Here are some caffeine peaks (below) and EthoxyEthanol peaks (above). The display is DOSY-like, yet the numerical treatment is a simpler and faster mono-exponential fit. Each column has been processed independently from the rest, and each column yields a different result:
Now let's apply a Whittaker Smoother, with a small value of lambda (100), along each ROW:
or a big lambda (500):
or a huge value (8000):
The smoother averages the results obtained from the different columns. The peaks are perfectly aligned.