In general... no.
There are cases where long parallel traces can cause unwanted coupling between signals. Consider for example a set of data or address lines in a memory array. These signals usually change state as a group... so it doesn't really matter if the edges get fuzzy. They are clocked some time after the transition... when all the fuzz and and ringing have died out.
Now clocks on the other hand are usually given more carefull consideration. Lowering the node impedance, guard traces and other tricks are often used. Jittery clocks are bad... oomkay?
Much depends on the type of logic family employed... not really the speed... but rather the rise and fall times. The steeper the edges, the more coupling you get. Old Schottly, "S" devices could be a PITA that way.
There are cases where long parallel traces can cause unwanted coupling between signals. Consider for example a set of data or address lines in a memory array. These signals usually change state as a group... so it doesn't really matter if the edges get fuzzy. They are clocked some time after the transition... when all the fuzz and and ringing have died out.
Now clocks on the other hand are usually given more carefull consideration. Lowering the node impedance, guard traces and other tricks are often used. Jittery clocks are bad... oomkay?
Much depends on the type of logic family employed... not really the speed... but rather the rise and fall times. The steeper the edges, the more coupling you get. Old Schottly, "S" devices could be a PITA that way.
Hi poobah,
IBM mainframes even up to the era of the 3090 (if I'm not mistaken) were built (orpartially at least) using wire-wrap technology. Done properly and with care, even fairly high speeds were achievable. The physical distance between traces on a pcb versus that of wire-wrapping for parallel lines would indeed be (possibly) smaller on wire-wrap, but not by much.
I agree of course there are limits - depends on the length, distance, rise/fall time of the signals as well as the voltage levels of the logic devices used. But wire-wrap serves pretty well up to several MHz. Just don't run things like hundreds of MHz or GHz on it...
Cheers!
IBM mainframes even up to the era of the 3090 (if I'm not mistaken) were built (orpartially at least) using wire-wrap technology. Done properly and with care, even fairly high speeds were achievable. The physical distance between traces on a pcb versus that of wire-wrapping for parallel lines would indeed be (possibly) smaller on wire-wrap, but not by much.
I agree of course there are limits - depends on the length, distance, rise/fall time of the signals as well as the voltage levels of the logic devices used. But wire-wrap serves pretty well up to several MHz. Just don't run things like hundreds of MHz or GHz on it...
Cheers!
Yes Sir!
Even into the mid 80's... I know Amdahl computers was "stitching" boards with small welded wires. Their approach to cross talk was to put every wire in straight line, shortest distance between to points. Their studies showed it had the least amount of overall noise coupling.
The boards looked like HELL though!
Even into the mid 80's... I know Amdahl computers was "stitching" boards with small welded wires. Their approach to cross talk was to put every wire in straight line, shortest distance between to points. Their studies showed it had the least amount of overall noise coupling.
The boards looked like HELL though!
rfbrw,
There is a thread here of guy who bought a National op-amp and chose the "MicroStar" package by mistake.
I used one of these before... something like 6 mil pads on 12 mil spacing. HE actually soldered wires on the bottom and went about his business.
I think I would have coughed another buck for right part!
There is a thread here of guy who bought a National op-amp and chose the "MicroStar" package by mistake.
I used one of these before... something like 6 mil pads on 12 mil spacing. HE actually soldered wires on the bottom and went about his business.
I think I would have coughed another buck for right part!
As Poobah has already said its down to edge speed, not clock speed.
Outputs are a lot faster now than in the 1980s, edge speeds on even basic logic chips can exceed 1GHz now.
High speed signals like that need good ground returns as you're in the realms of waveguides. On a PCB with plane layers (note: doesn't have to be a ground plane) the return current actually follows the route of the track above it. It does not take the shorted path back to the source.
This is why splits in plane layers can be a very bad thing and why decoupling close to the chip is needed.
This close coupled return path also reduces radiated emissions from the track. You're not making a big loop antenna out of it.
Outputs are a lot faster now than in the 1980s, edge speeds on even basic logic chips can exceed 1GHz now.
High speed signals like that need good ground returns as you're in the realms of waveguides. On a PCB with plane layers (note: doesn't have to be a ground plane) the return current actually follows the route of the track above it. It does not take the shorted path back to the source.
This is why splits in plane layers can be a very bad thing and why decoupling close to the chip is needed.
This close coupled return path also reduces radiated emissions from the track. You're not making a big loop antenna out of it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.