If you're unwilling to or can't calibrate your phylogeny by placing prior distributions on node ages in BEAST (for example), you can try using a few other approaches. Nonparametric rate smoothing is implemented in the R package ape. This is also implemented in r8s, along with PL and ML methods. I've had good results using treePL for some of my previous work.
Be aware that making a tree ultrametric is more than simply 'making the tips equal,' hence my model-based suggestions above.