Do anybody have experience about the accuracy of thermal simulation in natural convection.I am using Flotherm 7.2 for thermal simulation, but simulated result is not at all accurate with actual testing, the conditions are same as that of actual testing in thermal simulation.Mainly I am facing problem with Choke and Capacitor.Do any body have some idea regarding this. your response will be highly appreciated.

Regards, Jitender

I recall seeing a natural convection validation on their web page. Can you provide geometry for a comparison simulation? It would be nice to cross validate between simulation tools.

see flotherm applications page:

http://www.flomerics.com/products/fl...asestudies.php

Best Regards

Hey Titan, thanks, How can i upload geometry( PDML file of Flotherm)in this forum. i am really struggling to understand the behaviour of flotherm in case of natural convection. i performed electrical testing on board for some components like choke, capacitor, Mosfets, resistors etc. some of the components result are pretty accurate but some components accuracy too worse. the power loss provided by electrical are 100% correct to best of my knowledge.

regards, jitender

For electrolytic caps try a block with an orthotropic conductivity k_x = k_y = 0.35 W/mK and k_z (board normal) = 30 W/mK.

For chokes and the like, due to the vast difference in materials used to construct them, you might want to consider modelling as a few cuboids, some for the windings, some for the core, each with appropriate materials.

There's very little validation for models of such items as usually it's the high powered IC packages, or power converting PAs, that are the most thermally critical and thus have benefited from years of modelling practice experience.

For any more info I'd advise contacting your local Flomerics support engineer.

Hey Robin, thanks for invaluable suggestion. For electrolytic cap i used the value k_x = k_y = 0.35 W/mK and k_z (board normal) = 30 W/mK. but still result is more than 20% higher as compared to electrical testing. in fact i had contacted Flomerics supports also, but they give only rough idea in stead of precise ideas. so i am struggling to understand its unpredictable behaviour. are you also facing the same problem? i guess there is something wrong with software in case of natural convection.

regards, Jitender

Funny how the software is always blamed first. As I said before it's simply a question of modelling methodology as opposed to some fundamental failure of natural convection calculation in a software that's been around for 20 years.

If you can't figure out how to model the thing, ask the suppliers for each item for either a Flotherm model or some indication of their thermal characterisation.

hey Robin, thanks for comments. Generally i never conclude on such type of matter, but i am trying to solve this problem for last 8 months. Infact 2 times i contacted Flomerics support also. but instead of going in root cause of the problem, they just give over all ideas which i tried to implement.

Definately it worked,for most of the components, we came pretty close to actual testing result except from Capacitor and choke. any way i will continue my efforts.

thanks for suggestion, Regards, Jitender

You would be surprised to know that even a software which is here for around 20years could have issues. Sometimes things aren't just user errors.

Hi Jitender, given that the results are fairly accurate for some components and not others it suggests that how the code handles natural convection is not the issue, otherwise one would expect to see some trend, such as all component temperatures over-predicted. For components like capacitors connect to the board with just a few of wires, and a small air gap so care is needed to get the correct contact resistance between the package and the board. Modelling chip packages correctly from a thermal point of view is not easy. I would also go back and question the powers. Expereince has shown that its not easy to get time average power for normal operation from electrical engineers, who are used to working with max power specs, etc. Ask how the powers are calculated, for what condition they apply, etc. so that even if the numbers remain unchanged, your confidence increases. Hope this helps.

I also have problem about the acuracy of Flotherm

Being?

A couple of points.

1) I discovered a major error in Flotherm recently (6.1) which has since been fixed. If you ever get questionable results in Command Center always go to Tables view to check the figures. I have seen numerous errors in Command Center where the wrong figure is mapped to the wrong output variable cell in the Scenario Table. It can also be a result of a particular object (enclosure for example).

2) You might confirm that the material you are using in the model for the leads have accurate material properties.

3) Are you modeling the copper traces and planes in the pwb orthotropically? Could be a source of error if not.

For improved results I put orthotropic properties on the pwb cuboid (manually, not smart part) and then apply grid constraints to the pwb cuboid to have several cells through the through plane direction. How many depends on how many layers are in the board. These extra grid cells allows the orthotropic phenomenon to manifest in the numerical solution. It is a work around for not importing a board file. The Kxy and Kz values are calculated from the percent copper of each layer.

4) If the capacitor is large enough, do a bench test. Take your capacitor, and put a resistor inside it soldered to the capacitor leads. Using a known power dissipation in the resistor compare the measured temps to the model.

20 years and MCAD is still in the condition it's in (unuseable).