<p>pespin has uploaded this change for <strong>review</strong>.</p><p><a href="https://gerrit.osmocom.org/c/osmo-bts/+/25329">View Change</a></p><pre style="font-family: monospace,monospace; white-space: pre-wrap;">MS/BS Power Control Loop: Fix downscaling averaging bug<br><br>The bug showed up in previous commit and is fixed in this commit. It can<br>be seen how rounding error is carried over time in the average<br>measurement, and affects final values.<br><br>Change-Id: I680d1c94bd4bae179b14b26662a819fa1462a5c8<br>---<br>M src/common/power_control.c<br>M tests/power/ms_power_loop_test.c<br>M tests/power/ms_power_loop_test.err<br>M tests/power/ms_power_loop_test.ok<br>4 files changed, 11 insertions(+), 9 deletions(-)<br><br></pre><pre style="font-family: monospace,monospace; white-space: pre-wrap;">git pull ssh://gerrit.osmocom.org:29418/osmo-bts refs/changes/29/25329/1</pre><pre style="font-family: monospace,monospace; white-space: pre-wrap;"><span>diff --git a/src/common/power_control.c b/src/common/power_control.c</span><br><span>index b3066cd..0362675 100644</span><br><span>--- a/src/common/power_control.c</span><br><span>+++ b/src/common/power_control.c</span><br><span>@@ -36,6 +36,8 @@</span><br><span> </span><br><span> /* We don't want to deal with floating point, so we scale up */</span><br><span> #define EWMA_SCALE_FACTOR 100</span><br><span style="color: hsl(120, 100%, 40%);">+/* EWMA_SCALE_FACTOR/2 = +50: Round to nearest value when downscaling, otherwise floor() is applied. */</span><br><span style="color: hsl(120, 100%, 40%);">+#define EWMA_ROUND_FACTOR (EWMA_SCALE_FACTOR / 2)</span><br><span> </span><br><span> /* Base Low-Pass Single-Pole IIR Filter (EWMA) formula:</span><br><span> *</span><br><span>@@ -84,8 +86,8 @@</span><br><span> return Val;</span><br><span> }</span><br><span> </span><br><span style="color: hsl(0, 100%, 40%);">- *Avg100 += A * (Val - *Avg100 / EWMA_SCALE_FACTOR);</span><br><span style="color: hsl(0, 100%, 40%);">- return *Avg100 / EWMA_SCALE_FACTOR;</span><br><span style="color: hsl(120, 100%, 40%);">+ *Avg100 += A * (Val - (*Avg100 + EWMA_ROUND_FACTOR) / EWMA_SCALE_FACTOR);</span><br><span style="color: hsl(120, 100%, 40%);">+ return (*Avg100 + EWMA_ROUND_FACTOR) / EWMA_SCALE_FACTOR;</span><br><span> }</span><br><span> </span><br><span> /* Calculate target RxLev value from lower/upper thresholds */</span><br><span>diff --git a/tests/power/ms_power_loop_test.c b/tests/power/ms_power_loop_test.c</span><br><span>index b9d6867..5f83329 100644</span><br><span>--- a/tests/power/ms_power_loop_test.c</span><br><span>+++ b/tests/power/ms_power_loop_test.c</span><br><span>@@ -204,11 +204,11 @@</span><br><span> /* Avg[t] = (0.2 * 40) + (0.8 * 29.60) = RXLEV 31.68 (-78.32 dBm),</span><br><span> * but due to up-/down-scaling artefacts we get the following:</span><br><span> * Avg100[t] = Avg100[t - 1] + A * (Pwr - Avg[t] / 100)</span><br><span style="color: hsl(0, 100%, 40%);">- * Avg100[t] = 2960 + 20 * (40 - (2960 / 100))</span><br><span style="color: hsl(0, 100%, 40%);">- * Avg100[t] = 2960 + 20 * (40 - 29)</span><br><span style="color: hsl(0, 100%, 40%);">- * Avg[t] = 3180 / 100 = 31.80 */</span><br><span style="color: hsl(120, 100%, 40%);">+ * Avg100[t] = 2960 + 20 * (40 - ((2960+50) / 100)) <- HERE we lose 0.1: (2960+50) / 100) = 30.1</span><br><span style="color: hsl(120, 100%, 40%);">+ * Avg100[t] = 2960 + 20 * (40 - 30) <- HERE we lose 20*0.1 = 2.0! (upscaled, hence we lose finally 2.0/100=0.2)</span><br><span style="color: hsl(120, 100%, 40%);">+ * Avg[t] = (3160) / 100 = 31.60*/</span><br><span> apply_power_test(lchan, -70, good_lqual, 1, 9); /* RXLEV 40 */</span><br><span style="color: hsl(0, 100%, 40%);">- CHECK_RXLEV_AVG100(31.80);</span><br><span style="color: hsl(120, 100%, 40%);">+ CHECK_RXLEV_AVG100(31.60);</span><br><span> </span><br><span> mp->ewma.alpha = 70; /* 30% smoothing */</span><br><span> lchan->ms_power_ctrl.current = 15;</span><br><span>diff --git a/tests/power/ms_power_loop_test.err b/tests/power/ms_power_loop_test.err</span><br><span>index 