low noise amplifier for high impedance source

Part of an analytical FTMS, or a system for studying ion molecule reactions, or nuclear particles, or ...? Field? Inquiring (okay, mildly nosy :-)) minds want to know. It also affects the advice since we don't know the frequency range you care about - sub 10 kHz for large m/z on a small permanent magnet or tens of MHz for protons in a supercon, for example.

-- Regards, Carl Ijames carl.ijames at verizon.net

Carl Its an FTMS. The frequency of operation is 10 kHz to 1 MHz for the amplifier. We have a 7 Tesla superconducting magnet. thanks

I see the HRT isn't working.

--
Service to my country? Been there, Done that, and I\'ve got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida

** Groper LIAR Alert.

** But NOBODY in sight with a SINGLE BLOODY CLUE about electronics design.

Just this pimply faced wanker pretending to be what he is not.

How extremely f****ng odd.

....... Phil

Thanks, Archie. It's been a little while since I built my last one of those but they are great fun :-). Have you looked at the preamp that Gordon Anderson from PNNL published several years ago, maybe 1998? There was also a preamp poster at ASMS I think it was two years ago, where the author compared room temperature and crygenic operation. Might have been Pete O'Connor's group. Often the input is a dual JFET for best matching, in a differential cascode configuration. Something like the Siliconix (now Vishay) U401. They have a good app note on JFET noise modeling, as well, at

Comisarow did a nice job deriving the signal and noise behavior but if you want to make quantitative signal calculations don't forget that he made the infinite parallel plate approximation so you will only get some fraction of the signal he calculates (maybe a third? depending on aspect ratio and open or with end plates). One of Alan Marshall's students published a few papers on signal strengths in cylindrical and rectangular cells compared to the infinite paralel plate approximation back in the 1990's, too.

-- Regards, Carl Ijames carl.ijames at verizon.net

If I understood correctly, you apply potential across plates and measure current through your experiment? If that's the case, what is potential value? It will be very hard to measure so low signal and have wide bandwidth. I would suggest having a tuned circuit for frequencies of interest and in addition, lockin amp after fet preamp. Reduced bandwidth would help with noise problem. There are amps from Burr Brown (TI) that work in sub pA range, although questionable if it would work on 4K. Could keep them warm by heating PCB. You could use differrent preamp boards for different experiments or make it adjustable so you could tune it for frequency of interest. In addition, you could match impedance of your amp to experiment.

Phil, do you have Tourette sindrome? If not, you hardly have excuse for language used in posts and if you are not adding anything constructive or you are not interested in topic you could leave or just shut your cakehole. As support electronic tech. for research I know how often is difficult to solve problems in measuring small signals in scientific experiments, and everyone involved interprets problems from their own angle of view.

: Its noise isn't appreciable higher at 1 to 10kHz. I'll look : for my data and post it.

My voltage noise data at Id=7mA for three input loads (only 10ohm is of significance) can be glanced at

:> But the BF862 does seem to have a rather large gm for its :> low capacitance. When you say "contrary to Mother nature" ... : No, I was just being funny. I suppose it's really the older

I was wondering this remark, too. It approaches the un = sqrt( 8/3 k T / gm) figure quite nicely.

Regards, Mikko

"Sinisa Jezdic"

** Do you have ASD ?

Superfluous question, I know.

So Fuck Off , wog idiot.

......... Phil

Can you explain your plot? Does the 0.58nV/rt-Hz annotation mean your measured input noise with a 10-ohm source? That's pretty good, much better than Philips claims. The 1/f frequency looks like about 1.5kHz.

Judging from the 1M rolloff above 5kHz, it looks like your amplifier has about 32pF of effective input capacitance. That means the 50kHz 150k region is also somewhat attenuated.

Hmm, the Johnson noise for a 150k resistance is about 50nV, yet your bottom-middle annotation is 1.2nV, so I guess I really don't at all understand your annotation scheme.

What was your amplifier's voltage gain? We can't do much with your data plots without knowing that.

Yes, of course, 1/sqrt. My point was Philip's improved gm for their small JFET was a big factor in improving noise over the older JFET.

