2N2222 Zener noise

Started by Bitrex November 26, 2010
I suppose I could set things up to do the research myself, but since I'm 
feeling lazy I thought I'd check first: By how much do you have to 
reverse bias the emitter-base junction of a common transistor such as 
the 2N2222 before it starts getting appreciably noisy? Does it have to 
be above the 5-6 volt breakdown voltage, or will significant noise occur 
before that point? I'm looking for a noise source that will work with a 
3 volt supply.  Maybe a different transistor with a lower breakdown voltage?
On Nov 26, 9:26=A0am, Bitrex  wrote:
> I suppose I could set things up to do the research myself, but since I'm > feeling lazy I thought I'd check first: By how much do you have to > reverse bias the emitter-base junction of a common transistor such as > the 2N2222 before it starts getting appreciably noisy? Does it have to > be above the 5-6 volt breakdown voltage, or will significant noise occur > before that point? I'm looking for a noise source that will work with a > 3 volt supply. =A0Maybe a different transistor with a lower breakdown vol=
tage? You are snookered by the physics. Reverse-biased diodes break down by two different mechanisms depending on the voltage applied acorss the junction, the Zenner (quantum tunneling) mechanism at alow voltages - less than 5V - and the avalanche (impact ionisation) mechanism at higher voltages. http://ecee.colorado.edu/~bart/book/book/chapter4/ch4_5.htm Avalanche breakdown gives a noisier leakage current than quantum tunnelling - with avalanche breakdown there is always the chance that an electron will make it all the way across the gap without generating a second electron-hole pair, killing the leakage current until a cosmic ray or a radioative atom in the vicinity generates a new charge carrier pair to restart the avalanche process. Quantum mechanical tunnelling is also a random process, but it is a single random process, without the second stage of avalanche multipication to generate lots of outliers. With a 3V supply rail you are stuck with Zener breakdown. You may have to settle for a pseudo-random binary sequence generator, or amplifying the Johnson noise from a resistor - 1nV per root hertz for a 50R resistor at room temerature, rising as the square root of the resistance, bearing in mind that "low noise" integrated circuit op amps typically also generate 1nV per root hertz white noise at room temperature. -- Bill Sloman, Nijmegen
On 11/26/2010 6:26 AM, Bill Sloman wrote:
> On Nov 26, 9:26 am, Bitrex wrote: >> I suppose I could set things up to do the research myself, but since I'm >> feeling lazy I thought I'd check first: By how much do you have to >> reverse bias the emitter-base junction of a common transistor such as >> the 2N2222 before it starts getting appreciably noisy? Does it have to >> be above the 5-6 volt breakdown voltage, or will significant noise occur >> before that point? I'm looking for a noise source that will work with a >> 3 volt supply. Maybe a different transistor with a lower breakdown voltage? > > You are snookered by the physics. Reverse-biased diodes break down by > two different mechanisms depending on the voltage applied acorss the > junction, the Zenner (quantum tunneling) mechanism at alow voltages - > less than 5V - and the avalanche (impact ionisation) mechanism at > higher voltages. > > http://ecee.colorado.edu/~bart/book/book/chapter4/ch4_5.htm > > Avalanche breakdown gives a noisier leakage current than quantum > tunnelling - with avalanche breakdown there is always the chance that > an electron will make it all the way across the gap without generating > a second electron-hole pair, killing the leakage current until a > cosmic ray or a radioative atom in the vicinity generates a new charge > carrier pair to restart the avalanche process. Quantum mechanical > tunnelling is also a random process, but it is a single random > process, without the second stage of avalanche multipication to > generate lots of outliers. > > With a 3V supply rail you are stuck with Zener breakdown. You may have > to settle for a pseudo-random binary sequence generator, or amplifying > the Johnson noise from a resistor - 1nV per root hertz for a 50R > resistor at room temerature, rising as the square root of the > resistance, bearing in mind that "low noise" integrated circuit op > amps typically also generate 1nV per root hertz white noise at room > temperature. > > -- > Bill Sloman, Nijmegen >
Thanks for your reply and for the enlightenment regarding the difference between breakdown mechanisms. Instead of trying to amplify Johnson noise, perhaps I can use a spare opamp section of the project as a hysteretic oscillator and make a charge pump to kick the voltage up above the 5 volt threshold? I wonder if a low voltage op amp like the LM324 acting as a charge pump would be able to get the voltage high enough to initiate the avalanche breakdown mode - I'm not sure what the '324's output can swing at such low voltages, but I bet getting 5 volts with a 3 volt supply with such a circuit might be pushing it.
