On Tue, Mar 11, 2025 at 09:52 AM, John Woodgate wrote:
It's quite possible that the data sheet Figure 13 isn't based on the Spice model, but on measurements of actual devices? of superior, rather than average, performance.
That is possible, but I think not likely.? By my read, the whole purpose of that Application Note is to show you the results of SPICE simulations.? Therefore, I think none of its plots were from measurements.? Also, it states that "the characteristics are always typical values and do not represent the limits of process variation."? I think it suggests that everything in that AppNote is average, not superior performance.
?
I agree with MaticH that the comparison is rather poor.? It should be much better, if not exact.
?
But read on.? I think there are reasons to explain it.
?
AppNote AN90034 was published in April 2022.? The LTspice model used today is dated June or July 2023.? We know it is a different model file because it lists a different number of MOSFETs than the ones shown in Figure 1 of the AppNote.
?
AN90034 refers specifically to the LTspice model here:
? ?
But the one in MaticH's (and my) simulations is here:
? ?
?
I think there was a significant change when they went from V1.1 to V3, even though it was only one year.
?
The older ZIP file is no longer there.??I tried retrieving it from the Wayback Machine but they did not successfully save it.
?
So anyway, that is what I think happened.? Nexperia's model changed significantly between 2022 and 2023 and this explains why today's simulations differ so much from the plots in AN90034.
?
To make things even more interesting, Nexperia has yet another SPICE model for the same part, here:
That one is a non-encrypted generic SPICE model (Level=3 NMOS) and it should work in LTspice as well as most other SPICE programs.? It lacks the two thermal pins.? I did not try it, but I do not expect close agreement between that model, and the AppNote.
?
?
