[SI-LIST] Re: Ground Article
- From: Chuck Corley <corley@xxxxxxxxx>
- To: si-list@xxxxxxxxxxxxx, si-list-bounce@xxxxxxxxxxxxx
- Date: Wed, 18 Jan 2017 10:27:59 -0800
... But noisy power supply rails (feeding digital
circuitry) will not do you well when they also feed sensitive analog.
Thoughts:
Agree with you 100%. Clean power supplies are critical for low-noise
circuits. For 32+ years I've designed sensitive ADCs, DACs, amplifiers
and other analog and digital circuits in measurement instruments for HP,
Agilent, Keithley and National Instruments. When we've hunted down
noise sources on sensitive circuits it's usually from the power supply,
or from external noise coupling onto the power supply.
Basics: Typically for feeding ultra-sensitive analog circuits that want
very low noise levels I follow a switching power supply by a low-noise
LDO (Low Dropout) Linear regulator. The voltage output of the switching
supply needs to be high enough to accommodate the extra headroom needed
by the LDO.
- In addition to using an LDO to reduce power supply noise, there are
also special LDOs that are specifically designed for extra-low noise
output.
- The LDO should be far enough from switching supplies (and their
surrounding input and output caps and components) so that the switcher
circuit can't couple noise into the LDO circuit.
- In extreme low-noise cases separate smaller LDOs that independently
feed separate or isolated RF/analog amplifiers or ADCs/DACs may be
required to isolate noise between circuits. But ... be careful to not
go crazy and start doing this extra-extreme level when it's not really
needed.
- Once you make the supply clean, keep it clean! Don't put noisy
digital circuits on the same supply. Don't run that clean low-noise
supply on a plane right up a few mils under another noisy power plane on
an adjacent layer that could couple noise into it. Etc.
Split planes:
So far I still haven't seen a side-by-side comparison where split planes
actually helped (for cases when the board has actually implemented a
SOLID ground plane). Typically I see split planes used for isolation
purposes and then later they find it didn't help or made the noise worse
and then went back to solid planes. Just because someone's done split
planes on a design doesn't mean it actually helped. ;-) One of the
side-by-side tests I saw done was when I was at Cisco/Cerent, after
measurements showed the noise had actually increased by splitting the
planes they also went back to merging the planes back together. I think
split planes don't usually help because a true solid plane is an almost
magical thing, normally it's low inductance makes it hard to accumulate
enough of a voltage drop to create significant high frequency noise
between points on the plane. To be clear, I think it's possible noise
could be generated across points on a good solid plane, but it's fairly
hard to generate enough noise to cause a problem if it's a truly solid
plane.
Now if a plane is NOT solid (one example: Under BGA ICs where the ground
plane has been eaten into too far) then it's not too hard to get enough
voltage across the ground plane to create noise in sensitive circuits,
and that's when people I've talked with said split planes helped. But I
think split planes helped in those cases because they didn't have a
truly solid ground plane in the first place and that's what actually
caused the noise problem.
Chuck Corley
National Instruments
On 2017-01-18 08:50, Kevin G. Rhoads wrote:
On 1/18/2017 9:16 AM, Istvan Novak wrote:
in mixed-signal boards and on some boards
where naturally very noisy circuits (like high-current DC-DC converters)
have to coexist with sensitive circuits, like SerDes, PLLs, etc,
splitting the ground plane between the domains can improve the isolation
and thus improves performance, as long as we make sure that we dont
create SI and/or EMI problems by crossing the splits with signals that
otherwise need a solid reference.
First the disclaimer -- I mostly follow the SI list to learn.
Next the other disclaimer -- the board I speak of is a one-off for
university research use.
We had a multi-use board (i.e., K101) for instrumentation support on
sounding rocket science flights. Among other things, it digitized three
axis analog magnetometer data and formatted that to feed by synchronous
serial to NASA telemetry.
Four layer board
16 bit ADCs
10 MHz synch serial channels (NASA WFF 93 standard)
SOLID ground plane
Nominal +28 from NASA batteries through DC-DC for +/-12 and +5 in box
First gen used filtering -- digitized channels fairly noisy -- even
though the +5 was decoupled to death -- distributed decoupling across
the board -- both electrolytics and 0.1 microF-s
Second gen split the +5 supplies -- generating one from +12 via LM7805
and that one supplying the ADCs and +5 op-amp front ends.
MUCH less noise
So -- the SI issues are truly serious. I have nothing to add to that
except solid ground plane with separate power supplies can work well in
some situations. But noisy power supply rails (feeding digital
circuitry) will not do you well when they also feed sensitive analog.
If anyone wants to see the details (links are "public"):
Cascades 2 board photos
https://www.facebook.com/kgrhoads/posts/4915482088395
Cascades 2 instrumentation
http://www.facebook.com/album.php?aid=2050033&id=1338500958&l=c23c77bae3
Cascades2 schematics
https://www.facebook.com/media/set/?set=a.4916478913315.1073741831.1338500958&type=3
Sincerely
Kevin
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