Bill Sloman wrote:
> On Nov 26, 9:26 am, Bitrex wrote: > >>I suppose I could set things up to do the research myself, but since I'm >>feeling lazy I thought I'd check first: By how much do you have to >>reverse bias the emitter-base junction of a common transistor such as >>the 2N2222 before it starts getting appreciably noisy? Does it have to >>be above the 5-6 volt breakdown voltage, or will significant noise occur >>before that point? I'm looking for a noise source that will work with a >>3 volt supply. Maybe a different transistor with a lower breakdown voltage? > > > You are snookered by the physics. Reverse-biased diodes break down by > two different mechanisms depending on the voltage applied acorss the > junction, the Zenner (quantum tunneling) mechanism at alow voltages - > less than 5V - and the avalanche (impact ionisation) mechanism at > higher voltages. > > http://ecee.colorado.edu/~bart/book/book/chapter4/ch4_5.htm > > Avalanche breakdown gives a noisier leakage current than quantum > tunnelling - with avalanche breakdown there is always the chance that > an electron will make it all the way across the gap without generating > a second electron-hole pair, killing the leakage current until a > cosmic ray or a radioative atom in the vicinity generates a new charge > carrier pair to restart the avalanche process. Quantum mechanical > tunnelling is also a random process, but it is a single random > process, without the second stage of avalanche multipication to > generate lots of outliers. > > With a 3V supply rail you are stuck with Zener breakdown. You may have > to settle for a pseudo-random binary sequence generator, or amplifying > the Johnson noise from a resistor - 1nV per root hertz for a 50R > resistor at room temerature, rising as the square root of the > resistance, bearing in mind that "low noise" integrated circuit op > amps typically also generate 1nV per root hertz white noise at room > temperature. > > -- > Bill Sloman, Nijmegen >
Oh really, I've also been led to believe and educated, years ago, that zener mode and impact mode is a balance at 5 volts and anything under that goes into "impact" mode and above is zener. Looking at the behavior of the commonly known 5.1 Vref diode where a diode is in series with the zener stabilizes the effects between the two. I think you may want to do a better job of memorizing that document you red before posting your proclaimed expertise drivel. If it were me, I would've suggested to look at a "NOISE DIODE" and not throw a crap load of physics that you most likely barely understand. Things like SHot and Johnson Noise could also be a good help as references to loop up.. Now, I've seen some tricks done using a tunnel diodes in my day.. Jamie.
Bill Sloman wrote:
> On Nov 26, 9:26 am, Bitrex wrote:
[...]
> With a 3V supply rail you are stuck with Zener breakdown. You may have > to settle for a pseudo-random binary sequence generator, or amplifying > the Johnson noise from a resistor - 1nV per root hertz for a 50R > resistor at room temerature, rising as the square root of the > resistance, bearing in mind that "low noise" integrated circuit op > amps typically also generate 1nV per root hertz white noise at room > temperature. >
Dudes, dudes ... what is so difficult about making a higher voltage from 3V? -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
On Fri, 26 Nov 2010 03:26:21 -0500, Bitrex
 wrote:

>I suppose I could set things up to do the research myself, but since I'm >feeling lazy I thought I'd check first: By how much do you have to >reverse bias the emitter-base junction of a common transistor such as >the 2N2222 before it starts getting appreciably noisy? Does it have to >be above the 5-6 volt breakdown voltage, or will significant noise occur >before that point? I'm looking for a noise source that will work with a >3 volt supply. Maybe a different transistor with a lower breakdown voltage?
You win't find a transistor that zeners at 3 volts. Well, maybe an exotic microwave one. Do you need high-quality, flat, Gaussian noise? How about doing it digitally, with a pseudo-random shift register? John
On Nov 26, 2:30=A0pm, Jamie
 wrote:
> Bill Sloman wrote: > > On Nov 26, 9:26 am, Bitrex wrote: > > >>I suppose I could set things up to do the research myself, but since I'=
m
> >>feeling lazy I thought I'd check first: By how much do you have to > >>reverse bias the emitter-base junction of a common transistor such as > >>the 2N2222 before it starts getting appreciably noisy? Does it have to > >>be above the 5-6 volt breakdown voltage, or will significant noise occu=
r
> >>before that point? I'm looking for a noise source that will work with a > >>3 volt supply. =A0Maybe a different transistor with a lower breakdown v=
oltage?
