OK to have heatsink "live" in linear power supply?

[grantb5]

I'm building a kit which contains a linear +15/+5/-5/-15 PSU, basic old school 78xx and 79xx TO-220 stuff.

AC comes in, goes to transformer (and fuse), then diode bridge. 7805/7815 parts are side-by-side and have a heatsink that is large but not connected on the PCB to anything other than mechanical stability. No traces, not even ground. The 78xx regulators TO220 tabs attach to that heatsink with a screw each. I have no worries about that.  7915/7905 regulators similarly share a heatsink that is again "floating" other than being attached to the TO220 tabs with screws and again to the PCB with large but floating tabs. This means the heatsink for the 79xx regs will be at the unregulated negative rail.

Should I be worried about this? Is this dangerous?  Or maybe it could be "noisy"? I didn't design it, I'm just building and wondering if it's worth the trouble of isolating the 79xx's from the heatsink. The heatsinks are threaded so I would have to drill them out maybe.

Edit: the entire kit is housed in a metal case when completed (so one wouldn't come in casual contact with the PSU).

[Nauris]

Should not be any problem. It is low voltage after all, I remember seeing some device with heatsink at mains potential even.

[ArthurDent]

Although there is nothing wrong with having a 'hot' heatsink, I prefer to keep any larger piece of metal at mains ground or common potential and use an insulated mounting kit and heatsink compound for all regulator tabs that aren't at common potential. If you start probing around in the circuit and have forgotten that the heatsink isn't at zero potential, you may have a problem.

[Ian.M]

Yeah. However grounding it may not be so simple.  A mica washer or silpad introduces extra thermal resistance, so the negative regulator heatsink may need to be bigger to maintain the same max junction temperature as the positive regulator.  If you have to leave it 'live', a  warning sticker on the heatsink: "CAUTION: NOT GROUND" would be advisable.

[Zero999]

Yes, it's no problem. I've also seen a mains powered device with a life heatsink before and considered it to be perfectly safe, because it was inside an insulated enclosure with sufficient separation from anything anyone could touch.

[coppercone2]

I think its basically a bad design decision from a repair/modify/test stand point

You can always make things easier for people that need to work on your circuit physically. This is not one of those things.

When you design something, this consideration goes against performance considerations, pretty much always. Testability/repairability/anti-repairman-fuckupability is all different design things you can work on. Some people will actually make demands for these factors if you are selling them equipment (i.e. some people might demand plexiglass plates over entire PCB's so tech's can work on the wiring harness or do subassembly-level repairs without exposing themselves to any potentially hot pins (or at least minimize potential exposure). So the chassis begins to look like a out door cafe with umbrellas over all the tables in case it rains. Why? Because a small plastic bit is worth less to some companies then a tech's finger or a unit that has odd damage occur to it.

I would recommend using a pad and grounding the heat sink to circuit ground so if something goes bad it blows a fuse. When I see it I think 'mother fucker'.

[james_s]

It's certainly a common thing to see. When possible I prefer to keep heatsinks isolated as it makes short circuits less likely but sometimes maximizing thermal transfer is more important and anything you do to isolate the heatsink will reduce the thermal transfer.

[ejeffrey]

I definitely don't mind heat sinks that are at safe low voltages as long as you really make sure they can't inadvertently short to ground -- i.e., they should be mechanically anchored in place and there shouldn't be any way for the enclosure or internal wiring to come in contact with it.  These are good ideas anyway so it shouldn't be a big deal.

I personally would rather not have a heat sink at mains potential if possible just because the added risk for someone poking around inside.  It does depend on the device.  In a power supply is one thing.  An exposed power supply inside an otherwise low voltage instrument is different situation.

If you want to avoid this without sacrificing thermal performance you can use a small copper or aluminum heat spreader that is then isolated from the main heatsink.   This increases cost but it will also let you use a single large heatsink for multiple devices with different tab potentials, so may help you out in that way.

[james_s]

Adding a small heat spreader can solve one problem but creates another, it's yet another joint adding thermal resistance between the die and the air. It may improve transient response but still reduces the total heat transfer.

[duak]

 I would add an insulating pad because I tend to tinker with instruments and have encountered Murphy's Law all too often.  Barring that, I would mark it somehow - people just forget.

