I'm attempting to calculate the base-spreading resistance of BJTs from manufacturer-provided SPICE models, which use the Gummel-Poon model. This Agilent publication (Internet Archive link here, if needed) describes how to calculate the base-spreading resistance from model parameters. I've copied the equation here:
$$ \begin{align} r_{\mathrm{bb}'} &= \mathrm{RBM} + 3(\mathrm{RB}-\mathrm{RBM}) \left(\frac{\tan(z) - z}{z\tan^2(z)}\right),\\ z &= \frac{\sqrt{1 + \left(\frac{12}{\pi}\right)^2 \frac{i_{\mathrm{b}}}{\mathrm{IRB}}} - 1} {\frac{24}{\pi^2} \sqrt{\frac{i_{\mathrm{b}}}{\mathrm{IRB}}}}, \end{align} $$ where \$i_{\mathrm{b}}\$ is base current, RBM is minimum base resistance at high current, RB is zero-bias base resistance, and IRB is current at medium base resistance.
I've applied these equations to several NPNs and compared the results with the measured results from The Art of Electronics (3rd Ed.), and am getting very different results. For instance, OnSemi provides a spice model for the 2N3904, where the relevant parameters are: RB=5.8376 IRB=50.3624 RBM=0.634251. Using these values with the above equations (and base current 100uA) yields an rbb of approximately \$5.8\,\mathrm{\Omega}\$. The Art of Electronics in Table 8.1a, however, gives this as \$110\,\mathrm{\Omega}\$.
That IRB looks high.
What accounts for the difference in rbb? Is this a valid method of estimation?