> > > You are snookered by the physics. Reverse-biased diodes break down by > > two different mechanisms depending on the voltage applied acorss the > > junction, the Zenner (quantum tunneling) mechanism at alow voltages - > > less than =A05V - and the avalanche (impact ionisation) mechanism at > > higher voltages. > > >http://ecee.colorado.edu/~bart/book/book/chapter4/ch4_5.htm > > > Avalanche breakdown gives a noisier leakage current than quantum > > tunnelling - with avalanche breakdown there is always the chance that > > an electron will make it all the way across the gap without generating > > a second electron-hole pair, killing the leakage current until a > > cosmic ray or a radioative atom in the vicinity generates a new charge > > carrier pair to restart the avalanche process. Quantum mechanical > > tunnelling is also a random process, but it is a single random > > process, without the second stage of avalanche multipication to > > generate lots of outliers. > > > With a 3V supply rail you are stuck with Zener breakdown. You may have > > to settle for a pseudo-random binary sequence generator, or amplifying > > the Johnson noise from a resistor - 1nV per root hertz for a 50R > > resistor at room temerature, rising as the square root of the > > resistance, bearing in mind that "low noise" integrated circuit op > > amps typically also generate 1nV per root hertz white noise at room > > temperature. > > Oh really, I've also been led to believe and educated, years ago, that > zener mode and impact mode is a balance at 5 volts and anything under > that goes into "impact" mode and above is zener.
Then you had better go and get yourself a better education.
> Looking at the behavior of the commonly known 5.1 Vref diode where a diod=
e is in series with > the zener stabilizes the effects between the two. You would be thinking of the 1N821 through 1N829 6.2V voltage reference diodes, which did include a a forward diode to compensate for the temperature coefficient of a 5.6V "zener" diode. http://www.datasheetcatalog.org/datasheet/motorola/1N823.pdf http://206.209.106.106/datasheets/Zeners/AppNotes.pdf The breakdown voltage you get with a pure zener (quantum tunnelling) mechanism decreases with increasing temperature, while an avalanche breakdown voltage increases with increasing temperature. At 5.6V the balance between the two mechanisms favours the avalanche route enough that the voltage across the reverse biassed diode increases by the same 2mV/K that the voltage across the forward diode decreases. http://en.wikipedia.org/wiki/Zener_diode
> =A0 I think you may want to do a better job of memorizing that document > you red before posting your proclaimed expertise drivel.
Since you managed to confuse the the 5.1V "zener" diode, which has a roughly zero temperature coefficient, with the 6.2V reference diodes which do include a forward diode, your own advice can be seen as less than reliable.
> =A0 =A0If it were me, I would've suggested to look at a "NOISE DIODE" and=
not
> =A0 throw a crap load of physics that you most likely barely understand.
Google doesn't throw up much on noise diodes. The nearest thing to something useful came from here http://www.electronicspoint.com/noise-diode-t23382.html
> Things like Shot and Johnson Noise could also be a good help as > references to look up..
I did refer to Johnson noise. Google on that and you get to http://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise which does include a reference to shot noise. http://en.wikipedia.org/wiki/Shot_noise Neither is going to be of much help to the OP, who wants a solution to a problem - not the education that you obviously failed to absorb.
> =A0 =A0Now, I've seen some tricks done using a tunnel diodes in my day..
The real trick with tunnel diodes today is finding where you can buy one. The difficulty with tunnel diodes is that they are broadband devices. If you don't mount them in a properly designed transmission line environment, they will oscillate at a frequency your oscilliscope can't follow, and the voltage levels that you will see - at frequencies that your oscilliscope can follow - won't look anything like what you wanted and expected. -- Bill Sloman, Nijmegen
In article ,
 Bitrex  wrote:

> I suppose I could set things up to do the research myself, but since I'm > feeling lazy I thought I'd check first: By how much do you have to > reverse bias the emitter-base junction of a common transistor such as > the 2N2222 before it starts getting appreciably noisy? Does it have to > be above the 5-6 volt breakdown voltage, or will significant noise occur > before that point? I'm looking for a noise source that will work with a > 3 volt supply. Maybe a different transistor with a lower breakdown voltage?
