Low EMI Design

Posted by: koksalocal on 12/23/2008

1) We have an EMI related problem (high radiated emission) in our power supply that is a 15 W flyback(Top243, EF20 Core Tansformer). we are just wondering for what we should pay atention when designing transformer and which of the followings are advantageous for low radiated emissions?

1) Continious mode or Discontinious Mode
2) High Ip(low BP) or low Ip(High BP)
3) Low Lg or High Lg
4) or another

2)Does ultrafast diodes like mur420 at outputs create EMI problem?

3) What kind of diodes are advantageous for low emi?

- Schottky or fast switching?

Best Regards...

Comments

This is a big subject but let me do my best to give some pointers.

1) Continuous conduction mode forces the output diode to recover whilst in forward conduction which can create a high frequency oscillation as the diode recovers. The benefit of continuous conduction mode is lower primary currents and therefore higher efficiency and reduced differential mode (< 1MHz) conducted EMI. I would recommend staying with continuous conduction mode - 90%+ of the design I see are continuous conduction mode flybacks. Add an RC across the output diode (use an ultrafast type) to solve diode recovery ringing. Schottky diodes don't have the reserve recovery but due to leakage inductance on the secondary side there can still be ringing that an RC across the diode will help - in general they are lower noise. The use of schottky vs UF is generally determined by the output voltage (ie required diode voltage rating, <100V for schottky), cost and efficiency rather than EMI.

2) BP (flux density) - never considered this before - should have little effect on radiated EMI.

3) Larger gaps sizes do increase stray flux fields which may couple into input components. The maximum gap depends on the cross sectional area of the core. Can you share the transformer gap size you are using? I'd suggest 0.3 mm max for EF20

4) Try adding a copper band around transformer core - this will reduce stray magnetic field, electrically connect the band to primary return and this will then electrostatically shield the transformer - in compact designs the electric field causes more coupling into the input components.

Make sure that you are using good layout practices, specifically the high voltage nodes are minimized in area (drain node and clamp components) and the high current loops (clamp to transformer and secondary to diode to output cap) are small.

Cheers

PI-Chekov

Dear Chekov Thank you for your kindly reply. Please help for the followings. 1)Some documents on the web suggest that ultra fast output diodes may cause emi problems because of their abrubt recovery characteristics and advise soft recovery dioedes. Mur series diodes have very abrupt and fast switching properties. We have little bit doubtful about mur series diodes. What do you think about this issue? 2) Why does PI Expert Design Software force the the optimization to discontinuous mode? 3) As you wrote in your reply, in continuous mode, output diodes recover while the top switch is starting to conduct. that may cause excessive primary current and thus may create emi. Is that true? 4)Does shield winding help for low emi?. - Connection of shield winding - Placement of shield winding 5) If we compare the continuous and discontinuous mode for two designs with same output power and transformer core; Continuous Mode Design requires - High Primary Inductance and small Lg - Low Ip(Primary Peak Current) - High Peak Flux Density (Bp) and Discontinuous Mode Design requires - Low Primary Inductance and big Lg - High Ip(Primary Peak Current) - Low Peak Flux Density (Bp) Which one is advantageous for low emi? Best Regards...

Hello again,

Can you provide some more background to your design (output voltage, power, application, cost sensitivity etc) as the spec does come into the design choice.

1) Regarding UF diode. 90% of flyback designs use UF diodes and still meet radiated EMI requirements. The diode noise is relatively easy to deal with using an RC snubber across the diode unless your application requires much lower levels than EN55022 specifies.

2) PI Expert doesn't specifically force the design to discontinuous conduction mode unless necessary based on the user input. I just tried a 12 V, 1A design and the KP was 0.46 (continuous conduction mode). Can you provide the design file?

3) Ultra-fast diodes will create a current spike on the primary as the reverse recovery current is reflected back through the turns ratio. Ultra-fast types are needed vs slow recovery to keep this current low.

4) Yes - shields help EMI. PI Expert supports two types of shield windings. One that is placed first on the bobbin and one between primary and secondary. These reduce common mode coupling between primary and secondary, reducing mostly conducted EMI (150 kHz to 30 MHz).

5) Difficult to give definitive answer but in general discontinuous design will require more differential mode filtering on input stage, larger primary switch (or have lower efficiency), larger output diode, larger output caps, larger clamp dissipation but allows a fast recovery diode (150 ns) which will reduce high frequency EMI - but generally only at one point on the radiated EMI spectrum - put differently discontinuous conduction mode doesn't solve all radiated EMI issues.

Cheers

PI-Chekov