TI to acquire National Semiconductor

I doubt if he wants to come back. By the tone of his articles in ED, he sounds like retirement suits him.Still damned crotchety! Tom

We drove past his house last week. All the various rusty/dented VW beetles and vans are gone from his front yard and the surrounding street. What's going on here?

John

Does he go to the DeAnza flea market? It should be this Saturday. [I just do the Livermore one. DeAnza has no deals. ]

Pease was infamous for selling his books out of the boot on a bug.

Don't know... I haven't been to the flea market since Foothill closed down. It was getting junky; I think ebay killed it. I'll be skiing anyhow... almost last chance for the season. Although I have skiied on the 4th of July...

Yup, or a card table. He also sold the books, his and Jim Williams', out of his house. I have the autographed "Troubleshooting" book.

John

I checked with the guys at work. They tell me the ADCMP565 is a replacement for the MAX9690. I believe it has better ESD protection, we used to use 9690s in some test fixturing and they would blow up if you looked at them crosseyed.

It's still in production, and available.

Disclosure - I work for Analog Devices, in the group that did the ADCMP565, but aside from participating in design reviews I had nothing much to do with it.

Steve

I use the ADCMPs a lot these days, but none are drop-ins for the 9690. The 9690 was a classic single ECL comparator in SO8, like CMP01.

Maxim had them fabbed by some subcontractor in Minnesota, I think. There was a problem, some sort of charge migration, that made them all fail after a year or so in oparation. Maxim applied the easiest available fix: discontinued the part without notice.

They replaced it with the "drop-in equivalent" MAX9691, which is a

For modest accuracy, modest speed comparator apps, I use LVDS line receivers. For good stuff, ADCMP565/567/582. The 582 outputs sub-40-ps edges.

John

The fab was VTC. Plenty of parts were orphaned when that fab was shut down. Probably not charge migration since the part was bipolar.

Pinch resistor? The symptom was that the parts would fail when they got hot, and the fail temperature crept down over time, hitting room temp after a year or so. Baking at 100C for a couple of hours would fix them for a while by pushing the fail temperature back up.

The failure was that an output transition would happen normally or not happen at all, erratically, with higher duty cycle of failure at higher temperature. Sounds like something gone wrong in the latch enable circuit. I didn't use latch enable.

It confused me for a while, as my temperature cycling testing was also annealing the parts!

Any idea what physics could cause that?

John

I used to like VTC a lot--they had a beautiful 70 MHz OTA, for instance. I still have some left. (TI makes the OPA860, which they laughingly call an OTA, but doesn't have enough pins and it's nonlinearity is horrible.)

Cheers

Phil Hobbs

...its nonlinearity is horrible.) (It was an editing scar, not a braino, but is too illiterate-looking to leave alone.)

PH

The VTC fab was resurrected as PolarFab. I ran several of my chip designs there... many years ago now. I don't know if they still exist. ...Jim Thompson

Ya sure, ya betcha:

Heating up parts is always a good way to get rid of stored charge. This is routinely done after some types of failure analysis or focused ion beam type repairs. [Wrap part up in Al foil and bake in oven. No need to apply power.] But only for MOS as far as I know. I'm not really sure why that would effect a bipolar part. You have so many diode junctions in bipolar that I don't see a place to store charge. That is, the charge has a leakage path.

Generally not a problem in digital mos, but analog can often leave nodes that can get charged. You will see diodes or turned off fets acting as diodes in the layout to insure some nodes don't get charged up, or more correctly written they get discharged when the supply rail goes to ground. There are also "antenna rules" to insure some devices don't get charged during processing.

Hard to believe, but somebody wrote a wiki on this:

Maybe charges trapped in the oxide layers?

Almost certainly, but I can't imagine how that would affect a bipolar, ECL comparator. I was suspecting something like a pinch resistor or some other fet-like structure. But I'm not a semiconductor device person, so I thought I'd ask.

I thought the replacement 9691 part, with its antiparallel diodes cross the inputs, was really weird. It simply wouldn't work in any of our existing 9690 circuits.

John

That happens a lot. Philips (NXP) came out with a supposedly superior chip for the old 74HC4046 PLL, I think they called it 9046 or something. I believe it claimed less phase jitter. Long story short it had some peculiarties that made it not work in our otherwise rather mundane circuit, we went back to the 4046.

tion,

2. > >>>>> The 9690 was a classic single ECL comparator in SO8, like CMP01.

all

-ps

You can effect the matching of the comparator input pair with large reverse biasing, hence the clamp. That is if you yank them apart, the input referred offset can be effected.

Lots of ugly stuff can happen in IC bipolar comparators. They are not as simple as they look in the "representative schematic". Phase reversal is the most common bug.

Sure, but how am I supposed to use a comparator that leaks milliamperes of current between its pins? It shure wrecks ramp-type timing generators, and screws up things like an opamp driving one input and a dac driving the other.

Ironically, Maxim already had comparators without the diodes, that would stand lots of differential input, when they cut over to the

Well, don't do that.

John

It's really only a problem if you want to compare voltages that are different from each other.