Are MOVs noisy? Luxeon Rebel LEDs flicker in the 2V/microamp range, which I'm hoping is their MOV rather than defects in the chip slowly burning away. -- I will not see posts or email from Google because I must filter them as spam
John Larkin wrote:
> On Fri, 26 Nov 2010 03:26:21 -0500, Bitrex > wrote: > >> I suppose I could set things up to do the research myself, but since I'm >> feeling lazy I thought I'd check first: By how much do you have to >> reverse bias the emitter-base junction of a common transistor such as >> the 2N2222 before it starts getting appreciably noisy? Does it have to >> be above the 5-6 volt breakdown voltage, or will significant noise occur >> before that point? I'm looking for a noise source that will work with a >> 3 volt supply. Maybe a different transistor with a lower breakdown voltage? > > You win't find a transistor that zeners at 3 volts. Well, maybe an > exotic microwave one. > > Do you need high-quality, flat, Gaussian noise?
What on earth is flat Gaussian noise? A flat, ie. even distribution is not a Gaussian, so perhaps I'm misunderstanding this statement. Please enlighten... BTW, the new Agilent 3352x arb. generators have a nice feature that lets you adjust the bandwidth of the noise, which is Gaussian. They also have a LFSR noise generator, but it's not analog but rather on/off. I guess if it were to be analog, it would have to let you select some number of bits to gather for each output. This would give a flat analog noise. It would have been nice if you could select the distribution of the noise. Perhaps I'll send these ideas on to Agilent. They have been very nice to me over the years, taking many of my ideas directly to the scope and generator prod. devel. managers. It can also use the noise to modulate a PWM signal, so coupled with a little flyback pulse generator, can be used to simulate a lot of jittery physical phenomena, like laser pulses. -- _____________________ Mr.CRC crobcBOGUS@REMOVETHISsbcglobal.net SuSE 10.3 Linux 2.6.22.17
"Bitrex"  wrote in message 
news:XcCdnet8IpWVNHLRnZ2dnUVZ_rOdnZ2d@earthlink.com...
> On 11/26/2010 6:26 AM, Bill Sloman wrote: >> On Nov 26, 9:26 am, Bitrex wrote: >>> I suppose I could set things up to do the research myself, but since I'm >>> feeling lazy I thought I'd check first: By how much do you have to >>> reverse bias the emitter-base junction of a common transistor such as >>> the 2N2222 before it starts getting appreciably noisy? Does it have to >>> be above the 5-6 volt breakdown voltage, or will significant noise occur >>> before that point? I'm looking for a noise source that will work with a >>> 3 volt supply. Maybe a different transistor with a lower breakdown >>> voltage? >> >> You are snookered by the physics. Reverse-biased diodes break down by >> two different mechanisms depending on the voltage applied acorss the >> junction, the Zenner (quantum tunneling) mechanism at alow voltages - >> less than 5V - and the avalanche (impact ionisation) mechanism at >> higher voltages. >> >> http://ecee.colorado.edu/~bart/book/book/chapter4/ch4_5.htm >> >> Avalanche breakdown gives a noisier leakage current than quantum >> tunnelling - with avalanche breakdown there is always the chance that >> an electron will make it all the way across the gap without generating >> a second electron-hole pair, killing the leakage current until a >> cosmic ray or a radioative atom in the vicinity generates a new charge >> carrier pair to restart the avalanche process. Quantum mechanical >> tunnelling is also a random process, but it is a single random >> process, without the second stage of avalanche multipication to >> generate lots of outliers. >> >> With a 3V supply rail you are stuck with Zener breakdown. You may have >> to settle for a pseudo-random binary sequence generator, or amplifying >> the Johnson noise from a resistor - 1nV per root hertz for a 50R >> resistor at room temerature, rising as the square root of the >> resistance, bearing in mind that "low noise" integrated circuit op >> amps typically also generate 1nV per root hertz white noise at room >> temperature. >> >> -- >> Bill Sloman, Nijmegen >> > > Thanks for your reply and for the enlightenment regarding the difference > between breakdown mechanisms. Instead of trying to amplify Johnson noise, > perhaps I can use a spare opamp section of the project as a hysteretic > oscillator and make a charge pump to kick the voltage up above the 5 volt > threshold? I wonder if a low voltage op amp like the LM324 acting as a > charge pump would be able to get the voltage high enough to initiate the > avalanche breakdown mode - I'm not sure what the '324's output can swing > at such low voltages, but I bet getting 5 volts with a 3 volt supply with > such a circuit might be pushing it. >
It'd probably be easier to get a single transistor blocking oscillator to work, the charge pump gives just under Vccx2, if you need more you have to cascade doubler stages. For a one-off you could liberate the erase oscillator from a scrap cassette deck - the inductor is usually styled like an IFT, so nice neat finished job.
