Ping Hobbs: JFET Noise

If you have BF862-class or better JFETS, their flatband noise is of the sam e order as a good BJT stage, so you lose 3 dB or so that way.

For a single-ended circuit such as an ordinary TIA, my fave is a bootstrapp ed shunt-feedback loop on the JFET, running into a bipolar op amp. (The shu nt feedback senses I_D and gooses the tail current source to keep it consta nt.)

The benefit there is that the local feedback gets you a really low Zout sta ge with the same noise as the FET itself. With a capacitive transducer such as a photodiode, that gets rid of the omega C_d e_N noise current, almost irrespective of the BJT noise.

For a true diff stage, something like a Sziklai arrangement might work (NJF ET'S and PNP diff pair), where local feedback prevents the noise addition?

(I haven't thought about this very hard, so definitely consider this a mini mally-baked idea.)

Cheers

Phil Hobbs

Does it not depend on on the source Z?

** Yes, but the app requires high impedance - something JFETs excel at - with no input current required for operation, there is virtually no current noise.

Same goes for vacuum tubes which are still used in some condenser mics where the signal source is a 50pF capacitor.

.... Phil

Does what?

Cheers

Phil Hobbs

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203 Briarcliff Manor NY 10510

hobbs at electrooptical dot net

Phil, I'm having trouble visualizing that configuration. Can you send me a sketch? Thanks!

...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 

             I'm looking for work... see my website.

d

I've just finished bringing up my first proprietary product based on the id ea--a TIA that's shot noise limited above about 10 nA in a 1-MHz bandwidth. Thus I'd rathher not share the schematic, but here's the gist in a bit mor e detail.

You make a JFET follower with a BFS17ish tail current source in the usual w ay, except that you bypass the daylights out of its emitter.

That makes the current source much noisier, but that goes away when the loc al feedback loop gets closed. Then you bootstrap the JFET's drain with ano ther BFS17ish follower.

At this point, all the capacitances of the FET have been bootstrapped away, and its operating point doesn't depend on the instantaneous signal. All that's missing is a really low output impedance, which makes the follower a ble to bootstrap some gross amount of input capacitance with a gain like 0.

999. Then (key step) you put a resistor in the top transistor's collector, and f eed that back to the base of the bottom one (the tail source), so that the FET's bias conditions become independent of the load as well as the input s wing.

My current one has Zout of a fraction of an ohm, with a bandwidth of 500 MH z or so.

(I'm using a built-up circuit, so I get to use 10 ms time constants, but y ou have to DC couple stuff.) ;)

Cheers

Phil Hobbs

ar

end

idea--a TIA that's shot noise limited above about 10 nA in a 1-MHz bandwidt h. Thus I'd rathher not share the schematic, but here's the gist in a bit m ore detail.

way, except that you bypass the daylights out of its emitter.

ocal feedback loop gets closed. Then you bootstrap the JFET's drain with a nother BFS17ish follower.

y, and its operating point doesn't depend on the instantaneous signal. Al l that's missing is a really low output impedance, which makes the follower able to bootstrap some gross amount of input capacitance with a gain like

0.999.

feed that back to the base of the bottom one (the tail source), so that th e FET's bias conditions become independent of the load as well as the input swing.

MHz or so.

you have to DC couple stuff.) ;)

That's very similar to a White Cathode Follower (except for bootstrapping t he FET drain):

kevin

the FET drain):

g

Yes, that's the basic idea except that the third device improves it out of all recognition. The other shunt feedback thing I use is a PNP wraparound o n the bootstrap's drain resistor (like the usual PNP booster trick for 3-te rminal regulators, but with the drain bootstrap as well).

That works very well too, but is less flexible and harder to stabilize. You really can get gains of 0.999 driving some fairly serious capacitance, alo ng with sub-nanovolt noise, which makes a pretty nifty bootstrap.

Cheers

Phil Hobbs

the optimum design for low noise depends on the source Z.

But I see the question was addressed in the OP... I shouldn't post when 1/2 asleep. M

How much it depends sorta depends. ;)

The BF862 has subnanovolt noise, only a couple of picoamps of gate leakage (at 25C and ~ 2V g-s), and only a few picofarads Cdg. When you bootstrap away the capacitance, and maybe put a few FETs in parallel, you get a really nice wideband stage that works very well over a very wide range of impedances.

You can snoop V_gs with a FET op amp and a big resistor, and adjust the DC tail current to exactly I_DSS if you like, but it degrades the low frequency noise pretty badly. In this product I've got the FET biased near its zero tempco point, with a set-and-forget offset adjustemnt. Then the photodiode leakage is the dominant drift source.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

This is a toy diff amp made with Sziklai-type devices.

It has the drive of the bipolars and the noise of the JFETs.

Does need good PNPs though.

BTW "PUHFARRY" is the bare die model for the Intersil HFA3xxx PNPs.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

I'll give that a try. Just corresponded with X-Fab and they now have a circular-geometry JFET in process XB06 (which disconnects you from substrate issues).

I'm also amused that PSpice is again supported... for awhile it was decided no one could design a chip with PSpice... except that _I_ have... hundreds ;-) ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 

             I'm looking for work... see my website.

