LM27403 simulation in TINA is extremely slowLMLM

Sorry, I am having problems attaching files with Firefox.  Hopefully now OK with IE.

LM27403_SIMPLE.TSC

Ian,

I have confirmed that the simulation is quite slow. I will take a closer look at the model to see what could be causing the slower response. Please note that the two models are relatively similar, however, the 403 is a much more complicated model than the 402.

Hi Britt

Thanks for your help.

I have noticed a couple of interesting things.

First, if I make identical schematics for 402 and 403 with the same values everywhere (and with the current sensing defeated by shorting the sense pairs, although I don't know if that was necessary) for a typical buck, similar to one of my applications, and then simulate both, they stabilise to identical results. This is good for me, since it means that I can run the faster 402 simulation iteratively to design my circuits (I actually need several bucks in this design) in the knowledge that the behaviour of the 403 will be the same (although I could check this with an overnight run). I would just design the 402 in instead, except that the 403 is somewhat cheaper. My previous observation thet the 402 seemed less efficient was theresult of Webench producing different component values for the 402 and 403.
Second, it appears that the 403 simulation is only painfully slow for the first 100us or so (which takes 45m on my computer). During that time, the output voltage doesn't really go anywhere (although the MOSFETs are switching). After that, as the output rail starts to rise, the simulation speeds up considerably (but doesn't ever get as fast as the 402). In fact, it reaches 40% in that example I posted in about 1h. So it looks like it is having some (convergence?) problems before the converter really gets going.

So now I'm ready to do my designs, but I'm frustrated to notice that many of the CSDxxxxx N-channel MOSFETs which are suggested by Webench (which I am using to get starting points for my designs) do not seem to have SPICE models which work with TINA (and those that DO exist are encrypted so I can't make the others from them). Can I adapt the PSPICE models which exist for some of the missing parts to run with TINA? (those libs don't compile in TINA when imported). This is frustrating. In my desperation, I managed to persuade my circuits to simulate in LTSPICE (the 402 and 403 models aren't encrypted, so its not too hard) but of course I can't simulate the TI MOSFETs in LTSPICE either - but LTSPICE has a full library of competitors' MOSFET models, so I could use those. It seems that TI's decision not to provide usable models for its CSDxxxxx MOSFETs must be losing them design wins.

All the best
Ian

Ian,

Which CSD FETs are missing TINA-TI models? Let me know as we have most, if not all, of the CSD FETs in TINA-TI. The encrypted PSpice models for the CSD FETs will not work in TINA-TI nor in LTSpice. Unfortunately, LT does not allow us to use LTSpice (due to their EULA) for TI products.

If there is another part that you are having trouble getting into TINA-TI but have a PSpice model for it, please let us know. We can normally get the model into TINA-TI if it is a PSpice compatible model. The big problem is when the model developer uses the "built in" PSpice models for logic gates. These are incompatible with ANY software tool other than PSpice.

Let me know on the CSD FETs. We can usually get them to you very quickly.

Thanks for your patience and the update.

Thanks Britt, for your prompt reply.

Regarding the CSD models: I was messing around with Webench to get some suggested component values and MOSFET types as a starting point for my TINA simulations. I noticed that a few CSD MOSFETs didn't seem to have TINA models, but the only one I kept a note of was CSD17578. It's quite possible that I'm just not looking in the right place. Perhaps it would be best if I now redo all my target designs over the next day or two and make a definitive list of the CSD devices which I can't find TINA models for. This would be better than me bothering you for a few at a time.

On the subject of LTSPICE, it has a couple of features which are much better than TINA, which was the main reason why I bothered to convert the 402 and 403 models. The main one is the way it can dump a comprehensive efficiency analysis at the end of the simulation (or end the simulation as soon as the efficiency data is available). This not only states the overall efficiency, but also itemises the % and power lost to the various components (high-side and low-side FETs, inductor, controller, caps, sense Rs etc.). This is very useful and available instantly. On TINA (but I'm no expert) I have had to filter the input and output voltages with RCs to eliminate ripple so that I can cursor them and calculate the overall efficiency. I can likewise filter and monitor other circuit nodes to get the breakdown, but it's a lot of hassle. I mention this only for your wish-list!

But in fact, it didn't work out for me: it appears that LTSPICE can only do this trick by hooking special content which has been added to the spice models of the Linear converters and controllers, allowing detection of stable states etc. So that feature wasn't available when I had got the 402 and 403 models working in LTSPICE!

All the best
Ian

Ian,

Yes, everything you have said about LTSpice is true. Their built in models are unique.

Please see the attached .TSM file for the CSD17578Q3A. You may place it using the Insert-->Macro feature in TINA-TI.

The best place to search for SPICE models for TI products is here:

https://webench.ti.com/webench5/spicemodels/

CSD17578Q3A .TSM

Hi Britt

I have now run preliminary Webench simulations of all my converters, and have begun to do more detailed simulations in TINA. In the process I have been unable to find the TINA spice models for the following CSDs:
17575
17577
17578 (you provided this one yesterday
17579 (note that all of the above jave Q3A and Q5A variants, but I don't know if they have separate models)
17303
17309
18509

Also, I have noticed for the parts whose TINA spice models I DID find, some have 'transient' in the model filename and some don't. Are all of these models usable for transient simulation, or would I need the 'transient' version of them all (which mostly aren't listed).

Thanks for your help,
Ian

Thanks so much!