John Larkin wrote:
> On Sat, 27 Nov 2010 18:11:00 -0500, Jamie > wrote: > >> John Larkin wrote: >> >>> On Sat, 27 Nov 2010 17:09:19 -0500, Jamie >>> wrote: >>> >>> >>>> John Larkin wrote: >>>> >>>> >>>>> On Sat, 27 Nov 2010 09:38:58 -0800 (PST), Bill Sloman >>>>> wrote: >>>>> >>>>> >>>>> >>>>>> On Nov 27, 5:08 am, John Larkin >>>>>> wrote: >>>>>> >>>>>> >>>>>>> On Fri, 26 Nov 2010 03:26:11 -0800 (PST),BillSloman >>>>>>> >>>>>>> >>>>>>> >>>>>>> wrote: >>>>>>> >>>>>>> >>>>>>>> On Nov 26, 9:26 am, Bitrex wrote: >>>>>>>> >>>>>>>> >>>>>>>>> I suppose I could set things up to do the research myself, but since I'm >>>>>>>>> feeling lazy I thought I'd check first: By how much do you have to >>>>>>>>> reverse bias the emitter-base junction of a common transistor such as >>>>>>>>> the 2N2222 before it starts getting appreciably noisy? Does it have to >>>>>>>>> be above the 5-6 volt breakdown voltage, or will significant noise occur >>>>>>>>> before that point? I'm looking for a noise source that will work with a >>>>>>>>> 3 volt supply. Maybe a different transistor with a lower breakdown voltage? >>>>>>> >>>>>>>> You are snookered by the physics. Reverse-biased diodes break down by >>>>>>>> two different mechanisms depending on the voltage applied acorss the >>>>>>>> junction, the Zenner (quantum tunneling) mechanism at alow voltages - >>>>>>>> less than 5V - and the avalanche (impact ionisation) mechanism at >>>>>>>> higher voltages. >>>>>>> >>>>>>>> http://ecee.colorado.edu/~bart/book/book/chapter4/ch4_5.htm >>>>>>> >>>>>>>> Avalanche breakdown gives a noisier leakage current than quantum >>>>>>>> tunnelling - with avalanche breakdown there is always the chance that >>>>>>>> an electron will make it all the way across the gap without generating >>>>>>>> a second electron-hole pair, killing the leakage current until a >>>>>>>> cosmic ray or a radioative atom in the vicinity generates a new charge >>>>>>>> carrier pair to restart the avalanche process. Quantum mechanical >>>>>>>> tunnelling is also a random process, but it is a single random >>>>>>>> process, without the second stage of avalanche multipication to >>>>>>>> generate lots of outliers. >>>>>>> >>>>>>>> With a 3V supply rail you are stuck with Zener breakdown. You may have >>>>>>>> to settle for a pseudo-random binary sequence generator, or amplifying >>>>>>>> the Johnson noise from a resistor - 1nV per root hertz for a 50R >>>>>>>> resistor at room temerature, rising as the square root of the >>>>>>>> resistance, bearing in mind that "low noise" integrated circuit op >>>>>>>> amps typically also generate 1nV per root hertz white noise at room >>>>>>>> temperature. >>>>>>> >>>>>>> Certainly not white noise. Nearly all opamps have a 1/f noise corner. >>>>>> >>>>>> For bipolar op amps the noise corner is close to 10Hz. >>>>> >>>>> >>>>> Most are higher than that, if it's specified at all. And opamps can >>>>> have very non-Gaussian behavior, like temperature driven offsets, >>>>> popcorn noise, ambient RF rectification, PSRR issues. An opamp isn't a >>>>> very good noise source. A zener can make several hundred nV/rthz of >>>>> pretty high quality noise. A PN shift register is almost perfect. >>>>> >>>>> John >>>>> >>>> >>>> what constitutes high quality noise? I've never heard it put that way! :) >>> >>> >>> Good noise approximates a physical, random process. Different defects >>> matter to different people. The RF boys want spectral flatness. Crypto >>> people want zero predictability. Process people might want a nearly >>> Gaussian PD. Audio folks don't want audible defects, like hum or >>> popcorn noise or detected RF. >>> >>> >>>> Noise is noise, be it white, pink, brown and your analogy "Popcorn", >>>> first time I've heard that one ect ! >>> >>> >>> Look it up. >>> >>> John >>> >> That's interesting, I did look that up, and it appears to be another >> name of something I was already aware of.. I read a parcel on >> Wki and it mention behavior of defects found in transistors. >> >> I did have a batch of smt 2222's that would not work in linear >> applications, only switching because at a specific current (Ice) No >> mater the configuration, I was getting a small noise that could be seen >> via scope in the collector circuit. You go below or above this current >> point and it would go away.. When I set the scope up to expand the view >> on that section, I go what looked like a random oscillation. >> >> I used a spectrum analyzer on that point and got random noise that >> started around 100hz up to about 1 mhz.. It would change just by varying >> the