The pspline function uses P-splines (Eilers and Marx, Statistical
Science, 1981), which are a spline basis using a regular set of knots.
Looking at the code for pspline, which isn't so hard, let
dx = (max(x) - min(x))/ nterm where nterm is round(2.5 * desired
degrees of freedom)
The knots go from min(x) -df*dx to max(x) + df*dx.
The default df is 4.
Once the knots are known, the bs() function generates the pspline basis
for me.
Terry T