrier shows up to start a new avalanche.

At higher currents the avalanche hardly ever quenches, and the noise is just the statistical noise on a sustained avalanche process

-- Bill Sloman, Nijmegen

"Tim Wescott"

... Phil

"George Herold"

Hi Tim, I just had 12V zeners +30k ohm across 15 volts. Zero correlation between zener voltage and noise. (I only tested 5.)

White noise snapped on at just under 1uA at 5mV, fell back to 2mV at 30uA rising again to 6mV at 100uA - then back to 5mV at 250uA and down to 2mV at

Looks like the internal noise generator is a varying current source that is being shunted by the zener's falling dynamic resistance as average DC bias increases.

So there is a tug of war going on.

.... Phil

tor

to

t -

Hi Mook, OK here are the numbers from the 5 diodes I tested. Noise is measured with a SRS spectrum analyzer (which has this weird dBV units.)

diode voltage noise (V) dbV

George H.

out

at

is

Nice, Did you find a place with more low frequency noise?

Here=92s some 'scope shots with bias current listed and noise density (at 35kHz) (All but the last made with zener feeding a TIA opamp circuit.)

the following is starting to look asymmetrical.

The one below has more noise at lower frequencies.

It=92s hard to catch the character with a single =91scope shot. But you can see distinct levels in the following

And now just occasional spikes up.

This last was made with diode grounded and cap from mid point between diode and bias resistor feeding a gain of 2 opamp circuit.

Very distinct levels!

George H.

That makes sense.

-- My liberal friends think I'm a conservative kook. My conservative friends think I'm a liberal kook. Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software

actor

p.)

ry

to

age

nt

to

is to

ext -

Doesn't go to DC but can give you a LOT of bang for your buck: Use a soundcard, custom code, and analyze away.

You get a dual channel system, plus can use the dual channel source to 'sweep'

Ch1 monitor the goes-intos Ch2 monitor the goes-outas

Do some math and you get data over the range of 10Hz to 90kHz. Not an SRS, but a cheap competitor.

..

me 12V

factor

up.)

try

st

e to

d).

ltage

rent

he

e is to

text -

Hi Robert, You mean a frequency swept source from the sound card? That's very useful too!

George H.

m...

some 12V

(a factor

ve up.)

f: try

he

me

best

ise to

sed).

o.

voltage

urrent

the

m 80 to

ise is to

ed text -

is

I mean TOTAL control of the source. as in 'arbitrary' signal source. and two of them! Plus, my soundcard's [Creative Labs EMU1212] channels match to better than 0.4% and can calibrate to better than 0.05% hold all day. and those errors are mainly at the higher end, roll-off. Playing special tricks youi can easily get down into the 5-10 ppm. I have measured data that shows less than 1ppm, that's using ALL the bits. I routinely use it to find a system's transfer function and noise spectral density, and the operating ambient noise sources: AC mains, PC noise, monitor noise, etc. You'd be surprised how loud those are. That card is the closest I've ever been able to get to a Nyquist rate: useful to 95%, but in spec up to 93% For $100 and a little fussing you have better than $10,000 worth of instrumentation. If you avoid writing software like the plague, there are 'free' SW packages out there [I think] with decent UI's and control panels..