small low-capacitance Schottky diodes

We use some SC79s, but that package is actually too small, and tends to be hard to solder. The SOD323 seems about right. I like the SMS7621/BAT15 and SMS3922, which are super-low capacitance but maybe not enough voltage or conductance for you.

PMEG4005EJ is tiny and 500 mA, but it's 50 pF.

You might look at BAS86 mini-melf or the BAT74 dual.

I've characterized an Avago (now Broadcom) e-phemt as a diode. Numbers like 1 amp, 0.4 Vf, 1 pF. That's a big one.

Would use of the (pun) BAER die be of help?

Looking at RF parts, I see 200 mA and under 1 pf

but I'm not clear on package sizes (and unsure what the figure of merit is).

Am 08.12.2017 um 05:49 schrieb John Larkin:

That looks interesting as a low noise mixer or phase detector.

A normal ring mixer has to much ohmic resistance, at least 2 Schottkies in series in the ring, even more for high level. And some even have explicit resistors to generate some bias. (And I do not mean that half-thermal diode resistance)

With Schottkies you do not need to consider anything better than an AD797 as a post detector amplifier.

regards, Gerhard

Those are nice small sc-70 low-capacitance parts. Hmm. the sc-70-3 may be more suitable than a sod-232, and there may be more diode choices in that package.

I'd characterize the mmbd330 as 1.5pF and 20mA, and the mmbd770 as 1.0pF and 10mA. Similar to some other very low-capacitance diodes I have. The bat54 at 6pF is a low-capacitance part, but I wonder if there's a small part with 200 or 300mA of current capability and maybe 10 or 15pF of capacitance at 1.5 volts.

The inductor current is 40mA at our 2.4-volt 10mA design load current, but Rob has been running his tests to 100mA, which means 400mA inductor current. He often sees excellent efficiency from 10 to 40mA.

Yes, I noticed and saved away John's plot last year. It'd be nice to get a logarithmic IV plot for a diode- connected e-phemt. John's ATF-50189 is a rather large one, in a SOT-89 package, and rather expensive. But there must be nice small low-cost alternate choices.

I haven't been paying attention, does the e-phemt have a substrate diode?

But why? Because they have too much dead time?

I'm disappointed that controllers almost exclusively have fixed dead time, it's too much, and none have it adjustable into the negative range.

That's how you prevent body diode conduction and reduce switching loss. No need for diodes.

The LTC3810 application circuit is amusing. They toss in a B1100 across a

I measured. It does absolutely nothig for the thing it's intended to do, and only makes the circuit worse (incrementally so, for the reason you're concerned with).

A lot of LT apps (that one included) are utterly ignorant of EMC, also..

...But ah, you may ask: doesn't dead time make things worse? Only if you haven't matched the loop impedance to the switching impedance. Loop impedance is sqrt(L(loop) / Coss), and switching impedance is Vpk / Ipk. That, and you need a loop time constant pi*sqrt(L(loop)*Coss) comparable to the commutation time.

A long loop time constant does mean more energy storage, which means lower Fsw or a means of "stirring" that energy back into the supply (a "lossless" snubber). But somewhere between an unreactive loop (t_loop t_commutation) there is a loss minima and that is what you are shooting for.

t_loop also gives you the dead time vs. shoot through tolerance, i.e., how close to zero it needs to be to have seamless commutation without incurring excessive losses.

...

As for schottky, they exist on a continuum of junction size and width, with very little distinguishing things beyond that. The main thing left is high vs. low leakage.

So it would seem, to get low voltage drop, you need to take advantage of the low voltage requirement (was the 2.4V figure the PIV??). This alone will increase capacitance (thinner junction), but hopefully it will reduce the voltage drop more.

To that end, you need, like, a PMEG something or other, if they had a

How about using it in switch mode? Two phemts gate driven antiphase, with their drains alternately grounding the ends of a center-tapped transformer winding? That would have zero junction drop and very low Johnson noise resistance.

They don't seem to. Depending on what you connect the gate to, it can be a diode in either direction.

I'd expect that the ei curve would not be logarithmic, but that's a guess.

I use the ATF-50189s as switches, in laser drivers and such. Beautiful part... I hope they keep making it.

That's a thought, ATF-50189, Rds(on) settling in on 1.0 ohms, for Vgs greater than 0.8 volts. Oops, way too high for us!

That's funny. I cannot read John's article. It's here, but impossible to display.

Yes, it should work this way. One would probably try to slow down the transistors somewhat, or one would mix down the noise on all harmonics up to daylight.

There is an article from NIST where they have made a ring mixer from 2N2222 that seems to have excellent 1/f behavior. They write that the 2N2222s work as diodes, but they do work as switching transistors, which is probably the reason why they are good.

I assume that your phemts would be even better at 100 MHz because they are faster and not driven into saturation the bipolar way.

regards, Gerhard

A little gate current really turns it on. There might be some sort of bipolar mechanism kicking in.

I can give it 5V gate drive, but need Rds(on) in the 50 to 200 milli-ohm region.

UJTs are just JFETs that weren't made correctly (no pinchoff). A modest reduction in Rds(on) (maybe 2-3x) is typical, within gate current ratings.

Don't know how that affects speed, but it should slow it down. Like an IGBT, the minority carriers will tail off as drool.

Tim

I mis-spoke in earlier correspondence.

The surviving Philips body size is something they're calling an SOD1608, which measures 0.85mm x 1.65mm and is 0.5mm tall. So it's probably not physical body size that will be the issue.

The 100mA forward drop is 240mV typically and you couldn't ask for a lower forward overvoltage than this body size offers. The capacitance, though, is 60pF typically at 1V.

I'm looking through the lists for a lower current rating, assuming this capacitance will reduce accordingly.

RL

There's still distributor inventory for a variety of interesting NXP parts, like the PMEG2005 and PMEG4005, 500mA diodes. The '2005ej (20V, 66pF) comes in a sod-323 package and '4005et (40V, 43pF) in a sot-23 package. The '4005ej version and some others are still available from Arrow. I'll grab some of them.

The Avago e-phemts start to draw gate current around +0.75 volts, and they have a pretty low max allowed gate current, 12.5 mA I think. The gates are apparently fragile. We poke in the max allowed gate current when we really want to turn them on hard.

So drive it with 5V in series with say 4V/10mA = 390 ohms?

That might slow it down a bit, by our standards. Maybe put a small cap in parallel?

I seem to recall measuring Rds-on, but can't find the data. I'll do that again. The RF boys never spec stuff like that... just Smith charts and tables of s-params. Avago actually gives some DC curves, which is unusual.

Some of the terminals on some phemts are called RF IN and RF OUT and GROUND. They pretend the parts are amplifiers, not fets.

I haven't seen a speed reduction when the gate conducts. Maybe it's not a bipolar action, but Id jumps abruptly with gate current, more than I would have expected from the small increase in Vg.

I recently discovered that a guy lives across the street from me who is an engineer at Avago, for these very same parts. I'll ask him. He must have a hell of a commute down to Santa Clara from here.

Bob Pease used to live close to here too. He did a lot of driving.