bias slightly. >> >> I have since replaced that batch and all is well once again. >> >> MMBT2222 iirc... >> >> > > Popcorn noise is super low frequency and is usually seen in the form > of discrete/bimodal level shifts, like the output of a random pulse > generator. It's common for it to look like pulses, milliseconds in > duration, milliseconds to seconds apart, or random in time level > shifts. Sometimes a part will sit quietly for hours and then let out a > burst. The magnitude is in the ballpark of 10s of microvolts RTI. The > cause is supposedly ionic contamination in oxide layers or something. > It's named because it sounds like corn popping. > > Analog Devices has an appnote where they claim it's a thing of the > past. They should test some of their own DACs. > > John >
A lot of 1/f noise is actually popcorn in character. With peace to AD, it's getting worse rather than better as devices shrink--it's mostly local conductance fluctuations. Electromigration is another source of popcorn noise. 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
John Larkin wrote:
> On Sat, 27 Nov 2010 09:38:58 -0800 (PST), Bill Sloman > wrote: > >> On Nov 27, 5:08 am, John Larkin >> wrote: >>> On Fri, 26 Nov 2010 03:26:11 -0800 (PST),BillSloman >>> >>> >>> >>> wrote: >>>> On Nov 26, 9:26 am, Bitrex wrote: >>>>> I suppose I could set things up to do the research myself, but since I'm >>>>> feeling lazy I thought I'd check first: By how much do you have to >>>>> reverse bias the emitter-base junction of a common transistor such as >>>>> the 2N2222 before it starts getting appreciably noisy? Does it have to >>>>> be above the 5-6 volt breakdown voltage, or will significant noise occur >>>>> before that point? I'm looking for a noise source that will work with a >>>>> 3 volt supply. Maybe a different transistor with a lower breakdown voltage? >>> >>>> You are snookered by the physics. Reverse-biased diodes break down by >>>> two different mechanisms depending on the voltage applied acorss the >>>> junction, the Zenner (quantum tunneling) mechanism at alow voltages - >>>> less than 5V - and the avalanche (impact ionisation) mechanism at >>>> higher voltages. >>> >>>> http://ecee.colorado.edu/~bart/book/book/chapter4/ch4_5.htm >>> >>>> Avalanche breakdown gives a noisier leakage current than quantum >>>> tunnelling - with avalanche breakdown there is always the chance that >>>> an electron will make it all the way across the gap without generating >>>> a second electron-hole pair, killing the leakage current until a >>>> cosmic ray or a radioative atom in the vicinity generates a new charge >>>> carrier pair to restart the avalanche process. Quantum mechanical >>>> tunnelling is also a random process, but it is a single random >>>> process, without the second stage of avalanche multipication to >>>> generate lots of outliers. >>> >>>> With a 3V supply rail you are stuck with Zener breakdown. You may have >>>> to settle for a pseudo-random binary sequence generator, or amplifying >>>> the Johnson noise from a resistor - 1nV per root hertz for a 50R >>>> resistor at room temerature, rising as the square root of the >>>> resistance, bearing in mind that "low noise" integrated circuit op >>>> amps typically also generate 1nV per root hertz white noise at room >>>> temperature. >>> >>> Certainly not white noise. Nearly all opamps have a 1/f noise corner. >> >> For bipolar op amps the noise corner is close to 10Hz. > > Most are higher than that, if it's specified at all. And opamps can > have very non-Gaussian behavior, like temperature driven offsets, > popcorn noise, ambient RF rectification, PSRR issues. An opamp isn't a > very good noise source. A zener can make several hundred nV/rthz of > pretty high quality noise. A PN shift register is almost perfect. > > John >
I'm a big fan of shot noise for this. By my actual measurement it's Gaussian out beyond 7.1 sigma, corresponding to a threshold crossing rate of 10**-11 times the bandwidth. Another key virtue is that the noise PSD is derivable from the DC photocurrent by a trivial first-principles calculation. You do need a decent amplifier, and enough photocurrent to drop a volt or two across your (metal film) sense resistor, but it's easy to get way within 1 dB accuracy that way. Best of all, at least for electrooptics folk like yours truly, the frequency response is *exactly* that of your front end. 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
On Fri, 26 Nov 2010 15:23:51 -0800, Rich Grise wrote:

> What ever happened to Watson A Name's micropower switchers for LEDs?