When I was a young lad, I was tuning up a vacuum tube receiver IF strip.  In this case, the capacitors were variable and the adjustment screws were at B+, about 300 VDC.   What idiot would design something like this and then not mark it?  My plastic handled  screwdrivers were either too small or too large but I had an all-metal screw driver that fit just right.  I didn't know about the hazard the first time I touched the screw but because the screw was stiff I had to grab the metal chassis with the other hand...  It took me a few minutes to find the screwdriver because I threw it somewhere.  Worst thing about it, is that I did the exact same thing about 30 minutes later.  Nowadays, I would definitely mark the adjustment with some sort of warning.

[ejeffrey]

Adding a small heat spreader can solve one problem but creates another, it's yet another joint adding thermal resistance between the die and the air. It may improve transient response but still reduces the total heat transfer.

Not if done right, at least not compared to using an insulating washer between the package and the heat sink.  Yes, you have replaced one junction (package to heat sink) with two (package to heat spreader and heat spreader to heat sink), but the two can both have potentially much lower thermal resistance.  The package to heat spreader can have less thermal resistance because you don't use an insulating pad.  The heat spreader to heatsink can have much larger surface area. 

[Zero999]

I think its basically a bad design decision from a repair/modify/test stand point

You can always make things easier for people that need to work on your circuit physically. This is not one of those things.

When you design something, this consideration goes against performance considerations, pretty much always. Testability/repairability/anti-repairman-fuckupability is all different design things you can work on. Some people will actually make demands for these factors if you are selling them equipment (i.e. some people might demand plexiglass plates over entire PCB's so tech's can work on the wiring harness or do subassembly-level repairs without exposing themselves to any potentially hot pins (or at least minimize potential exposure). So the chassis begins to look like a out door cafe with umbrellas over all the tables in case it rains. Why? Because a small plastic bit is worth less to some companies then a tech's finger or a unit that has odd damage occur to it.

I would recommend using a pad and grounding the heat sink to circuit ground so if something goes bad it blows a fuse. When I see it I think 'mother fucker'.

Ease of servicing is a valid reason for having an isolated heat sink, but quite often there are other factors such as size and cost, which are much more important. This is especially true for a product produced in large volumes, at low cost, which isn't designed with serviceability in mind, as it goes against making it as small and cheap as possible. You may disagree with this, but nowadays few power supply designs are built for repair: bad PSU, junk the board and replace it with new one!

[coppercone2]

if that's the angle your going for.. I thought it might be a home project etc where cost is much less important then arguing with counter trying to justify their job even moreso then trying to ensure reasonable savings (ensure company survival)

If you are putting a kit together for yourself.... you might not want to have a big corporate mindset . Home built thing, no boss to argue with, no real mass production cost savings.. might wanna make it serviceable. I doubt its a big business if you are selling 78 series linear regulators based power supply even if your designing the kit.. bet you won't sell more then 1000 of those and IMO with a damn kit people put together, the people will be happier if they don't blow up their kits if its some novice with non-jacketed multimeter probes monkeying around the circuit (your probobly gonna get at least one person using nail heated by a plumbers torch as a soldering iron probing it with loose wires from a harbor freight DMM connected in ammeter mode, the type of person that will defiantly try to abuse your warrente). Dealing with this one person might be more costly and time consuming and generally unpleasant then just absorbing a bit of cost. If I was in business for myself I would do as much as possible to remove customer interaction unless its for features that can make more sales.  I.e. pre-solder the QFPNL on the PCB. Add some probe points rather then expect people to probe SOIC traces with some kind of fancy needle sharp probes in a place where you think they might tend to. Reduce skill level for manufacture and testing as usual to reduce fails and complaints.

Also some people might see a thermal pad/grease/etc as part of a kit as a 'cool' assembly step. Like if you read about how its made and you wanna buy it for some kid, its kind of interesting how you need to align up a little gasket for a electrical part. I always found screwing the PCB into the chassis and the mechanical assembly steps kind of interesting. Adds a bit of variety. So long your not having to screw down greasy thermal pads every day its kind of interesting. Only thing I would avoid is glue.... . The skill level for using it neatly without destroying furniture etc is too high. It's easier to learn how to solder to nasa standards lol

[Zero999]

I have not made a linear power supply project, since I was an apprentice. The last one I made, was a bipolar tracking power supply. I fixed the LM217 and LM337 directly to small pieces of copper, with screws, nuts and an insulated sleeved washer, and fixed them to the case, with a thin piece of insulating material sandwiched in between. It worked quite well and the fact the pieces of copper float at different voltages didn't make it difficult to debug or repair the power supply.