DcDc converter Module level EMI Measurements

1
Dc-Dc converter Module level EMI Measurements

• Peter Miller

2
Why is this isan issue?

• EMI is an issue with a high power Dc-Dc
  converter.
• Current EMI specs do not cover (Bidirectional)
  Dc-Dc converters as they assume only 1 battery.
• The EMI spec (including how its measured) can
  have a large impact on the converters cost.
• Different ways to measure Dc-Dc EMI for each OEM
  will increase the cost impact even more
• Not a major technical issue - we just all need to
  agree on the same (reasonable) way to test
• The ultimate test will always be when its wired
  correctly into the real car

3
Key Assumptions

• Basic test methods will not change (eg CISPR-25)
• They are well proven
• Radiated emissions are not a function of battery
  voltage and so limits as well as test methods
  need not change
• Conducted emissions (assumed to basically be a
  simpler way to measure low frequency radiated
  emissions) are also largely independent of supply
  voltage (low frequency radiated emissions to
  1st order are proportional to current)
• LISN will be left at 5uH/50 Ohm.
• Only reasonable value that is readily available.

4
What needs to be defined?

• Dc-Dc converter EMI depends upon its input
  voltage and output load (resistance impedance)
• This is particularly true lt30MHz where power
  switching is main cause of EMI.
• As frequency increases most converters EMI
  becomes less variable to input voltage and output
  load resistance.
• To keep test time reasonable, only one typical
  condition should be tested.
• We thus need to define input voltage to converter
  and output load.

5
Motorola Proposal

• Keep test setup as defined in CISPR-25 (and many
  other standards)
• Including use of 1 or 2 LISNs as required.
• Run converter at 1/15 full power
• Defines output load resistance
• At these powers a simple free air cooled resistor
  is a suitable load (testing at full load may mean
  dissipating 1kW or more which is much more of an
  problem - for example EMI from a cooling fan
  could mask real EMI issues)
• Input power can be reasonably supplied by a
  battery for the duration of the test (again this
  could be a problem at full load).
• Avoids possible issues with converter stability
  if run at full power from 5uH source.
• Set input voltage range to allow operation from
  batteries (eg 36.5/-2.5V) batteries are a low
  impedance source that do not generate EMI
• If converter is designed to be 4-wire connected
  to a battery then add a 10mF capacitor (
  -20100 tolerance, with good high frequency
  performance) in parallel with the load to
  approximate this.

6
Please note

• For the commonest converter topology (buck
  converter 42-14V) in normal operation
• output current ripple amplitude is independent of
  load
• Input mark-space ratio will change with load
• but, switching edge rates are largely independent
  of load.
• EMI is therefore not very sensitive to the load.

VoutVinDc DcDuty cycle (0-1) so Vout lt
Vin Switches operate in antiphase (or S2
is replaced by a diode).
Iout
Iin
I
Time
7
Other background information

• Typical dual voltage architecture
• Dc-Dc converter operation
• Other views on 1/15.
• Diagrams of test setups.
• Conclusions

8
Dual 14/42V Architecture
42V
Vehicle loads
Jump start post
Alternator incl. Regulator Rectifier functions
or ISA
DC-DC Converter
14V
42V
Vehicle loads
12V Battery
Remember, this is just one example architecture.
36V Battery
Energy Management Software
9
Operation into batteries

• In a 42V PowerNet application the Dc-Dc converter
  drives a complex load that may consist of just a
  battery, it may be inductive, a motor, capacitive
  or resistive.
• Stability must be demonstrated into effectively
  any load
• At low loads (a fully charged battery)
  subharmonic oscillation may result in audible
  noise
• Driving a flat battery requires a constant
  current drive (which can again result in
  subharmonic oscillations).
• Dc-Dc operation mode ( thus EMI) depends upon
  battery state, load, etc.

10
Driving batteries continued.

• 4 wire connection to batteries (as shown below)
  will help with EMI by using battery as part of
  EMI filter.
• Current module level test specifications do not
  normally allow this set-up for testing.

Short wiring
11
Issue - voltage drops on leads

• High currents (particularly on 14V side) require
  short thick cables.
• Convertor output is constant power, thus the
  lower the input voltage the higher the input
  current!
• E.g. 12V source, 40mOhm interconnect resistance
  for 500W output 50Amps will be drawn 2V drop on
  cables and converter sees 10V at its terminals.
• Cu wire 25mOhm mm2/m _at_120oC
• Changes of input current will alter EMI,
• thus changes in interconnect resistance will also
  change EMI.

12
Contradictory issues

• To be realistic Dc-Dc converter must be connected
  to batteries.
• But, battery state and voltage may have a large
  impact on EMI.
• Standard solution is to use a LISN
• This is 5uH/50Ohm impedance which was designed to
  approximate leads and battery.
• 1m long (in total) 0.7mm2 cable is 25mOhm and
  0.33uH.
• Battery is 1mOhm and has complex variation of
  impedance with frequency (but much lower
  impedance than cable).
• 5uH and 0 (?) Ohm of LISN is very different from
  cable battery
• At 100KHz, LISN Z3 ?, 0.33uH25m ? 0.23 ?.

13
What impact does this different impedance have?

• At 14V, 700W is 50Amps.
• If converter runs at 100KHz and filtering in
  converter causes 20dB attenuation of current at
  100KHz (so current now 5Amps)
• 3 ? impedance and 5Amps would cause 15V drop
  (hard on a 14V supply!) gt converter may not
  function at all.
• 0.23 ? impedance would cause 1.15V drop (note
  with 1.15V drop converter will need more current
  due to its constant power operation, it will
  actually draw 55Amps seeing 12.7 Volts).

14
But I cannot buy a 0.33uH LISN

• If converter is run at 1/15 full power this will
  give approximately the same voltage on the
  converter input (fed via 5uH) as it would see at
  full power with a 0.33uH inductance.
• This also makes the use of a resistive load more
  practical (700/1550W).
• But isnt EMI related to current so wont this
  setup (with its lower currents) give a much lower
  EMI than that measured in the car?
• Its already been shown that converter power does
  not have a big impact on EMI, at least for the
  commonest topology.

15
Standard EMI Test Setup

• This has a 1.5m harness between LISN and
  converter.
• In a vehicle would expect much shorter leads to
  battery, for example 0.1m.
• 1/15 current in 15 length gives same EMI as
  total current in normal length.
• As EMI? IL
• If current did depend upon load results are still
  reasonable.

16
Proposed solution

• The same as standard test setup with extra
  parameters defined
• Converter is run at 1/15 of its max. output power
• Load is resistive
• Where load is also a battery then a 10mF (
  -20100 tolerance, with good high frequency
  performance) should be placed in parallel with
  load to simulate this.
• Voltage range allows battery power.

17
Conclusions

• Proposed test set-up
• Requires no changes to existing standards
• The fact that EMI will be tested at 1/15 load and
  the load is a resistor ( capacitor) can be
  placed in the dc-dc converters specification.
• Should give reasonably repeatable results (even
  site-site) - as converter operational state is
  well defined.
• Is a reasonable approximation to use in the
  vehicle
• the vehicle test will trap any issues with the
  real setup in the car.
• is not expected to cause a significant cost
  impact (over having to meet vehicle EMI
  requirements)
• As test setup models filtering due to
  batteries.
• Additionally propose that jump-start post (if
  present) is not in use during testing as wiring,
  etc undefined during real use.