Silly resistor values

Ah, yes :-)

Not in ultrasound machines.

Easy: Have a script plot sheet numbers into all the blocks. Then you just turn to page 37 if you want to see the first eight channels of a beamformer. I've done lots of "dead tree" design reviews in the 80's and

More successful is usually the informal method: An evening in the pub with the CEO and the CFO. Especially the CFO will catch on to the idea, fast.

Why? Power supplies make a mess of schematics and add little information.

^^^^^^^^^^^^^^^^^^^^^^^^^^^^ Exactly.

I'd have to see how that annotation works. I'm still not so sure about flipping back and forth when tracking a signal through the system. We'll see.

Question: How do you deal with connectors? I'm thinking along the lines of a board mounted cable connector. DO you put them on separate pages even if they only have one source/sink? I have a switching network that would reduce to the control inputs if I made it a hierarchical block.

Don't think we have a CFO, or he's the same guy as the COO (the IT people are next to useless here). The owner is pretty hands on, so he's the one I have to get interested. Have never been out to a pub with any of them though. I don't suppose distys count? ;-)

And in the end, it won't work either. Just like Obama himself.

For example, there are intricate phase-locked balancing mechanisms that make sure that any ever so slight sag is 100% synchronous to the Doppler PRF. One must understand that in order to debug noise issues and so on. Just a big block saying "100V/500mA supply" doesn't help. If you didn't use such load balancing but wanted to just make it's "stiff enough" you'd need a Mack truck to deliver the machine ;-)

It's much worse in a flat sheet structure because you'd have no overview how cable harnesses are organized. Because they can't easily be shown. I get a lot of flat sheet structures from clients for review and optimization and this easily adds a few hours to the final bill.

Usually those are on the top page, together with a reasonable amount of functional blocks. Then reviewers or new engineers can see where which lines go. Best to draw in all wires and buses and not just place little islands "netlist-style". Much easier to comprehend.

What? No pubs, no outings, no pool parties? Hmm, that was always different at the places I worked at. A couple days ago we had the dog of one client's president here, and later that night when they picked him up we had a few beers together.

I don't suppose distys count? ;-)

Nope :-)

it's

already

price

largely

I use a copy because my boss tells me to. The rest doesn't matter.

Then it's not a "power supply". What do you do for logic/analog supplies (and ground)? Do you wire them explicitly or do you use an implicit global?

Cable harnesses? Hmm, no cable harnesses on the boards. ;-)

Are you saying that you show physical relationships on schematics? I do with decoupling, sorta.

The problem is that I have a pile of connectors and few busses. The connectors are all differential audio so putting them into busses is a problem. I've placed them in their own blocks but the top page is still going to look like a bowl of spaghetti.

Last Nevember (still nice weather ;) they have a party for the families in the parking lot. Other than that, not so much. They did have extra tickets to Braves games a couple of times. The last time they provided transportation too, but July Saturday afternoons get a bit warm in Atlanta.

We have one who likes to stay in town at the university visitor's center. He's had us out for pizza and beer (I did eat the pizza). ;-)

They all supply power :-)

What do you do for logic/analog

GND always global, digital supplies mostly global except for phase-noise sensitive stuff, analog supplies usually explicit.

I meant a true and full top-down architecture. The top sheet would be the complete ultrasound machine and those things usually do contain cable harnesses. Just like cars do.

I don't know what you mean by physical. I do try to be clear enough on schematics so readers can say "Ah, that's the big cable over yonder". After all, I often design stuff for companies who do not have an analog engineer on staff which is why they "rent" me. But their engineers must still be able to comprehend a detailed module or system spec.

Well, how many is a pile?

Well, that's a start. But don't leave it all just with the employer, we organized most of the outings ourselves. Beach trips, keggers, pool parties at someone's home and so on.

I'd have gone for the brewsky. At my first employer it has happened that many of the cars were still in the lot after midnight because people took cabs home, on account of too much booze.

If my product was a power supply, no, I wouldn't make its output a global. If it used supplies for its internal circuitry, yes, it would be a global. I have that instance in the widget I'm designing now. It has an output that supplies power to a DUT. That output isn't a global. ;-)

I make them all global, though we show the pins on the symbols. I'm trying to get them on separate pages even. The layout guy doesn't like that (he's also the librarian).

