30 second settling time. 1nV/ rt Hz at .03Hz should do that 30 pV nicely looking at DC!
I am working on a low noise amplifier that should do better than 200 pV/rt Hz. But The chips seem to have a corner frequency of 2000 Hz, so it will be a complement to the low noise preamp I just built.
http://www.st.com/web/en/resource/technical/document/datasheet/CD00066247.pdf
^ You've got to be working with some seriously low impedances if that chip is your weapon of choice.
But what has such a low 1/f corner? Hard switching Schottky or JFET chopper?
I have mixed down white noise from 50 MHz down to DC and that was flat down
to "DC" as far as I could see. (MCL diode ring mixer).
Probably also works the other way around?
200pV/sqrtHz: That's about what my 20 averaged ADA4898 opamps deliver.
Corner is 20 Hz or so. Probably my FFT analyzer limits this; it might be lower.
20 op amps is where soldering starts to get boring. The better solution would
be 2 amplifiers with only 10 op amps each and then doing cross correlation
in a stereo FFT analyzer. Calculating the 3 cornered hat like signal source
analyzers do for measuring phase noise.
Last edited:
It wouldn't surprise me to see a reed relay chopper for a DC instrument. But I also recall the technique was to balance the input against a known source and when the difference trimmed out you knew the attenuation values, thus the input. By balancing out you would yield a very high or even infinite input impedance.
Yeah, I still have 2 trays of mercury wetted reed relays, they must be mounted
with the right side pointing up. You feel the mercury flowing around when you
shake them. No problem waste then after all??
with the right side pointing up. You feel the mercury flowing around when you
shake them. No problem waste then after all??
They say 30 sec time constant not settling time and 8 pV peak to peak not rms and besides explicitly say equivalent noise resistance of .03 Ohms.
You might be better off paralleling some SSM or THAT dual transistors and run them hot in a composite amp. The 1/f current noise is possibly bad on that RF amp. You could also try those crazy JFET's (forgot the number), I think with care they would do .25nV but the heat would start to be a problem.
EDIT - I still compute from their numbers tau of 30s is ~.005Hz filter and noise of .03 Ohms is 22pV so that's 1.5pV rms or give or take 8pv p-p
Last edited:
"Input 1. Approximately 30K ohms, 10 seconds after a step change in the input increasing with time to more than 10 M ohm."
Sure sounds like an auto balancing bridge amplifier. Noise of a .03 ohm resistor is around 20-25 pV/rt Hz. So at 30 seconds that would be 125 pV/ rt hz. Really nice but not magic.
Your numbers have me scratching my head 22pV/rt_Hz * sqrt(.005Hz) = 1.55pV rms or ~8pV p-p as stated. Seconds are inverse of Hz.
They explicitly imply that adding ANY resistance at the input above the stated enr (.03) compromises the noise.
Last edited:
1/30 S = .0333... Hz
Sqrt(.0333 Hz) = .18257 rt Hz
121 pV/rt Hz x .18257 rt Hz = 22 pV
Yes 22pV rms would be ~130pV p-p vs the 8 in the data sheet. A 30 second tau is .005Hz but your on the right track now.
Last edited:
Yes I missed tau so F = 1/(2 pi tau) "The equivalent noise resistance of the P12 is only 30 milliohms" or 22 pV/rt Hz
1/(2 x pi x 30 S) = .00531 Hz
Sqrt(.00531 Hz) = .0728 rt Hz
302 pV/rt Hz x .0728 rt Hz = 22 pV
Still wrong, Ed V/rt-Hz times rt-Hz is volts rms
22 pV/rt Hz x .0728 rt Hz = 1.6pV rms or close to 8pV p-p, you said it right up there the P12 is an amplifier, albeit a very low frequency one, with 22pV/rt-Hz noise down to sub Hz frequencies.
- Status
- Not open for further replies.