essage
sage
ps.com...
gain
is to
*...*G_2it and
e adds and
stages
use as
it.
to have
the signal
ps
)Yeah, I think subtracting them is right. The circuit looked like this,
V(+)--RRR--+--K|--+--K|--+--RRR--V(-) | | | C GND C C C | | +--RR--+--RR--+ | OUT
where K| are zeners
George H.
essage
sage
ps.com...
gain
is to
*...*G_2it and
e adds and
stages
use as
it.
to have
the signal
ps
)"Yup, I've noticed the small asymmetry. It gets progressively worse at
Ahh, Thanks John I think that explains it. I was trying to maximize the noise which happened with high voltage zeners at low bias current (10-20 uA or so) If I ever build one again I'll try higher bias currents. Of course if I ever build it again I would change some other things too. Are you ever embarrassed by the circuits you designed 5-10 years ago? (OK reset that to the beginning of your circuit building career which for me started (in earnest) 9 years ago.)
George H.
message
G_1*...*G_2
and
adds and
stages
as
have
signal
Well, I've been designing circuits since I gave up chemistry in about
1956 or so. Sure, I've done some terrible stuff, partly out of ignorance and partly because parts used to be a lot worse and a lot more expensive than they are now.This is sort of a golden age now, with so many cool parts that can implement all sorts of signal processing and algorithms that were only theoretical not too long ago, or at least theoretical unless you had a military-scale budget and a supercomputer.
John
message
message
...
gain
is to
G_1*...*G_2
handle it and
stage adds and
Cascaded stages
should use as
downstream
it.
advisable to have
the signal
stage
amps
level)
While the noise output is highest in the knee, the asymmetry is also highest there as well. Subtracting two of them is a great idea to reduce the asymmetry some.
.
nd
few
I think the other poster might want to brush up on input referred noise. If you do it right, the first stage dominates.
George Herold Inscribed thus:
Just curious ! Have you tried an LED as a noise source ?
--
Best Regards:
Baron.
Or the classic noise generator used in W.W.II. A photo tube & lamp. Different lamps had different characteristics, but some were capable of wiping out the entire usable RF spectrum from LW to VHF and was used in airborne jamming.
Fair Radio still had new surplus units, well into the early '70s.
-- You can\'t have a sense of humor, if you have no sense!
message
ssage
ups.com...
gain
is to
1*...*G_2e it and
ge adds and
d stages
d use as
mit.
e to have
the signal
emps
l)
t eJosephKK wrote,
" While the noise output is highest in the knee"
Joseph, do you have any idea why this is? I would naively expect that the noise would only depend on the =91avalanche factor=92 (how many more electrons are 'released' for each electron emitted into the junction.) And I=92d expect that to be mostly constant for a given zener. So there must be something else going on.
George H.
n message
message
roups.com...
le gain
ed is to
G_1*...*G_2
dle it and
tage adds and
ded stages
uld use as
eam
se it.
ble to have
Wes the signal
age
al
amps
vel)
at
ew
est
are
It's mostly my past ignorance that bothers me. (And of course I'm still ignorant of so much...) Just looking back on the zener noise circuit. Indeed I get about 300nV/rtHz of noise from the zener. (A number I think you quoted earlier.) The 'mistake' I made in the circit was using an opamp with not enough slew rate. At the highest gain level the noise signal turns into a triangle wave...Sigh. Of course I don't think anyone but me has noticed this and it hardly matters.
George H.
...
t eHi Baron, How are you thinking of using the LED? You can get full shot noise in the current from any diode. I've also used LED's shinning on photodiodes to get full shot noise. The advantage of an avalanche zener diode is that you get =91gain=92 in the diode. Each electron makes many more. You can also get a lot of shot noise from a PMT used in current mode. There the gain is even bigger 10^6 or so.
George H.
message
message
gain
is to
G_1*...*G_2
it and
adds and
stages
use as
it.
to have
the signal
I was discussing ignorance with one of my guys yesterday. We're trying to specify and buy a crystal oscillator to be used as the timebase for an a/d conversion of a spin resonance effect. Spins may last seconds, and multiple shots may be signal-averaged over up to an hour, and we'll be doing FFTs on the averaged sample data. So how do we specify the oscillator? We concluded that 1) we're not smart enough to really understand the problem, especially in the way that OCXO makers specify their parts and 2) if one of us did know this much about this or similar subjects, we'd probably be too specialized to design the whole system.
Can anybody recommend a source for OCXOs that doesn't want an insane amount of money for them? 64 or 128 MHz would be nice, but we could work with 5 or 10 or some such if that was a standard, affordable part. Under $50 would work.
John
e in message
in message
legroups.com...
tiple gain
need is to
se G_1*...*G_2
handle it and
n stage adds and
scaded stages
should use as
stream
rease it.
isable to have
GBW
roves the signal
stage
total
rst
ing amps
level)
that
slew
kes
n yargest
ere are
at
t eWhat does the first "O" stand for? I had a job interview once at Frequency Electronics. They make CXO's and atomic clocks. I had an idea of how (maybe) to make a better temperature control. But I learned that temperature drifts are not the big problem. The 'uncontrolled' frequency drifts come from material either leaving or sticking to the Xtal. Can you synchronize the oscillator to the GPS? We have =91free=92 atomic clocks orbiting over our heads.
George H.
message
message
multiple gain
need is to
G_1*...*G_2
handle it and
stage adds and
Cascaded stages
should use as
downstream
decrease it.
advisable to have
improves the signal
stage
amps
level)
"O is for "oven."
This particular app needs low phase noise and good stability for, say, a hour at a time. Oven crystals run at their "turning point" temperature (the flat spot on the f-vs-t curve) so are very stable there. Longterm aging isn't an issue. Using GPS usually means locking to a 1 PPS source, which is great for longterm stability but not much help with phase noise. And GPS would sure blow the $50 budget.
