RE: [Vo]:Caveats to using SPICE for thermal analysis
From: Alan Fletcher Sent: Sunday, June 02, 2013 12:27 PM Subject: Re: [Vo]:Caveats to using SPICE for thermal analysis David has been concentrating on the control aspects, I have been doing RC modelling similar to that described in your very helpful paper. I did a detailed mesh model of a heat exchanger ... but gave it up because it was too sensitive to the parameter for heat transfer from water to metal. See http://lenr.qumbu.com/rossi_ecat_oct11_c.php and http://lenr.qumbu.com/rossi_ecat_oct11_spice.php I'm now working on an RC model of the eCat .. progressing from a lumped model to a ladder model. Wow, I'm pretty impressed Alan, I don't think I've ever seen that a thermal SPICE model of such complexity. Most of the significant thermal analysis work I've been involved with has been done by MEs using dedicated FEA applications, and the SPICE modeling was typically used only by EEs for heat-sink selections and other random little tasks. I think about two or three dozen elements is the most I've seen actually used for thermal modeling before. Have you looked around at any open-source FEA tools that might be used for your project? If there are any decent ones, I suspect they'd lighten your workload considerably as compared SPICE, but I'm not familiar with what's actually available, so this is just speculation. Anyhow, best of luck with your efforts, and thanks for sharing. -Robert
Re: [Vo]:Caveats to using SPICE for thermal analysis
From: Robert Ellefson vortex-h...@e2ke.com Sent: Monday, June 3, 2013 12:05:18 AM I did a detailed mesh model of a heat exchanger ... but gave it up because it was too sensitive to the parameter for heat transfer from water to metal. See http://lenr.qumbu.com/rossi_ecat_oct11_c.php and http://lenr.qumbu.com/rossi_ecat_oct11_spice.php I'm now working on an RC model of the eCat .. progressing from a lumped model to a ladder model. Wow, I'm pretty impressed Alan, I don't think I've ever seen that a thermal SPICE model of such complexity. Most of the significant thermal analysis work I've been involved with has been done by MEs using dedicated FEA applications, and the SPICE modeling was typically used only by EEs for heat-sink selections and other random little tasks. I think about two or three dozen elements is the most I've seen actually used for thermal modeling before. Have you looked around at any open-source FEA tools that might be used for your project? If there are any decent ones, I suspect they'd lighten your workload considerably as compared SPICE, but I'm not familiar with what's actually available, so this is just speculation. Anyhow, best of luck with your efforts, and thanks for sharing. I installed Elmer, but couldn't get it to work. (My ignorance, I think). The hotcat is easier, because it's radially symmetric -- just a cylinder in a cylinder in a ... I'll go as far as I can with Spice (which works) and then give Elmer another try.
Re: [Vo]:Caveats to using SPICE for thermal analysis
From: Robert Ellefson vortex-h...@e2ke.com Sent: Saturday, June 1, 2013 8:59:31 PM Hello Vortex-L participants, First, I’d like to introduce myself, since this is my first time posting to the list. Hi ! Welcome on board. I have skimmed a few recent threads discussing thermal modeling using SPICE that David Roberson (and others?) has been posting about, and finally found a point I might add. David has been concentrating on the control aspects, I have been doing RC modelling similar to that described in your very helpful paper. I did a detailed mesh model of a heat exchanger ... but gave it up because it was too sensitive to the parameter for heat transfer from water to metal. See http://lenr.qumbu.com/rossi_ecat_oct11_c.php and http://lenr.qumbu.com/rossi_ecat_oct11_spice.php I'm now working on an RC model of the eCat .. progressing from a lumped model to a ladder model. The heat from the resistors is indeed represented by a step function (actually, a quick rise, a 150 sec hold, and a quick fall with an off time of 300 seconds) the overall time constant being about 1000 seconds. Some elements of the model can be calibrated from data given by the experimenters -- others could be obtained from data they have, but have not released. (Basically, we need the rise time when the ecat is first turned on, and the fall time when it is turned off.) My present results are interesting, but inconclusive ... I'll try and get some plots later today. We all seem to be using ltspice http://www.linear.com/designtools/software/?gclid=COjAl8aKxrcCFSFyQgodoUUAHQ#LTspice I'm happy to share my models/schematics with anyone who wants them. (Warts and all) If anyone wants a specific sub-task, it's writing a nonlinear model for the radiation component. Power is known to vary as (T1^4 - T0^4) : I have a model that works for a voltage source, but goes bananas (technical term : laymen don't have to use it) for a current source. There are limitations to using capacitive elements to represent thermal mass with transient inputs, particularly with discontinuous input functions. In addition, appropriate element sizing and granularity is important; too few elements or the wrong size elements will see results diverge from real responses. I found one decent appnote that discusses some of these points in detail, as applied to semiconductor packages: www.onsemi.com/pub/Collateral/AND8218-D.PDF If I find any more write-ups, I’ll post them. It’s been too long since I was directly involved in SPICE thermal modeling, but I do recall a number of warnings from experts about divergences, subtle and not. Hope this helps, Robert Ellefson
