TD amp

.

Only $97.

Joe

I was at Haltek (or Halted?) some years ago, one of the now-gone silicon valley surplus dealers. The real estate values drove all them out.

They had a big bin full or TDs and didn't know what they were so wanted 10 cents each. I bought a bunch, but should have bargained for the lot.

I think that Steve Jobs offered Halted some Apple stock in return for some parts, and they refused.

The germanium TD fab process was insane.

I think someone still makes germanium "back diodes", basically a variant of a TD.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

I have a tunnel diode from ebay somewhere

Yikes. I paid 10 cents for mine.

But modern ICs are mostly faster these days.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Lower power, too. Old Ge TDs have about 200 pF junction capacitance, so to get any speed you need a very high peak current, at least 100 mA.

Cheers

Phil Hobbs

A few of them got risetimes in the 10s of ps, but as you say, at high currents. And all you got was a few tenths of a volt step.

The best bang per buck these days, for sub-ns steps, is probably some cmos gates and some laser drivers. Or SRDs.

Somebody orter write a book about making fast pulses, a modern version of Millimicrosecond Pulse Techniques.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Step recovery diodes have the problem that you have to set up the forward current, and then reverse bias the diode, and you don't get the step-recovery until you have swept out the charge carriers, and that takes a while.

ECL was much more predictable. You needed wide band transistors to get any sort of voltage swing out of the transitions, but it was all a lot more designable.

Fast CMOS puts a lot of noise on the power rails, and that can be a real nuisance. I used a Percival distributed amplifier to get a pair of complementary 800psec wide +/-5V pulses which was reliable enough to ship, but a simpler circuit with bigger transistors gave us 500psec, so we settled on that.

I suspect that the Percival circuit would gone faster with slightly bigger wide-band transistors, but 500psec was good enough.

Distributed amps are great for making big fast swings, but

They are AC coupled Inefficient Hard to bias Cost hundreds of dollars each

This uses an HMC659 distributed amp (slow, only 15 GHz) to make 8 volt pulses into an e/o modulator. The chip costs us $300 each.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Took, not too. My typing is really bad lately. I blame the chair.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Yes, I find it is so difficult to get it at the right height for my toes to reach the keyboard.

Ctrl/alt/shift should be pedals, like on a pipe organ.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Mine wasn't.

Mine wasn't.

Mine wasn't.

Mine certainly didn't. It was cheaper than the simpler big transistor device that replaced it.

None of your claims are backed up by the Wikipedia entry

Not the most informative of images, and the data sheet wasn't all that informative either

"Right"? You mean that your second attempt worked well enough to let you sell it. Granting bizarre the claims you've just made about distributed amplifiers this is improbably quick. The Analog Devices part must be well designed.

The design is available for purchasing or licensing, in which cases higher resolution images are available.

Yes. Part of the iteration came from change of philosophy from the customer.

We had to invent a new way to bias distributed amps for pulse work. Like most RF parts, the data sheets and appnotes and eval boards assume RF or dc-balanced digital data, 8b10b or some such. And the data sheet abs max numbers are not to be trusted.

It's interesting to bench test $300 parts to see what the abs max numbers actually are.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

There are an infinite number of ways of making a distributed amp. You didn't invent a new way of biasing distributed amps - you invented a new way of biasing the part Analog Devices sells as a distributed amp

Your reading comprehension doesn't seem to be all that good.

Senior people do seem to take pleasure in blowing up expensive parts. Mostly it's because they don't know exactly what they are doing and don't want to spend the time it would take to find out, but some of then have a problem admitting that they don't know exactly what they are doing, even to themselves.

My bias loop would work on other parts, and the HMC659 really is a distributed amp.

When they say ABS MAX VOLTAGE the words are pretty obvious, except that it's not clear which voltage they mean. Most RF part data sheets really mean power supply voltage, where the part output swings to 2X the supply.

Sine waves are boring.

At $300 each, and difficult to desolder and replace on a board, it's not as much fun somehow.

Yikes. Word salad.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Your sentence processing skills aren't great. I wasn't denying that that the HMC659 was a distributed amplifier - merely pointing out that it wasn't the only one around. Your new biasing scheme may work for some of the others, but perhaps not for all of them.

But you can decompose more complicated repetitive waveforms into sums of sine waves. Paul Dirac exploited that.

It's a fairly explicit insult - people with better language processing skills will get the message. Of course it is not explicitly aimed at you

- the former boss I had in mind was in many respects an excellent boss, if a menace in the lab. He died more than ten years ago.