[SI-LIST] Re: Off-topic? Questions about measuring power supply ripple/noise (PARD)
- From: "istvan novak" <istvan.novak@xxxxxxxxxxxxxxxx>
- To: <AD123338@xxxxxxxxxxxxxxxxxxxxxxxxxx>, <si-list@xxxxxxxxxxxxx>
- Date: Thu, 15 Aug 2002 06:44:34 -0400
Adam,
The procedure you want to follow may depend on whether you do
design validation in the lab, or want to create production testing for
the manufacturing floor.
The 20MHz scope bandwidth is the typical value people use on
production testing. It has a couple of reasons I can think of:
- in a production environment it would be difficult to make
sure that >20MHz noise contents is NOT picked up from the
environment
- in the old days, 20MHz covered generously the possible
switching frequencies even if you applied the 100xswitching-frequency
rule
A well-designed switching regulator will generate an approx.
sawtooth-like periodic output ripple (corresponding to the
switching frequency), plus a transient noise
part, which in case of random load, looks like a random noise.
The random noise with DC load should be much smaller than the
periodic ripple.
Not well-behaved and/or faulty switching regulators may
exhibit a) deterministic or random subharmonic oscillation
(if this happens, it is easy to catch the different ripple
frequency and magnitude with a 20MHz scope), or b)
periodical or slightly random self oscillation at a frequency
unrelated to but lower than the switching frequency
(should be equally easy to catch with a 20MHz scope),
or c) strong high-frequency ringing sitting on the
switching edges (its frequency may be much higher than 20MHz.,
so catching it will not be possible with a 20MHz scope).
In lab validation environment it should be possible to should
be possible to solder coax cables with short pigtail to the
output points, and to use a wide-band scope to check for all
possible problem, and to check the transient behavior if
your electronic load has pulsed-load capability, or to
connect the coax cable(s) to a network analyzer to
measure the small-signal output impedance.
You are right, the 10uF + 0.1uF capacitors mostly do
not represent actual usage conditions. In best, those are
there to suppress the extra noise you may pick up from the
loose cable bunch you have at the power-supply's output.
A well-designed regulator should not require any external
capacitor to meet the specification within its stated regulation
bandwidth. Outside of the regulation bandwidth we have to
rely on external bulk and bypass capacitors anyway.
Catching the peak-to-peak noise in infinite persistence is a
good way, but first you have to make sure that the scope probe,
the probe cable and possible ground loops associated with the
scope and DUT will not produce too much erronous noise pickup.
you can check it by using the same exact setup you plan on using
for the measurement, except connect both the hot and ground
points of your probe to the ground point on the DUT. Taking the
reading in this setup will give you the background noise reading.
You may find that switching AC loads (for instance lights, or
motors) or ESD discharge in the neighborhood would generate
big background noise reading.
Regards
Istvan Novak
SUN Microsystems
----- Original Message -----
From: "Dixon, Adam" <AD123338@xxxxxxxxxxxxxxxxxxxxxxxxxx>
To: <si-list@xxxxxxxxxxxxx>
Sent: Thursday, August 15, 2002 12:29 AM
Subject: [SI-LIST] Off-topic? Questions about measuring power supply
ripple/noise (PARD)
>
> This is my first post and may be slightly off-topic. I am trying to
better
> understand the hows and whys of measuring power supply ripple/noise. I
have
> seen this referenced as "Periodic And Random Disturbance" or PARD in
several
> power supply suppliers' application notes. Any comments or recommended
> reading/references are greatly appreciated. I searched the SI-list
archives
> unsuccessfully. I am working with 90W to 300W multiple-output PC power
> supplies (+3.3V, +5V, +12V,...) and several of the industry standard power
> supply design guides (ATX, SFX, TFX, WTX form factors) all state to use a
> 10uF electrolytic and 0.01uF ceramic capacitor on each ouput and to
measure
> with an oscilloscope set to 20MHz bandwidth. My questions are (comment on
> as many as you like, please!):
>
> 1) Why 20MHz scope bandwidth? I saw one Intel application note
indicating
> that bandwidth should be approx. 100X the switching frequency -- is this
the
> reason? If I use full bandwidth, I see a higher peak-to-peak noise value
> but the 20MHz value seems to be an industry standard.
>
> 2) The 10uF and 0.01uF "load" capacitor values do not generally correlate
> with the actual application loads. Two or three of the outputs have 200uF
> or more of bulk/electrolytic capacitance and another few uF's worth of
0.1uF
> and/or 0.01uF bypass/decoupling capacitors. Is this a legacy issue to
> provide a common measurement configuration? The electronic loads are used
> in constant current mode with the ability to dynamically load one or more
> outputs simultaneously.
>
> 3) What is the best way to capture the maximum peak-to-peak noise? I am
> using a TDS7xx scope and don't see a way to create a trigger condition for
a
> max/min condition. I am wondering about the validity of setting the
display
> for infinite persistence mode and capturing a large number of samples on a
> slow timebase (100uS to 1mS?) to get a good view of how the output varies.
> Does this sound valid? I haven't seen anything in print which discusses
> this.
>
> 4) Probing where the electronic load is connected is difficult due to the
> distance between the +/- connections. One application note recommended
> removing the plastic insulator of the scope probe to get the shortest
> possible ground connection in order to minimize measurement error from
noise
> coupling. I have hacked up a passive Tek probe ground wire to shorten it
to
> approx. 2" including clip -- any thoughts about improving the probing?
>
> 5) One last scope question: If I am observing a +5VDC output, what is
the
> difference between using AC couple mode versus DC couple with a +5V offset
> applied? Is there a significant difference internal to the scope that
> favors either mode? I guess I will go to the lab and try both but any
> comments are welcome.
>
> Thanks for reading!
>
> Sincerely,
> Adam Dixon
> NCR Corporation
> adam.dixon@xxxxxxx
> (770) 623-7093
> ------------------------------------------------------------------
> 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 archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> Old (prior to June 6, 2001) list archives are viewable at:
> http://www.qsl.net/wb6tpu
>
>
------------------------------------------------------------------
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 archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