Any theories about optimum depth of pinch-off and noise? ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 

             I'm looking for work... see my website.

Some JFETs are more quadratic than others. ;)

The BF862's transconductance looks more like a laser diode curve than a parabola--zilch up to some threshold, and nearly constant slope above there. The noise depends basically on the transconductance, being ideally sqrt(2/3) of the calculated Johnson noise of 1/g_M iirc.

So I think it depends a fair amount on the actual behaviour of your FETs.

Since I don't have the luxury of matched devices to track out these variations, I generally sort of pour on the coal--8 to 10 mA I_D for AC coupled things, and near I_DSS for DC-coupled things. However, they work pretty well down to 3 mA or thereabouts.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

That's my whole chip power budget :-( ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 

             I'm looking for work... see my website.

Well, no big surprise. But you don't have to run the BJTs very hot to get improved transconductance overall--running the JFETs and BJTs at about the same current works fine. Your devices are probably much smaller anyway, and of course the parasitics are smaller.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Phil, could you explain for non-native speaker what you mean by "bypass the daylights out of current source's emitter"?

I tried to reconstruct an LTspice simulation after Phil's description, just for fun. I never really grokked why one would want unity-gain amplifiers in a low-noise situation, but indeed, it's easy to tweak the gain to unity, the noise is sub-nanovolt and the output impedance is below 1 ohm. I did not bypass the tail current source's emitter though. With a little more work, it could make a nice FET probe pre-amp.

The .asc file is included below.

Jeroen Belleman

Version 4 SHEET 1 3304 808 WIRE -80 -80 -112 -80 WIRE 128 -80 0 -80 WIRE 288 -80 128 -80 WIRE 128 -32 128 -80 WIRE 288 -32 288 -80 WIRE 128 80 128 48 WIRE 176 80 128 80 WIRE 352 80 240 80 WIRE 128 96 128 80 WIRE -272 160 -320 160 WIRE -144 160 -192 160 WIRE 80 160 -144 160 WIRE 32 224 -80 224 WIRE 288 224 288 48 WIRE 288 224 32 224 WIRE 32 256 32 224 WIRE 288 256 288 224 WIRE -80 272 -80 224 WIRE 128 304 128 192 WIRE 224 304 128 304 WIRE 128 336 128 304 WIRE -80 384 -80 336 WIRE 32 384 32 336 WIRE 32 384 -80 384 WIRE 64 384 32 384 WIRE 288 400 288 352 WIRE 352 400 352 80 WIRE 352 400 288 400 WIRE 384 400 352 400 WIRE 32 448 32 384 WIRE 128 448 128 432 WIRE 288 448 288 400 WIRE -80 560 -112 560 WIRE 32 560 32 528 WIRE 32 560 0 560 WIRE 128 560 128 528 WIRE 128 560 32 560 WIRE 288 560 288 528 WIRE 288 560 128 560 FLAG -320 160 0 FLAG -112 560 0 FLAG -112 -80 0 FLAG 384 400 out FLAG -144 160 in SYMBOL njf 80 96 R0 SYMATTR InstName J1 SYMATTR Value jbf862 SYMBOL npn 64 336 R0 SYMATTR InstName Q1 SYMATTR Value bfr92a SYMBOL npn 224 256 R0 SYMATTR InstName Q2 SYMATTR Value bfr92a SYMBOL res 112 -48 R0 SYMATTR InstName R1 SYMATTR Value 220 SYMBOL res 272 -48 R0 SYMATTR InstName R2 SYMATTR Value 33 SYMBOL res 112 432 R0 SYMATTR InstName R3 SYMATTR Value 50 SYMBOL voltage 16 -80 R90 WINDOW 0 -32 56 VBottom 2 WINDOW 3 32 56 VTop 2 SYMATTR InstName V1 SYMATTR Value 10 SYMBOL voltage -96 560 R270 WINDOW 0 32 56 VTop 2 WINDOW 3 -32 56 VBottom 2 SYMATTR InstName V2 SYMATTR Value 10 SYMBOL res 272 432 R0 SYMATTR InstName R4 SYMATTR Value 3k3 SYMBOL res 16 432 R0 SYMATTR InstName R5 SYMATTR Value 1k7 SYMBOL res 16 240 R0 SYMATTR InstName R6 SYMATTR Value 10k SYMBOL voltage -176 160 R90 WINDOW 0 -32 56 VBottom 2 WINDOW 3 32 56 VTop 2 WINDOW 123 60 56 VTop 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V3 SYMATTR Value SINE(0 100m 10meg) SYMATTR Value2 AC 1 SYMBOL cap 240 64 R90 WINDOW 0 0 32 VBottom 2 WINDOW 3 32 32 VTop 2 SYMATTR InstName C1 SYMATTR Value 1u SYMBOL cap -96 272 R0 SYMATTR InstName C2 SYMATTR Value 1u TEXT -448 -144 Left 2 !.lib Philips.njf TEXT -448 -96 Left 2 !;tran 1u TEXT -448 -56 Left 2 !.ac dec 100 1k 1G

"the (living) daylights out of" is an english idiom to do something to excess in an aggrssive manner.

--
This email has not been checked by half-arsed antivirus software