From: Phil Hobbs
Subject: Re: reverse-biased diode noise source?
Date: Mon, 16 Sep 2002 12:53:40 -0400
Organization: IBM T. J. Watson Research Center
NNTP-Posting-Date: 16 Sep 2002 16:53:42 GMT
X-Mailer: Mozilla 4.61 [en] (OS/2; U)
Now that amplifiers are so quiet, it's easy to make a well-calibrated
low frequency noise source by shining the light from a LED on a
photodiode--_NOT_ a phototransistor or photodarlington. Drive the LED
from a quiet current source (e.g. a battery and metal film resistor or a
well-designed current source with a big voltage drop across its (metal)
sense resistor). Provided the incident light is really unmodulated, the
photocurrent exhibits exactly full shot noise. Shot noise is white and
Gaussian to very high accuracy, so you can calibrate the noise power
spectral density by measuring the DC and applying the shot noise
i sub N = sqrt(2*e*I sub dc) (1-Hz BW)
where e is the charge on the electron (1.602e-19 coulombs).
This method is much better than the Zener approach, because it's based
on fundamental physics rather than device parameters that have to be
calibrated from unit to unit. Since it's easy to generate a
photocurrent that has exactly full shot noise--shine a flashlight on
your photodiode--it's easy to test by comparison whether your LED drive
current source is quiet enough.
Tom Faloon wrote:
> Hi Walter,
> Any silicon or germanium rectifier diode, transistor base emitter junction,
> or zener diode will act as a noise generator when reverse biased to
> breakdown (avalanche) region.
> However, this is generally an unwanted effect, and manufacturers do not
> specify characteristics.
> It is possible to buy diodes with published characteristics, but these are
> specialist devices,
> and cost and minimum order quantity put them in reach of business users
> You can obtain useful output from everyday devices, but you will not be able
> to predict results accurately in advance, or assume repeatability from
> component to component.
> In the simplest application, simply reverse connect the diode to a supply,
> via a current limiting resistor, increase the supply voltage until reverse
> current flows,
> and a noise voltage will then be produced across the diode.
> You can expect a fairly 'white but not perfect' noise spectrum over a wide
> frequency range.
> Non perfection is most likely to be due to secondary noise functions in the
> diode, the
> most pronounced being flicker noise. (1/f) which will be present below about
> 100 kHz.
> The bandwidth will depend on the components and layout used. (Junction
> capacitance, lead inductance, PCB track capacitance & inductance.)
> Bandwidths into the GHz region are possible.
> The noise level generally increases with reverse current. Level will depend
> on the device used
> As a rough figure think fractions of a nanoVolts / root Hz. (Perhaps X 10 to
> Useful devices.
> A standard rectifier diode. (Not always practical, because most diodes have
> a fairly high breakdown voltages.)
> Base emitter junction of a bipolar transistor is more practical. (Reverse
> breakdown is often around 5V)
> (Leave the collector unconnected.
> Zener diode, but use devices above about 6 or 7 Volts.
> (We call all these devices Zener diodes, but in fact they operate in two
> different modes. Devices operating below about 5 Volts rely on the Zener
> effect. Higher
> voltage devices use the avalanche effect.)
> Tom Faloon
> remove 'z' from address to email direct.
> Walter Harley wrote in message
> > I see many references to the fact that a reverse-biased zener diode
> > functions as a wideband noise source. However, I can't find much
> > information about how to put this into practice. How much noise does it
> > generate, for how much current? Does it have to be in avalanche, or
> > does it generate noise even at leakage-current levels? Is there any
> > difference between different diodes, e.g., different voltages? What is
> > the frequency spectrum like?
> > Thanks,
> > -walter