Ok, to "solve" the DOF noise, I did some test, with different PT samples, PASS and Random sampler.
In this first test, I used:
(*) Default Random sampler = Alpha 3.
(*) MSI = 15
(*) PASS= 250
with different PT samples
(*) PT Samples: 2 = 6 min
(*) PT Samples: 5 = 10 min
(*) PT Samples: 10 = 15 min
(*) PT Samples: 25 = 32 min
Obviously, more samples -> high rendering time -> higher quality.
But the noise inside the DOF, especially near the bright pixel, doesn't' change too much, even after 32 minutes.
In the second test, I tried to change the Random sampler:
(*) MSI = 15
(*) PASS= 250
with same PT samples (2 and 5) an, as Random sampler, Alpha 3 vs 5D High
(*) PT Samples: 2 - Alpha 3
(*) PT Samples: 2 - 5D High
(*) PT Samples: 5 - Alpha 3
(*) PT Samples: 5 - 5D High
In this case, no big differences...
Third test, I tried with the same time (32 minutes) but with different PT samples ( 5 vs 25) No big differences. Again, noise in the DOF:
So, I don't know how to improve the noise in the DOF. Maybe only by waiting? But the rest of the image was very clean... :-(
In attachment the uncompressed image *.TIF