Oh, we don't do higher level packaging in the schematics. That's done on mechanical drawings. Not my job. ;-)

Physical relationships of components. Maybe even dimensions. A cable would "look" as it does in real life, with all the branches and twists.

We do all of that in the (3D) mechanical drawings.

More than 30 on the back (most are stacked). Another bunch of knobs, displays, LEDs and stuff on the front. We're reusing most of the mechanical stuff from another project so all that baggage comes along too.

So far, I've put all of the stuff that goes to the DUT on one hierarchical block and stuff to various other test equipment on another two. That sorta works but the top page is nothing but lines. It's not as bad as I thought it would be though.

The Braves tickets were left-overs from customers. Our customers are sorta into sports, since they're football coaches. ;-) I suppose we gotta find a busybody to get all that stuff together.

Not allowed anymore. :-( IBC is as strong as it gets. Over the last couple of years I've volunteered to be the DD for such things more than once.

You can do that but when several of the supplies are synchronized to signals that originate in the innards of the machine this can get old. Makes it difficult to understand for anyone but the original design engineers.

Cable harnesses aren't exactly mechanical in nature. One must understand where which one goes and where it branches. When looking at wiring diagrams for some cars it becomes obvious pretty quickly how that kind of visibility can be screwed up.

Ok, I rarely place dimensions in schematics, only where it helps folks find a branch. Or to show what stuff is underneath a shield which reduces the frustration level of younger service guys.

With decoupling I do that as well. Except for mundane circuitry such as glue logic where there is just a bunch of 0.1uF on 3.3V in a corner of the sheet.

Those could still find room on one B-size sheet, unless too many of them are 50-pinners or larger.

It can make sense to put the connectors on the 2nd hierarchy level, especially if they are on a common large panel which is a separate sub-assy in production.

Mine would only have not so useful things to offer. Scalpels, syringes, catheters and so on. But there was that steel company that had managed to find a brandy with the exact same spelling of the company's name. So they bought cases of that.

I remember a high-tech place near Munich, Germany, where we had some die bonding done. In the hallway there was a vending machine. No cans but bottles like in the old days. Five columns to choose from:

That gave me a chuckle.

Cable *harnesses* are quite mechanical in nature. A cable on a schematic may be just a description of pinouts, but a harness is a bit more. I'd rather the mechanical drawings be done separately where the CAD tools are designed for this.

I don't put them in the corner of the sheet unless they're in the corner of the board (placement doesn't matter). Otherwise, I show them next to the device they're decoupling. What I want to do is have a homogenous "gate" for each device with only the power supplies. I can then place that "gate" on a separate page and get the power supplies completely off the schematic page.

At .1" spacings I don't think I could fit them all on. Most (the ones in question here) are 1/4" stacked jacks, so have about 20 pins per pair (only three used, one of them ground). They're supposed to be XLRs but they won't fit so I'll convert to XLRs on the patch cables. It's a COTS cable so it should be fine.

Today I increased the page size for the top level schematic from

Decoupling is my other problem here. The FPGA is an F324 (18x18-1mm pitch) which has no Central Park for decoupling. Altera says to use

I bet those gifts were appreciated. ;O)

I was once snowed in, in a fire house in downtown Poughkeepsie NY. We had 24" of sloppy snow and the police were just sending people around the city in circles. Anyway, the firemen cleared everyone out of cars, which were left as they sat, into the fire station. When I was about to put *a* quarter into the coke machine, one of the firemen told me to try the Dr. Pepper. Yep, a 3/4/5.

That's common in Europe. Even IBM had beer in the cafeteria in Germany and Wine in France. Such things were an instant termination offense in the US. That and drugs were about the only way to get fired.

I meant the cables in there :-)

For clarity of the schematic I try to keep them together just like they are in the harness.

Ugh, I never do that. Power symbols go right where the A-device is. Placing the decoupling caps there is ideally the right method but sometimes they do get in the way with digital stuff.

Ok, if that's 30 stacks and not 30 connectors that could get too dense.