Whatever happened to Watson A. Name? -- "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." (Richard Feynman)
On Nov 28, 3:55=A0pm, Winfield Hill  wrote:
> On Nov 28, 2:26=A0pm, whit3rd wrote: > > > On Nov 28, 8:28=A0am, Bill Sloman wrote: > > > > =A0Avalanche breakdown is noisier than quantum mechanical > > > tunnelling (the Zener mechanism). With avalanche breakdown, you have > > > not only the shot noise from the separate charge carriers making thei=
r
> > > independent way through the device but you also have the statistical > > > uncertainty in the avalanche multiplication > > > Not the way I see it; the advantage is that avalanche IS GAIN > > applied to the shot noise. =A0That gain makes the signal big enough > > to dominate any additions in later amplifiers. =A0 Without that gain > > in the avalanche process, your later amplifiers are equal > > and hard-to-characterize additional sources of noise. > > =A0Simply imagining shot noise with gain can't and doesn't work. > =A0Consider a current of 10uA, that's 6.25M electrons per 10ns, > =A0and the random shot-noise variation would be 2500 in 10ns, or > =A00.04% noise. > > =A0Now imagine that each electron is multiplied by 100. =A0Oops, that > =A0can't be right, because now we'd have 625M electrons per 10ns. > =A0OK, so scale back to 62500 electrons per 10ns, and let each of > =A0those turn into 100 by avalanche gain. =A0Now the random "shot" > =A0noise would be 250 "events" per 10ns, or 0.4%. > > =A0But in reality the discrepancy is worse than that, much worse. > =A0We've learned that in avalanche, each starting electron actually > =A0starts a "microplasma" channel that conducts a channel current > =A0e.g., 20uA, lasting say 3 to 50ns, and ending abruptly, in under > =A0a ns, as the small nearby capacitance gets discharged and the > =A0local voltage no longer supports avalanche. > > =A0Some plasma channel electron-atom collisions emit photons, > =A0some of which create photoelectrons starting more channels > =A0nearby. =A0It's all highly chaotic. > > =A0So we end up with an intermittent noisy "oscillation" of current > =A0pulses of quasi-random length. =A0When you look at the waveform > =A0at longer timescales you see the superposition of thousands of > =A0events, which simply looks noisy. =A0But with the right lashup and > =A0equipment you can view individual channels starting and stopping, > =A0sometimes on top of each other. > > =A0This topic, with ASCII waveforms, bench circuit-setup details, > =A0theory and literature references, was extensively covered here > =A0on s.e.d. many years ago. =A0You can read all about it.
Yes! The old discussion was very nice. I liked the article by K.G. McKay "Avalanche Breakdown in Silicon" Phys. Rev. (94) pg 877. (May 15, 1954) It would be fun to see these different channels start to conduct as the voltage is raised. There are some temperature effects which also sound interesting. (Wouldn't it be nice if these old articles were now part of the public domain?) George H.
On Nov 28, 2:26=A0pm, whit3rd  wrote:
> On Nov 28, 8:28=A0am, Bill Sloman wrote: > > > =A0Avalanche breakdown is noisier than quantum mechanical > > tunnelling (the Zener mechanism). With avalanche breakdown, you have > > not only the shot noise from the separate charge carriers making their > > independent way through the device but you also have the statistical > > uncertainty in the avalanche multiplication > > Not the way I see it; the advantage is that avalanche IS GAIN > applied to the shot noise. =A0That gain makes the signal big enough > to dominate any additions in later amplifiers. =A0 Without that gain > in the avalanche process, your later amplifiers are equal > and hard-to-characterize additional sources of noise.
Simply imagining shot noise with gain can't and doesn't work. Consider a current of 10uA, that's 6.25M electrons per 10ns, and the random shot-noise variation would be 2500 in 10ns, or 0.04% noise. Now imagine that each electron is multiplied by 100. Oops, that can't be right, because now we'd have 625M electrons per 10ns. OK, so scale back to 62500 electrons per 10ns, and let each of those turn into 100 by avalanche gain. Now the random "shot" noise would be 250 "events" per 10ns, or 0.4%. But in reality the discrepancy is worse than that, much worse. We've learned that in avalanche, each starting electron actually starts a "microplasma" channel that conducts a channel current e.g., 20uA, lasting say 3 to 50ns, and ending abruptly, in under a ns, as the small nearby capacitance gets discharged and the local voltage no longer supports avalanche. Some plasma channel electron-atom collisions emit photons, some of which create photoelectrons starting more channels nearby. It's all highly chaotic. So we end up with an intermittent noisy "oscillation" of current pulses of quasi-random length. When you look at the waveform at longer timescales you see the superposition of thousands of events, which simply looks noisy. But with the right lashup and equipment you can view individual channels starting and stopping, sometimes on top of each other. This topic, with ASCII waveforms, bench circuit-setup details, theory and literature references, was extensively covered here on s.e.d. many years ago. You can read all about it.
