diff --git a/TEST_PLAN.md b/TEST_PLAN.md
index d4b0d0e..b1fb9ad 100644
--- a/TEST_PLAN.md
+++ b/TEST_PLAN.md
@@ -168,19 +168,50 @@
 
 ## 13. Settings — Audio Section
 
+### 13a. Sample Rate Tests
+
 | # | Test | Expected Result | Pass/Fail |
 |---|------|-----------------|-----------|
-| 13.1 | Sample Rate dropdown shows current value | e.g. "48000" |
-| 13.2 | **AUDIO:** Change Sample Rate (e.g. 44100) | POST succeeds, JACK restarts |
-| 13.3 | **AUDIO:** NAM survives the change | NAM CPU > 0% after restart |
-| 13.4 | Buffer Size dropdown shows current value | e.g. "512" |
-| 13.5 | **AUDIO:** Change Buffer Size (e.g. 256) | POST succeeds, JACK restarts |
-| 13.6 | **AUDIO:** NAM survives (CPU > 0%) | No DI signal |
-| 13.7 | **AUDIO:** Change both rate + buffer together | Both apply, NAM survives |
-| 13.8 | **AUDIO:** Change to unsupported rate (192000) | May fail gracefully or work |
-| 13.9 | **AUDIO:** Change buffer to 1024 | Larger buffer, lower CPU, higher latency |
-| 13.10 | **AUDIO:** Change back to 512 | Restores original |
-| 13.11 | **AUDIO:** Verify no persistent "peaking/poping every second" | Clean audio through NAM |
+| 13.1 | Note current Sample Rate in dropdown | e.g. "48000" |
+| 13.2 | **AUDIO:** Play guitar — note the tone and latency | Baseline reference |
+| 13.3 | **AUDIO:** Change Sample Rate to **44100**, Buffer 512 | POST succeeds, ~6-10s JACK restart |
+| 13.4 | **AUDIO:** NAM survives (CPU > 0%) after restart | No DI signal |
+| 13.5 | Play guitar — does the tone sound different from baseline? | Pitch should be same, slight feel difference possible |
+| 13.6 | **AUDIO:** Change Sample Rate to **48000**, Buffer 512 | POST succeeds, NAM survives |
+| 13.7 | Play guitar — compare to baseline | Should match baseline |
+| 13.8 | **AUDIO:** Change Sample Rate to **96000**, Buffer 1024 | POST may take longer, higher CPU |
+| 13.9 | **AUDIO:** NAM survives (CPU may be higher) | Expect higher CPU % at 96k |
+| 13.10 | Play guitar — any difference in clarity/feel? | Higher sample rate may sound cleaner |
+| 13.11 | **AUDIO:** Change Sample Rate to **192000**, Buffer 1024 | May fail if device doesn't support it |
+| 13.12 | If 192k fails, verify rollback works | Still running at previous rate, NAM CPU > 0% |
+| 13.13 | **AUDIO:** Return to **48000**, Buffer 512 | Back to baseline |
+
+### 13b. Buffer Size Tests
+
+| # | Test | Expected Result | Pass/Fail |
+|---|------|-----------------|-----------|
+| 13.14 | Note current Buffer Size in dropdown | e.g. "512" |
+| 13.15 | **AUDIO:** Change Buffer to **1024** | POST succeeds, lower CPU expected |
+| 13.16 | **AUDIO:** NAM CPU drops | e.g. ~18-25% |
+| 13.17 | Play guitar — feel the latency | Higher latency noticeable |
+| 13.18 | **AUDIO:** Change Buffer to **256** | POST succeeds, higher CPU expected |
+| 13.19 | **AUDIO:** NAM CPU rises | e.g. ~60-80% |
+| 13.20 | Play guitar — feel tighter latency | Lower latency, snappier feel |
+| 13.21 | **AUDIO:** Change Buffer to **128** | May xrun on complex models |
+| 13.22 | **AUDIO:** NAM CPU may hit 90%+ | Listen for pops/glitches |
+| 13.23 | **AUDIO:** Change Buffer to **64** | Likely xruns immediately |
+| 13.24 | **AUDIO:** Return Buffer to **512** | Back to stable range |
+
+### 13c. Combined Changes & Stability
+
+| # | Test | Expected Result | Pass/Fail |
+|---|------|-----------------|-----------|
+| 13.25 | **AUDIO:** Change rate + buffer at once (e.g. 44100 + 256) | Both apply, NAM survives |
+| 13.26 | **AUDIO:** Change back to 48000 + 512 (one POST with both) | Both apply, NAM survives |
+| 13.27 | **AUDIO:** Rapidly change buffer 3x in a row (512→256→1024→512) | Each change works, NAM survives all |
+| 13.28 | **AUDIO:** Rapidly change rate 3x in a row (48000→44100→48000→44100) | Each change works, NAM survives |
+| 13.29 | **AUDIO:** Leave running at 48000 + 512 for 30 seconds | No pops, no dropout |
+| 13.30 | **AUDIO:** Verify no persistent "peaking/poping every second" | Clean audio through NAM |
 
 ---