Sinogram clipping (flattened peaks) at low kVp in XCIST simulation

[xie-pixel]

Hi, I am using XCIST / CatSim to simulate CT scans with a simple phantom containing water and Fe inserts. I observed an unexpected issue in the sinogram when using lower tube voltages (80 kVp and 100 kVp):
The sinogram values (.prep) corresponding to the metal (Fe) regions appear clipped/flattened at the top

Additional observation:
Even when increasing the tube current to 500000 mA, the clipping still persists, suggesting that this issue may not be caused by photon noise alone.

My Questions：
（1）Is this behavior expected in XCIST simulations at low kVp?
（2）Could this be caused by:
photon starvation (I ≈ 0)?
internal lower bound / clipping on detected intensity before log transform?
numerical stability handling (e.g., log floor)?
（3）If so, where in the pipeline is this handled? (e.g., detector model, log conversion, preprocessing)
（4）Is there a way to:
disable or adjust this clipping?
access the raw (unclipped) projection values?

[MingyeWu]

It's caused by the -log process in PrepView.py:

    ###--------- log
    prep = (phantomScan-offsetScan)/(airscan-offsetScan)
    
    if getattr(cfg.physics, "disable_low_signal_correction", False):
        ### simple low-signal correction and -log
        smallValue = 1e-12
        prep[prep<smallValue] = smallValue
        prep = -np.log(prep)
    else:
        ### signal-domain low-signal correction and -log
        prep = LowSignalCorr(cfg, prep)
    
    prep[prep<0] = 0

If p-value (the -log) is over 12, because exp(-12) = 6.1e-6, there will be nearly no primary X-ray penetration. To avoid -log on zero, we uses a very small value (1e-12) OR process it with the low signal correction. In reality, those pixels will only have scatters, crosstalk, and e-noise, no primary signal. They will be defined by low signal correction, MAR, or other interpolation-based or advanced methods.

[xie-pixel]

Thank you very much for your reply. I completely understand, and I really appreciate you taking the time to answer