Solution in a larger domain is not extracted correctly

[SonyPony]

Using the kwave_solution, I tried to compute the wavefield in a homogenous SOS map (1 m/s) in a 96x96 domain, which resulted in a positive peak at the source location (first image).

However, if I run a simulation with the same parameters in a larger domain (512x512), the resulting wavefield has a negative peak at the source location (second image). I expected the same wavefield as in the 96x96 domain, but with "more" waves.

[SonyPony]

@astanziola Did you manage to find the bug?

[astanziola]

I gave it a try this weekend, and I don't see why it shouldn't work.

At first, I was thinking that is just a numerical issue in the way we evaluate the dt here, but seems weird (those are not very long time-stepping simulations).

Another thing that is know is that the accuracy of the single-frequency solution of PSTD simulations (that is, k-Wave) accumulates errors that only vanish for dt -> 0. But I do believe that this is compensated for in k-wave, see eq. 2.16 and relative discussion in the k-wave manual

I'm sure it is a silly error somewhere, I'll keep digging 😐