|
3
Gordon Drive, P.O.Box 1347 Rockland, Maine 04841 U.S.A.
|
|
© 2004 Avocet Systems, Inc.
|
Call
Us Today at 207-596-7766 ("Picton Press")
|
|
Avocet Systems, Inc. : The Complete Solution for Embedded Systems
Development Tools
|
|
|
Embedded Update
The European CE Standard (2)
The last issue of the Embedded Update talked about the new European CE standard
for controlling emissions from electronic devices. It sparked quite a few replies.
Thanks to everyone!
Most were of the following nature: "I pity any small company trying to get this
CE mark. We ended up assigning an engineer half time for a year just to figure
out what it was all about."
Others replied that compliance testing costs around $10,000 US, though re-engineering
to make a failed product meet the specs can run much more.
Two long replies we're reprinting in full, with permission from their authors.
The first is from Sy Wong, who is "active in pointing out to whoever will listen
the herd instincts of current computer designs such as the 1970-era concepts
used in current processor architectures, operating systems and the lack of rationale
with VHDL. His favorite subject is the Safety-Subset of an ISO standard as a
simple unified programming and hardware description language (HDL) to support
embedded systems developments with public domain free software tools." - His
words. Sy's comments are more historical, but are fascinating none the less.
----- Sy Wong's Comments --------
I know of no certifications for anything but had some experience with shielding.
My first brush with shielding was at the GE High Voltage Lab in the 40s to develop
instrumentation for measuring corona discharge with photomultipliers. After
ascertaining that line filters were satisfactory, I set up the photometer instruments
in a screen room with myself in it. First I must measure how well the room screens
by measuring from top to bottom with a special scope and somebody let go a 5-mega
volt discharge to the top of the screen room. I was somewhat paled by the 100,000-volts
measured. We had a special pulse scope with deflection sensitivity something
like 25kv per inch.
The next was at the Institute for Advanced Study on electrostatic memories using
cathode ray tube secondary emission on the CRT face of about 1 mv. We had to
use double copper and mu-metal shields specially heat treated plus careful grounding.
It certainly worked. The chief engineering was a radar man from Hazeltine and
knew all about shielding, things like opening shapes, circulating currents,
etc.
Finally in 1955 I landed the job of project responsibility for a computer contract
with NSA at Philco, also a radar manufacturer during WWII. The surface barrier
transistors of that time were very fast, with small voltage swings (about .3
volt) at relatively low current. The customer had a very experienced project
manager wanted proof that the computer will be immune to EMI. We showed him
our demonstration box with a preset counter that ran from a battery in a shielded
screen box that lights will stop blinking if stopped on comparison errors. Without
the box, the counter stopped often. He was not convinced.
I then put the shielded box in another screen shield box with a single contact
point between the two and let him point the output feed of a magnetron airborne
radar transmitter to the screen box. It kept on working but he was not satisfied.
I then handed him a Tesla coil that was used to test vacuum systems. He sparked
all over the screen box and It kept on working but he was still not convinced.
He said that a Tesla coil had very little power.
He looked around the lab, and found a high voltage power transformer with big
bushings for the radar set. He pulled a metallic arc on the screen box and it
kept on working. He had to acknowledge that I was right that good shielding
practice works, especially when induced circulating currents are controlled.
I believe that EMI source and receiver may be reversed but the principle of
careful shielding and grounding should be the same. The radiation tests for
military security in anechoic chambers have been well established for many years.
You can make a cookbook of good "overdesigns" that have a high probability of
passing tests but what is just enough seems to involve too many variables to
be possible. Cost constraints makes the design much harder. Another way is to
examine the standard to see if it is realistic or necessary for given circumstances.
Arbitrary laws are unconstitutional in U.S.A.
------- Comments from Cobus De Beer -----------
Mr. De Beer had a more technical reply. You can contact him at Cobus de Beer,
TSP (Pty) Ltd, P. O. Box 9, Irene 1675, Pretoria, South Africa, E-mail: CdB.TSP@pixie.co.za
I do some work as a consultant to a company that sells suppression components.
When they get a system from a customer that they cannot suppress they use my
services. I learned how to suppress systems while working as a designer of military
equipment and there the specifications are a lot tougher than in commercial
equipment.
