AnsweredAssumed Answered

Modeling SCR/IGBT Parts w/ User Defined Material

Question asked by Mark Snyder on Oct 12, 2011
Latest reply on Oct 14, 2011 by Bill McEachern

Like • Show 0 Likes0
Comment • 7

LONG POST TO FOLLOW

My company uses a great deal of IGBT and SCR's where the junction temps is critical to the survival of our power conversion units. (I have attached a cut sheet to one of our more common parts)

My quandary is in how (in the heck) to model these parts in FLOW because we do not have any specific material data from the component manufacturer. In most cases, they provided us with (highly suspect) thermal resistance coefficients between junction and case and between the case and the heatsink.

For those of you still reading this....good on you and thanks..... I can empirically measure the temp at the base of the component sitting on top of the heat sink and the case of the device. But, I have to backward calculate the temp of the junction which is not hard. But, I want to be able to use flow to predict the temp of the case because these little buggers and others like them are mounted such that convective flow heats the component mounted above it.

SO.............

My original thought was to take our existing SWx models, used as space claims, and make them multi-body. From there I can apply a material to the metal base (no big deal) but I have no idea what the material of the case is. This means I would have to keep futzing with a user created material until I get a case temp in FLOW that matches what we measure real world.

Can anyone else think of a better (spelled e-a-s-i-e-r) way to do this?

Bill McEachern Oct 12, 2011 12:07 PM

Correct Answer

What I do is take the specified resistance, the amount of power (thermal) and calculate the case temp. I run some quick conduction simulations to figure out the equivalent conductivity in the componet to produce the desire temp maximum and then use that for the component value. I am doing this in Simulation as opposed to flow - just cause it is faster and I am working on a conduction probelm at the moment. Not positive this is the way to go as the equivalent k seemed a bit high (106 W/mK). I normally use 18W/mK for IC's as a general rule........but then again it was pumping out 50W so hopefully the manufacturer did realize good conductivity was important. They said the equivalent resistance was based on the two resistor model.

See the reply in context

Attachments

62511210_071003.pdf148.4 KB

No one else had this question

Outcomes

Bill McEachern Oct 12, 2011 12:07 PM

Actions

Mark Snyder @ Bill McEachern on Oct 12, 2011 1:16 PM

Hi Ya Bill,

Ever get the feeling that it is just you and me here?

Are you saying that you apply (1) custom material to the entire SCR/IGBT based on that equivilant conductivity (k)?

Actions

Bill McEachern @ Mark Snyder on Oct 12, 2011 1:21 PM

Yes. If you do not have detailed knowledge of what the construction is on the inside what else can you do?

Actions

Mark Snyder @ Bill McEachern on Oct 12, 2011 1:45 PM

Well, given that my initial plan stated above was to futz with a user created material until I back in to the tested temp value.....I would say I was already thinking along the same lines as you suggested. Only difference was that I was going to create the part as multibody so the base could be one material and the case antother. Not sure of the practical benefit there. Yet.

Actions

Bill McEachern @ Mark Snyder on Oct 13, 2011 6:24 AM

Hi Mark,

As I recall for these types of components they are a metal plate (Al typ?) with an IC mounted onthe plate. That maybe a good way to go but it would depend on what they give you. If they just give you the temp rise per watt to the bottom of the metal plate from the junction....well then a single mat model tuned to give the right junction temp at the specified base plate location temp then it doesn't really matter. You could always assume the base plate is al. and then tune the IC package properties to get the same thing.

Actions

Mark Snyder @ Bill McEachern on Oct 14, 2011 10:55 AM

Hi Ya Bill,

They typically give you (2) thermal resistances. (1) from the case of the scr to the heatsink and (1) from the case to the junction. It may go without saying, but there is no way to actually measure the temperature of the junction. The case (I believe) is the base portion of the heatsink.

So, what do you think of this:

1. Model the SCR as a multibody part.

2. Assign Aluminum to the base/case(part that sits on the heatsink) and make it a volumetric heat source

3. Assign the default insulator material to the top of the SCR and see where we stand.

Mark

Actions

Bill McEachern @ Mark Snyder on Oct 14, 2011 11:07 AM

The JEDEC model for a 3D representation of a 2 resistor model would be a stack of resistor bottom, high conductor, resistor top (insulators around the sides depending on what is going on/tool in use). You would add the heat source to the high conductor part - say 2 or 3 orders of magnitude higher conductivity that the most conductive resistor - so it is at more or less constant temp through out (like a node in a 2 resistor model). The bottom goes to board, top to whatever is on top of it, if anything.

