Coherence Azimuth Windows Size out of range


I am processing PALSAR-2 level 1.1 pair in SNAP, after creating stack by “Automatic Coregistration” I encounter error when making interferogram. I use the default parameters and just input the stack, but the error message shows:
A problem occurred during the target product initialisation. Type: OperatorException Message: Operator 'CreateInterferogramOp': Value for 'Coherence Azimuth Window Size' is out of range (1,75].

I check the two intensity bands in stack and they display correctly.
Is there anything I missed?

I am using SNAP 2.0 with S1TBX 2.0.3 on Windows7 x64 with 32 GB RAM.

Thank you

Please try with SNAP 3.0 and report back.

Hi Allen,

The error was caused by your “Coherence Azimuth Window Size” is larger than 75 pixels. This limitation has been removed in the new release. You can try smaller window size if you are still using an old version.


Hi junlu, I use default size which is 5, and I also tried smaller value but still failed.

And I tried process again with latest SNAP3.0 and problem is solved, thanks.

For Sentinel-1 data I attempted azimuth window of 1 and received this message:
Error: [NodeId: Coherence] Operator ‘CoherenceOp’: Value for ‘Coherence Azimuth Window Size’ is out of range (1,90].
Azimuth window of 2 works. If 1 is not valid, I suggest changing the message and help.

I would not use anything smaller than 3 for the coherence window.

Value for ‘Coherence Azimuth Window Size’ is out of range (1,90].

1 is not accepted, although still noted in the range for v7.x and v8. I’d like to run 4x1 or 5x1 and live with noise.
Why the ] or just typo?

Help page for coherence estimation appears to have azimuth and range reversed, and 1 is not valid for range or azimuth:

In case of ESA’s ERS and Envisat sensors, the factor 5:1 (azimuth:range) or similar ratio between the factors is chosen to obtain approximately square pixels (20x20 m^2 for factors 5 and 1).

The above is noted for a multilook post processing step.

For the Sentinel-1 SLC, multilook default is range:azimuth, 4:1, and for coherence, 10:2. Opposite listed.

The value of 1 should be allowed @junlu. The noise and bias in low coherence areas will be massive. What’s your use-case for the coherence if I may ask?

I am working with hyper-temporal coherence stacks, where noise can be temporally filtered, or use other derivatives less sensitive to the noise. @mengdahl

Dynamic range is lost as the tiny coherence windows produce large biases at low coherence values. Perhaps bias does not matter in your application if you’re mostly looking for high-coherence targets.