having floating copper pour connected to thermal pad on IC invites noise problems?

Hello,
We have an offline LED driver IC which is of the switched linear regulator type. The IC has a thermal pad on it. However, its datasheet does not say to which pin the thermal pad should be connected. If we connect the thermal pad to a bottom layer copper pour on the PCB, then we can’t leave this pad floating. –Because as you know, floating copper pours on PCBs generate all kinds of noise problems?
Do you agree that leaving the copper pour which connects to the thermal pad ‘floating’ will invite all manner of noise problems…specially since on our tiny PCB, this ‘floating’ pad will be near the microcontroller and all the PCB traces leading to the microcontroller.
The LED Driver IC is DT3007B by Seoul semiconductor, and we have requested the datasheet from info.europe@seoulsemicon.com but get no reply. The datasheet is nowhere to be found on the web.

The IC is featured on this app note…
https://www.seoulsemicon.com/_upload/Goods_Spec/[SPEC]SMJQ-XC95W4PX_Rev1.0.pdf

Assuming it is similar to the DT3001B, the data sheet says:

Note : Exposed pad on bottom surface is thermal pad. Therefore,
It must be not connected to any electrical node. (Electrically Isolated)

Click to expand...

If you are worried about it floating, it should be safe to connect it to ground (VSS or whatever) through a 100K resistor. I think you may be a slightly paranoid about noise problems though, microcontrollers are pretty resilient devices as long as you follow standard design guidelines. I've got PICs here with I/O pins connected to outdoor wires several metres long and they behave perfectly even in thunderstorms.

Brian.

Connect the thermal pad to your biggest ground. It is most likely electrically isolated from the rest of the chip. It is only there to remove the heat from the device and therefore needs a thermal sink of its own - thus, connect to ground.

I'm not IC fab expert but I'm having trouble figuring out how connecting a thermal pad to ground would be problematic. I'm betting it's fine to ground the thermal pad, if you're not sure then just try it.

Thanks,,problem is we've already had 1000 PCBs made up with it floating...and the thermal copper is on the base of the PCB...we cannot add a wire there as it would rise the PCB off the heatsink on to which it is mounted.

- - - Updated - - -

DT3001B datasheet
http://www.seoulsemicon.com/_upload/Goods_Spec/Acrich2-Applicationnote.pdf

..Thanks, Betwixt kindly found the DT3001B (sister chip of DT3007B)app note and on page 21 it indeed says the thermal pad should not be connected to any electrical node. As Betwixt kindly says, maybe with a 100k its ok.
Does a floating piece of copper act like a Yagi antenna and re-radiate noise all over the PCB?

treez said:

...and on page 21 it indeed says the thermal pad should not be connected to any electrical node.
Does a floating piece of copper act like a Yagi antenna and re-radiate noise all over the PCB?

Click to expand...

If it is electrically isolated, why would it matter if you connected it to any node?

Also, it doesn't act like a Yagi antenna, but more like a microwave patch. But this patch is not energyzed, so no problem.

If the datasheet says "don't connect", why would you place a 100k resistor? Can you assure that it doesn't affect circuit operation?

You are free to doubt the justification of this statement, but what should be the consequence? It would need reverse engineering of the chip topology to specify a substantiated connection alternative. If we expect that the thermal pad is connected to the chip substrate, it might basically be at the same potential as V-. Shorting both terminals might have nevertheless negative effects, e.g. increasing the latch-up susceptibility.

Thus, unless you get other information form the IC vendor, follow the datasheet recommendation.

The data sheet isn't clear whether it means it IS electrically isolated or it SHOULD be electrically isolated but it does say it must not be connected to any electrical node so I guess there is an internal connection, even if to a non-conductive substrate. Bear in mind that these devices run directly from rectified 110/220V AC, there is no 'ground' as such so the pad could have high voltage on it or be susceptible to internal breakdown if tied to another conductive path.

Treez - yes, but as stated it would work as a patch antenna with a resonant frequency in the GHz region so you need not worry about 50/60Hz! I would keep a sensible guard area around it though in case it did adopt a high voltage.

Brian.

I would keep a sensible guard area around it though in case it did adopt a high voltage.

Click to expand...

Thanks, I am glad to hear this from other people, as I always associate floating bits of metal on PCBs as potentially floating up to a high potential…….unfortunately, the floating bit of metal, since its needed for cooling, is quite extensive……it goes under half of the microcontroller, and is under many of the traces which run into the micro. I wonder if it may be re- inducing its (possible) high voltage into the micro input pins?
I am not sure how high in voltage things can be induced, I always think its mains peak where that is the highest voltage around?

We really need to see the PCB and the schematic but I appreciate it is confidential to your company.
Certainly, running large areas of copper carrying high voltages right underneath any MCU is a recipe for disaster, especially if it's VSS line is grounded. There would be a risk of elevated AC line voltages (spikes etc.) flashing to neighboring tracks. The actual voltage, provided it isn't high enough to flash over, isn't the problem, it's the current it carries and rate it changes that causes the problem both by induced and capacitively coupled signals.

The best scenario would be a double sided board with vias under the DT3007 so the second copper layer increased the heat dissipating area and ideally, a different double sided area under the MCU for good electrical screening and low impedance grounding. I'm not sure if your present board is single, double or multi layer or if there is an option to change it.

Brian.

I would keep a sensible guard area around it though in case it did adopt a high voltage.

Click to expand...

This could actually be the reason why some units work when first powered up, but then don’t work on the second and thereafter powerings up.
Maybe at the first power up, the floating copper area is not charged up to high voltage…but by the time the second power up comes, it is charged up to high voltage and interferes with the microcontrollers start up routine etc?
The other thing is it could be interfering with the eeprom in the micro. The micro does store stuff in eeprom throughout the operation.
Thanks, our board is 4 layer...we do have a ground plane directly under the micro, on the sub-top layer.

- - - Updated - - -

Years ago I worked in a radar power supply place in London. One day the 14kV PSU output terminal was 1 metre away from a metal nut which was not quite earthed to a metal enclosure because the screw it was round was badly rusted and the nut was not contacting the metal enclosure, but was 1mm away from it. There was a loud snapping sound every second as the metal screw got induced up to high voltage and repeatedly flashed over to the earthed enclosure through the air.
There was no changing voltage there…just a high voltage terminal, and it surprised me how the nut that was 1 metre away could get induced up to such high voltage.
So I think we can assume that our floating bit of copper is going to be induced up to mains peak voltage.

- - - Updated - - -

I am thinking we can do a test with some copper tape to ground the floating PCB copper....just to see what effect it has.