Floating heatsink in offline Flyback SMPS?

As there is no RF hot tab pressing against the heat spreader ( via insulation ) - there is no RFI issue

a lot easier to pass RFI if the heat spreader is left floating in this case ( as proven by RFI results in document supplied by you )

ToP switches etc are notable for the tabs being at rectified mains 0v.

Thanks, i saw the EMI results but it didnt say whether or not heat spreader was connected or not for them. Even with the quiet thermal node of the switcher IC, there is still a pair of switching nodes close to that metal and there will be noise coupling to the floating metal. I am surprised they dont tie it down?

On page 82 of this below (another innoswitch cct), the heat spreader is connected to primary side ground.
It makes you wonder why its floating in the aforementioned document.
Floating metal is always bad.


Its often that in testing/board-bring-up for example, one connects say a flying twisted pair scope probe wire to a PCB's nets and
finds that the PCB then doesn't work or malfunctions. Its the same principle.
The flying wire probe acts like an antenna, and so too does any heat spreader etc etc.
They all absolutely have to be tied down, even if that's by a Y capacitor rather than a resistor, or screw etc etc what have you.

I once worked on a 240W push pull 18-36vin, 24v, 10A out, approx. 150kHz or so.
It was in a huge great "total surround" chunking great aluminium block heatsink...must have cost a fortune to machine it out from solid...
...it was in two halves, which screwed together...looked way over-engineered.
Anyway, SMD power fets and diodes were all gap padded to the block...and so switching nodes were very well capacitively coupled
to the aluminium block. The thing would not work at all if the PCB ground was not y - cap'd to the aluminium block.

Not only that, but the multi-layer SMPS PCB (which obviously sat in the block) had to have an entire inner layer devoted to
a chassis plane...….so all the rest of the PCB was effectively capacitively coupled to the chassis plane since all the
PCB layers were very close to the chassis plane...so effectively this chassis plane formed a whopping great stray Y capacitor
between all the SMPS circuitry and the chassis.
This is common practice with these types of SMPS. You can see VicorPower recommend this technique with some of their SMPS
modules. You look at it and wonder why an entire SMPS PCB plane is devoted to chassis...but then you realise that noise problems can
be enormous when an SMPS PCB sits tightly surrounded by a dirty great metal block that's floating.

Anyway, we had to stay late and unpayed because we weren't allowed home because
this thing wasn't working (a previous engineer had designed it and left it)….and so I saw
this junior EE sneakily solder all the screw rest rings (which obviously connect to chassis) to the PCB ground !!!

...He knew full well how to get it working, and did so like this as it was his quickest (and sneakiest) way of getting it working
so he could go home!
Muggins here stayed late and de-soldered all has "jack connections" and did it properly with Y capacitors.
No bad to the gaffer though, he stayed with me and soldered them too.