Miniaturization

Getting up to the point where you can use the machine is 99% of the time and work. It's simpler to just send the data out and have a real board made. And you aren't limited to two-sided and thick traces, and get nice holes drilled where you want them.

We had one of the milling machines at both of the last places I've worked. Neither has been in the time I've worked there.

f manufacture of radios etc. PCBs were made by metal deposition from an arc with a mask. The only thing humans needed to do was add mains lead, put va lves in and screw the panel into the case. Capacitors were made by milling the plastic PCB thin before deposition. Valve holders weren't separate, the y were pins the machine inserted into PCB holes. The video shows the stages of manufacture for a fairly small domestic reaction set. The whole thing w as controlled with electromechanics, and various limit switches detected pr oblems & stopped the production line.

de that way. I must have saved a link but I'm damned if I can find it.

CBs.

It's jobs like that that sort the determined to learn from the not.

Reminds me of a colour TV LOPTF that had arced over & carbonised. Removed e very bit of carbon tracking, but that left not enough insulation. For some reason I don't remember just patching it wasn't workable so I glued on a st ack of chopped up bits of record, making an old fashioned EHT insulator. Th e volts refused to take the long path and the customer was happy. People ar e all scared of liability now, you can't even replace a fibre cordgrip with FR4, despite its superior characteristics. That was mild compared to one f ix I did.

NT

It's hard to imagine a hobbyist chip that would be better than ones you can buy off the shelf. I think most hobby boards are either to reduce cost by DIY or to produce a special device that is not readily available. I can't think of anything not available at a chip level and certainly the cost factor is not in favor of DIY chips.

You can do deep sub-micron lithography directly to the photoresist on the wafer with a pretty ordinary scanning electron microscope if you modify it to gate the beam on and off. If you are only making a few chips that could be fast enough to tolerate. Still I am more interested in economical access to a proper commercial fab with high yield and repeatable device parameters. Perhaps some of the Chinese fabs might be more open to this sort of arrangement, especially if it is part of the deal they get to pinch any IP worth pinching from the designs.

I agree that sampling front-ends would be interesting (preferably on a

I think that in spite of limited experience, hobbyists could come up with some very important things, simply because they would not be taking their instructions from risk-averse managers and marketing departments that mostly come from a business background, have no passion for electronics itself, and are both unaware of the possibilities offered by a fab process for products that have not already been done by others. Some of the hobbyists would also be ex-chip-designers like me.

Also doing chip design is explicitly forbidden by the LTSpice licence agreement.

There are a few packages around. I haven't checked these out but there is "electric", "magic" and a few others. For simple stuff you can even generate a "CIF" file which is a very simple format (x and y coordinates of polygons), and get someone to read it into Cadence when you're finished.

There might also be an argument for designing one's own chips to avoid component obsolescence. Given that many analogue chips are sold for 10 to 50 times the cost of wafer fab, you could do a lifetime buy of wafers much more feasibly than of packaged parts, and if you need to make more wafers later then depending on how ordinary the process was, that might also be more feasible than scouring the dodgiest parts brokers for more of whatever Maxim part you can't get this time.

RF, ADC, DAC, and fancy analogue stuff. FPGAs already exist for doing custom digital.

Cheers

Phil Hobbs

Besides, the most expensive part of an I/C is the _package_. ...Jim Thompson

E-beam writing in a SEM is heartbreakingly slow. I used to use a Leica e-beam system with 30-nm resolution, but iirc that was something in the tens of microamps at 70-100 kV. SEMs are typically nanoamps to hundreds of nanoamps at lower voltage. Schlooowww.

Still I am more

You can do all of those things at board level. A custom chip only becomes interesting (even as a hobby) when you need more stuff than you can conveniently fit on a board, or when you need something with very high performance. Hardly the "crappy transistor" thing we started out with. You'd think that if there were really hobbyists out there who knew enough about device-level design to do a custom chip at the transistor level, they'd be doing super-fancy discrete designs at board level now. Are they? The only people I know who do that are SED folks--Joerg, John L, one or both Tim W's, and me. (That's assuming that seven transistors is a severe understatement.) ;)

I'm sure I've missed a few--how many of you are there out there who do nontrivial discrete transistor designs at board level? ("Nontrivial" meaning roughly, "complete HF receiver and up".)

Again, there are a lot of things you can do with discretes and matched BJTs at board level. I'm in violent agreement about the importance of a fire in the belly and time to spend, but I haven't seen anybody doing device-level design as a hobby since ham radio went commercial. Some of the old QST stuff was pretty good.

Well, that would sort of help. ;)

I sort of doubt whether LT would care if you were just doodling.

That's a bit of a stretch, I think. Fun to think about, though.

Cheers

Phil Hobbs

I suspect that "license" restriction is strictly to avoid liability when your processed IC doesn't match the simulation ;-)

I design the chip circuitry using PSpice.

Many of my clients check my work using LTspice.

Then my layout contractor uses...

$1499 ...Jim Thompson

Mike Engelhardt's tale is that when they decided to give away their internal design tool for free, they wanted to make sure that others didn't use it to design chips to compete with LT's.

Potentially nice layout tool.

Cheers

Phil Hobbs

I don't know how they would tell. A netlist leaves no fingerprint.

My buddy in Columbus has been using that for many years for his side work. He also does a lot of military, but they provide him with Cadence tools.

...Jim Thompson

Disgruntled ex-employees is the usual route, I think. All they need is to file a complaint "on information and belief" and they'll get court-ordered discovery, at which point the other party is screwed. Unless you're a Democratic politician or one of their cronies, destroying or failing to produce evidence will get you in a _lot_ of trouble.

Cheers

Phil Hobbs

The safety of the "one-man band" ;-)

There are an awful lot of people using LTspice to simulate non-LT products, to the point that one of my major Spice modeling clients _requires_ that my models be compatible with PSpice, LTspice, TopSpice and TINA. ...Jim Thompson

I use LTspice all the time, and I rarely use LT stuff because it's so expensive. They do make nice chips, though.

Cheers

Phil Hobbs

A word of caution. In Tools, Control Panel, Spice, set Solver = Alternate. Otherwise your result will be of dubious accuracy. ...Jim Thompson

Some of their little switchers are affordable, and having the models and an example sim are real selling points.

I don't know of any other semi manufacturer who up-front provides LT Spice models; they do other goofy stuff. Some will quietly help you convert their models for use with LT Spice.

I haven't found anything they sell to be affordable. They have great products and support, both LT Spice and (equally good) FAEs but their costs just aren't in line with others. I'd have to have a *really* good reason to spend $1 on an LT part than, for instance, TI.

Mostly cutting Rubilith. "Tape" was only for patching mistakes.

But there was a "floor plan"... I did my first several chip layouts on a quadrille pad ;-)

Give us a demo >:-} ...Jim Thompson

RF, ADC, DAC already exist on chips. They also exist on chips with digital to do various things. What would you want to design a chip to do? Chances are it's already out there.

Designing your own chip is no guarantee against obsolescence. You need a foundry line to make a chip and they close too. The point is electronics advances rapidly. It is hard to stand still in the current.