shorting of analog and digital grounds
antima
component is recommended for shorting analog and digital grounds
1)inductor
2)zero ohm resistor
3) diode (schottky)
Richard Henry
"antima" <ras_...@yahoo.com> wrote in message
news:1138086262.3...@g44g2000cwa.googlegroups.com...
Con't forget capacitor, and that the inductor might be implemented as a
ferrite bead.
Lay out pads that can accept at least one type of any of these. Label it
Z1.
anshu
conclusion
Richard Henry
"anshu" <ras_...@yahoo.com> wrote in message
news:1138089308.6...@g49g2000cwa.googlegroups.com...
> thanks for ur respose can u elaborate ur ans.i am unable to come to
> conclusion
>
To put it simply: Lay out pads for a 1206 part.
Judging where to put the part or parts (electrically and physically) is
another issue.
Roger Hamlett
"antima" <ras_...@yahoo.com> wrote in message
news:1138086262.3...@g44g2000cwa.googlegroups.com...
The diode would only apply, if there was a possibility of a potential
difference. If so, how would the rest of the connections cope?. Inductor,
or resistor, would be 'normal' methods. However you could use either
method, and end up with problems. If (for instance), you have analog
signals 'crossing' to an ADC, and you joined the grounds to a point in the
digital curcuit, where there are high currents flowing (perhaps near to a
switch mode regulator), you would never get good results.
So start instead, by asking yourself, what signals need to cross between
the two areas?. What filtering ineeded on these signals?. Which components
they go to?. What other potential ground routes exist (ground loops must
always be considered)?. How is the supply got to the analog part?.
Best Wishes
Pooh Bear
antima wrote:
It can be any value resistor really in theory. If well designed there
should be no net current flowing there. In practice I might use 100
ohms or even simply track the two together at a suitable Kelivin
connection.
Graham
John Larkin
Use a single, common ground plane and you won't have this dilemma.
Splitting grounds rarely makes sense.
John
Richard Henry
"John Larkin" <jjla...@highNOTlandTHIStechnologyPART.com> wrote in message
news:qrlct193hgcr22dsn...@4ax.com...
Unless dealing with high-quality video, or high-quality audio, or noisy
analog loads, or noisy digital loads.
Joerg
>>Use a single, common ground plane and you won't have this dilemma.
>>
>>Splitting grounds rarely makes sense.
>
I'll second John's comment. He works near a hill that spews lots of EMI
from antenna towers. My dose of EMI reality comes from huge freight
planes calling in their approach right above the lab.
>
> Unless dealing with high-quality video, or high-quality audio, or noisy
> analog loads, or noisy digital loads.
>
Even then. I have never seen split grounds work well and this is after
about 20 years in medical ultrasound. Those machines always have to see
and hear the grass growing just a tad better then the competition. Plus
they had to perform in the vicinity of strong RF emitters which can
become next to impossible with split ground planes.
Regards, Joerg
Pooh Bear
John Larkin wrote:
It certainly does in audio.
Graham
John Larkin
It does allow you to avoid thinking about a few things. But magnetic
field pickup and emi issues are still there in full force. And al
least some ground loops will persist. So you may as well save a plane
and do it right.
Besides, audio isn't "electronics" as far as I can tell.
John
Pooh Bear
John Larkin wrote:
Having digtial ground currents flowing in the audio ground is a
non-starter. It's audible ! No matter how good your ground plane it, it
still has resistance.
> Besides, audio isn't "electronics" as far as I can tell.
>
Is that some kind of elitist snobbery ?
Graham
Jim Thompson
<rabbitsfriend...@hotmail.com> wrote:
John is still miffed about being called on his defiance of the Laws of
Thermodynamics ;-)
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |
It's what you learn, after you know it all, that counts.
John Larkin
<rabbitsfriend...@hotmail.com> wrote:
No, it's an opinion that the things that apparently matter in audio
are subjective and non-quantifiable.
John
John Larkin
No, I'm a great fan of the Three Laws. I just can't get anybody else
to respect them. I've spent enough years in steam engine rooms to
appreciate how easy it is to accumulate entropy.
We'll just have to wait until gasoline is $12 a gallon and electricity
is 40 cents a kwh, then thermo will get a little more respect.
Do you really believe that a resistive electric water heater is 100%
energy efficient?