8ab6419..8f58882 100644</span><br><span>--- a/tests/power/ms_power_loop_test.err</span><br><span>+++ b/tests/power/ms_power_loop_test.err</span><br><span>@@ -23,8 +23,8 @@</span><br><span> (bts=0,trx=0,ts=0,ss=0) Lowering MS power control level 14 (2 dBm) => 15 (0 dBm): ms-pwr-lvl[curr 14, max 0], RSSI[curr -40, avg -47, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span> (bts=0,trx=0,ts=0,ss=0) Keeping MS power at control level 15 (0 dBm): ms-pwr-lvl[curr 15, max 2], RSSI[curr -75, avg -75, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span> (bts=0,trx=0,ts=0,ss=0) Raising MS power control level 15 (0 dBm) => 13 (3 dBm): ms-pwr-lvl[curr 15, max 2], RSSI[curr -90, avg -78, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span style="color: hsl(0, 100%, 40%);">-(bts=0,trx=0,ts=0,ss=0) Raising MS power control level 13 (4 dBm) => 11 (8 dBm): ms-pwr-lvl[curr 13, max 2], RSSI[curr -90, avg -81, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span style="color: hsl(0, 100%, 40%);">-(bts=0,trx=0,ts=0,ss=0) Raising MS power control level 11 (8 dBm) => 9 (12 dBm): ms-pwr-lvl[curr 11, max 2], RSSI[curr -70, avg -79, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span style="color: hsl(120, 100%, 40%);">+(bts=0,trx=0,ts=0,ss=0) Raising MS power control level 13 (4 dBm) => 11 (8 dBm): ms-pwr-lvl[curr 13, max 2], RSSI[curr -90, avg -80, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span style="color: hsl(120, 100%, 40%);">+(bts=0,trx=0,ts=0,ss=0) Raising MS power control level 11 (8 dBm) => 9 (11 dBm): ms-pwr-lvl[curr 11, max 2], RSSI[curr -70, avg -78, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span> (bts=0,trx=0,ts=0,ss=0) Keeping MS power at control level 15 (0 dBm): ms-pwr-lvl[curr 15, max 2], RSSI[curr -50, avg -50, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span> (bts=0,trx=0,ts=0,ss=0) Keeping MS power at control level 15 (0 dBm): ms-pwr-lvl[curr 15, max 2], RSSI[curr -50, avg -50, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span> (bts=0,trx=0,ts=0,ss=0) Raising MS power control level 15 (0 dBm) => 13 (4 dBm): ms-pwr-lvl[curr 15, max 2], RSSI[curr -110, avg -92, thresh -75..-75] dBm, C/I[curr 14, avg 14, thresh 12..16] dB</span><br><span>diff --git a/tests/power/ms_power_loop_test.ok b/tests/power/ms_power_loop_test.ok</span><br><span>index 8c92b03..27992b1 100644</span><br><span>--- a/tests/power/ms_power_loop_test.ok</span><br><span>+++ b/tests/power/ms_power_loop_test.ok</span><br><span>@@ -62,7 +62,7 @@</span><br><span> Avg[t] is RxLev 29.60 (expected 29.60)</span><br><span> lchan_ms_pwr_ctrl(RxLvl=-70 dBm) returns 1 (expected 1)</span><br><span> MS current power 11 -> 9 (expected 9)</span><br><span style="color: hsl(0, 100%, 40%);">- Avg[t] is RxLev 31.80 (expected 31.80)</span><br><span style="color: hsl(120, 100%, 40%);">+ Avg[t] is RxLev 31.60 (expected 31.60)</span><br><span> lchan_ms_pwr_ctrl(RxLvl=-50 dBm) returns 0 (expected 0)</span><br><span> MS current power 15 -> 15 (expected 15)</span><br><span> Avg[t] is RxLev 60.00 (expected 60.00)</span><br><span></span><br></pre><p>To view, visit <a href="https://gerrit.osmocom.org/c/osmo-bts/+/25329">change 25329</a>. To unsubscribe, or for help writing mail filters, visit <a href="https://gerrit.osmocom.org/settings">settings</a>.</p><div itemscope itemtype="http://schema.org/EmailMessage"><div itemscope itemprop="action" itemtype="http://schema.org/ViewAction"><link itemprop="url" href="https://gerrit.osmocom.org/c/osmo-bts/+/25329"/><meta itemprop="name" content="View Change"/></div></div>
<div style="display:none"> Gerrit-Project: osmo-bts </div>
<div style="display:none"> Gerrit-Branch: master </div>
<div style="display:none"> Gerrit-Change-Id: I680d1c94bd4bae179b14b26662a819fa1462a5c8 </div>
<div style="display:none"> Gerrit-Change-Number: 25329 </div>
<div style="display:none"> Gerrit-PatchSet: 1 </div>
<div style="display:none"> Gerrit-Owner: pespin <pespin@sysmocom.de> </div>
<div style="display:none"> Gerrit-MessageType: newchange </div>