Maker Pro
Maker Pro

Cooling of SO8

H

Heindorf

Jan 1, 1970
0
Hi,
I'm using a TC4420-SMD as a gate driver ( 15V, 100kHz, Qg=0.8µC, Rg=5
Ohm). Now the board is finished and the creature gets pretty hot despite
a little heatsink glued on top.Is there a pin-compatible IC available
with perhaps a higher power dissipation and/or lower output impedance?
Is there a better way to cool down the case ( no fan)?
Rolf
 
P

Paul Burke

Jan 1, 1970
0
Heindorf said:
Hi,
I'm using a TC4420-SMD as a gate driver ( 15V, 100kHz, Qg=0.8µC, Rg=5
Ohm). Now the board is finished and the creature gets pretty hot despite
a little heatsink glued on top.Is there a pin-compatible IC available
with perhaps a higher power dissipation and/or lower output impedance?
Is there a better way to cool down the case ( no fan)?

The usual way with little SM devices is to track out the ground pins to
an appreciable area of copper on the PCB, the leads and solder being
much more thermally conductive than the plastic case.

Paul Burke
 
H

Heindorf

Jan 1, 1970
0
Paul said:
The usual way with little SM devices is to track out the ground pins to
an appreciable area of copper on the PCB, the leads and solder being
much more thermally conductive than the plastic case.

Paul Burke

Hello,
that's not a bad proposal and I'll consider this at any rate the next time.
Currently there is little space available and the board is finished. BTW the
frequency is 50 kHz ( two stages alternately switched on) and I calculate
approx. 0.3W for each driver. With 25°C and the small heatsink it's tolerable,
I think. But at 70 °C I've doubts.
Rolf
 
F

Fritz Schlunder

Jan 1, 1970
0
Heindorf said:
Hi,
I'm using a TC4420-SMD as a gate driver ( 15V, 100kHz, Qg=0.8µC, Rg=5
Ohm). Now the board is finished and the creature gets pretty hot despite
a little heatsink glued on top.Is there a pin-compatible IC available
with perhaps a higher power dissipation and/or lower output impedance?
Is there a better way to cool down the case ( no fan)?
Rolf


The Microchip TC4422 is pin compatible and has somewhat lower output
resistance. Some similar parts by IXYS are also pin compatible and have
slightly better output resistance.

http://ww1.microchip.com/downloads/en/DeviceDoc/21420d.pdf

http://www.ixys.com/99054.pdf

I assume you understand that total gate drive power is fixed for a given
switching frequency, gate charge, and input voltage, irrespective of total
gate resistance. All you can really expect to do is offload some of the
heat away from the gate driver IC and into the gate resistor. You could
keep the TC4420 part and just increase the gate resistance value, but then
you would have slower switching performance. Alternatively you could
replace the TC4420 with a lower resistance part like the TC4422 and then
increase the gate resistor value a corresponding amount. In this way
switching performance will not be altered (for the better or worse), but
some of the heat will be offloaded from the driver IC and more into the gate
resistor.

Alternatively you could replace your MOSFET(s)/IGBT(s) with lower gate
charge device(s), or decrease the switching frequency, or decrease the gate
drive voltage...
 
A

Anders F

Jan 1, 1970
0
Heindorf said:
Hello,
that's not a bad proposal and I'll consider this at any rate the next time.
Currently there is little space available and the board is finished. BTW the
frequency is 50 kHz ( two stages alternately switched on) and I calculate
approx. 0.3W for each driver. With 25°C and the small heatsink it's tolerable,
I think. But at 70 °C I've doubts.

I don't know the device. But generally you should go for power packages for
power devices. Ie. SO8 with an exposed thermal pad under it (which you
solder directly to a copper fill with thermal vias to a plane). 0.3W is
nothing - if you have problems even with a heatsink I guess you calculated
wrongly ;-)

/A
 
H

Heindorf

Jan 1, 1970
0
Fritz said:
The Microchip TC4422 is pin compatible and has somewhat lower output
resistance. Some similar parts by IXYS are also pin compatible and have
slightly better output resistance.

http://ww1.microchip.com/downloads/en/DeviceDoc/21420d.pdf

http://www.ixys.com/99054.pdf

I assume you understand that total gate drive power is fixed for a given
switching frequency, gate charge, and input voltage, irrespective of total
gate resistance. All you can really expect to do is offload some of the
heat away from the gate driver IC and into the gate resistor. You could
keep the TC4420 part and just increase the gate resistance value, but then
you would have slower switching performance. Alternatively you could
replace the TC4420 with a lower resistance part like the TC4422 and then
increase the gate resistor value a corresponding amount. In this way
switching performance will not be altered (for the better or worse), but
some of the heat will be offloaded from the driver IC and more into the gate
resistor.

