Input stage mess

"Ken Smith" a écrit dans le message de news:dmo7vt$b6o$ snipped-for-privacy@blue.rahul.net...

wrote:

He he, 100 x TL74 : say 10pF/opamp and we're back to square one :-) And it'd be more like 6400 x TL74. Ouch...

Not even. Your noise floor depends on the total gm and you'll need to run a lot of juice into the fet. The supply can't handle that and if you AC couple it you'll have to provide a constant current source.

Also note that a BJT is even less an option here because of the emitter shot noise current: neglecting Rbb' (which we obviously can't) 200pV/rtHz will call for at least IC = 2.q.Ut^2/en^2 = 5.4mA and an input shot noise current being at least sqrt(2.q.IC)=42pA/rtHz, hence an input DC blocking cap impedance = 2000/42=48 ohm. For a 2nV/rtHz noise @ 0.1Hz, this is 33000uF which is a pretty big cap, with its own issues :-)

Good try anyway :-)

Fred,

let me ask a slightly different way...

if it is hard to build a circuit that has sufficently low self noise comapred to the power supply to make the measurment..

why is it that the application ciruict can have sufficently low self noise such that the noise from the power supply is the dominnat problem?

In other words, I don't doubt that your video display or whatever is very critical and it's easy to see small amounts of noise on the screen....

but how can this video circuit itself have noise that is so much lower than the power supply when the power supply noise is so low we can't measure it.

Mark

I wonder what the math would be like on a varicap LC oscillator driving a spectrum analyzer or equivalent?

Or a varicap parametric amplifier? They were a fad for a while, before fet opamps got good. I think TI even made a monolithic varactor-based opamp once.

John

"John Devereux" a écrit dans le message de news: snipped-for-privacy@cordelia.devereux.me.uk...

writes:

at

Hi John, turning your drawing upside down you can see it's exactly the same as the more usually represented

BFC || |\\ ----||--+-----------|+\\ || | | >--+------ | .-----|-/ | | | |/ | | | ___ | | +----|___|--' .-. .-. 10 | | | | 1K| | | |1 '-' '-' | | --------+-----'

with the opamp supplies common point displaced across the input. Unfortunatly it will suffer exactly the same shortcomings which are too high noise current through the blocking cap, huge blocking cap value and *poor* behaviour of these caps at such a low level, cap charging pb,... And changing the opamp supply return point, you've just added another pb, which is a too low input impedance (1R or so) which will make a superbe voltage divider with the cable inductance, loosing about 10dB of the precious signal at the upper corner of the bandwidth (0.5uH @ 1MHz = 3R).

The AD797 has also 0.8nV/rtHz noise which calls more for 16 opamps than 8.