guard rings

Looks similar to this where they state it's 3ns.

Looks like a first order capacitor charge to me. Don't have any idea how well that actually fits because it's just a sketch. That pulse generator has to be producing a current pulse.

Am 10.04.21 um 18:48 schrieb snipped-for-privacy@highlandsniptechnology.com:

HP/Avago used to write that in the data sheets or an app note.

30 pF @ 0V to 15 pF at some Volts.

I heard that could happen (even in a usable way) with some PIN diodes.

It will need some recovery time also, albeit less. I did some push-pull frequency doublers recently. 100 -> 200 MHz was easy, but 200 -> 400 MHz had "effects". In the end it was good enough, but just barely. It looked like trr. 13 dBm in, still NOS Avago HSMS-282F.

Gerhard

Capacitance doesn't sound like "reverse recovery" to me. I can simulate the capacitance effects but the current spikes are huge in real life. Diode models don't include the fun stuff.

I'll just have to try other diodes.

PINs seem to have soft recoveries, unless you do the Grekhov trick.

Any estimate of peak current? They spec < 100 ps Trr, at 5 mA.

What's the chance somebody there would even know? Like for old designed stuff, do any current employees even know how or why the designs exist the way they do, or undestand the test parameters being used to check parts? There's a vague claim that institutional knowledge is lost in 20 (or 30?) years. In the world of computers, it's way faster than that. I've come across a bunch of products where the manufacturer really has no idea what the specs even are, or how the thing even works.

I was recently looking for a fast single LVDS receiver chip, and tried OnSemi. Their web site is basically useless. If I search for "lvds receiver" I get over 4000 irrelevent hits. I emailed their tech support and they found one quad part.

I found a good single, FIN1002, by googling. It's an old Fairchild part, so maybe the Fairchild people who knew about this stuff are gone. The TI/Burr Brown situation is similar. Nobody knows anything.

Try Octopart. They give Datasheets, Distributors, MOQ, Price. You can give a part number or a general description:

The FIN1002 is slow. 400 MHz instead of 600 MHz:

Findchips cannot find the FIN1002, but gives a substitute that doesn't exist in Octopart:

Not very useful...

I started running across Trr specs for SBDs by Matsushita/Panasonic some years ago. In fact, it was hard to determine whether some parts were schottky or not, just from their data sheets. Even harder when, in some specific cases, the title missed this info, as a typo.

I assume that it simply indicates the parts' performance when inserted into a Trr test jig, with defined If, di/dt, etc.

The part capacitance, and possibly physical package constraints, generate a recognizable current reversal and recovery current peak that can be related to the conventional diode recovery measurement form/profile.

When pulsing parts, it's probably best to characterize performance in-circuit, rather than depend on data sheets to give all the answers, as these are characterized for conventional use.

I'm unaware of guard ring parasitic conduction as ever being considered as a practical occurance in SBDs. Their effects predominate in reverse bias, after all.

RL

Well, you wouldn't have to look farther than wikipedia to get a mention of Trr occurance at higher SBD forward voltage drops.

Higher voltage SBDs, with higher normal forward voltages, have to resort to different methods of guard ring formation to avoid this issue.

RL

Huh? High isn't how I'd characterize Schottky forward voltages, and guard rings are generally to prevent surface tracking, breakdown, in reverse bias. Why and when do they act in forward bias?

I wonder if this ring contributes to Schottky diode reverse leakage?

boB

Possibly but I thought part of the reason for the guard structure was to reduce leakage due to surface effects.

piglet

There's no reference to sources for the Wikipedia information, so I can't help you there.

If guard rings form a parasitic PN junction, they'll act as PN junctions do, when stressed. While their function is to reduce reverse leakage, they also contribute to ruggedness, as reverse breakdown can be localized there, where the structure has modest avalanche energy absorbtion capability. The schottky metal barrier has no ability to break down without popping.

You get the same thing when trying to use mosfets as rectifiers. If the Rds of the fet exceeds the internal parasitic diode Vf, there's reverse recovery to deal with, when Vds is inverted.

I note the forward voltage drop of Larkin's SBD is 0.6V at the rated 200mA - so the parasitic diode will be expected to do something around there.

Note that in the Trr test jig, stresses are manually adjusted to present a peak Irr of 10mA. There's no reference to a controlled di/dt.

Higher voltage parts use a different fab method /metalization material to form guard rings - in some cases multiple concentric guard rings.

RL

I'm looking for single-channel LVDS input, cmos output, and minimum prop delay with modest overdrive. Looks like there are a few parts around 1.5 ns. I'm using these as fast/cheap comparators.

We already use the TI SN65LVDS2 and that looks like about as good as it gets.

I once designed a 6 volt, 120 amp, 3-phase switching regulator using Motorola schottky diodes. The guard rings acted like step-recovery diodes, making 200 amp, 2 ns pulses. It was ugly.

It's making 2 ns 200 amp pulses? So, that's how Highland Technology was founded?

Joe Gwinn

I'm baffled by Fairchild. It seems to be that the korean packaging plant that was setup by the real Fairchild took over the name. No idea where the insides were fabbed or who really ran the show until On bought them. They seemed real friendly with samples in the past, honoring requests for rediculous quantities of stuff though.

Can and do these places run out of masks or lose the 3 ring binder with production steps? Like they get dropped or scuffed up? How would they make new ones? How is the original artwork even stored for a 74LS00 or NE556?

At that lower voltage and higher current, the leakage inductances are pretty dominant - so the quick stuff is often lost in snubbing.

RL

No, but we used to make CAMAC nuclear instrumentation modules, and decided to make our own CAMAC crate with power supplies. They show up on ebay now and then. Search for Highland Camac.

We did get kick-started in fast stuff by a guy from Los Alamos. He showed up one day, threw a LeCroy TDC module on my desk, and said "If you can do this, we'll buy them." I designed an equivalent, LeCroy heard somehow and cut their price in half on the next bid, and Harry disqualified them.