A simple reverse breakdown eb junction, running at about 32uA is measured to be producing noise with a 10dB/decade slope through the audio band. This measured both before and after a x30 amplifier stage.
Measurement was courtesy of an Audio Precision analyser.
Was expecting a level spectrum 'white' noise characteristic and had subsequent filer to follow for 'pink' characteristic. Wanted both to be available.
You probably have to buffer the diode. Maybe a J-fet.
Depending on the model of AP, you have to watch the loading. Note directly related to your problem, but sometimes the full power bandwidth test has issues as the AP switches amps since the loading changes.
Scopes are wonderful things and I'm using mine, but you can't tell pink fro m white visually on a good day. Getting an FFT from a recording or scope du mp takes time and the AP seems to show a white noise spectra where it does exist, on other references, so far.
I'd already doubled the current once (to 60uA), producing a slight flatteni ng above 20KHz. Taking your suggestion - 300uA produced flattened response down to about 40Hz, which might be good enough for anything but this applic ation.
The kicker, is that the flattening occurs by suppressing the LF end of the spectra, not raising of the HF spectra; so that the net amplitude of the wh ole thing is suppressed; in this case by almost 20dB at 200Hz. That's hard to get back.
I'm going to have to find a noisier source that will require only the singl e gain stage intended. Either that or turn the pink noise filter into it's inverse to produce white noise from the original pink. Sound messy to you?
What does it look like on the 'scope? As JL said you can sometimes get signals that looks like random telegraph noise. (Random plus/ minus steps in the voltage.) I think that might have a 1/f character. At different currents the steps go away and you get pulses... usually with an RC tail. So try a different current. (it might be more or less.) What sort of frequency range are you looking for? A 20V zener, and ~100k hom resistor, biased from ~30V makes lotsa noise up to a few MHz. (but keep it out of the light!)
rom white visually on a good day. Getting an FFT from a recording or scope dump takes time and the AP seems to show a white noise spectra where it doe s exist, on other references, so far.
Hmm is it a digital 'scope with a built in FFT?
ning above 20KHz. Taking your suggestion - 300uA produced flattened respons e down to about 40Hz, which might be good enough for anything but this appl ication.
e spectra, not raising of the HF spectra; so that the net amplitude of the whole thing is suppressed; in this case by almost 20dB at 200Hz. That's har d to get back.
gle gain stage intended. Either that or turn the pink noise filter into it' s inverse to produce white noise from the original pink. Sound messy to you ?
Is this a one of, or something for production?
the problem I found with most zenering type noise sources is that the noise varies wildly from piece to piece. (factors of 2 or more.) The 20V (and a lso higher) zeners don't seem to have this issue. (noise is pretty much th e same from piece to piece.) My "guess" is that this is becasue the whole diode breaks donw in the 20V zener and the pulse height I get depends only on the diode capacitance.. which is farily constant.
My point was that a low current zener breakdown can make strange pulses and spikes and even sawtooth oscillations, or non-Gaussian probability distributions. A scope can spot things like that.
What sort of noise density are you seeing? My zener (in the pdf I posted) was making around 300 nV/rthz, and the b-e junction of the BFT25 was a lot less.
Digital pseudorandom sequence generators are more predictable. Or official, tested noise diodes. Anything else is risky.
--
John Larkin Highland Technology Inc
www.highlandtechnology.com jlarkin at highlandtechnology dot com
Precision electronic instrumentation
Oops... ok only to ~500kHz. Here's a link to some 'scope shots.
formatting link
(I hope that works.) Tek001 is the FFT at low frequency.
002 is FFT at high frequency
003 is averaged signal (triggered near top of peaks.)
004 is a single trace.
Grin.. the 20 V zener gives volt level pulses with no gain. I'd have to stick it into a spectrum analyzer (which is at a trade show) to get real numbers.. but something like 1V/(1MHz)^1/2 ~1mV/rtHz. (OK maybe a factor of ten smaller than that.)
Hmm, Well the spectrum is pretty flat down to low frequencies... I don't really know where it get's 1/f-ish. (The specturm analyzer is on a truck somewhere between here and Denver.) But certainly good to 10 Hz. (Maybe I'll remember to look at the LF part of the spectrum when it returns.)
Re: the Gaussian amplitude distribution. Well this will certainly not be gaussian at high frequencies. (time scales on order of the pulse width.) But at low frequencies there is no problem... It's like shot noise, at short enough time scales it's non-gaussian too... delta functions and all that.
And who cares about gaussian anyway?... as long as it's white. Non-gaussian might be a "feature", less dynamic range needed for the same noise level.
The truly white waveforms have no dominant shape until they're pinked, at which time the LF content begins to dominate. If it weren't filtered for audio bandwidth before further manipulation, pink could be quite difficult to handle. While white, it looks oddly the same, no matter what time reference is used on the scope.
I have a number of digitally generated white noise source files to fall back on. They all have their problems. The aim is the eventual production of something with a controlled crest factor - I know it won't be completely pink when finished, but that's what the standard prescribes.
Perhaps it's time for standard writers to simply specify 'the waveform', rather than describing how it is supposedly achieved. RMS and peak values of .wav or other files are easily established, maybe generating pink faces.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.