BTW, at low currents theoretical JFET gm values all track the drain current, not the current density, and are therefore theoretically unaffected by the JFET's die size. Perhaps it's only for currents approaching Idss that the die size becomes an important factor. But larger dies usually means higher Idss, which means higher gm, so...

--
 Thanks,
    - Win

Winfield Hill wrote: : Can you explain your plot? Does the 0.58nV/rt-Hz annotation mean your : measured input noise with a 10-ohm source? That's pretty good, much : better than Philips claims. The 1/f frequency looks like about 1.5kHz.

No, each annotation refers to the gridline underneath it. The white noise is between the 0.58nV and 1.2nV gridlines, pretty close to the specified 0.8 nV/rtHz.

: Judging from the 1M rolloff above 5kHz, it looks like your amplifier : has about 32pF of effective input capacitance. That means the 50kHz : 150k region is also somewhat attenuated.

Thanks for noticing this. I didn't pay that much attenuation to the

1Mohm trace (it tells hardly anything about the current noise which I'd be interested in), but now I see there is something funny in that measurement. The 130kohm trace shows (in another plot) -3dB point perhaps at 100kHz whose implied Cin=12pF is not compatible with the 1Mohm figure.

: Hmm, the Johnson noise for a 150k resistance is about 50nV, yet your : bottom-middle annotation is 1.2nV, so I guess I really don't at all : understand your annotation scheme.

It's a parallel connection of the 150k load and the 1M gate bias resistor, i.e. 130k actually. The trace lies between the 73nV and 37nV gridlines.

: What was your amplifier's voltage gain? We can't do much with your : data plots without knowing that.

That's about 175 V/V, the annotations take this into account (i.e. they are input-referred).

An interesting detail is the surprisingly high high-pass frequency, which is due to the dc-blocking cap at the output. I tried the ultra-high capacitance density ceramic from Murata, the 10uF GRM31 type. Did not expect much of it, but their impedance plot

didn't look too bad so gave one a try. It took some amount of detective work and a hint from a colleaque before I noticed that their capacitance value collapses when there is a dc biac voltage present across them. . Be warned!

Regards, Mikko

Mikko S Kiviranta wrote: ...

Do you have any high capacitance X5R/X7R capacitors and can you post diagramms like the above for these X5R/X7R capacitors?

Thanks

--
Uwe Bonnes                bon@elektron.ikp.physik.tu-darmstadt.de

Institut fuer Kernphysik  Schlossgartenstrasse 9  64289 Darmstadt
--------- Tel. 06151 162516 -------- Fax. 06151 164321 ----------

Hello, my pour Student At first: you need to rceive enough information about physics of noise for JFETs. We have two separate processes and different solutions,namely: for low current noise you have to reduce area to get small input current so current noise varys with root (i can not remember exactly)from Input current. So you should take transistor with Igs Hello everyone,

Hi Sinisa, thanks for the suggestions. The induced current is in higher pA range. I agree that using tuned circuit will help if i am trying to detect signal due to few ions (infact Gerald Gabrielse does use tuned circuits to pick up signals induced by single protons over at Harvard). However for my application which is a mass specectrometer you only know a range of frequencies where your unknown sample ions will have - in essence we want to do a broadband (10KHz to 1MHz) detection.

There are amps from Burr Brown (TI) that work

Yes, i was suggested using JFETs and warm these up to 70 kelvin (just to turn them ON) inside the cryogenic vacuum system. I am going to try it as i realize, getting bigger GaAs MESFETs (with low 1/f noise) which work fine at 4 kelvin is not working out.

You could use differrent preamp boards for

We did thought about it, if i can find a way to tune the preamplifier impedance which sits in a vacuum system, externally. Still, it will be something i want to try in future, for now i want to see what is the best i can do in broadband detection mode using low noise JFETs. thanks again. Regards, Arch

Uwe Bonnes wrote: : Do you have any high capacitance X5R/X7R capacitors and can you post : diagramms like the above for these X5R/X7R capacitors?

For a 2.2uF 10V X7R cap in 1206 case, the type AVX CM316X7R225K10AT see

, AVX_CM316X7R_2U2_Unbiased.gif and AVX_CM316X7R_2U2_5VBiased.gif .

The 5V dc bias has no obvious effect.

Regards, Mikko