John Larkin wrote in news: snipped-for-privacy@4ax.com:
I worked on a device we ended up not getting contracted to complete. But the intended use was as an EMP pulser to set off car bombs approaching the green zone in Iraq. They were going to be set up along the new jersey walled road that leads to the gate.
I would be all for throwing an error for nets with the same name on different pages, or even the same page without an off-page connector. Nothing should be global by default.
What matters to us is not minor conveniences, but getting boards right first pass. Explicit, visible offpages are safer than any other way to name nets. It's a lot easier to add jumpers than to cut inner layer traces and planes.
We do print and read the schematic net list to look for obvious mis-spellings and single-node nets and such. It's very rare for us to have a layout error that makes rev A unsellable; that hasn't happened in years.
We do pull out a few unused uP and FPGA pins to test points, which can be scoped or kluged to. My latest innovation is to not solder mask vias, which makes them easy to probe or solder to. We *do* make mistakes, so it's better for them to be easy to hack.
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
I had one guy who named a bus A[0:15] on one sheet and A[15:0] on another, a uP connecting to a 64K eprom. So we wrote an application that reordered the code image to correct for the address remapping.
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
PADS Logic invents a unique name for any net that does not have a named offpage, like $$$23456.
I can add a named offpage to a net, with no other offpages, if I want to force a net name, so we can see it when we route or something. It works.
Again, what really matters is getting the first board right. We don't prototype products and we assume we'll sell rev A, the first etch. I know companies that formally define *five* PCB iterations and always use those five, and often more. Three year developments keep lots of engineers employed.
How many layout iterations do people here average?
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
John Larkin wrote in news: snipped-for-privacy@4ax.com:
Depends on what is being laid out.
RF may need a massage.
Logic may as well, if a lot of real estate is involved and trace length timings become an issue.
Sometimes signal timing can cause a respin requisite.
Power supplies can have hard start issues resolved with circuit changes. We had some FETs that we ended up putting a singular toroid core right onto the lead of the FET before insertion into the PCB.
That and adding things like transzorbs or the like. Sometimes the need for a circuit element in a design does not rear its head until after an initial iteration.
We design review the heck out of things, schematic and layout, and have checklists for both. We leave stuffing options or jumper hooks if we think there is any risk. It's a friendly sport to find mistakes in other peoples designs and layouts, so we manufacture very few dumb errors.
We do occasionally lay out a little test board for some tricky new subcircuit, but we don't prototype whole products.
My new guy just got his first PCB design all built. We made the classic mistake of swapping the V+ and V- power pins on two opamps, which is klugable but not pretty. He won't do that again. We should have caught that during a review, and my layout guy should have noticed it.
There are *so many* possible mistakes.
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
mandag den 15. april 2019 kl. 16.09.55 UTC+2 skrev snipped-for-privacy@notreal.com:
rote:
in message
rant of
es usually
wouldn't run it. It took a few goes looking over it to notice I'd connected the power rails together via 2 base junctions. Oops. A base-ic error.
to
rts of
ence
s;
them
ool
en if
like
on
but afair you have to put labels on the pins you connect to a bus, the bus is just shorthand for getting a bunch of signals off a page without having to add a port for each and everyone
obal
in Kicad net labels default to hierachical, i.e. when you label a net the net name is in the netlist is /sheetname/netname
Is that an ebay bargain? Since you're only building one circuit, you might consider the NTE equivalent, NTE458, which you can buy from NTE direct:
formatting link
The generic subs aren't much of bargain if they're way out of spec and you have to buy a bunch of them. Apparently people have been plugging the NTE458 into 2SK117 circuits for quite a while, with no complaints.
I don't want to clutter up pages with unnecessary off-page connectors but I do like to name every net. It makes finding one's way around a layout much easier. I also try to get the layout engineer to have at least one segment of every (non-power) net on surface wiring, often fanning out from QFP/QFNs.
The checking in the designer forward annotation doesn't allow things like single pin nets.
I add a resistor to every unused PIO. I find it's a lot easier to wire to a resistor than a pad.
But it makes a mess of the schematic if you do a lot of them.
If you decide that there will be five iterations, you're lucky of you only have five, sure. our production groups a similar thing but it gets worse than that. THe reason they have several spins built into the schedule is more because the customer more or less demands it (they want a place to insert changes, too).
I generally do one but I only do prototypes. I do have three revs of my last design but they're really three different designs that only look similar. OTOH, my prototype run is probably larger than your normal production run. ;-)
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.