Yes! Anyway I’m trying another kind of processing: 30 May - 17 Jul - 10 Aug
I think that I had such a difference between the pairs I showed you because they are not coregistrated between them. So the first pair has got a baseline, the second has got another baseline.
I totaly agree with you. The software still under development and I believe this errors should be tackled by developers before moving forward in a new options. So single interferogram should be an indicator of the study area except of atmospheric error. However, in SNAP in addition to the atmospheric error the accumulation of other sources of error makes or gives this trend of elevation patterns . Hope so to get an opinion of @mengdahl
You can try the Ellipsoid Correction instead of the range-doppler and compare the results.
This way you will have two reference datasets so to say.
Also a small hint for the projection, try to use metric projections (e. g. UTM) for the terrain correction to get more real result (instead of degrees). This is just a suggestion, depends on your goal how you want to display the results.
yes, a DEM for reference is automatically downloaded by SNAP for the terrain correction process.
As your result depends on the quality of the DEM (I agree to @Suprd), you can also trie the SRTM 1Sec or run it without any reference information using the Ellipsoid correction. I expect the best results with the SRTM 1 Sec however. Chosing a UTM projection (automatic zone) will be of favour if you want to compute a hillshade of your result. This often reveals the true quality as it highlights smaller glitches due to low coherence or radar shadow areas.