John
Hi George, The reason I asked was because I once built a device for optimising the noise bandwidth of VHF receivers. Not being able to afford the specified noise diode I tried a red led. I was quite surprised that it produced as much noise as it did. It seemed to work quite well. I still have that bit of kit kicking around.
Thankyou for your interesting notes.
--
Best Regards:
Baron.
...
t a rBaron, I was reading about zeners in AoE (2nd edition section 6.14) during lunch today. And realized I'd been making a big mistake in my understanding of the noise from zeners. I had been thinking it was current noise. But it's voltage noise that is sensed. Ahhh, it=92s the I-V curve of the zener that turns the current noise into voltage noise! As you get closer to the knee the curve flattens out and you get a larger voltage noise for the same amount of current noise. (See figure 6.22 in AoE) (Of course as you reduce the current you have less current noise, but apparently the slope of the I-V curve is the larger effect.) I think this also explains the asymmetry that is observed.
So I think I now see how you used an LED as a noise source. I=92ll have to try it some time.
George H.
Quite simply it has to do with the curvature of the curve at the knee. It is an unstable area in the operating characteristic, thus high noise. However the changing slope perturbs (skews the third order and fourth order moments of the probability distribution) the amount of positive values compared to the negative values. But i would have to study up quite a bit to back that up with real math.
Thanks for the response Joseph, I 'think' I figured this out today. You're right it has to do with the curvature of the I-V curve. All diodes have shot noise in the current. (And zeners have more.) From the I-V curve you can determine the conversion of current to voltage and thus the amount of voltage noise. And the asymmetry is simply the result of the curvature. A noise fluctuation that decreases the current has a bigger effect on the voltage than one that increases the current.
George H.
In the zeners I played with, as the current went down the noise began to look asymmetric, then started to look more like random discrete spikes, at at low currents turned into a noisy sawtooth oscillation. That's not shot noise. It's more like random spots of negative resistance. The voltage builds up into the junction capacitance, and some small zone gets a thermal electron that starts a local avalanche breakdown, which discharges the junction cap below the avalanche voltage threshold. Something like that.
At high currents, there are many, many avalanche breakdown pulses added up, so the central limit theorem kicks in to make the noise gaussian.
The voltages involved are way bigger than shot noise would produce...
300 nV per root Hz with 1 mA into a few ohms of zener dynamic impedance is roughly 3000 times the shot noise level.John
o o
hat
ore
nd
o - mCool, I meant to ask you about the oscillation part you mentioned earlier. I=92ll certainly look into it. To be honest I never measured the I-V curve of a zener before yesterday. I used a 1n5245
15V zener and found about a 1 M ohm impedance up to about 14.2 Volts, at which point it dropped to maybe 20 ohms or so. I had a crappy meter measuring the voltage and not much resolution.I (uA) V (V)
7 6.85 9.1 9.0 12 12.1 14.2 14 21 14.2 51 14.28 100 14.28 813 14.30Sorry, I should have put more current through it but it was the end of the day, and I had to get home.
The circuit I used to make a noise source uses a 20 V zener in series with a 1Mohm resistor across the +/-15 Volt power rails. A
1uF cap picks off the noise voltage at the junction. So It looks like I was running at a point =91above=92 the knee. Call it 10uA of current which gives sqrt(2*e*I) ~ 1.8 pA/rtHz of noise. Times 1 meg (I=92m assuming the same resistance as above which could be a mistake.) is 1.8 uV/rtHz. OK that is more noise than I thought I had? Clearly more measurements are in order. But it=92s going to have to wait for a few weeks. I=92ve got an important conference/show in two weeks and I have lots to get ready.then John wrote,
=93> The voltages involved are way bigger than shot noise would produce...
Hmm, maybe we=92re working in different ranges? I was working at 10uA with a bandwidth of a few hundred kilo hertz max. (the 1uF cap was followed by a 8 MHz GBW opamp with a gain of 20.)
And I wouldn=92t be surprised that the current noise is way above the =91single electron=92 shot noise. As long as the time scales one is looking at are long compared to the transit time of charges through the diode, then the effective charge of the electron will be multiplied by the Multiplication factor.
I noise =3D sqrt( M*e*I) The only expression I found for M was the following.
M =3D 1/{1-(Va/Vbr)^n} where n is 2-6. (not very useful given the range of n)
from here.
This sounds like it will be fun to measure. First the current noise then the voltage noise and then see if they fit with the DC I-V curve.
George H.
"John Larkin" wrote in = message news: snipped-for-privacy@4ax.com...
also
that
more
knee.
and
text -
You might be interested in early work on avalanche diodes as noise sources. Haitz and Voltmer, "Noise of a Self-sustaining Avalanche Discharge in Silicon: Studies at Microwave Frequencies," J. Appl. Phys., vol 39 (June 1968), pp. 3379-3384.
(In that paper they were running mA of current and getting noise at Ghz frequencies that was, say, 20 dB above kT.)
Haitz, "Noise of a Self-sustaining Avalanche Discharge in Silicon: =20 :Low Frequency Noise Studies," J. Appl. Phys., Vol 38 (June 1967), pp. 2935-2946.
I believe these draw on the theory of Hines in: "Noise Theory for the Read Type Avalanche Diode," IEEE Trans. of Electron Devices, vol. ED-13 (Jan. 1966), pp. 158-163.
Also, see this article by Motohisa Kanda of NBS: "An Improved Solid-State Noise Source, IEEE Trans. on MTT-24 (Dec. 1976), pp. 990-995. (This covered 2 - 4 GHz)
--=20 Regards, Howard snipped-for-privacy@ix.netcom.com
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.