Re: China resumes export of Nexperia chips, with some restrictions

That doesn't automatically make them good ideas. People buy stuff from confidence tricksters, but you have stayed in business for quite a while so your products clearly work well enough.

But there's a difference between working well enough, and working really well, as I got to find out at Cambridge Instruments when I had to rework bits of the their GaAs single-crystal pulling machine because some of the parts had gone obsolete in the ten years since it was first put together. There was a lot of bad design - or non-design - in the bits I had to modernise. It hadn't stopped the machine from making 95% of the single crystal GaAs pulled on the western side of the iron curtain.

I invented my current mirror version of the Baxandall class-D oscillator to get a better sine wave source for the LVDT-based weighing system. The output amplitude was lot more stable than what we'd got from the system we had to replace. Nobody noticed. Getting rid of a 741 and its pop-corn noise was a lot easier, and that did get noticed.

No idea. Popcorn nose is down in the microvolts.

The problem was that the 741 data sheet was really old, and didn't include any pop-corn noise specification.

As semi-conductor processing got better, it became possible to keep it pretty low, and when things went wrong and bunch of more modern op amps failed to meet their pop-corn noise spec, they'd get packaged up as

The problem with the crystal puller was that it used a couple of kilowatts of noisy RF heating to keep the GaAs surface exactly at its melting point - 1260 Celcius. If the temperature was exactly right the crystal we were growing got steadily heavier and we weighed the crystal continuously to work out whether the crystal was growing at exactly the right rate.

Putting the weight signal through a 741 meant that the pop=corn noise turned off the heating completely for about a minute at a time and turned it on hard for about the next thirty seconds or so.

Putting in a better op amp meant that the heating ran at around a fairly steady 30% of full output all the time. It took a a day or so to grow a crystal, and the environment was lot more peaceful after I'd done that. People did notice.

If I'd been aware of the problem, I'd have put in something seriously better, like an LT1028, which I'd designed into several circuits by that time, but I'd merely chucked out the 741 because it was poorly specified.

My definition is that important people buy our stuff because they can't find anything better.

The new thing involves burst digitizing thousands of big current spikes. I can only speculate on the physics right now.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

But the original 741s tended to have popcorn noise. Millliseconds or seconds of silence then a burst of offset pulses, hundreds of microvolts. It was caused by crud in oxide layers. Current production parts are probably OK.

I like the OPA197.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Am 05.11.25 um 16:39 schrieb john larkin:

IIRC popcorn noise was a problem at RCA. They even had an app note on how to measure it in their 1974??? data book. The 1st data book I ever bought, almost still at school...

Gerhard

Starting as a freshman EE student, I had a real job too. TI and Motorola reps would show up with a car full of data books and a briefcase full of samples. Great times.

We actually have a real-life TI rep again now. We have sandwiches in our conference room.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Samsung S23 phone. It has great cameras. Incredibly dumb otherwise.

It does funny stroboscopic things shooting my Tek 11802 sampling scope (raster scan CRT) but using night mode fixes that. Longer exposure I guess.

The rainy/snowy season is just starting.

We could meet in the library, or in GPS, the local dive bar.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

It was a great deal easier to use. The 709 needed three frequency compensation components and the input transistors blew up if you put more than few volts between the inputs. The 741's performance wasn't noticeably better.

Things like the uA715 had more bandwidth, the uA725 had more gain and the LM308 had a higher input impedance.

The LM307 was National Semiconductor's equivalent to the 741. It was a better part, but didn't sell anything like as well.

Part number?

And get packaged up and sold as 741 parts if they aren't.

Manufacturers test for pop-corn noise these day and modern data sheets specify performance in the 1/f noise region.

Burr-Brown made nice op amps, but they were never cheap. I preferred the Linear Technology parts.

Niche markets work like that. People buy the least worst product they know about.

If there's only one product in the market it can still sell, even if it is pretty horrible. The Lintech electron beam testing electron microscope was a classic example. It had an eighteen month order book when Schlumberger hired one of Lintech's engineers and got him to build them a debugged version - Mike Engelhart of LTSpice fame was part of his team. Once the Schlumberger machine hit the market, Lintech never sold another machine, and closed down after they'd shipped all their existing orders. Schlumberger had a long order book by then, and Samsung got desperate enough to order a Lintech machine, but the order arrived a few hours after the company had shut down. Cambridge Instrument bought up Lintech's assets and eventually delivered a Lintech machine to Samsung.

But you can't tell us enough to let us speculate too.