Fail radiated emissions EN55015 - better approach

[Jester]

I have been tasked to get an LED array board (Luminaires) to meet EN55015 (not my design). The board passes conducted but fails radiated at 60MHz and it's harmonics (by about 3-5dB). I started near field probing the board which prompted me to think the design is far from ideal. See schematic below, basically 5 groups of 6 LED's each with a buck converter powered by an off board CV 24V power supply.

I'm inclined to eliminate all the buck converters and simply run all the LED's in series with a CC 100V supply or perhaps two series groups with a 48V CC supply. This particular power supply could be used:
https://www.mouser.com/ProductDetail/MEAN-WELL/ELG-75-C700A?qs=%2Fha2pyFaduizdTvi8J3Fhw%252Bfcj%2FjSgK19Nxe1mlgglIb5Q3f1LLwVw%3D%3D

It claims to meet EN55015 (although it warns that at attached circuit will influence results), never the less the power supply will be running at about 50% of it's rated output so that should help.

This approach will lower the overall cost, improve reliability (less components) and hopefully reduce emissions.

Does anyone see any down-side to this approach?

Also thoughts on need for on board filtering  (the power supply will be within a foot of the board).

[OM222O]

yes, just use a CC circuit. 100v might be a bit too much but it's the cheapest option if you want to directly rectify from mains.
there are tons of CC led drivers these days and if you match your voltages they won't be dissipating much power anyways, so no need for heat sinking.

[Jester]

yes, just use a CC circuit. 100v might be a bit too much but it's the cheapest option if you want to directly rectify from mains.
there are tons of CC led drivers these days and if you match your voltages they won't be dissipating much power anyways, so no need for heat sinking.

I measured the LED voltage at about 2.93-2.96V, so about 90V required, the option B has a pot to adjust the current, 10V of headroom seems reasonable and will allow some control over the intensity.

Not sure on need aboutr on board filtering, safer to make provision and not populate than the other way around I guess.

[OM222O]

you said 6 groups of 5 leds, that is just about 18 volts, so use a 24v input voltage which is a standard automotive voltage.
then use 5 cheap linear LED drivers which allow you to "program" the current as well for which you can use a pot. It's extremely easy and there is zero switching and will pass both radiation and conduction tests. I emphasize again: use linear drivers, not switching ones!

power dissipation would be a bit high (about 1.8 watts at 300mA which was on the schematic) or about 3 watts at 500mA (assuming a worst case scenario?) but it's nothing too crazy. you can reduce it by simply adding a few diodes from the supply input to the input of each driver which should drop the supply by about 0.7v to lower heat dissipation of the driver chip. alternatively use a zener diode in reverse orientation, then you can use whatever value that seems suitable for the leds (maybe a 3v one to drop the supply down to 21v which then puts half the load on the diode) this way you can go passive and not require any heat sinking too.

[ap]

From a qick look it sounds like the max switcher frequency is 1MHz. This means that what you see as non-compliance is a higher harmonic. The 4.7uF capacitor is certainly not able to dampen any such frequency, so I would start there with some ceramic capacitor in parallel and dampen high speed edges (which cause you problems). Say 100nF as a starting point. See what happens, appliy it to other places. In the end a trial and error approach will bring you there.