I was wondering if anyone could help with the selection of the transistor Q1 in the noise generator circuit below?
It's an NPN, used with the EB junction reverse biased to breakdown. Noise is generated by this breakdown, and amplified by Q2.
I'm interested in knowing what transistor parameters matter here. I guess I don't care about the Beta degradation, as it will never be used to amplify. The goal is to have a reliable noise output over a (say) 10 year product life. Alternatively, I'm interested in knowing if there are any particular types of transistor to avoid for this application.
What is the frequency range of use, how flat should it be, how much degradation can you accept?
Regarding the reversed biased transistor, you might consider a zener diode. When you use a zener diode with voltage >> 6V, avalanche breakdown dominates (which is good). It also helps to reduce aging by switcbing it of when not in use.
Probably you have to do some filtering to flatten the spectrum because of 1/f noise component in zener diode current.
We had an instance of a similar circuit in an old product return from the field (after several years of service) because it was consistently failing a statistical test on the output noise.
When checked, the DC bias seemed ok, and the output still looked "noisy" on an oscilloscope.
I was wondering whether this could be due to some design flaw, or whether it was just a random failure.
For some years now, I've been interested in making a "tracking generator" (broadband noise source) so that I could sweep filters using a spectrum analyzer instead of the sweep generators that I have been using for a goodly number of years.
Do you have any clue what the slope of a 10 volt zener is from, say, a Megahertz up to UHF? Does current change alter the slope, and if so, do you know how?
My thought is to get noise with the zener and then a few MiniCircuits mic amps to get some decent voltage/power out. My original thought was to go with some of those noise diodes from Noise Inc. that they were selling as seconds that didn't quite meet spec, but they killed that program a few years ago.
Thoughts???
Jim
--
"If you think you can, or think you can\'t, you\'re right."
--Henry Ford
"John Larkin" wrote in message
news:a74tf39it6nkf0fte5ffc6tg8gorjtf4k6@4ax.com...
> A 10-volt zener, running at 1 mA, ain\'t bad. Figure about 300 nV per
> root Hertz noise, mostly symmetric.
>
> John
>
I never design off of "samples". In general the company makes samples from the 3 sigma high end of the distribution curve.
Some of my filters have a 10 kHz bandwidth and that would entail a SRD fudamental frequency of 100 Hz or so to see the fine detail. Just FYI, SRD multipliers have a power fall of 1/n, where n is the order of multipication. That ain't "flat" by my definition. And it goes to hell in a handbasket for frequencies above 10 MHz or so. That dog won't hunt.
Yeah, and I've tried in here to find a tracking generator for the HP 140T and the HP 8555A that I just acquired and I've gotten a couple of "yeah, I've got one around here if I can find it" tracking generators to 1.2 GHz, which is just fine with me, but nobody has "found one" yet, after six months.
Can I booger one with a mixer and all that jazz, of course, but I'm six months behind on ten projects and trying to swap money for time.
The old 2N2222 or the newer SMD BFW92A seems to be the favourite in some articles I have. I have not yet built one, but it's on my to-do list quite soon.
Gee, John. Didn't mean to pi$$ you off. No offense intended. I find, in general, that if I need something that I can't find quickly on the market at a decent price that I'll probably wind up producing it as a product at some point in time.
Well, I must admit that I've only spent five years doing nothing but SRD multipliers right out of school in the '60s. HP taught us that we could incrementally beat the equation 1/n for spot frequencies if we were willing to do a little of the old echo it off the cavity wall, but the best I was ever able to do was 1.5/n in a relatively broadband multiplier at 3 Gig.
In an untuned comb generator with a limiter, anything is possible; I'm not sure how herotek is getting their flatness.
I am one with Joerg who is probably the biggest tightwad on this ng. He's the best at making the eagle scream I've seen in a while and I am following suit. I'll continue to look for the mating unit to my analyzer. The reason I wanted a broadband noise generator is that it would work both with the HP and with the old Nelson-Ross that has been my standard for thirty years or so. Say what you like about "modern" equipment, some of that old RF stuff (especially Rhode & Schwarz) is still the standard of excellence.
An SRD comb generator converts the input (usually a sine wave) into a train of very narrow inpulses. The Fourier transform of an ideal impulse train is a series of uniform-amplitude spectral lines from the
0th harmonic (DC) to forever, spaced at the pulsing frequency. That's what an SRD comb generator does. Since the impulses aren't perfect, the frequency spikes fall off at high frequencies, in a roughly sinx/x envelope or something like that.
At low frequencies, program a pulse generator to make, say, a 1-volt,
10 microsecond wide pulse at 100 Hz. That will create a frequency comb with a line every 100 Hz and a sinx/x envelope with the first null at
100 KHZ, which will be plenty flat to 10 KHz or so. Use that to drive your filter. The spectrum analyzer will be a lot happier than if you used noise, and you'll get clean plots fast.
THen I humbly admit my lack of skill at the Fourier process. My notes indicate that a Fourier transform of an ideal impulse train has a rolloff as a function of frequency. I'll have to go back and see what I missed that day of class forty-five years ago. And remember, I'm a physics that got converted to engineer after I discovered that the engineering school was still dicking around with tubes and the physics department had the only classes in solid state devices. It wasn't until I went back into the EE graduate program ten years later that the engineering school had begrudgingly accepted the transistor as relevant.
Funny, looking back...I had quite a few graduate units under my belt and was thinking seriously about finishing up the thesis when my advisor, and an RF Circuits graduate prof, started talking about microstrip. Understand, with my SRD and other microwave work, I'd been working with microstrip for nearly ten years. First day, he walks in and starts the class off with, "Presume you have an alumina substrate a quarter of an inch thick..."
"But Professor Pucksuckle, the thickest alumina you can buy is 50 mils and most of us are working with ten to fifteen mil alumina."
"It doesn't matter. This is MY class."
"But you couldn't afford alumina of that thickness, and even so, the flatness would be horrendous. My God, not even Wheeler or Stinehelfer would consider that thickness to be reasonable."
"Are you questioning my design capability, Mr. Weir?"
"Yes sir, I am. Even with G-10 board at 62 mils we have problems with the equations not working out exactly for the even and odd modes. How you intend to work with it with a dielectric factor more than triple is beyond me."
"Mr. Weir, you and I need to meet after class. I will not have my class interrupted by you and your fallacious information."
That was my last day in grad school, and I don't regret it in the least. After all, was my company going to pay me another dime if I had my Masters? Hell no, it was MY company, and is to this day.
I'm not interested in anything below 5 MHz. or above 500 MHz.. That's my usual domain after I left the microwave world for my own company.
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.