Switcher question: heatsink causing problems?

This is a common issue also with RF transistors, usually solved by insulated mounting on the heat sink. A heat sink that is floating on a node that contains spikes will be hard to muffle other than with full shielding. I never float them for another reason: Even if it's low voltage one accidental slip with a tool can send sparks flying and fry a lot of circuitry.

Hi. You need be carefull with the connection of the metal diode with he heat sink because it becomes a large antenna emitting noise ! I recomend to insulate the diode with a mica insulator, but it is important to remenber that you put a thermal resistence of about 1 degree celsius per watt. Another way is to put two diodes in parallel to divide the thermal load. To reduce EMI emission the ideal is to insulate diode and ground the heat sink

snipped-for-privacy@yahoo.com wrote:

I would worry more about the power mosfet (integrated in lm2678 in this case), since more currents are going through there. The output of the lm2678 and the anode of the diode are connected, you should have a big copper pad for them anyway. Even if you have good grounding for the lm2678, you still need good cooling for the output. Heavy currents are flowing from input to output, not to ground.

Good design of power pads under the bulky components (L & C) is better than heatsink sticking up anyway. If you layout your circuit properly, you should solve both problems at the same time.

I'm not sure what you're trying to say. The anode is connected to ground, not the output as you said. I know I need good cooling; I'm just saying it's probably better to have the heatsink at ground potential than connected to the switch output. By "connected", I mean internal to the part.

Are you saying the PCB traces could replace the heatsinks? That doesn't work on TO-220 devices.

Yes, I meant the cathode (the positve side of the diode), it's been a long time since I last deal with these terms.

Many regulators have the hot side for the tab, since that is more important to cool than the ground.

Thick traces to big pads. Many circuit boards are cooled by copper pads only.

Why not, it depends on how you mount and route it.

If you do need a heatsink for the TO-220 Schottky, then you should electrically insulate the heatsink from the TO-220 package. You should probably buy a "heatsink mounting kit" that includes, in addition to a mica insulator, an insulating "lipped washer" (can't remember the proper name, right now) that fits into the hole in the TO-220 package. Alternatively, maybe you could use a plastic bolt and nut, and a mica insulator. Remember to use heatsink compound, too.

How many amps is the Schottky conducting? Maybe it shouldn't be getting as hot as it is. Have you looked at its waveforms? Maybe a snubber-type network would help to keep it cooler. (Or maybe something else in the circuit also needs tweaking.) In a 56W boost-mode switcher that I made, the TO-220 Schottky barely gets warm to the touch, without any heatsink. Depends on the average current, of course. But any high-frequency ringing, etc, could heat it much more than necessary. Just a thought.

Good luck.

- Tom Gootee

"He who lives in a glass house should not invite he who is without sin."

I think he meant that you could possibly bend the TO-220's leads by 90 degrees and then bolt the TO-220 to a large copper pad on the PCB, which would act as its heatsink.

- Tom Gootee

"He who lives in a glass house should not invite he who is without sin."

Yes, insulate the heatsink.

No for the plastic screw. Even if you insulate the heat sink, you want to conduct the tab to a copper pad on the pcb, through a metal screw.

For our 12V 3A smps, only the mosfet (external, so-8) get warm. It does not need a heatsink either.

According to calculations by National's SMS v6.24, in my configuration the IC dissipates 730mW, and the diode 890mW. Now technically, I might not need a heatsink under these circumstances. But if a semiconductor is too hot to touch for even 1 second, I prefer to use a heatsink. Cool silicon is happy silicon. Now for the copper pad....there isn't enough room to do that. I've simply bolted on a heatsink. And the parts are vertical from the PCB. And yes, I'm going to use the insulator, or get insulated diodes. I could have some extra noise and this could be causing more heat. I'll check it out.

linnix, I don't know what you're talking about regarding the positive side needing cooling and is "more important to cool than the ground." The LM2678's tab is grounded. All 78xx devices I've seen are the same. The 317 might have the output connected to the tab if I remember correctly.

No.

The plastic bolt was in case he didn't have one of those "lipped (non-conductive) washers", so the bolt would not inadvertently connect the heatsink to the TO-220, by touching the edges of the hole.

If the heatsink is insulated, and it's an upright-mounted through-hole TO-220 Schottky, he's not going to connect the heatsink bolt to a PCB pad, in order to connect the TO-220's "tab" to the PCB. That's already connected to the PCB, through one of the TO-220's leads.

- Tom Gootee

"He who lives in a glass house should not invite he who is without sin."