John
Jim Thompson
The transfer to air is the hard part. A small percentage goes out as
some useless wavelength stuff... so maybe 98% goes to heat.
You really should do an exact calculation for the heat-pump case.
Figuring in compressor inefficiency, and TWO transfers to air, ought
to make it a real wiener.
Nice in theory, but crap in practice.
Want to analyse something useful... analyse an evaporative cooler...
water changing states is MARVELOUS ;-)
Pooh Bear
John Larkin wrote:
I don't subscribe to 'audiophoolery'.
I can however clearly hear 'digital birdies' when digital ground currents get
into the analogue path.
For those only operating with ~ 12 bits of resolution it may not matter much
however.
Graham
John Larkin
<rabbitsfriend...@hotmail.com> wrote:
I do NMR gradient amps with 20 bit DACs, a couple ppm wideband noise,
and switching regs on the same ground plane as the low-level stuff.
All you have to do is keep the power loops fairly local and don't span
low-level loops all over the place. Splitting planes just gives you
two planes at different potentials, and, unless the circuit is very
simple, too many parts have to span both planes. Splitting causes
about as many problems as it cures.
I'll occasionally nestle a low-level circuit in the middle of a
c-shaped cutout of the ground plane, so plane currents don't make
microvolt drops within the sensitive area. You can get similar results
by creating a square of ground surrounded by a break in the plane,
with just some small connecting tabs left; the resistivity of the tabs
is enough to greatly reduce the current through the sensitive region,
but the thing will still be emi tight.
But external magnetic fields will still create potentials across a
plane, so it's best to keep loops small and handle delicate stuff
against one ground via, or differentially. Fans are killers, both from
mag fields and (for dc fans) nasty pulse currents.
John
Pooh Bear
John Larkin wrote:
You'll note I'm sure from my comments that having 2 planes at different potentials
is *exactly* what I want to avoid. The whole point of the 'zerohm link' or even
low-R between AGND and DGND is to ensure both planes are equipotential ( where it
counts ). Clealry they can't be equipotential everywhere due to copper resistance.
'Power loops' can't be that simply defined to be honest.
> I'll occasionally nestle a low-level circuit in the middle of a
> c-shaped cutout of the ground plane, so plane currents don't make
> microvolt drops within the sensitive area. You can get similar results
> by creating a square of ground surrounded by a break in the plane,
> with just some small connecting tabs left; the resistivity of the tabs
> is enough to greatly reduce the current through the sensitive region,
> but the thing will still be emi tight.
That sounds not entirely unlike what I have done.
> But external magnetic fields will still create potentials across a
> plane, so it's best to keep loops small and handle delicate stuff
> against one ground via, or differentially. Fans are killers, both from
> mag fields and (for dc fans) nasty pulse currents.
I *always* keep loops small ! I recall some colleagues thinking I had 'done magic'
when re-tracking a pcb to avoid 50Hz pickup. They simply didn't 'get it'.
Graham
Pooh Bear
John Larkin wrote:
You'll note I'm sure from my comments that having 2 planes at different potentials
is *exactly* what I want to avoid. The whole point of the 'zerohm link' or even
low-R between AGND and DGND is to ensure both planes are equipotential ( where it
counts ). Clealry they can't be equipotential everywhere due to copper resistance.
'Power loops' can't be that simply defined to be honest.
> I'll occasionally nestle a low-level circuit in the middle of a
> c-shaped cutout of the ground plane, so plane currents don't make
> microvolt drops within the sensitive area. You can get similar results
> by creating a square of ground surrounded by a break in the plane,
> with just some small connecting tabs left; the resistivity of the tabs
> is enough to greatly reduce the current through the sensitive region,
> but the thing will still be emi tight.
That sounds not entirely unlike what I have done.
> But external magnetic fields will still create potentials across a
> plane, so it's best to keep loops small and handle delicate stuff
> against one ground via, or differentially. Fans are killers, both from
> mag fields and (for dc fans) nasty pulse currents.
I *always* keep loops small ! I recall some colleagues thinking I had 'done magic'
Robert Lacoste
message de news: qrlct193hgcr22dsn...@4ax.com...
> Use a single, common ground plane and you won't have this dilemma.
>
> Splitting grounds rarely makes sense.
