OffTheShelf DCDC module is getting hot even on low power?

[zenerbjt]

Dear Engineers,

We are using a Traco THN 20-4811WI DCDC module on our PCB to supply 5V, 5.5W. Vin is 48VDC.

With the PCB in open-air-lab-ambient at 25degC, we record (with Flir TG167 Thermal cam) case temperature of 62degC on the DCDC module case. (This is a believable reading as the case feels quite hot to the finger).

Now, I have done discrete 48Vin to 5vout, 6W Sync Buck converters before, and none of the components are anything like this hot.

Do you think there is some problem with these DCDC modules?

(As a comparison, when we use the JTK3048S05 DCDC module for the same thing, the case is much cooler…about 32degC.)

Traco THN 20-4811WI DCDC module datasheet:
https://www.farnell.com/datasheets/2706771.pdf

JTK3048S05 datasheet:
https://www.xppower.com/portals/0/pdfs/SF_JTK30.pdf

 

[stenzer]

Hi,

this sounds strange. To increase the housing to additional ~40 °C, the converter has to dissipate 2.27 W (17.6 °C/W for natural convection). As the converter has an efficiency of about 90 %, it has to source a power of about 22.7 W, and that's more than the rated 20 W and far away from the stated 5.5 W. How moch current does the converter sink at 48 V?

Please share a schematic of your circuitry.

BR

 

[zenerbjt]

Hi, I am seeing 130mA output current indicated on the 48V DC Bench supply (not using a battery yet)

 

[stenzer]

So, you are sourcing 6.24 W. For an efficiency of 89 %, the converter should only dissipate about 0.6 W. This would lead to a temperature increas of ~12 °C. What is the actual output voltage and current (output of your converter)?

BR

 

[FvM]

The reported numbers don't make much sense.
Pin = 48*0.13 = 6.24W
Pout 5.5W
Pv = 6.24 - 5.5 = 0.74 W
Expected Heating 0.74 * 17.6 = 13K

 

[zenerbjt]

Thanks, we will double check measurements. The 48V,0.13A at the input is read off the Lab PSU LCD screen. I measured 5V on the output with the scope, but will add a sense res to check the current at the output.

Strangely, our other DCDC module (the one that runs much cooler) , the JTK3048S05, does not give a Thermal impedance figure in its datasheet.

 

[stenzer]

Hi,

as you measured the output voltage by a scope, have you noticed any kind of oscillation, which does not correlate with the converter switching frequency?

As you measured the temperature by a thermal camera, have you set the correct emissivity? The inverter from traco comes in a metal housing which requires a low emissivity setting, whrereas the inverter from xppower has a black housing (high emissivity).

Have you double checked the temperature by e.g. a K-Type thermo couple, which is often included with a multimter? Usually objects with a temperatures of 60 °C and upwards are too hot for touching them for long.

BR

 

[zenerbjt]

temperatures of 60 °C and upwards are too hot for touching them for long.

Thanks i agree, i am used to touching 60degc with my "finger thermometer", and the Traco case feels like 60degc.
I took the temperature on emissivity of 0.95, and took the temp on the side of the Traco module, where it is "slightly shiny black".

 

[stenzer]

Ok, so please provide a sketch of your circuitry and measurement results of the output voltage and current.

BR

 

[zenerbjt]

Hi,
Will get the measurements soonest. However, i wonder if the Traco Module has been damaged by an input overvoltage transient. Maybe somebody set the Vin to 76V say, and then suddenly connected the 76V to the connector, and the Vin to the Traco rang up above 100V.

Somebody might have accidentally set it to 76V or so, thinking it was one of our 24S input circuits, or something.

The attached is the circuit. (PDF and LTspice sim)

 

[stenzer]

Hi,

you can easyly test your circuitry without the ferrit bead: is the converter still getting hot.
You also can replace the ferrit bead by a low valued resistor to measure the sourced current.

I do not see any input protection e.g. by a TVS which would protect the converter aigainst transient voltages. There may be stored a lot of energy in your input inductor, which is pretty large with 12 µH. Of course the stored energy depends on the sourced current (W = 0.5 \[\cdot\] L \[\cdot\] I²). And a possible voltage increase is given by the change of the current through the inductor (V = L \[\cdot\] \[\frac{\mathrm{d} i}{\mathrm{d} t}\]).

BR

 

[dick_freebird]

"Natural convection" is modified by the surrounds;
if the converter is mounted on a plane which
impedes airflow but does not provide good
conduction then the rating-basis airflow that
they call "natural" may not be what you get.

Output instability of even a few mV can result
in load-energy "sloshing" as current is repeatedly
sourced and sunk from the load as it follows the
LF sinusoid on top of the load. This can tank (heh)
efficiency. Might check the loop comp and play
with stability / load transient response tradeoffs.
For our DC-DC POL parts we would tune the loop
right to the edge of stability to demonstrate
our transient response, but then if someone
went and added some more passives it might
go a tad unstable, or not - depends some on
the not-so-explicit filter cap attributes, ESL
and ESR zeroes, how big and where.

 

[Easy peasy]

Often the no load temp and the full load temp are very similar on these modules due to the circulating currents to provide ZVS over wide load range ....

 

[zenerbjt]

Often the no load temp and the full load temp are very similar on these modules due to the circulating currents to provide ZVS over wide load range ....

Thanks, thats what i was thinking, but then i saw the attached, where it says that the efficiency at 25% load is above 85%. So even with circulating currents, -its still running significantly too hot.

One point regarding the circulating currents though, the capacitor we placed at the input to the DCDC module has an ESR of 0.685 Ohms, (as in attached schem) so would not be optimised for working with such circulating currents...would you agree?
(please ignore the "48-101vdc" input voltage, it will be 42-52v for us here)

 

[Easy peasy]

the input cap has little / nothing to do with the circulating currents in the converter power stage ...