Actions

7 Replies

Bill McEachern Oct 12, 2011 12:07 PM
Unmark Correct
Correct Answer
What I do is take the specified resistance, the amount of power (thermal) and calculate the case temp. I run some quick conduction simulations to figure out the equivalent conductivity in the componet to produce the desire temp maximum and then use that for the component value. I am doing this in Simulation as opposed to flow - just cause it is faster and I am working on a conduction probelm at the moment. Not positive this is the way to go as the equivalent k seemed a bit high (106 W/mK). I normally use 18W/mK for IC's as a general rule........but then again it was pumping out 50W so hopefully the manufacturer did realize good conductivity was important. They said the equivalent resistance was based on the two resistor model.
Like • Show 0 Likes0
Actions
- Mark Snyder @ Bill McEachern on Oct 12, 2011 1:16 PM
  Mark Correct
  Correct Answer
  Hi Ya Bill,
  
  Ever get the feeling that it is just you and me here?
  
  Are you saying that you apply (1) custom material to the entire SCR/IGBT based on that equivilant conductivity (k)?
  Like • Show 0 Likes0
  Actions
  - Bill McEachern @ Mark Snyder on Oct 12, 2011 1:21 PM
    Mark Correct
    Correct Answer
    Yes. If you do not have detailed knowledge of what the construction is on the inside what else can you do?
    Like • Show 0 Likes0
    Actions
    - Mark Snyder @ Bill McEachern on Oct 12, 2011 1:45 PM
      Mark Correct
      Correct Answer
      Well, given that my initial plan stated above was to futz with a user created material until I back in to the tested temp value.....I would say I was already thinking along the same lines as you suggested. Only difference was that I was going to create the part as multibody so the base could be one material and the case antother. Not sure of the practical benefit there. Yet.
      Like • Show 0 Likes0
      Actions
      - Bill McEachern @ Mark Snyder on Oct 13, 2011 6:24 AM
        Mark Correct
        Correct Answer
        Hi Mark,
        
        As I recall for these types of components they are a metal plate (Al typ?) with an IC mounted onthe plate. That maybe a good way to go but it would depend on what they give you. If they just give you the temp rise per watt to the bottom of the metal plate from the junction....well then a single mat model tuned to give the right junction temp at the specified base plate location temp then it doesn't really matter. You could always assume the base plate is al. and then tune the IC package properties to get the same thing.
        Like • Show 0 Likes0
        Actions
        Mark Snyder @ Bill McEachern on Oct 14, 2011 10:55 AM
        Mark Correct
        Correct Answer
        Hi Ya Bill,
        
        They typically give you (2) thermal resistances. (1) from the case of the scr to the heatsink and (1) from the case to the junction. It may go without saying, but there is no way to actually measure the temperature of the junction. The case (I believe) is the base portion of the heatsink.
        
        So, what do you think of this:
        1. Model the SCR as a multibody part.
        2. Assign Aluminum to the base/case(part that sits on the heatsink) and make it a volumetric heat source
        3. Assign the default insulator material to the top of the SCR and see where we stand.
        
        Mark
        Like • Show 0 Likes0
        Actions
        Bill McEachern @ Mark Snyder on Oct 14, 2011 11:07 AM
        Mark Correct
        Correct Answer
        The JEDEC model for a 3D representation of a 2 resistor model would be a stack of resistor bottom, high conductor, resistor top (insulators around the sides depending on what is going on/tool in use). You would add the heat source to the high conductor part - say 2 or 3 orders of magnitude higher conductivity that the most conductive resistor - so it is at more or less constant temp through out (like a node in a 2 resistor model). The bottom goes to board, top to whatever is on top of it, if anything.
        Like • Show 0 Likes0
        Actions

Modeling SCR/IGBT Parts w/ User Defined Material

Attachments

Outcomes

Related Content

Recommended Content