Yup. Your average bandgap comes in at about 100 nV in 1 Hz, and micropower ones are worse. Multiply that by Vout/1.22V, and you get into microvolts pretty fast.
The reason is that the voltage is formed by adding a B-E drop to 10 times a
60-mV delta V_BE. That 23 dB is a killer.Yup, Good Medicine.
Cheers
Phil Hobbs
In this particular case the main purpose of designing the microphone was to get very low noise. The capsules rolled off below 300 c/s at
6dB/octave (20dB/decade), so the associated circuitry had the corresponding degree of boost built-in to give a flat response down to at least 30 c/s. That meant that any circuit noise was going to appear 20dB worse at 30c/s than it actually was, so minimum LF noise was really critical in this case.Even if 30c/s isn't particularly audible, it could cause intermodulation if you were listening to very faint sounds at high gain.
Ever measured a gas tube like an OB2? That'd be an interesting comparison
Zeners are pretty noisy.
We use a pseudo-random shift register as the noise source in some products. My guys went overboard and gave it a repeat period of some centuries.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
Maybe you could use your tritium light as random source: panteltje.online/pub/tritium_light_movie_mvi_3243.avi
To download in Linux use: wget panteltje.online/pub/tritium_light_movie_mvi_3243.avi
I've used a blue LED pilot light as a source for a cheap and cheerful audio test noise generator. Followed it with a couple of stages of audio op-amps and an optional state-variable filter to give wideband, notched, band-limited, high-pass and low-pass outputs. Handy for checking loudspeakers and other electro-acoustic devices.
Am 25.12.25 um 01:58 schrieb john larkin:
No, Zeners are not noisy at all, you can see it in <
The BZX84C2V7 delivers maybe 5 dB more noise power than 1nV, let's say 2.5 nV /sqrt Hz. It starts getting worse as soon as the avalanche effect takes over with rising voltage. You can follow that in the BZX84 series from 2V7, 3V3, 4V7, 6V8...
Zener and avalanche effect have different TCs; that's why reference diodes usually are around 6-7.5Volts.
Sorry, the trace colors/styles are selected by gnuplot.
Prof. Zener once sued the industry that they should not use his name for avalanche diodes, because it was not "his" effect. I respect that act of scientific honesty.
They settled to call them Z-diodes and to pretend it was for the form of the characteristic curve, but the ghost was out of the bottle already.
That does not really help. Single bit outputs of a LFSR are highly correlated as can be imagined easily by considering long runs of D-FFs unbroken by XORs. Making the shift register long to span centuries increases the probability of long 1 or 0 runs.
Using always only the same half percent of a polynomial (from reset) may create noticable DC bias. In a spread spectrum system you would see a residual carrier then, giving your foe a hint where jamming you works best.
There is some discussion on this in Robert C. Dixon: Spread Spectrum Systems With Commercial Applications . Wiley-Interscience isbn 0-471-59342-7
Someone here uses a news reader that does not respect that text in <> should not be modified. URLs don't like it to be broken by CR LF.
Zener or Johnson noise is plenty random. Scramble it digitally to remove residual statistical imperfections.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
Everybody calls them zeners, sometimes even "bandgap zeners." Design engineers, diode makers, Digikey, LT Spice.
I measured 340 nV/rthz on a 10-volt 1N758.
Sure. An infinite series of random bits will eventually have an arbitrarily long run.
I think my FPGA guys XOR'd a few different roughly 48-bit pseudorandom sequences to get the super long period. That may help.
Any bias is easily fixed, but reaslistically there are lots of sources of offset.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
Wouldn't trust any direct connection between an integrated circuit pin and a sensing point, particularly in the presence of noise.
Also don't trust internally generated reference currents. Why? Probably because it's just one more added complication and source for error, over which the end user has no control.
RL
In lots of reg chips, the connection is internal, out of your control.
If it don't work, don't use it.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
That's not my department !
. . . says Werner von Braun.
RL
One experiment is worth a thousand expert opinions.
. . . says Werner von Braun.
John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics
Not if E. Laureti is running the experiment.
In fact you need a expert to run your experiment if you want to get reliable results. And well-run experiments aren't cheap. If you can find one reliable expert, it's a lot cheaper to buy his or her advice than it is to run the experiment.
The history of physics is full of deceptive and badly executed experiments that mislead people big-time.
The published histories of physics tend to leave them out. It's like evolutionary history - you concentrate on the success stories, because the failures didn't leave any descendants.
Not quite. Tom Lehrer:
.
He did in fact say this.
Joe
Have something to add? Share your thoughts — no account required.
Ask the community — no account required