Max current for a JFET connected as diode (2023 Update)

Quantity, not quality.

See AoE3 page 361. The idea is to use depletion fets as current limiters. The LND150s add a couple of k equivalent resistance, but there are hunkier parts, like DN2530.

Ixys make a self-protecting SSR that can be fun as a current limiter.

Even with reduced beta, f_T is often higher in reverse mode, I've heard.

Cheers

Phil Hobbs

Getting a 12-GHz f_max transistor to oscillate at 2 kHz requires a real gift. ;)

Thermal, I gather?

There's a TI app note on using GaN FETs as ideal diodes. (GaN FETs are also 2-D electron gas devices, like pHEMTs.)

Cheers

Phil Hobbs

On a sunny day (Fri, 15 Apr 2022 10:58:34 -0400) it happened Phil Hobbs snipped-for-privacy@electrooptical.net wrote in snipped-for-privacy@electrooptical.net:

Looks like it is oscillating in the GHz range sort of self-quencing. bottom wave is full of RF? Wire length -> wave length etc, maybe resistor is inductive too.

2K was the sinewave generator. The oscillations are the fuzz on the negative swing.

I was recently playing with ideas about an ideal full-bridge rectifier. There is an LTC chip to do that with four mosfets, but I need more voltage than that chip can handle.

EPC ganfets as diodes might be a compromise on Von, without elaborate gate driver stuff.

The Q that you'd need for the aliased waveform to be that clean would be fairly impressive.

Cheers

Phil Hobbs

Ah, right, didn't pay enough attention.

A transformer might be a win--you save a volt when the on-time is short enough, and don't lose anything much when it isn't.

Cheers

Phil Hobbs

It was a breadboard with dangling wires, not very tight.

Is that SiGe? I'd like to get a good SiGe to play with.

That BFU does look nice. Lots of voltage, which we need sometimes.

Yeah, I just didn't read it right. Moving on, nothing to see here, folks,.... ;)

Cheers

Phil Hobbs

No, it's follow-on to the late lamented BFG series. The BFG403 was even nicer than the BFG25A, which was nicer than the BFT25A, which was nicer than the OG MRF9331 that Win and Paul enthused about in AOE II. (I have a reel of BFG403s too.)

The SiGe you want is either BFP640 or BFP650. The BFP780 was recently discontinued as well. Get your transistors while they're hot, folks!

Actual measured parameters: beta ~ 250, V_A ~ 300 V at AC, essentially infinite at DC due to thermal changes of V_BE.

You will definitely want to put a BLM15BA050SN1D bead in the base though!

Cheers

Phil Hobbs

The 'max rating' is in reference to device deviations from its specifications, and zener ratings assume you care, critically, about the threshold voltage, not leakage.

Most of the use-a-transistor-as-diode discussion is about typical actual leakage, but NOT in a test scenario where breakdowns and wacky bias conditions occur, so some of the observations won't be complete in the sense of including age and 'abuse' conditions.

If it were practical, I'd consider that you want an input resistor, and a clamp-to-ground, i.e. antiparallel diodes, into an inverting node of an op amp; since the inverting node is always pseudo-ground, those clamp diodes have zero applied bias, and leakage is a near non-issue (unless they're photovoltaic?).

If it's coming from outside the box, you _do_ want that on the inputs. In high performance applications you often have to stand on one leg a bit to get there. For instance, the protection diode can be an RF Schottky plus a lightly-biased zener, to get both low capacitance and good clamping. Other times, I bootstrap the protection circuitry such that a hard fault gets caught but the capacitance and leakage mostly go away.

A simpler example is protecting a noninverting buffer using a series pairs of Schottkies from the input to both supplies (4 diodes in all), biased with a resistor to the output.

The diodes connected to the input have nearly zero volts across them, and hence nearly zero leakage, but transients get dumped to the supplies harmlessly. (One might need a couple of series resistors to make this really robust.)

Cheers

Phil Hobbs

I have used the 2n4117 or mmbt4117 trick in the past and learned that they are only "diod-y" up to about 1-2mA forward current. After that they look more like resistors.

All diodes get ohmic at some current, and often have a zero-TC point around the transition. That can be useful.

You get low leakage from small junctions, and small junctions get ohmic at lower currents than big ones. That's one reason that small-signal RF transistors tend to have low leakage.

If you connect collector to base and use that config as a diode, you get better, beta-enhanced, forward conduction and not extremely worse leakage, but low breakdown voltage. Connecting b to e is interesting too; that gives reverse-beta gain and more voltage.

Idmax could well blow out the gate. Try it.

Yup. (I have a couple of reels of BFT25As.) I haven't tried using the reverse beta for anything--interesting idea.

One time I was working on a current-feedback loop working in the low nanoamps, with a few BC-connected BFT25As to provide emitter degeneration for the main device.

I found that the bandwidth tanked below a few nanoamps, because the BC-connected transistor is actually the world's simplest feedback amplifier, and (despite being a 5-GHz transistor) it was running out of f_T.

Using the BE diodes by themselves fixed it.

Cheers

Phil Hobbs

I have a couple of hundred too. There is probably a decent replacement somewhere. Reverse beta is about 4.

Phemts and GaN parts make interesting diodes too.

Jfets are terrible diodes, and the PAD parts are jfets.