some pics

The superfast ones have been getting easier to find lately--I scored an SD-48 (30 GHz) for super cheap a few months ago. (Not that I actually need it for anything at the moment, but it's fun to have.)

Cheers

Phil Hobbs

Huh, OK. (As always, much thanks) Depleting it doesn't change the resistance that much... That's great.. I think my model of the world needs an update.

I reversed bias the PD today, power was linear with LD current up to ~30+ mA... where I think I bumped into the opamp current limit, (which is fine.)

George H.

Back reflector, (I was reading Donati at lunch today, :^)

George H.

In the late '70s we had a problem with a high current contactor driver. They were failing short, causing all sorts of grief. One particular black eye caught the attention of the *big* cheeses at IBM. One of the failing parts was driving a contactor that controlled inrush limiters on a mainframe. When it failed it turned on, connecting the inrush limiters continuously. They're only supposed to be connected for a second or so. Like a DE locomotive, these resistors are in the upper plenum of the mainframe where the supply fans do a double duty. When it failed, the fans blew billowing smoke all over the corporate offices of a major customer (a bank). That got the attention of the residents of Armonk.

I was tasked to figure out why they were failing. Well, the root cause was that the free-wheeling diodes were placed at the contactors, not on the board (the diode on the driver wasn't even used - the designer didn't want to bring 24V onto his board "just for that"). What interested me, though, was the only parts that failed were Motorola. The Sprague and TI (I think it was) never failed. I used the Tektronix SPS systems, we were talking about recently here, with the 7912 scan converter tube to catch the parts going into avalanche and then secondary breakdown. By digitizing the voltage and current I could easily tell the difference.

The Sprague parts were by far more robust than the others. In the circuit, I couldn't get it to go into secondary breakdown before it avalanched. The TI part would go into breakdown but I had to give it more kick than possible in the real circuit. The Moto part was just horrible. It would breakdown almost by looking cross-eyed at it.

When we looked at the die photos it was clear, why. The Sprague part was as you suggest (lots a little fingers in the base region), while the Moto part was just a big hunk of transistor. TI was somewhere between. When we talked with the manufacturers, Sprague was a bunch of power device guys who did a monolithic design. The Motorola part was done by IC guys and they just scaled the output transistor up to get to the current.

Of course the solution was to drop Motorola from the purchased parts list. No one wanted to fix the real problem (a common issue).

We bought two Nu Focus 25 GHz o/e's, one singlemode and one multimode. Super wide wavelength range, 400-1650. It's hard to find fast detectors in the visible range. Schottky InGaAs, only 0.2 a/w at peak wavelength.

Interesting. That's almost as wide a range as germanium (350-1800 nm typically).

Germanium has really good carrier mobility, but for reasons I don't fully understand Ge diodes are super leaky and therefore super noisy at low signal levels.

Cheers

Phil Hobbs

Cheers

Phil Hobbs

Hm. I probably don't understand just how much area ratio is needed, or is achievable from such a fractal.

The smallest scale factor, of course, only needs to be as small as the epi's lateral resistance dictates. I suppose that wouldn't be

That must be fun... the best "conductor" you can get, really just looks like a mushy dielectric...

Well, how different? Between these two examples, and an average RF part like PD75006? They're all the same (ballpark) voltage, so the doping can't be too different.

Obviously, the basic construction is different, which leads to differences in capacitances, series resistances, and power dissipation. The practical frequency range of these goes from ~10s of MHz (IRF540) to 100s (2N7002) to

These examples are 60, 100 and 65V, so the doping profiles can't be /too/ different.

There are also high voltage RF transistors, and high voltage switching transistors. These are probably an even better comparison, what with Superjunction parts having vanishingly low Crss (well, at modest bias). Of course, all that doping and built-in-potential does other things, too. Offhand, I see ARF1501 working in the 10s of MHz (but perhaps not 100s), and something like SPA06N60C3 will do quality switching at perhaps a MHz, but would be capable of RF much higher if it had comparable power dissipation.

How many Si transistors are actually diffusion limited? I wouldn't think the SPA thing would be circuit limited, based on the parameters in the datasheet (very grossly, 1 / [2*pi*(Qg(tot)/Vg(on))*RS] is around 100MHz). But I doubt it could actually go that fast for other reasons (like package parasitics, but perhaps diffusion as well?).

Tim

Fermi level pinning in Ge screws up the typical, semi-conductor/metal workfunction relation. My very limited understanding of fermi level pinning is that the metal (electron) level goes to the middle of the gap... That might make it leakier.

George H.

For the fireworks show, when those tantalums go off 8-)

An EE friend called me a couple days ago to say hello. Also to ask about tantalums. He had one on a power rail, properly derated and all. One day while he was just sitting there drinking a coffee ... TCHK ... *PHOOOMP*

How boring, parts that just lay there and work.

Would you rather have Democrat parts that just lay there and don't work?

Someone ordered a batch of cheap (multicomp) tants. Pfffft.. the smell of tantalum in the morning. I thought it was put in backwards, till I read the charred label. I set up a string of 10 (35V tants) one blew at 16V, and two more at about 20V.

George H.