4
\$\begingroup\$

I am working on a high power LED board project. It requires to have 100 LEDs, of different types and colors. There will need to be 17 individual strands (need to be able to dim them separate.) Max current of any strand will be around 1A. It is planned on powering this from a single 24VDC supply. This can be done using a 2Oz copper board, the back will have a full aluminum heat sink, and active cooling is allowed. Board is pretty large, around 1' x 1'.

My first thought was to definitely use switching LED drivers, to avoid having to burn off excess power as heat. But the idea of using 17 separate switchers seems like a EMC nightmare that may not have a easy solution.

I thought about using a TPS92512 driver, which allows me to drive all of their clocks together from an external crystal to help with EMC. But 17 switchers on 1 board still doesn't seem feasible to me.

So, should I not even consider using switchers, and just figure out the thermals with linear drivers instead? Or is there any other option I may be missing?

Edit: Did some quick math. If they were all driven linearly we would have to dissipate around 150 watts total..

Since the most a single linear driver can dissapate is around 5 watts, that approach isn't an option.

Only other thing I can think of is to use a FET to drive each strand, and use several high power resistors per strand to help spread out the heat across the board.

Edit 2: Some needed information. LED strands can not be combined

(number of LEDs in each strand, TOTAL voltage drop across strand, voltage needed to dissipate assuming 24VDC supply, Current req'd)

  1. 7 LEDs, 20.2V, 3.8V, 1.2A
  2. 7 LEDs, 20.2V, 3.8V, 1.2A
  3. 2 LEDs, 5.8V, 18.2V, 1.2A
  4. 7 LEDs, 20.2V, 3.8V, 1.2A
  5. 7 LEDs, 20.2V, 3.8V, 1.2A
  6. 2 LEDs, 5.8V, 18.2V, 1.2A
  7. 4 LEDs, 13.6V, 10.4V, 0.3A
  8. 8 LEDs, 22.4V, 1.6V, 1.8A
  9. 4 LEDs, 12.4V, 11.6V, 0.8A
  10. 10 LEDs, 21V, 3V, 1.2A
  11. 10 LEDs, 21V, 3V, 1.2A
  12. 4 LEDs, 8.4V, 15.6V, 1.2A
  13. 8 LEDs, 16.8V, 7.2V, 0.76A
  14. 4 LEDs, 8.4V, 15.6V, 0.76A
  15. 4 LEDs, 8.4V, 15.6V, 0.76A
  16. 8 LEDs, 14.8V, 9.2V, 1.62A
  17. 4 LEDs, 12.8V 11.2V, 0.78A

Another thought I had:

We can drive strands 1, 2, 4, 5, 7, 8, 9, 10, 11, 13, and 17 directly using the 24VDC input, using a FET and resistors to limit the current. The rest of the strands we can use one switching driver each like I originally thought. We would then only need 6 switchers, and would dissipate only around 60W from the strands directly driven via the 24VDC supply.

\$\endgroup\$
10
  • 2
    \$\begingroup\$ You may be better of with them not all running in sync, it spreads the spectrum out rather than generating sharply defined harmonics. \$\endgroup\$ – Phil G Jul 19 '19 at 21:23
  • 2
    \$\begingroup\$ I don't really see why you're so worried. A lot of the EMV you'll be producing will be pretty easy to filter out on the supply side, and probably never leave your board. \$\endgroup\$ – Marcus Müller Jul 19 '19 at 23:58
  • \$\begingroup\$ You will need a MCPCB to have acceptable temp rise for 150W/144 sqin. or 1W/sqin . 2oz copper needs about 0.2W/sqin. Unless you have done an advanced thermal via and heatsink design with forced air. \$\endgroup\$ – Tony Stewart EE75 Jul 20 '19 at 9:31
  • \$\begingroup\$ Can the LEDs be multiplexed? \$\endgroup\$ – po.pe Jul 25 '19 at 13:34
  • 1
    \$\begingroup\$ @VoltageSpike Sure. They just can't be combined, as I need to be able to control each seperately \$\endgroup\$ – Kyle Hunter Jul 25 '19 at 20:00
1
+50
\$\begingroup\$

You have 24V DC coming in, but you can efficiently convert that to other voltages using stepdown switchers. For instance you have a number of strands at 5.8V and 8.4V - for these you could step down to about 9.5V or so and them use simple linear constant current drivers per strand - or even have two step down regs at say 9.5 and 6.8V or so. This approach will dramatically reduce both emissions and heat dissipation in the linear sources, with minimal complexity.

Obviously there are tradeoffs for the number of step-down regulators and power dissipation. I leave it to you to explore the permutations to arrive at an acceptable solution.

\$\endgroup\$
1
  • \$\begingroup\$ I like this approach a lot. It requires the least amount of switchers, and there is not much heat dissipation from the linear driving either. Time will tell how this ends up working out! \$\endgroup\$ – Kyle Hunter Aug 1 '19 at 13:15
5
\$\begingroup\$

17 different switchers might sound like a lot, but it is "just" 24 dB more emissions than having only one. Well designed switcher is easily 30 dB from the EMC limits, so there shouldn't really be a problem. Of course you need to design them well, but it's still not as difficult task as you might think.

Without doing more math, to me the EMC trouble seems easier to solve than the heat trouble you'll have if using linear drivers.

\$\endgroup\$
1
  • \$\begingroup\$ Emc trouble always looks easier to solve until you hit it ... \$\endgroup\$ – danmcb Aug 3 '19 at 10:24
3
\$\begingroup\$

Using DCDC buck converters is definitely the way to go. You can find inexpensive types in SOT-23-6.

This one might be just the ticket: https://www.diodes.com/assets/Datasheets/AP63200-AP63201-AP63203-AP63205.pdf - it uses spread-spectrum to reduce peak EMI and controls drive to reduce ringing at the inductor.

EMI Tips:

Use shielded inductors for the DCDCs.

Consider a common-mode filter between the regulators and the LED array. You could do something as simple as split the power supplies onto a separate board and run its cables to the LED array through a common ferrite core.

If you use the separate-wires approach, you can use the low-side of the LED string for current sensing for your LEDs.

\$\endgroup\$
1
\$\begingroup\$

The 17 switchers design should be fine. This is purely for driving LEDs, so you don't have any ultra sensitive components to worry overly about.

Besides just following good design and layout practices there are two simple things that come to mind for me:

Firstly you can run them on a couple of different clocks. That would spread the noise out instead of having one big spike. Two or three should be enough I think?

Secondly paying close attention to your returns. Route them properly avoiding sharing copper as much as possible where appropriate. Basically don't just via them to a plane blindly because it's convenient. I think this is often the area people screw up, I've certainly done so.

As you said it's a large board, so there is plenty of space to lay things out well.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.