I fully support John & Joerg advice : In 99.999% of the designs I've worked
on a plain common ground plane was the best solution. And that includes
projects in the GHz range or with high resolution ADCs. That doesn't mean
that digital currents in a ground plane can't deteriorate the performance of
sensitive analog systems, but just that splitting the ground plane is
usually not the best solution : It is usually far better to avoid the issue
with a good PCB placement, just putting the digital stuff (and digital power
supplies) on one side of the board and the analog ones in the other...
For me the big issue with splitted ground planes is magnetic coupling : If
you have two ground planes, one for section A and the other for section B,
connected in one point X then ALL signals and power tracks interconnecting
sections A and B should be routed exactly through point X too, or you will
have a marvelous current loop, that could either grab parasitic signals or
generate some spurious...
Friendly,
--
Robert Lacoste
ALCIOM - The mixed signal experts
www.alciom.com
Spehro Pefhany
<see-my-...@www.alciom.com> wrote:
>"John Larkin" <jjla...@highNOTlandTHIStechnologyPART.com> a écrit dans le
>message de news: qrlct193hgcr22dsn...@4ax.com...
>> Use a single, common ground plane and you won't have this dilemma.
>>
>> Splitting grounds rarely makes sense.
>
>I fully support John & Joerg advice : In 99.999% of the designs I've worked
>on a plain common ground plane was the best solution. And that includes
>projects in the GHz range or with high resolution ADCs. That doesn't mean
>that digital currents in a ground plane can't deteriorate the performance of
>sensitive analog systems, but just that splitting the ground plane is
>usually not the best solution : It is usually far better to avoid the issue
>with a good PCB placement, just putting the digital stuff (and digital power
>supplies) on one side of the board and the analog ones in the other...
Or slot it with a global keepout over the slot to keep current where
you want it and signal traces away.
Best regards,
Spehro Pefhany
--
"it's the network..." "The Journey is the reward"
sp...@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
soar2...@yahoo.com
run any high speed signal lines over the slot, especially digital
signals. Doing this will force the return currents to flow around the
slot, turning it into an excellent radiator (read "antenna").
Tom
Spehro Pefhany
Ken Smith
John Larkin <jjla...@highNOTlandTHIStechnologyPART.com> wrote:
[... splitting ground plane ...]
>It does allow you to avoid thinking about a few things. But magnetic
>field pickup and emi issues are still there in full force. And al
>least some ground loops will persist. So you may as well save a plane
>and do it right.
If you split a ground plane to avoid thinking about things, you are doing
it for the wrong reason and are likely to have trouble. The reason to
split the plane is because you have to think about more things by the
nature of the design.
If you have 20A pulses of current and 10uV low frequency signals on the
same PCB you need to think about a lot of things.
--
--
kens...@rahul.net forging knowledge
Pooh Bear
Ken Smith
John Larkin <jjla...@highNOTlandTHIStechnologyPART.com> wrote:
[...]
>I'll occasionally nestle a low-level circuit in the middle of a
>c-shaped cutout of the ground plane, so plane currents don't make
>microvolt drops within the sensitive area. You can get similar results
>by creating a square of ground surrounded by a break in the plane,
>with just some small connecting tabs left; the resistivity of the tabs
>is enough to greatly reduce the current through the sensitive region,
>but the thing will still be emi tight.
[.. faints and recovers ..]
you split planes
>But external magnetic fields will still create potentials across a
>plane, so it's best to keep loops small and handle delicate stuff
>against one ground via, or differentially. Fans are killers, both from
>mag fields and (for dc fans) nasty pulse currents.
BTW: Digital sections almost always drive you to having more layers than
the analong needs. If you fill the unused space by making more ground
planes, you can improve matters.
You can bring all the switcher stuff together on a local plane above the
system ground and then take it to system ground at one place. Ideally,
the one point of connection is the same as where all the input and output
filters come together.
John Larkin
(Ken Smith) wrote:
>In article <nu6et1t5koanr2p9v...@4ax.com>,
>John Larkin <jjla...@highNOTlandTHIStechnologyPART.com> wrote:
>[...]
>>I'll occasionally nestle a low-level circuit in the middle of a
>>c-shaped cutout of the ground plane, so plane currents don't make
>>microvolt drops within the sensitive area. You can get similar results
>>by creating a square of ground surrounded by a break in the plane,
>>with just some small connecting tabs left; the resistivity of the tabs
>>is enough to greatly reduce the current through the sensitive region,
>>but the thing will still be emi tight.
>
>[.. faints and recovers ..]
