shorting of analog and digital grounds

[antima]

presently i am working on mixed signal design ,i want to know wich
component is recommended for shorting analog and digital grounds
1)inductor
2)zero ohm resistor
3) diode (schottky)

 

[Richard Henry]

presently i am working on mixed signal design ,i want to know wich
component is recommended for shorting analog and digital grounds
1)inductor
2)zero ohm resistor
3) diode (schottky)

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]

thanks for ur respose can u elaborate ur ans.i am unable to come to
conclusion

 

[Richard Henry]

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]

presently i am working on mixed signal design ,i want to know wich
component is recommended for shorting analog and digital grounds
1)inductor
2)zero ohm resistor
3) diode (schottky)

Worry more about 'where' you are going to join the grounds, than 'how'.
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]

presently i am working on mixed signal design ,i want to know wich
component is recommended for shorting analog and digital grounds
1)inductor
2)zero ohm resistor
3) diode (schottky)

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]

presently i am working on mixed signal design ,i want to know wich
component is recommended for shorting analog and digital grounds
1)inductor
2)zero ohm resistor
3) diode (schottky)

Use a single, common ground plane and you won't have this dilemma.

Splitting grounds rarely makes sense.

John

 

[Richard Henry]

Use a single, common ground plane and you won't have this dilemma.

Splitting grounds rarely makes sense.

Unless dealing with high-quality video, or high-quality audio, or noisy
analog loads, or noisy digital loads.

 

[Joerg]

Hello Richard,


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]

Use a single, common ground plane and you won't have this dilemma.

Splitting grounds rarely makes sense.

It certainly does in audio.

Graham

 

[John Larkin]

It certainly does in audio.

Graham

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]

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.

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]

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.


Is that some kind of elitist snobbery ?

Graham

John is still miffed about being called on his defiance of the Laws of
Thermodynamics ;-)

...Jim Thompson

 

[John Larkin]

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.


Is that some kind of elitist snobbery ?

Graham

No, it's an opinion that the things that apparently matter in audio
are subjective and non-quantifiable.

John

 

[John Larkin]

John is still miffed about being called on his defiance of the Laws of
Thermodynamics ;-)

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]

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

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 ;-)

...Jim Thompson

 

[Pooh Bear]

No, it's an opinion that the things that apparently matter in audio
are subjective and non-quantifiable.

John

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]

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

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]

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.

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]

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.

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