3M Getting Into PCB Materials

Cool. It would definitely be slick if there were some kind of board you could use where for circuits up to XX (XXX?) MHz you didn't have to worry the layout as much. Probably save development costs overall

Embedded capacitor, and ever embedded resistor, boards have been around for decades. Ceramic caps cost ballpark 1 cent each, and I bet those exotic boards are going to be really expensive. I do picosecond stuff, and multilayer pours with a modest scattering of 0603 bypass caps works fine. I've never designed a board that had too few bypasses. I knwe one guy who designed big logic boards and didn't believe in using bypass caps; his stuff worked.

(I also saw an Anritsu board being pick-and-placed with over 3000 caps.)

Big chips, CPUs and FPGAs, mostly have a lot of on-die capacitance to handle the fast stuff. They don't need heroic bypassing. They do sometimes need microfarads of bypass to cover gross current jumps, but those don't need to be very close. Nanofarads won't help much.

I wonder if all those religious charter schools will be teaching "faith-based" design, as well...;-)

I actually don't believe in electrons, I find them to be a heretical concept.

There is a lot of silly, and expensive, folklore in electronics design.

Unless you are designing tubes, like CRTs or PMTs or something, in which case electrons are pretty real. Ditto flash and DRAMs. Shot noise is calculated assuming that electricity comes in charge units of

Science, especially physics, is sometimes useful in electronic design.

The design of power distribution systems is complex and to do properly requires you to have data that you can only get by measuring yourself. So people over design using rules of thumb like "one cap per power pin as close as you can get it". Far over designed, but no one was ever fired for using too many decoupling caps.

I've also heard people mistakenly talk about how to do PDS design such as cap resonances. One group says one thing, another group says something different and very, very few of them ever take actual measurements.

Likely not but too few may not be good for your employment situation. "The datasheet sez..."

I remember looking at what seemed to be a very good book on PCB design in a university bookstore a long time ago. Very "evidence based." IIRC it was probably from around the early 1980s, unfortunately I don't recall the title, or how much relevance it would have today.

There are probably a lot of books about board design today but this one seemed particularly good - it might be a longshot but anyone have an idea of what one I might be remembering?

I think the cover may have been brown, orange, and white or something. Very early 80s looking, cover fonts like a DEC technical manual.

Like the charge also will "jump" the Esaki gap.

Well, it won't have been

which wasn't published until 1993. Win Hill said nice things about it here in the late 1990's, and I bought a copy. It's incoherent rubbish. There is useful information embedded here and there in the book, but it isn't easy t o find, and none of it was stuff I hadn't been able to find out earlier, fr om better written sources.

Howard Johnson seems to have gone around for some years giving expensive le cture courses based on bits of the book, and I've never heard anything posi tive about any of them.

Howard Johnson and Martin Graham's chapters are ostensibly evidence based - each one flogs a simple idea to death - but as a pedagogic exercise it is down there with E C Snelling on ferrites in providng loads of information i n the most inaccessible format that one could devise.

I attended a course by Lee Ritchey with a book titled "Right the First Time". He demonstrates ideas with analysis, simulation and then shows the results of testing on hardware. This is where I saw the data that convinced me that decoupling caps are over used and incorrectly used.

Yeah, I looked at that one once and wasn't particularly impressed.

I'd be leery of any book that describes it as "black magic." It's applied physics, not flippin' Hogwarts Academy.

Howard Johnson needs to stick to the black magic of good grilled cheese sandwiches

ere in the late 1990's, and I bought a copy. It's incoherent rubbish. There is useful information embedded here and there in the book, but it isn't ea sy to find, and none of it was stuff I hadn't been able to find out earlier , from better written sources.

e lecture courses based on bits of the book, and I've never heard anything positive about any of them.

ed - each one flogs a simple idea to death - but as a pedagogic exercise it is down there with E C Snelling on ferrites in providng loads of informati on in the most inaccessible format that one could devise.

Grilled cheese: Extra sharp cheddar, good rye bread, and a pile of caramelized onions between the bread slices. Then it has to be toasted enough so the cheese melts, runs down the sides and get's fried too.

Hmmm time for lunch.

George H.

Could it have been "Printed Circuits Handbook" by Clyde F. Coombs? Mainly, it was about fabrication... went through several editions.

Gee, I think there are a few "schools" in the electronic arena that have been doing that for decades. I'm talking about some of the traveling road show consultants.

Jon

The concept that a grid of decoupling capacitors on some real digital board can set up an LC resonance that builds up to several hundreds of mV seems ludicrous. The caps are all different, general layout variations will vary the series L's, so the thing can't possibly act like all the caps resonate at the same frequency. This seems like something that you could simulate with SPICE, but never see in the real world. And, of course, the logic chips will be acting as a big resistive load, gobbling up the transients as fast as they can make them. No question at all that bad decoupling could cause issues in mixed signal boards, though.

And, that's just the most prevalent craziness that some folks (usually some kind of consultant offering paid seminars) have been spouting.

Jon

I don't know how to reply to your comments. The resonances between the caps and with the power planes creates impedance peaks. This is real and can be measured. Also, I don't get why you think the caps are all mounted differently. They should be mounted with the shortest possible path to the power planes. Anything else limits their function.

The impedance peaks are the reason why some people use multiple values of caps to provide low impedance across the band. But when I say it can be measured, in a real design it is the series resistance of the caps that limit the resonances. So they are never very high to begin with just as the impedance dips are never very low.

You mean HoJo, the Black Magic guy? I started highlighting the ludicrous mistakes in his book, but gave up halfway through. He challenged people to debate him, but you have to pay to get into his seminar first. You can debate me if you pay me $400 an hour.

I've added SMA connectors to the power pours of multilayer boards and did TDR and TDT measurements. Basically, a power pour near a ground plane is a perfect capacitor, and adding ceramic caps anywhere makes it a bigger perfect capacitor. That's good enough.

There a zillion silly appnotes about Spice modeling PCBs and bypasses. Many are written by people who want to sell a lot of capacitors.

You really can get nasty resonances if you use daisy-chained traces for power distribution, but as you say, not on pours.

"Doctor, Doctor, it hurts when I go like this!"

"Well, don't go like that."

Cheers

Phil Hobbs

A lot of lore is left over from 2-layer boards full of TTL, with power runs along X and grounds on the other side, running Y.

Those were ugly days. We seldom do 2-layers any more, and those are little test things or adapters or something, rarely things that go into products.

We sometimes have a power pour, then a fat trace connecting it to another patch of pour. That's fine as long as both pours have some caps.