Gain peaking in driven shield

All sorts of warning flags should immediately pop up at you when you connect a capacitor from output to noninverting input. You might want to look up "negative immitance oscillator".

You may be able to achieve some success with a driven shield, but you're going to need a much better model of your input, and tune your circuit to those particulars - at least if you want to achieve semi-high performance.

HTH...

ld

the

Thanks Cassiope, Though I hope you don't think I added the capacitance. It just came along with the shield. I think I have a pretty good idea of what the input looks like. (There's not a whole lot to it.) I did a bit of modeling with LTspice last night, but I don't know how to model the driven shield. (I just reduced the Cig capacitance from the starting value of 60pF down to 6pF or something.) Perhaps the gain peaking is why I've only seen driven shields on real high impedance inputs. >100Meg

George H.

The example I've seen (in a Texas app note) had a few ohms in series with the op amp output.

Surely the effective capacitance from inner conductor to shield, which you call Cis, is essentially nil because of the driven shield masking the C - i.e. the inner conductor does not see a potential difference between itself and the outer conductor.

I'm about to try this myself for a pair of cables about 30cm long. I chose an AD8626 because -

- the input current is very small

- bandwidth seemed decent, a few MHz

- it can drive a few hundred pF

- it's unity gain stable

- it has low Cin

- it can use +/-15V supplies and drive as near the rails as my signals are likely to go

I'm connecting the inverting input directly to the output (a voltage follower) putting a 1 ohm resistor in series with the output, which will drive the shield.

If you need more info let me know and we can email directly.

You're trying to measure Johnson noise, right?

One problem with any such guarding/bootstrapping scheme is that there is a room-temp amplifier that picks up the signal and drives the guard, and it has noise of its own. In some cases that makes adding the guard a losing battle.

The OPA134 may be a bit slow, especially driving Csg.

What sort of coax are you using? Can you reduce its capacitance?

John

Hi Nemo, There are three layers. On the outside is ground, next is the shield, the center of the coax is the signal line from the resistor. Driving the middle shield, reduces the capacitance between the inner signal line and outer ground. However it adds the capacitance from the shield to the inner signal line. I'll look up the AD8626. And try adding a bit of R in the output line. Do you have a TI app note number? What sort of source impedance are you looking at? 100k looks OK, but 10k ohms shows gain peaking.

George H.

he

Yup, The gain peaking is a real pain.

There is a cable which has ~12" of RG174 (? not sure that's the right number, small coax) and in the probe there is ~8" of SS coax from Lakeside cryo. I'm pretty much stuck with the cable I have.

If you go here,

George H.

Closer to 13.

How about a BF862 jfet? 0.8 nv/rthz and very low current noise. No opamp can do that. Cheap, too.

You can make low-C coax, like they use in scope probes, with a tiny wire inside a biggish shield, with some low density spacer like spiral monofilament or something.

If you want to demonstrate Johnson noise vs temperature, it would be a lot easier to go hot!

John

I don't think that your circuit diagram shows all the necessary impedances where they ought to be. The OPA134 has finite output impedance, probably of the order of 100R, and your second circuit diagram should look more like

-- Bill Sloman, Nijmegen

ield

f the

und

in

g a

er

Opps, my mistake!

Hmmm, Roll my own? That's just going to be way too much time for very little 'gain'. I've basically got to put this to bed for the time being and get on with other things.

Yup, the news letter doesn't show the data for Johnson noise , but we change the temperature from 77 K to 373K.

If I get the gain peaking under control, then it will be fun to see how the driven shield changes as the Johnson noise decreases below the amp noise.... I wonder if I'll get what look likes a peak in the noise spectrum that is from the driven shield amplifier leaking it's noise in through the shield capacitance. Fortunately the whole driven shield idea is not a big part of measuring noise.

George H.

ld

the

=A0 =A0 =A0|

=A0 =A0 =A0|

D =A0GND

Hi Bill, That is about right. I was playing around with LTspice last night and was able to get what looked like nice flat 'gain' out to near 1MHz or so, with the simpler voltage follower circuit. I tried the above circuit so I could try rolling off the gain by putting some C in parallel with R3... But this just seems to shift the gain peak to lower frequencies.

George H.

On May 5, 9:26=A0am, George Herold wrote: [....]

If the cable is very long, the simple capacitance model of the cable fails. You need to start dealing with the inductance too.

I have used the LSK170 to make a very low noise non-inverting gain to make this work a little better. The noise put onto the cable by your shield drive starts to win at some high frequency.

--------------- ! ! [R] U1 [V] ! !\ ! !---+--------! >----+--+---- in ---->! !/ ! !---+--[R]------------- ! R2 [R] ! R1

IIRC R1=3D300, R2=3D30 U1=3Dgain of 100

It really matters that the gain be 1.00 and the phase be 0.00 at the frequency you need to measure. In my case, I have the option of changing the circuit between measurement bands. The circuit had a roll off at about F*100 and F/100

ield

f the

und

in

g a

er

how low, doesn't something like ADA4898 get close?

-Lasse

Can't see George's posts because he uses gmail but this may help: Older passive car antennas (sans integrated preamp) had very low capacitance coax in order not to detune the first AM filter and not to snuff out too much signal. But I am not sure where that can be bought without the antenna. Autoparts stores used to have it but that was >20 years ago.

It is nothing more than RG/62, 93 ohm coax.

The stuff I had was a whole lot higher than 93ohms. About 1/4" diameter, super-thin inner conductor. But I had bought that in Europe and it became hard to obtain after the advent of active antennas.

Technically it wasn't pure-bred coax from a performance point of view because the inner conductor wasn't fully enclosed by dielectric. It had some slack in there and was a wee bit longer per ft than the jacket, in order not to break that thin wire.

I used to repair car radios when I was a kid. (late '60s) The coax said RG/62. A Delco engineer told me it was 93 ohms. RG/62 has a very thin center conductor in a polypropylene tube, and the center conductor is a very thin wire that zig-zags it way inside of the polypropylene tube.

The dielectric is mostly air. The polypropylene tube was intended to keep the braid's shape by pressing it against the outer jacket. Just like some CATV hardline with ceramic rings between conductors, or air filled hardline used for high power AM broadcast transmitters.

Ours looked similar, center conductor sort of meandering through there. But it was over 150ohms, we had very fancy gear at the RF institute where I had a side job designing stuff. I don't remember why I measured it but instantly wanted to buy some for myself. The institute couldn't sell me any but told me where to get it, and it was within bicycling distance.

I believe the two 150ohm coaxes from this company are not twinax but not sure:

That's basically an LT1028 clone, with huge front-end bipolars. The price is a lot of current noise, 2.4 pA/rthz. Dump that into 10K and you have 24 nV/rthz.

John

Yeah, I've only got about 0.5m of cable at a few MHz max. Once the cable becomes a tranmission line you can't 'do' the driven shield trick. Can you?

Wow, not sure I really get that circuit. You've got the Jfet as a 'follower' and I assume I should hang the shield on the bottom of R1... (Correct me when I'm wrong) along with some resistance (?) to sink the current. The amp U1 is feeding the bottom of a voltage source and R which looks like a current source. But I'm not sure what R2 is doing?

The phase shift at high frequencies is certainly troublesome... Playing around with LTspice tonight it seemed that the peaking happened when I made the Cig capacitance too small. (Cig is only shown in Bill Sloman's correction to my circuit diagram.)

Perhaps if the gain is less than 1.00 the gain peaking will be smaller? I'm being too greedy by driving the shield too hard.

George H.