Scott,
I suspect the way the unit work is that there is a low noise high gain amplifier that is switched between the input and a reference voltage. This will produce an output that will be zero volts if the reference is equal to the reference. Just for arguments sake let us pretend that if the input signal is larger that the reference that the amplifier produces a positive voltage. This is used to increase the reference voltage. If the input is lower than the reference the reference is stepped down. At balance the amplifier will draw no or only leakage current from both the input and reference voltage.
Now what topology would you use for this amplifier?
I would condider an op amp style with a capacitor in the feedback loop.
Apologize for cell phone typos.
I suspect the way the unit work is that there is a low noise high gain amplifier that is switched between the input and a reference voltage. This will produce an output that will be zero volts if the reference is equal to the reference. Just for arguments sake let us pretend that if the input signal is larger that the reference that the amplifier produces a positive voltage. This is used to increase the reference voltage. If the input is lower than the reference the reference is stepped down. At balance the amplifier will draw no or only leakage current from both the input and reference voltage.
Now what topology would you use for this amplifier?
I would condider an op amp style with a capacitor in the feedback loop.
Apologize for cell phone typos.
Ed you are looking at the wrong thing. First of all the expected crest factor of random noise is 3.1 so rms to p-p conversion is roughly 6X. I fixed the 1.6pV so 1.6pV X 6 = 9.6pv close enough to 8. The question is not how long to wait, as you know there is no DC,
but how low a BW bounds your measurement error. In this case the amplifier BW limited to .005Hz has an uncertainty of 8pV p-p (+-4pV) because it has a 22pV/rt-Hz input noise.
Not sure, with any chopping or nulling scheme you need some amplification of the error somewhere. To get 22pV/rt-Hz I figure some reactive gain like in the transformer input amp so maybe a relay chopper and an AC amplifier with transformer feedback. It's not that large a leap from 60pV to 22pV. I don't think they used 2000 1nV amps in parallel.
Scott
The first issue I would look at is the source impedance of the reference. Insuspect it would have to be .01 ohms or less. Next would be the sampling capacitor. If this capacitor fed a transformer the I don't think magic would be required to get the rest of the performance. The capacitor probably does require some magic.
Now if the uncertainty is 8 PV how many time constants are required to settle to that level? Obviously this is a function of full scale settings in addition to the noise level.
I will try to set up the circuit equations and see what I think the amplifier contribution can be. Not sure these quick assumptions are actually considering the right issues. As even using actransformer front end I have trouble believing it needs to have a primary impedance of milli ohms.
Currently getting ready for the AES this weekend.
The first issue I would look at is the source impedance of the reference. Insuspect it would have to be .01 ohms or less. Next would be the sampling capacitor. If this capacitor fed a transformer the I don't think magic would be required to get the rest of the performance. The capacitor probably does require some magic.
Now if the uncertainty is 8 PV how many time constants are required to settle to that level? Obviously this is a function of full scale settings in addition to the noise level.
I will try to set up the circuit equations and see what I think the amplifier contribution can be. Not sure these quick assumptions are actually considering the right issues. As even using actransformer front end I have trouble believing it needs to have a primary impedance of milli ohms.
Currently getting ready for the AES this weekend.
International Climate Science Coalition (ICSC) "...moving the debate away from implementation of costly and ineffectual 'climate control' measures.”
Yep, that figures.
This is an old trick, like Tobacco Industry investing in "Cleaner Air Policy". Steal the language and words of your opposition, even invent titles similar and use them nefariously. No, that does it for me as if you have right on your side you have no reason to do that. I suggest "Merchants of Doubt" by Oreskes and Conway, that will give you a more complete and historical account of the science of global warming. BTW, the same tactics that was honed by the Tobacco Industry is now being employed - you didn't have to prove tobacco was not harmful, all you needed was to sow doubt and it was hugely profictable. In the United States nearly 25% of the population still express doubt whether there is a connection between smoking and lung cancer. Maybe the most success disinformation campaign ever and is now being repeated.
I have zero interest in politics and this is not a political issue to me. I am neither left or right and I have never voted for any politician or political party and never will. Yes, I have a number of friends who are/have been involved in climate science, collecting data etc and they are neither political and not rich either.
(13.5V - 6.8V) / 1KOhm
I got them from digi-key, but it looks more like a bandgap,
not a subsurface Zener. The 1/f corner should be at 100 Hz.
regards, Gerhard
Attachments
Yes, still on my list of things to try. I figure 40mA Id gets just below .3nV on one.
It's the demodulated noise/signal that shows on the meter and the uncertainty is random noise.
I don't think any conventional straight forward technique with massive parallelism can practically reach .022nV noise.
Mouser now sells these for about $20 although they don't stock many. They are indeed very quiet about 10 dB better then the Toshibas at similar currents last time I measured them.
Currently playing with them in a MC phono experiment since I had a couple hundred made some years back.
IF3601 and paralleled devices
Cooling will help a bit, although the optimal temp is far from 0 Kelvin. Usually about 100 K (77 K is already too cold). A trifle impractical. Very effective for minimizing gate leakage, although Vdg must be low.
Danyuk did find the noise corner on the 862 rather low, and I mentioned I stacked some, glued together, and with a wire as the heat sinking for the gate leads. Forget a planar array, I think, although stacks of circular arrangements might work.
I was kindly provided with a couple by Wayne C. and Nelson P., haven't (of course) had the time to play with them. Ah the cruelty of the 1/4 power law for voltage noise!
Cooling will help a bit, although the optimal temp is far from 0 Kelvin. Usually about 100 K (77 K is already too cold). A trifle impractical. Very effective for minimizing gate leakage, although Vdg must be low.
Danyuk did find the noise corner on the 862 rather low, and I mentioned I stacked some, glued together, and with a wire as the heat sinking for the gate leads. Forget a planar array, I think, although stacks of circular arrangements might work.
Master Wayne Sir,
call me daft, but do your lines mean it was a custom-order part a few years ago, and at current a limited serial production ?
(if you are willing to respond and familiar with the term going-Dutch, have a heart and please don't include a '$-figure per couple hundred' in your answer)
- Status
- Not open for further replies.