like, its basically the literal definition of the word 'common'
right thats look laid out for cosmetics, not performance
if thats some opamp shit that just makes me sad
because fuckit
like an autoroute test
heh
Posted by renesis at 11:44 | permalink | 0 comments
i didnt say you should have planes
im saying one bigass plane for everything is a pretty shit grounding strategy
but probably turns out better than the average guys murdering his grounds
as far as what you can 'get away with', there are plenty of products with completely shitty grounds that work
well like if you have an analog section that all references the same ground, you probably want those all common to a point before hitting the main ground and definitely before other subsystem grounds
and in general, shorter paths are more helpful, unless youre trying to get some isolation via a longer path
like, ideally everything has zero impedance to ground
and all noise/interference shows up on every subsystem that same, no phase or amplitude changes from parasitics
*the same
Posted by renesis at 11:39 | permalink | 0 comments
right but point is, there are guys who designs systems like this who wont fuck with ground planes because the performance can be pretty random
versus just keeping ground paths short and isolated and keeping signals crossing shit at right angles
no i dont necessarily think so because there are transient situations which can basically fuck anything, and your design should be able to deal with these within reason
jero32: i think even in white paper producing pockets of industry, analog design is essentially dead
big plane works better than shitty unplanned grounds, so fuckit everything big planes from now on
and then they fail at the EMC lab and end up trying 20 diff things anyway
ground strategies matter a ton for immunity and emissions
Posted by renesis at 11:34 | permalink | 0 comments
and i know analog guys whos job is basically to minimize noise to like .00001% of signal who wont even fuck with ground planes
the branched ground plane thing takes advantage of the low impedances and spread out paths in planes, and the path isolation to the reference in star grounding
cheater: no way
cheater: i was going to say no way but i tagged with %
so yeah drop a zero
and 120dB is not a crazy target
right which is the edge of capability
and about what measurement systems will self test at
120 is pretty reasonable unless power amps
and then its time to tag the spec with test conditions
Posted by renesis at 11:29 | permalink | 0 comments
star and branched grounds are best
a ground plane is a really, really shitty ground plane
random parasitics, random current flow
you can cut up the ground planes, and actually control current flow possibilities, and control noise and interference sinking/sourcing
because it has no thickness on some conceptual level?
i dont really understand the question
jero32: yeah not it says seperate ground loops, longer loops
i am arguing that you can slice up a ground plane in a way that actually maintains ground path lengths, but also isolates ground paths from each other
you cannot argue with the benefits of star grounding
this is like trying to argue against 4 wire sense
its just some shit that works
i dont think most engineers think about it enough to do something that makes sense
anyway, kevtris does the branched ground planes on a lot of his boards
Posted by renesis at 11:24 | permalink | 0 comments
which keeps the cap vcc common with the ic ground
which is like, the point of the cap
i dont think that is a big deal so much
like, youll maybe have a/c current flow vs offset a/c current flow in a trace
parasitics shouldnt care so much about offset
current flow
if the cap is not inline with the supply, then current to and from the cap may full reverse
i mean, if your dropping shit to giant common ground planes chances are you dont really give a shit about this stuff to much anyway
its mostly all related to controlling current flow via parasitic RLC in ground paths
Posted by renesis at 11:19 | permalink | 0 comments
jero32: i think if there are large current transients on the pin, there 1b will cause more supply voltage transients
say like, a digital ic power input
oh wait i didnt look at 1a very close thats actually going through the bottom of the board
i dont understand why they dont just connect the capacitor across the supply pins
trick question
i think both of those are wrong
the via common with the ground pin should be left of the pin
the cap should be closer to the pins, basically in contact with the via, the vcc side of the cap should be tried with a short a trace as possible
star ground, mfkr
drop everything to common ground plane is not star ground
i dont see why it shouldnt
that keeps the cap ground common with the ic ground
Posted by renesis at 11:14 | permalink | 0 comments
rab: yeah thats prob the width of the gap
looks like mouse encoder bits
Posted by renesis at 10:22 | permalink | 0 comments
Top | Add to Technorati Favorites