Can't you put some on the other side and get away with 0603?

One quarter? That must have been in the 70's. Poughkeepsie sound like IBM.

I remember the good old days. Weller station to your right, scope in the middle, beer to your left.

Not sure I follow.

That makes a mess of multi-gate devices. Which one do you put the power pins on? Power pins (and decoupling) also gums up the logical flow of the schematic.

Fifteen pairs of 1/4" connectors, plus a half dozen stacked RJ45s. It works out fairly well down a level in the hierarchy since they're grouped logically.

Where? No Central Park in an FBGA. Altera suggests 0402s and 0201s in between the vias but 0603s won't fit.

It was a fire house. Yes, March '78. I did a 19 year sentence in IBM P'ok (and another 14 in BTV). ;-)

Never drank when I worked. When I lifted the first beer it was an acknowledgement that I was done for the day.

There will be buses and wires that I deliberately run in parallel, usually with a few dashed line ovals around on the outline layer (not the net layer), plus some text such as "AC controller harness". Then everyone will know that all those lines are inside that big blue hose over there.

In logic sections I don't use power symbols, except in critical areas like LO phase shifters. In the analog sections I make enough room for them because much is decoupled individually and on another net because of series filtering. On a multi-gate device power is always on unit A.

This is one reason I cannot use GPL software such as gEDA or Kicad because they do not handle this correctly and the communities there do not seem to grasp te importance of it.

That would fit on one sheet.

On the opposite side of the circuit board.

Sometimes when working very late I'll have a nice brewsky. Only the good stuff. But only one, and I usually nurse it a long time. Otherwise the senses dull.

Isn't that the point?

Jim

Perhaps I'm not seeing it because I don't build schematics at higher level assemblies. On board schematics I'll use some hints (like running differential pairs together as long as possible - to the point of looking strange) but usually use only attributes.

THen when you swap gates in layout you redraw the schematic to follow the "A-gate"? That's what we do and it's *fugly*. It's better than putting them on all gates, but not much. I'd rather sequester them on a separate page and have all the "op-amps" look the same. DRC will quickly tell me if any are left out.

How do they do it?

of

little

The connectors alone will barely fit.

The FPGA is on the opposite side. ;-) The vias are on the side I'm talking about. ;-)

One is (never) enough.

I usually get it about right when drawing the schematic. Many of my layouts are noise-critical so I kind of picture it in my head while drawing. Else I can always re-annotate. Then either leave the power on the C-device or clean it all up.

They don't. IMHO gEDA's library editor doesn't have the proper means to do it correctly in the first place. Then on re-annotate it really screws up the instantiations, or slots as their community calls them. KiCad has an excellent library editor, much better than gEDA. But on re-annotate ... poof.

I mentioned it in both forums and in the gEDA group one of the leaders kind of flew off the handle, not seeing why this is so important and berating me not to bring this up all the time. Oh well, then I just walked away. But I occasionally check back in because the idea of gEDA is great.

That truly being between a rock and a hard spot ;-)

Well, then there is the occasional Schnaps :-)

Choosing the gates is up to the layout guy. It's not just a matter of back annotation, rather redrawing the schematics to show the power on the new 'A' gate - messy. If it were just a matter of back annotation it would be easy.

I guess I'd have to see it.

Not so great if it doesn't work. ;-)

Yep. I really liked the FPGA packages like the old FG680 where the balls were only five or six rows deep, with only power and ground on the inner couple of those. The center square was wide open for decoupling and termination. Not so with the newer packages.

One of the few drinks I didn't much like (as opposed to rum and gin, which were enough to make me barf). I never drank much of the hard stuff.

I'm missing someting. Why is the power on the A gate? That just leads to problems like this.

I'd expect it to be on a separate symbol or "gate" or whatever you call them. Another option is that power is connected automagically, but then you need a way to override that when you are doing fancy things.

Simulators often use $G_VDD and $G_GND... global power, hidden pins... makes it rough for layout people. Just going thru such an episode myself with a library connived (by a PhD with crossed eyes and his head up his ass) with pin order not following the standard for digital subcircuits :-(

...Jim Thompson