10 SMPS fails EMC
10 SMPS fails EMC
(OP)
Hello,
I work in a company that makes LED lighting products that run off car batteries at about 13.5V and we are in Washington.
There are 5 members of staff in the Electronics Dept.
Eighteen months ago, the Electronics Dept was asked to design and build a 10W SMPS which would power a series string of LEDs at 350mA.
We still have failed to successfully do this –mainly because we are failing the EMC test.
The PCB must have dimensions of 84mm by 15mm, due to the enclosure in which it must fit.
-There are two mounting holes of diameter 4.5mm which occur in the longways midline of the PCB at each end, each whose centre is 15mm from the respective short edge of the rectangular PCB.
-obviously there are two cylindrical pillars coming up from the enclosure which "mate" with these mounting holes.
So far we have done double-sided PCB designs.
The enclosure in which this PCB is to be fitted is actually a heatsink for the LEDs.
There is only 5mm of clearance above, and 5mm clearance below the PCB so tall components cannot be used.
Anyway, the problem is that we cannot fit in the sufficient amount of EMC filtering inductors and capacitors that would be needed to get this SMPS to pass the EMC test.
Have we reached the "boundary of impossibly small size" for a 10W SMPS here?
Is this beyond the wit of human-kind?
So far we have implemented continuous mode boost converters and these have failed the EMC.
Does any reader know of an "EMC quiet" topology that would allow us to be able to fit it into this small PCB ?
I am wondering if a transformer isolated push-pull converter would be more quiet with its LC output filter?......though having said that, the input current does have a discontinuity as the primary side "swaps coils" (other transistor starts conducting).
One bad EMC problem with the continuous mode boost is that the Schottky diode has capacitance and conducts a very high amplitude current spike through the MOSFET when the MOSFET turns ON......therefore, I am wondering if discontinuous mode boost should in fact be more "EMC quiet" in spite of its higher peak currents?
The other day, I was in the works conveniences and I overheard two senior members of staff in the organisation saying how useless our Electronics Dept was that five of us could not design a 10W Switch mode power supply in eighteen months.
Any tips on how to "design small" and pass EMC like this greatly appreciated.
I work in a company that makes LED lighting products that run off car batteries at about 13.5V and we are in Washington.
There are 5 members of staff in the Electronics Dept.
Eighteen months ago, the Electronics Dept was asked to design and build a 10W SMPS which would power a series string of LEDs at 350mA.
We still have failed to successfully do this –mainly because we are failing the EMC test.
The PCB must have dimensions of 84mm by 15mm, due to the enclosure in which it must fit.
-There are two mounting holes of diameter 4.5mm which occur in the longways midline of the PCB at each end, each whose centre is 15mm from the respective short edge of the rectangular PCB.
-obviously there are two cylindrical pillars coming up from the enclosure which "mate" with these mounting holes.
So far we have done double-sided PCB designs.
The enclosure in which this PCB is to be fitted is actually a heatsink for the LEDs.
There is only 5mm of clearance above, and 5mm clearance below the PCB so tall components cannot be used.
Anyway, the problem is that we cannot fit in the sufficient amount of EMC filtering inductors and capacitors that would be needed to get this SMPS to pass the EMC test.
Have we reached the "boundary of impossibly small size" for a 10W SMPS here?
Is this beyond the wit of human-kind?
So far we have implemented continuous mode boost converters and these have failed the EMC.
Does any reader know of an "EMC quiet" topology that would allow us to be able to fit it into this small PCB ?
I am wondering if a transformer isolated push-pull converter would be more quiet with its LC output filter?......though having said that, the input current does have a discontinuity as the primary side "swaps coils" (other transistor starts conducting).
One bad EMC problem with the continuous mode boost is that the Schottky diode has capacitance and conducts a very high amplitude current spike through the MOSFET when the MOSFET turns ON......therefore, I am wondering if discontinuous mode boost should in fact be more "EMC quiet" in spite of its higher peak currents?
The other day, I was in the works conveniences and I overheard two senior members of staff in the organisation saying how useless our Electronics Dept was that five of us could not design a 10W Switch mode power supply in eighteen months.
Any tips on how to "design small" and pass EMC like this greatly appreciated.





RE: 10 SMPS fails EMC
Benta.
RE: 10 SMPS fails EMC
Consider a 4-layer board, watch the length of your traces (shorter is not always better), and be willing to move a component to a different location/orientation even if it means making a larger board. Manufacturing it is more expensive, but don't discount metal shields for the nastiest cases.
Sounds like you guys are shooting for an LED foglight design. A word of advice... heat's a bitch, and unless you are meticulous about measurements and keeping the LEDs within spec, you're lights could fail much faster than anticipated. If you haven't already, consider electronic damping of power when die temps get too hot.
Dan - Owner

http://www.Hi-TecDesigns.com
RE: 10 SMPS fails EMC
The key here is to take a design and slightly modify it to meet your needs - not embark on a master's degree of learning how to build robust switching supplies. Leave that to companies that do ONLY that and do it well.
Maxim, Linear Tech, National, Micron, ST, and ON, all have chips to do just what you want.
National has a design lab where you dial a design even. Push the button and receive all the parts and a board.
Keith Cress
kcress - http://www.flaminsystems.com
RE: 10 SMPS fails EMC
5mm is a lot of room for components if your issues are in the MHz range so there may be solutions available without a complete redesign.
John D