MOSFET variable resistors--not like JFETs

Are you sure? Seems like I demo 'd that here with that very fet. Looking at your drawings, I believe you usually AC couple to the fet. I'll see if I can find my setup. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

      Remember: Once you go over the hill, you pick up speed

Pretty sure. I tried 50% first, then gradually cranked it up. The best was using two FETs in series, with gates in parallel and AC coupled to the drain of the top FET--abou 150% feedback. It made over 20 dB difference to the SINAD.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

I did some playing with NEC phemts as variable-resistance attenuators. They are very fast but nonlinear for swings above a few 10s of mv. I didn't try ac drive into the gates, which would probably defeat the speed virtues of these parts.

Too bad so many RF attenuators aren't suited to operate all the way down to DC.

John

Yo've got to be carefull with capacitive loading, as the feedback inceasingly enforces contstant current behavior, not resistive.

RL

Not at 2 kHz.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

...or more accutately, small signal amplitude and low dv/dt.

RL

Couldn't find what I previously tried, so I threw together a quicky behavioral controller around a 2N7000...

Seems optimum when adding 0.45*V(DRAIN) to the control voltage.

I swept IDRAIN from -100mA to +100mA stepping VCONT from 4.5V to 10V in 0.5V increments.

Now that I go back and look at your drawing I realize that you didn't ADD drain effects to the control voltage. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

      Remember: Once you go over the hill, you pick up speed

Oooops! Yes you did "add". Don't know why the discrepancy. Maybe the cascode? My experience with cascodes suggests VGupper needs to be about 133% (or greater) of VGlower. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

      Remember: Once you go over the hill, you pick up speed

Okay, so what do you think is actually going on? ISTM that the quadratic model doesn't apply in the low-ohms region with these devices, but I'm not entirely sure why, nor do I know how to calculate it analytically, which would be comforting.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

Thanks. I tried it both ways, as shown in the two sketches. The double stack wasn't really intended as a cascode, just resistors in series to get a higher resistance range. Doubling the resistance also makes the drain swing increase, which made it simple to get average dV_GS/dV_DS values above 100%. It seems to continue to improve up to about 150%.

The application was in the R_EE' compensation circuitry of an improved laser noise canceller, but I'm not actually going to do it this way--I'm going to use vernier-connected 1k dpots and op amps instead. The 1k variant of the AD8403 gets me better bandwidth anyway.

Just thought it was an interesting puzzle--one more FET mystery to add to the pile. Bipolars surely do follow their simple models a lot more closely than FETs do--for both noise and DC. If it weren't for the BF862 and some of the really quiet RF devices, I'd probably never use another discrete small-signal FET.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

[snip]

FET's don't follow the "school-book" equations very closely at all... particularly modern FET's with short channels. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

      Remember: Once you go over the hill, you pick up speed

The phemts, like NE8509 and FP1189, are astonishing as switches. You can drive the gates from off to enhanced in a few hundred ps with an EL or EP gate, and they seem to switch as fast as you can drive the gate. They also seem to be much more consistant than messy old jfets.

John

I'm only used to optimizing or fighting the effect in power circuits.

A common inrush limiting technique for capacitive loads is to reflect the load's dv/dt as a controlling fet's Crs.

RL

I'm going to be using some of those gizmos soonish--I expect to have a gig coming in to design a ~100 MHz preamp for detecting single molecules going by. I'll have about 60 electrons per molecule to work with, so the transistors are going to be soldered straight to the microprobe tips. Should be interesting. (Hopefully I'll be able to argue the client into slowing the molecules down a bit so I have more charge to work with.)

First milestone, as always on this sort of job, is to nail down a specification that both the client and I can sign off on.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

gives?

Hey, I wanted that job... but I never found the time to pursue it. You have all the fun.

All the compound semi fets have a *lot* of gate leakage current. Like microamps. The little ones get down to absurd noise figures in tuned RF circuits, like 0.4 dB, around 28K equivalent temperature.

Let's see, 60 electrons into 1 pF is 10 microvolts. For a 100 MHz amp, that's a noise density of about 1 nv/rthz for a 1:1 s/n ratio. Better slow them down.

John

.....

Phil, your numbers sound just like for an ion trap mass spectrometer. I worked in one of them some time ago, at LBNL. We could measure the speed and charge of the same ion about 100 times before losing it, therefore improving the measurement S/N.

Thanks, Jure Z.

gives?

This is actually a different outfit--it's one I've worked with before, in the Far East. The gate leakage and ionic current issues are going to have to be ironed out in the spec resolution phase, along with the ridiculously low available charge.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net

I designed a really nice (I think) preamp for a fourier-transform mass spec. The company (IonSpec) was acquired by Varian, and Varian was acquired by Agilent, and they killed the project. Pity.

An FTMS can detect ballpark 100 molecules orbiting in a magnetic field, in the very roughly 1 MHz range, with a mass resolution around

1 PPM. I wanted to detect one molecule, which might just barely be possible if the vacuum is good.

John

Traps are cool.

I asked the client about making the molecule cross the probes many times, but it's happening in solution, so there's no simple way to make the signal repetitive. (Brownian motion would mess up the phase, for a start.)

I'm expecting to do a few days reviewing their experiment and making suggestions, before we get to the actual circuit design. The original request was for 100 MHz BW, 1 nA peak current, 20 dB SNR. That's within

10 dB of the shot noise of that 1 nA, which in that bandwidth is probably impossible. (Not that there's anything wrong with that...) ;)

If we can figure out a way to do it, it'll wind up being a nice FET probe, anyway!

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net