diff --git a/Source/hoc-clk/sysmodule/src/aotag.cpp b/Source/hoc-clk/sysmodule/src/aotag.cpp
index 5180afc5..35cd372c 100644
--- a/Source/hoc-clk/sysmodule/src/aotag.cpp
+++ b/Source/hoc-clk/sysmodule/src/aotag.cpp
@@ -81,15 +81,31 @@ namespace aotag {
         .tsample_ate = 39,
     };
 
-    // These coeffs are likely NOT accurate. Probably need to dump them from L4T?
+    /*
+        Calculate AOTAG coeffs:
+        offset_C = 5000
+        halfA = low millis * 2 - offset_C * 2
+        halfB = high millis * 2 - offset_C * 2
+
+        count_A = (aotag_half_A + |thermb|) × 8192 / therma
+        count_B = (aotag_half_B + |thermb|) × 8192 / therma
+        therma_target = (soctherm_half_B - soctherm_half_A) × 8192 / (count_B - count_A)
+
+        thermb_target = (soctherm_half_A - D×2) - therma_target × count_A / 8192
+
+        alpha = therma_target x 1,000,000 / therma_raw
+
+        beta  = thermb_target x 1,000,000 - thermb_raw × alpha
+    */
+
     static const struct FuseCorrCoeff tegra_aotag_coeff = {
-        .alpha = 4496200,
-        .beta = -4496200,
+        .alpha = 1290000,
+        .beta = 40500000,
     };
 
     static const struct FuseCorrCoeff tegra210b01_aotag_coeff = {
-        .alpha = 4496200,
-        .beta = -4496200,
+        .alpha = 1290000,
+        .beta = 40500000,
     };
 
     struct aotag_sensor_info_t {
@@ -219,7 +235,6 @@ namespace aotag {
         thermb = div64_s64_precise(temp, CALIB_COEFFICIENT);
 
         calib = ((u16)therma << SENSOR_CONFIG2_THERMA_SHIFT) | ((u16)thermb << SENSOR_CONFIG2_THERMB_SHIFT);
-
         *calibration = calib;
     }