hilarious circuit

Well, it’ll boost _a_ voltage, all right—in SPICE, probably into the megavolts.

Cheers

Phil Hobbs

It's good that it actually doesn't boost. D1 is rated for 20 volts.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

D1 connected to wrong side of L1?

I would not bother .

Connect 24 V to output Remove chip and other components for savings :-)

Their graphic design and art department produced the schematic. It appears it wasn't proofread by an engineer.

The Golden Rule of Application Notes: Ultimately, the prevalence of this rumor highlights a fundamental truth about engineering documentation: Application notes are templates and educational tools, not drop-in manufacturing files. Experienced analog engineers always treat app note schematics as conceptual guides, verifying the pinouts against the official datasheet and simulating or breadboarding the circuit before finalized manufacturing.

Thanks for pointing that out. It's forehead-slappingly obvious now that you point it out. You have permission to make fun of me for not seeing it immediately.

It might work as a way to destroy unneeded IRLML6344s?

Kragen

Not a problem. The fet never turns on.

The MODPEX operation seems to be, thankfully, out of business. But Symmetry still offers training courses for switching regulator design, for $1000 per day per person.

I think the last thing to fall to AI will be circuit design.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

John Larkin wrote: |-------------------------------------------------------------------------| |"[. . .] | | | |[. . .] | |>Thanks for pointing that out. It's forehead-slappingly obvious now that| |>you point it out. You have permission to make fun of me for not seeing | |>it immediately. | |[. . .] | | | |[. . .] | | | |I think the last thing to fall to AI will be circuit design." | |-------------------------------------------------------------------------|

Recall an earlier thread ("Re: AI for FPGA design") in sci.electronics.design and comp.arch.fpga 13 months ago - e.g. news: snipped-for-privacy@irrt.De . . . |------------------------------------------------------------------------------| |"[. . .] | |"A lady claims via LinkedIn that an AI service produced a bad Verilog code, so| |she concluded that an AI is not going to threaten her job, and I wrote to her | |that she deserves a refund." | | | |[. . .] | | | |I asked Ms. Sharada Yeluri for permission to republish from that LinkedIn | |thread. [. . .] | | | |[. . .]" | |------------------------------------------------------------------------------| (S.

AI can probably do FPGA design pretty well, as it helps coding Python or C.

But analog circuit design is a different world.

It doesn't help that AI reads what are often very bad data sheets.

John Larkin does have issues with data sheets. My suspicion is that he doesn't read them all that carefully, and isn't careful enough to extract all the information provided, or pay attention to the information that isn't being provided.

A data sheet that doesn't set an upper limit on popcorn noise (1/f or flicker noise) is effectively warning you that the device generates a lot of it.

Data sheets often hide part defects. Deliberately.

Test stuff that really matters.

Does popcorn noise still happen? I think it was caused by oxide contamination and that's been fixed. Popcorn is distinct from 1/f.

BITD datasheets were engineering documents, but for the last 30 years or more they’ve been marketing documents.

First one I really noticed was the noise plots in the LT1028 datasheet, which cut off at 10 kHz iirc. (This for a 100 MHz GBW, unity gain stable part.)

The reason of course was to disguise the horrible noise peak at 300 kHz. After howls of anger and snorts of derision from customers, they eventually came clean about it.

There are lots of other examples. So I almost always do a quick proto of any new chip that doesn’t come with a trustworthy recommendation.

Cheers

Phil Hobbs

I thought that way until 1987, until I swapped out a 741 op-amp (which doesn't have a 1/f noise spec) in favour of a marginally more modern part that did.

The GaAs crystal puller it was controlling went from bang-bang operation

- full on for about 30 seconds then off for about 60 seconds - to running quietly at about 30% of full output. There was a about 30 kW of heating involved, and it took about a day and half to pull a crystal, so the operators who had to baby sit the machine for the whole period were a lot happier.

My guess is that a lot of what was being packaged up and sold as 741s were actually better parts that had failed their 1/f noise screening.

Popcorn noise is a sort of 1/f noise. There are apparently others.

Regular noise is an aspect of thermodynamics - entropy and disorder. If you let unintended impurities get into your thin base layer, and they move around you've got a different kind of noise source.

The relevant theory talks about a drunkard's walk, which is about a graphic as "pop-corn noise".