>you split planes
>
No, at least that's not what we call a split plane; a split plane has
different nets, by our definition (ie, by PADS.) I meant to just cut a
little C-shaped insulating path into the plane. It's still fully
connected along one edge. You can align the orientation of the C with
respect to known or suspected ground plane current directions. All
we're doing is steering circulating currents around the tender parts.
You can similarly do link-shorted gaps to break thermal paths, to
reduce thermal gradients in sensitive circuits like thermocouple amps.
It's still all the same node.
>>But external magnetic fields will still create potentials across a
>>plane, so it's best to keep loops small and handle delicate stuff
>>against one ground via, or differentially. Fans are killers, both from
>>mag fields and (for dc fans) nasty pulse currents.
>
>BTW: Digital sections almost always drive you to having more layers than
>the analong needs. If you fill the unused space by making more ground
>planes, you can improve matters.
I can't recall ever having more than one ground layer in my stuff, or
pouring ground regions on other layers.
>
>You can bring all the switcher stuff together on a local plane above the
>system ground and then take it to system ground at one place. Ideally,
>the one point of connection is the same as where all the input and output
>filters come together.
Again, I don't do that. It would tend to make the switcher ground have
a lot of ac noise relative to real ground, which could radiate ringy
things. I just always use a single ground plane.
Oh, another heresy: I bolt the pcb ground plane to chassis ground in
as many places as possible.
John
John Larkin
I do that, and worse, routinely. Splitting planes generally makes the
emi situation more hazardous, and emi can wreck a 10 (or 0.1)
microvolt signal just as badly as a ground loop, but it's a lot less
analyzable. We go for max emi integrity first (single ground plane,
hard-bolted to the chassis often) and then work around any
low-frequency loop problems locally.
I know people who use separate analog and digital ground planes with a
single-point connection, then pepper the board with big plane-to-plane
bypass caps everywhere to kill hf plane potential differences!
John
John Larkin
That's not usually a real problem. For normal (say, 1 ns or slower)
risetimes, a reasonably-sized slot is invisible. And more importantly,
a thin slot in the ground plane is usually spliced over by the
presence of nearby (by design!) power planes. With thin dielectrics,
the ground+power plane structure is really a single ac-equipotential
structure. HoJo and his apologists conveniently ignore plane-plane
capacitance when they analyze "return currents" and bypassing, and
proceed to make some mighty silly recommendations. The gap is actually
shorted by adjacent planes.
Just TDR a trace that runs over a ground-plane slit on a multilayer
board. Even at 20 ps risetime, not much happens.
John
Joerg
>
> Want to analyse something useful... analyse an evaporative cooler...
> water changing states is MARVELOUS ;-)
>
And mind the legionaire's disease that spreads when the water didn't
change state for a long time ...
Regards, Joerg
Jim Thompson
That occurs in water re-use situations. I've not heard of
Legionnaire's from an ordinary "swamp cooler".
Joerg
>
> That occurs in water re-use situations. I've not heard of
> Legionnaire's from an ordinary "swamp cooler".
>
Maybe. When I walked into our local HW store one day they just turned on
the swamp cooler. One guy mumbled "Oh rats, they didn't drain it last
fall" and one heck of a stench filled the air.
Regards, Joerg
Jim Thompson
Usually there's so much mineral content that the sump won't support
any form of life ;-)
When I had one (~15 years ago) I had a bleed so that there was always
a fair amount of fresh water in circulation.
Joerg
>
> Usually there's so much mineral content that the sump won't support
> any form of life ;-)
>
Not out here. They keep the tap water at high chlorine levels to avoid
forms of life in there. Last time I measured it was almost 2ppm. Problem
is, when it sits in that pan on a hot roof the chlorine is gone within a
week.
Regards, Joerg
Joerg
> That's not usually a real problem. For normal (say, 1 ns or slower)
> risetimes, a reasonably-sized slot is invisible. And more importantly,
> a thin slot in the ground plane is usually spliced over by the
> presence of nearby (by design!) power planes. With thin dielectrics,
> the ground+power plane structure is really a single ac-equipotential
> structure. HoJo and his apologists conveniently ignore plane-plane
> capacitance when they analyze "return currents" and bypassing, and
> proceed to make some mighty silly recommendations. The gap is actually
> shorted by adjacent planes.