On Nov 28, 8:26=A0pm, whit3rd  wrote:
> On Nov 28, 8:28=A0am, Bill Sloman wrote: > > > =A0Avalanche breakdown is noisier than quantum mechanical > > tunnelling (the Zener mechanism). With avalanche breakdown, you have > > not only the shot noise from the separate charge carriers making their > > independent way through the device but you also have the statistical > > uncertainty in the avalanche multiplication > > Not the way I see it; the advantage is that avalanche IS GAIN > applied to the shot noise. =A0
Not just gain, but statistically variable gain.
> That gain makes the signal big enough > to dominate any additions in later amplifiers. =A0 Without that gain > in the avalanche process, your later amplifiers are equal > and hard-to-characterize additional sources of noise.
If you've got shot noise on a current, you need to have it develop about 50mV across your load resistor (at room temperature) to equal the Johnson noise in the load resistor. As you increase the resistance further, the voltage noise generated from the shot noise rises in direct proportion to the resistance, while the Johnson noise rises as the square root of the resistance. It not that difficult to organise your circuit so that the shot noise is dominant, and also swamps the Johnson noise in the input stage of you first amplifier. -- Bill Sloman, Nijmegen
On Nov 28, 8:28=A0am, Bill Sloman  wrote:
> Avalanche breakdown is noisier than quantum mechanical > tunnelling (the Zener mechanism). With avalanche breakdown, you have > not only the shot noise from the separate charge carriers making their > independent way through the device but you also have the statistical > uncertainty in the avalanche multiplication
Not the way I see it; the advantage is that avalanche IS GAIN applied to the shot noise. That gain makes the signal big enough to dominate any additions in later amplifiers. Without that gain in the avalanche process, your later amplifiers are equal and hard-to-characterize additional sources of noise.
On Nov 28, 6:19=A0pm, Jan Panteltje  wrote:
> On a sunny day (Sun, 28 Nov 2010 08:21:05 -0800 (PST)) it happenedBillSlo=
man
> wrote in > <44952edc-e552-42b8-a17d-38da60548...@n10g2000yqd.googlegroups.com>: > > >> I bought 20 BC557B transistors for 1 Euro last week, that is retail. > >> Want the address? > > >Sure. E-mail it to me. The BC557 - ft 100MHz - isn't a broad-band > >transistor, but the right sort of hobby-supplier - one with ham-radio- > >oriented customers - should have the BFR92 and BFT92 in stock. > > Check your email.
Done. Thanks. Your first supplier - your source of cheap BC577's - doesn't have any broadband transistors. The second has BFR91 and BFR96 parts in the old pill-box package, which is easier to work with than the SMD-packaged BFR92 or BFT92. Tomorrow I'll see what I can get them to send me. -- Bill Sloman, Nijmegen
On a sunny day (Sun, 28 Nov 2010 08:21:05 -0800 (PST)) it happened Bill Sloman
 wrote in
<44952edc-e552-42b8-a17d-38da60548729@n10g2000yqd.googlegroups.com>:

>> I bought 20 BC557B transistors for 1 Euro last week, that is retail. >> Want the address? > >Sure. E-mail it to me. The BC557 - ft 100MHz - isn't a broad-band >transistor, but the right sort of hobby-supplier - one with ham-radio- >oriented customers - should have the BFR92 and BFT92 in stock.
Check your email.
On Nov 26, 3:26=A0am, Bitrex  wrote:
> I suppose I could set things up to do the research myself, but since I'm > feeling lazy I thought I'd check first: By how much do you have to > reverse bias the emitter-base junction of a common transistor such as > the 2N2222 before it starts getting appreciably noisy? Does it have to > be above the 5-6 volt breakdown voltage, or will significant noise occur > before that point? I'm looking for a noise source that will work with a > 3 volt supply. =A0Maybe a different transistor with a lower breakdown vol=
tage? You can amplifiy the shot noise of a forward biased diode. (Or light on a photodiode) George H.