Thermal Design - Vs Board Area Question

Thread Starter

toffee_pie

Joined Oct 31, 2009
231
Folks,

I hope everyone is keeping well and best wishes for 2022. :)

I am wondering can someone give me some guidance on this and I know thermal design is opening a whole can of juicy worms but I am using a particular switching converter IC and my power dissipation is not as good as I expect - its about 1Watt off the normal and I am getting 80 ish Percent efficiency as opposed to 90 - I am using a few mosfets but the power they are using doesn't account for the extra loss I am seeing.

The datasheet has this graph and gives an example of expected Power dissipation from the datasheet and resulting copper required.


I am using around 1.2 Watts, this figure goes off the screen for a R theta value, it will be around 42 C/W - does that mean I need sufficient copper (ground) on my pcb to compensate for this figure?

I currently have 45cm2 of copper on my pcb ( solid ground fill) - 2 layers, I got the figure from Zuken design force on the area fill section. The graph makes no sense for low R theta values as the required copper will be huge

I am trying to see if increasing my copper area will improve the efficiency I am seeing, its not to be taken lightly as its a new board spin - from 2 layers to 4.

or am I talking nonsense...
 

Attachments

Last edited:

LowQCab

Joined Nov 6, 2012
1,913
Start-out with your Filter-Components, they will make the most difference in Heat creation.
Design your Filter-Components (and Board) to handle at least twice the Current that You expect to deliver.
If You even start getting remotely close to Inductor-Saturation, the Heat will sky-rocket.
Also pay careful attention to your Capacitor-Specifications,
all of the minutia adds-up to a substantial difference.
Use the most expensive, "zero-recovery" Diode that You can find, and seriously over-rate it.

An Infa-Red Camera will show You where the most heat is coming from.
.
.
.
 

Thread Starter

toffee_pie

Joined Oct 31, 2009
231
Thanks, I am a bit stumped on this one however.

I got the DCDC converter and had a previous mock pcb done, purely to measure ripple and it was good - I was playing around with area fills around the converter - the ripple was 20/30mV AC so happy days, that is 24V in and 12V/1A out (12R load)

My finished design has a similar ripple, also EMC etc is all good but like I said the efficiency is about 10% down of what I expect, today I was wondering about that mock up board I done (which just had the converter and input and output caps) - surely that would have a much better efficiency...
the answer seems to be no, that was only marginally better - I am thinking its to do with the copper area used - my evaluation board uses 4 layers and its efficiency is 92% which is what I would expect - it has much more cm^2 of copper ground than I do on 2 layers.

But I am out of time to move to 4 layers and my boss doesn't think its the problem, that heat needs to go somewhere anyway doesn't it...

I could bolt a heatsink onto my mock pcb board, should be doable - and see if that drops the heat down and efficiency up - if it does it would strengthen the argument to get more layers

ugh
 

LowQCab

Joined Nov 6, 2012
1,913
More Copper is always better, no matter where You put it.
Don't just think in terms of "area", but thickness as well.

For the same amount of physical volume,
a Fan blowing directly on the Board will remove more Heat than a finned Heat-Sink.
.
.
.
 

Thread Starter

toffee_pie

Joined Oct 31, 2009
231
More Copper is always better, no matter where You put it.
Don't just think in terms of "area", but thickness as well.

For the same amount of physical volume,
a Fan blowing directly on the Board will remove more Heat than a finned Heat-Sink.
.
.
.
So, the situation is - the DCDC I am using is pants, I am a bit surprised - it runs far too hot (Thermocouple bonded to the top of the chip with Loctite and activator) - we are talking 90-120C from what tests I have done -- with 10V / 12V out @ 12R and varying input from 12 to 24V, It is cooking hot.