Alternatively you could replace your MOSFET(s)/IGBT(s) with lower gate
charge device(s), or decrease the switching frequency, or decrease the gate
drive voltage...

Hi,
I think I understand the relation between drive power and gate charge /
frequency.Usually we use the 4421/22. The choice fell upon the 4420 because of
its availability in SO8.During all this part searching I forgot the
pin-compatibility of both.I'll keep it in my memory in case of emergency.I
didn't know that IXYS makes also such parts.I'll check it out. Ok, thank you, I
think the 4422 will do it.
Rolf
 
L

legg

Jan 1, 1970
0
The Microchip TC4422 is pin compatible and has somewhat lower output
resistance. Some similar parts by IXYS are also pin compatible and have
slightly better output resistance.

http://ww1.microchip.com/downloads/en/DeviceDoc/21420d.pdf

http://www.ixys.com/99054.pdf

I assume you understand that total gate drive power is fixed for a given
switching frequency, gate charge, and input voltage, irrespective of total
gate resistance. All you can really expect to do is offload some of the
heat away from the gate driver IC and into the gate resistor. You could
keep the TC4420 part and just increase the gate resistance value, but then
you would have slower switching performance. Alternatively you could
replace the TC4420 with a lower resistance part like the TC4422 and then
increase the gate resistor value a corresponding amount. In this way
switching performance will not be altered (for the better or worse), but
some of the heat will be offloaded from the driver IC and more into the gate
resistor.

Alternatively you could replace your MOSFET(s)/IGBT(s) with lower gate
charge device(s), or decrease the switching frequency, or decrease the gate
drive voltage...
10 out of 10.

RL
 
R

Rich Grise

Jan 1, 1970
0
Hi,
I'm using a TC4420-SMD as a gate driver ( 15V, 100kHz, Qg=0.8µC, Rg=5
Ohm). Now the board is finished and the creature gets pretty hot despite
a little heatsink glued on top.Is there a pin-compatible IC available
with perhaps a higher power dissipation and/or lower output impedance?
Is there a better way to cool down the case ( no fan)?
Rolf

You didn't say what kind of glue. You can get thermally conductive epoxy,
but it's a little expensive.

Good Luck!
Rich
 
T

Terry Given

Jan 1, 1970
0
Rich said:
You didn't say what kind of glue. You can get thermally conductive epoxy,
but it's a little expensive.

Good Luck!
Rich

My experience with "thermal conductive" epoxy suggests it should be
called "less thermally insulating" epoxy. Its a lot better than an
ordinary epoxy, but absolutely ratshit compared to, say, clamping the
parts together - by an order of magnitude at least.

Cheers
Terry
 
H

Heindorf

Jan 1, 1970
0
Terry said:
My experience with "thermal conductive" epoxy suggests it should be
called "less thermally insulating" epoxy. Its a lot better than an
ordinary epoxy, but absolutely ratshit compared to, say, clamping the
parts together - by an order of magnitude at least.

Cheers
Terry

Hi Terry,
do you mean clamping together without anything in between?And how do you clamp
an S08 heatsink in a space limited board?
Regards
Rolf
 
T

Terry Given

Jan 1, 1970
0
Heindorf said:
Terry Given schrieb:




Hi Terry,
do you mean clamping together without anything in between?And how do you clamp
an S08 heatsink in a space limited board?
Regards
Rolf

Hi Rolf,

I should have been more specific - my experience relates to attaching
to-220 devices to a heatsink using thermal epoxt - the results were
disastrous, mostly because no-one involved bothered to read the glue's
thermal resistance and do some elementary calculations. Normally a
metal-to-metal contact is about 1K/W, with the glue it was easily 20K/W,
and highly inconsistent. My first comment on seeing this was "thats
gotta be crap" and it was. If you want a good thermal "adhesive" try solder.

I have done quite a few (20k) PCBs using SO-8 mosfets (eg o/p rectifiers
in a 3V 10A smps) and only ever use the PCB to cool the devices (inside
IP68 plastic enclosures no less). Lots of meaty copper is the trick -
none of this thermal relief nonsense, reflow ovens dont care about that.
(I had 1-4-4-1Oz Cu on my 4-layer pcbs)

HFPC '96 or '97 had a couple of great papers on using pcbs as heatsinks.
There is a fair bit of thermal capacity, but adjacent layers work well
even without direct connections, but as stated here already the direct
pin-to-Cu connections are the best path by far.

Conversely I once used 3xSOD87 82V zener diodes as a leakage clamp in a
flyback, and had tiny pads connected with short 10mil tracks. The diodes
desoldered themselves, precipitating the destruction of the switching
device. Adding a hefty chunk of copper dropper the thermal resistance
down from ~350K/W to ~30K/W - they ran fairly cool, and stayed attached.

Cheers
Terry
 
Top