Compliance testing is done in accordance with rigidly laid down procedure. The
most important aspect of the testing is that the equipment must be used as it
will be used by the consumer. The applicable limits depend on the type of equipment
being suppressed (or tested) for compliance. Power tools have different limits
from computing or video equipment. Medical equipment generally has the most
stringent limits. The limits for consumer equipment is also stricter than the
limits for industrial equipment. The only way to be sure as to which category
a piece of equipment belongs to, is to get a ruling from the certifying authority.
We find that we often disagree with the authorities as to which set of limits
apply because we disagree as to the category that the equipment belongs to.
The requirements fall into two broad categories, radiated emission and conducted
emission. It is of little use to attempt radiated emission certification without
knowing that the equipment satisfies the limits for conducted emission because
any suppression components fitted to correct conducted emission problems normally
alter the radiated emission characteristics.
Suppression design is best done experimentally because the stray capacitance
and inductance of the circuit is of critical importance and they are difficult
to estimate to allow the analytical design of a suppresser.

The circuit above is the first step in the design of a suppresser for conducted
emission. C1 is a non-polarized capacitor with a mains rating. (X - Capacitor)and
L1 is a current compensated inductor. It is wound on a ferrite core (not powdered
iron) and is selected for maximum inductance at the maximum load current of
the equipment under test. (BTW this discussion is for electronic equipment not
for power tools. A power tool normally only uses C1) The filter is added to
the equipment and a sweep is performed to find the emission spectrum. If the
emission is above the limit then C1 is increased until the emission on the L
and N lines have equal magnitude. Any further increase in the value of C1 will
have very little effect on the attenuation of the filter. This filter is effective
from low frequencies up to about 5MHz. Above 5MHz the (parasitic) inductance
of C1 becomes important and the effectiveness of the filter decreases. If you
need more attenuation the inductance of L1 must be increased. Use the smallest
value for L1 that produces the required low frequency attenuation. Higher values
result in less effective filtering at high frequencies because the interwinding
capacitance of L1 increases with increasing values of L. Typical values of C1
is 0.1uF to 1uF with typical values for L1 ranging from 500uH to 30mH. More
attenuation can be obtained by adding two filters in series, one for lower frequencies
and on designed for higher frequencies.
If more low frequency attenuation is required then section as follows can be
added between the filter and the equipment.
This time L2 must be an iron powdered inductor and rated for the maximum load
current drawn by the equipment. Again C2 is increased until the emission in
the two conductors are balanced. These sections are effective below 2MHz.
If there is troublesome emissions between 5MHz and 30MHz then the solution depends
on the nature of the equipment. If the equipment has an earth then Y-capacitors
are added to the first filter.
These capacitors introduce earth currents and their maximum values are determined
by safety regulations. Typical values are 4n7F. These capacitors have a great
effect on the 10MHz to 30MHz band and reduce emissions substantially in this
band Values are normally limited to less than 22nF by safety regulations. Medical
equipment generally prohibit the use of Y capacitors for safety reasons. The
capacitors must have a Y rating and are specially manufactured for this application
because if they fail to short circuit then the mains may be connected to the
equipment case.
This type of filter is useful for most electronic equipment. Switch mode supplies
normally require the addition of the L section before the current compensated
choke filter. Switch mode supplies are normally easier to suppress when the
earth lead is not present. A small value capacitor (5nF as starting value) on
either side of the current compensated filter in parallel with C1 (on the one
side) helps greatly to reduce the high frequency noise. Since the emissions
is measured between the mains leads and earth, it is desirable not to contaminate
the earth line with noise.
The filters shown above will not be effective in circuits that contain triacs
or SCRs. The noise must first be suppressed right at the triac.
The filter components L1 and C1 should be added where indicated. L1 must be
an powdered iron inductor.
Radiated emission is controlled through packaging and the use of proper layout.
Double sided PCBs are a lot (20 to 80dB) quieter than single sided boards. Earth
planes make a big difference. Current loops must be as small as possible. Four
layer boards are another 20 to 40 dB quieter than double sided boards. Decoupling
of fast ICs is of critical importance. Fast digital lines must be terminated
(like transmission lines).
In summary: Filters are best designed in the lab but it is a tremendous help
if the PCBs include room for all the components that might be needed.
The safety design of the product influences the design of the EMI filters.
When designing filters aim for at least 3dB below the limit to ensure a first
pass success at the certification test facility. If you are within 2dB of the
limit they will insist on testing more units to ensure compliance. ( Filter
components normally have +-20% tolerance not +-5% like other components.)
|
|
|