I have a few proto board, one of this is a bare bones DCDC I was using to play around with - I bolted on an aluminium enclosure to that as a conductor with a 12R load, so 24V in and 12V out - it worked and transferred the heat across but the efficiency did not improve at all - so moving to 4 layers wont help - that chip just runs like a sauna all the time

I would not have known this without doing testing, which is the whole point of it. Its a TI part and I would not recommend anyone else use it - coincidentally the evaluation board stopped working during my testing also - probably fried after all the going.

I need to redesign my scheme now to bypass that DCDC, I can still use it to drive other components but not the load itself
 

LowQCab

Joined Nov 6, 2012
1,913
A Full and detailed description of what You are trying to accomplish would go a long way
towards getting useful solutions.
.
.
.
 

Janis59

Joined Aug 21, 2017
1,383
Knowing how much "about" are any thermal calc of weakly explored materials, I would reccomend to srew tighhtly on Your pcb material of right size the ceramic or better metallic "box" resistor of some 5 to 20 W set the voltage to generate exactly one Watt, and then just MEASURE the temperature.
 

Thread Starter

toffee_pie

Joined Oct 31, 2009
231
A Full and detailed description of what You are trying to accomplish would go a long way
towards getting useful solutions.
.
.
.
honestly, its nothing complicated but essentially we have a DC input - (12-30v) and a 12v output - the 12V is coming from the dcdc - the output is switched via optical means but the dcdc is basically just going out to the load - 12Watts

it cant handle the heat so its getting bypassed - sometimes these modern chips can be too clever for their own good
 

LowQCab

Joined Nov 6, 2012
1,913
"" I am using a few mosfets ......... ""
What does this mean ?

Apparently You are already utilizing a SMPS to supply your smaller DIY SMPS,
why not just purchase a second SMPS Module and call it done ?
.
.
.
 

enggricha

Joined May 17, 2014
89
I see " output is switched via optical means " - are you using a optocoupler for switching mosfets? optocouplers can be slow and depending on your mosfets the switching losses can amount to 10% loss. But as @LowQCab mentioned without knowing what you are exactly trying to do, all this could be a moot point.
 

Thread Starter

toffee_pie

Joined Oct 31, 2009
231
"" I am using a few mosfets ......... ""
What does this mean ?

Apparently You are already utilizing a SMPS to supply your smaller DIY SMPS,
why not just purchase a second SMPS Module and call it done ?
.
.
.
I need FETs in my design for switches, they have nothing to do with the DCDC IC in question, which runs at 90+ Celcius at my rated load, even at 3Watts its burning hot -- I came across a white paper which mentioned this IC during thermal analysis - -unfortunately for me I saw it after I had selected it for my product.

The way I got over this was that I used this IC to power other components in my design and not drive the load -- talking IREDs, Comparators and such -- its getting to 37 Celsius doing this but of course I am losing my fixed output I expected from this product but I did a work around solution

Be mindful of highly integrated Ti solutions.
 
Last edited:

Thread Starter

toffee_pie

Joined Oct 31, 2009
231
"" I am using a few mosfets ......... ""
What does this mean ?

Apparently You are already utilizing a SMPS to supply your smaller DIY SMPS,
why not just purchase a second SMPS Module and call it done ?
.
.
.
The ship has sailed for making further design changes, but as per my message above I bypassed the volcano IC and got a work around solution but lose my fixed output voltage.
 

Thread Starter

toffee_pie

Joined Oct 31, 2009
231
I see " output is switched via optical means " - are you using a optocoupler for switching mosfets? optocouplers can be slow and depending on your mosfets the switching losses can amount to 10% loss. But as @LowQCab mentioned without knowing what you are exactly trying to do, all this could be a moot point.
I cant disclose much -- switching losses are minimal, talking 300mV tops. I have done my investigation and the IC in question runs far too hot, even expanding the number of layers of the PCB for ground planes will not help much - so I am worried about long term reliability - the fact an evaluation board fried during testing probably was a heads up. I was writing a labview code to control programmable bench supplies for my testing and it just stopped -- that is a 12W load - what it was rated for
 
Top