JFET used as a diode

You can still get the Skyworks ones. Broadcom did an NXP and killed off all the old Avago ones, as you know. But on the plus side, CEL has a whole bunch of new ones!

Yup. The Footprints multiplexer was built of 96 ordinary display LEDs from Chicago Miniature Lamp. Low femtoamp leakage near zero, and less than 50 fA from -5V to + 0.5V bias.

The cute thing was putting them under a plastic cover (painted black on the outside and white on the inside, then illuminating them with four other LEDs with PWM drive. That gave me an adjustable 0-5 pA bias current with reasonable uniformity, so that I could use a diode MUX with an inherently AC-coupled transducer, namely PVDF pyroelectric film.

Capacitance was much higher than a BFT25A's, of course.

Cheers

Phil Hobbs

Why not short base-emitter and expliot the reverse beta? I would expect that should reduce the base current by 5 so

It depends if you care more about a 40-mV change in V_BC or about super low leakage. The CB junction is the low-leakage one.

I have a production design (for a safety system) that has a TIA with a

Cheers

Phil Hobbs

Huh, linear down low but then changing to logarithmic. That's a cute trick.

George H.

There's leakage from collector to emitter?

I was thinking it would allow a part with 5mA I_B to pass 20mA safely.

That's probably correct; the Rbb would also be slightly less troublesome. Since emitter current is zero when it's open, there's not a lot of reason not to connect it to the base (it'd float near that potential anyhow, if one were inclined to let a terminal float).

Because the emitter is highly doped, and the B and E share a surface of the silicon, B-E junction leakage is higher than B-C. The prescription for C-B shorting is best at lowering the parasitic forward resistance, not the reverse current. They both give near-ideal diode behavior, but in different ways (different I-V quadrants).

There was a book called, "What They Don't Teach You in Harvard Business School." This is like, "What They Don't Teach You in AoE."

It's also faster (near zero t_rr), because it's never saturated.

Tim

Falls apart at low current though, on account of beta rolling off. The 'diode-connected transistor' is the world's simplest feedback amp. I've seen them oscillate, too--that was a weird one to debug.

Cheers

Phil Hobbs

I'll never get any sleep if you keep inventing horrors like that.

The Siliconix PAD1 asnd JPAD10 were exactly that. They were part of range that also offered higher reverse currents at lower prices.

Calogic now seems to make them

I did a proof of concept for a long range IR remote control for Samsung some years back--it got them a factor of three increase in range, which they were happy about. The main issue, interestingly, was avoiding the forest of optical spurs put out by electronic-ballast fluorescent lights. The biggest headache was the spectral lines from the mercury emission, which (near the ends of the tubes) have important harmonics of

The point of this story is that the low frequency bias loop used five diode-connected BFT25As in series, to give a wide operating range without adding much shot noise. It worked fine at high currents, but just crapped out below about 5 nA--the diodes stopped behaving like diodes.

It turned out that the 5-GHz BFT25As were slowing down so much at those low currents that they ran out of gain at 1 MHz. I switched to using the CB junctions instead, and the circuit started working.

With a sub-optimal layout, a diode-connected BFT25A can oscillate above about 100 uA I_C.

Cheers

Phil Hobbs

Here is the original readable version:

Pease Bounding and Clamping

Phil,

I sent you a copy earlier, but you didn't respond. I assume you didn't receive it or you are too busy.

There is a lot of useful information in the file that may be of interest to others, so I decided to go ahead and upload it and post the url here.

I didn't receive it--thanks for persisting.

Cheers

Phil Hobbs

Thanks for the reply. Just so you don't start worrying about other emails y ou may have missed, I'm pretty sure the problem is on my end. This has happ ened once or twice before where my email server simply refused to send emai ls to particular addresses. I don't know why.

Steve Wilson

I did some work on Fourier Transform Mass Spectroscopy. The old electronics had been design by a couple of chemists, and the spectrum was dominated by junk from fluorescent lights. The 30 dB excess wideband noise floor did make that look not as bad as it really was.

Sadly, Agilent acquired Varian and killed off both the NMR and FTMS operations.

Yeah, that's why I sell a lot of photon budgets. You don't know how you're doing till you know how good it _could_ be.

Cheers

Phil Hobbs

I had about convinced myself that we could detect a single molecule in orbit in the machine, but then Agilent killed the whole thing.

Once we did single molecules, that would be the end of it. No point in getting better from there.

The 'diode-connected transistor' is the world's simplest feedback amp. I've seen them oscillate, too--that was a weird one to debug.

ng

ch

he

of

hat

se

ve

"Molecule" is rather broad concept. Being able to detect a helium molecule

- actually an ionised helium molecule - is nice, but the two stage mass sp ectrometer that I worked on (briefly) in 1992 was supposed to detect ionise d protein molecules, and their fragmentation products.

Very precise molecular weights - good enough to allow you to keep track of the mass defects in the atoms in the molecule - were promised (and probably delivered).