>
I think what Spehro meant by global keepout is truly a 100% keepout. No
parts, traces or planes crossing that slot.
Still, I am not a believer in slots. I usually take them back out of
existing designs. Slots can make nasty antennas and can lead to egg in
the face when an unexpected redwood forest shows up on the analyzer at
the EMC lab (that is usually when my phone rings off the hook).
Regards, Joerg
Spehro Pefhany
<notthis...@removethispacbell.net> wrote:
>Hello John,
>
>
>> That's not usually a real problem. For normal (say, 1 ns or slower)
>> risetimes, a reasonably-sized slot is invisible. And more importantly,
>> a thin slot in the ground plane is usually spliced over by the
>> presence of nearby (by design!) power planes. With thin dielectrics,
>> the ground+power plane structure is really a single ac-equipotential
>> structure. HoJo and his apologists conveniently ignore plane-plane
>> capacitance when they analyze "return currents" and bypassing, and
>> proceed to make some mighty silly recommendations. The gap is actually
>> shorted by adjacent planes.
>>
>
>I think what Spehro meant by global keepout is truly a 100% keepout. No
>parts, traces or planes crossing that slot.
Just planes- more like what John was talking about.
>Still, I am not a believer in slots. I usually take them back out of
>existing designs. Slots can make nasty antennas and can lead to egg in
>the face when an unexpected redwood forest shows up on the analyzer at
>the EMC lab (that is usually when my phone rings off the hook).
>
>Regards, Joerg
>
>http://www.analogconsultants.com
Too bad most of the time the layout can't be dictated by electrical
considerations. There's always some high voltage or current stuff, a
heatsink or a terminal block or a big chunk of routing that forces
things to be sub-optimal from an electrical pov. Well, maybe it isn't
too bad. If it was easy anyone could do it.
Ken Smith
John Larkin <jjla...@highNOTlandTHIStechnologyPART.com> wrote:
[...]
>different nets, by our definition (ie, by PADS.) I meant to just cut a
>little C-shaped insulating path into the plane.
Its been a few years since I did my own layout. I was using Orcad PCB.
I'd do this:
********************************
********************************
**** ********
**** ****************** ******** <-- Sometimes this side wouldn't
**** ****************** ******** be here because it ran to the
**** *******GND1******* ***GND2* PCB edge
**** ****************** ********
**** ****************** ********
**** ****************** ********
**** ********
********************************
********************************
Then just before making plots I'd drop in copper
********************************
********************************
**** ********
*********************** ********
*********************** ********
************GND1******* ***GND2*
*********************** ********
*********************** ********
*********************** ********
**** ********
********************************
********************************
I'd then be careful never to do a rules check on it after this. If I ever
neede to make a change, I'd bust the planes apart again.
[...]
>>You can bring all the switcher stuff together on a local plane above the
>>system ground and then take it to system ground at one place. Ideally,
>>the one point of connection is the same as where all the input and output
>>filters come together.
>
>Again, I don't do that. It would tend to make the switcher ground have
>a lot of ac noise relative to real ground, which could radiate ringy
>things. I just always use a single ground plane.
No, it makes the ground of the supply less noisy and less ringy. The
copper under the switcher is an added plane. Its impedance is as low or
lower than the one that extends over the whole PCB.
At UHF frequencies, the capacitance between the two planes effectively shorts
them together. At the frequency the switcher runs at, it ensures that all
the AC currents remain under the switcher and don't go crawling all over
everything else.
>Oh, another heresy: I bolt the pcb ground plane to chassis ground in
>as many places as possible.
Do you used to work for Power10? No, on second thought don't tell me.
They tied things to the chassis in many places. As a result they
introduced about 1.5Vp-p spikes into all the sheet metal on a 19 inch
rack. If you hooked scope ground to one back corner of their chassis and
the tip the the far front corner you'd see all this crap. We had to
create isolated rack mounting hardware and stuff to use them.
Joerg
>
> I can't recall ever having more than one ground layer in my stuff, or
> pouring ground regions on other layers.
>
That's because you only design high-end stuff. There are situations
where it all has to be on single layer phenolic because it's a nickel
cheaper than double sided FR4.
>
> Oh, another heresy: I bolt the pcb ground plane to chassis ground in
> as many places as possible.
>
Same here. Then I make sure that the plating and surface coatings are
compatible so that a low impedance path stays low impedance over the years.
Regards, Joerg