feat(ui): Smooth efficiency average display using Exponential Moving Average (EMA)

[chillingbee]

Markdown

Problem

The efficiencyAverage in the web interface currently calculates the 1-minute average (or 1-hour average when modified) by dividing the live power consumption (info.power) by the hashrate. Because info.power is updated and telemetry is fetched every second, the displayed efficiency value fluctuates rapidly by ~0.15 J/TH or more due to minor real-time power draw spikes. This causes unnecessary visual flickering in the UI for a value labeled as an "average".

Solution

This introduces a frontend-side Exponential Moving Average (EMA) filter to smooth out the real-time power fluctuations for the efficiencyAverage display.

1. Added smoothedEfficiencyAverage to home.component.ts to persist the filtered state.
2. Applied a 99% historical / 1% new value split (old * 0.99) + (new * 0.01) to damp out the per-second noise.
3. Changed the backend data source for the average from hashRate_1m to hashRate_1h to provide a true, rock-solid long-term efficiency metric.

Result

The efficiency average indicator is now visually stable and reflects the true hardware performance without nervous second-by-second jumping, while still accurately adapting to frequency or voltage changes over a few minutes.

diff --git a/main/http_server/axe-os/src/app/components/home/home.component.ts b/main/http_server/axe-os/src/app/components/home/home.component.ts
index b5314a74..7a5d4cdf 100644
--- a/main/http_server/axe-os/src/app/components/home/home.component.ts
+++ b/main/http_server/axe-os/src/app/components/home/home.component.ts
@@ -175,6 +175,7 @@ export class HomeComponent implements OnInit, OnDestroy {
    public asicDomainsAmount: number = 0;
    public efficiency: number = 0;
    public efficiencyAverage: number = 0;
+   public smoothedEfficiencyAverage: number = 0;
    public expectedEfficiency: number = 0;
    public activePoolUserAddressPart: string = '';
    public activePoolUserSuffixPart: string = '';
@@ -871,12 +872,24 @@ export class HomeComponent implements OnInit, OnDestroy {
 
          this.efficiency = this.calculateEfficiency(info, 'hashRate');
          
-        // MANUELLE BERECHNUNG FÜR DEN STUNDEN-DURCHSCHNITT:
+        // BERECHNUNG MIT GEGLÄTTETEM FILTER (DÄMPFUNG)
          if (info && info.power > 0 && info.hashRate_1h > 0) {
-            // Leistung (W) geteilt durch Hashrate (GH/s umgerechnet in TH/s -> durch 1000)
-            this.efficiencyAverage = (info.power * 1000) / info.hashRate_1h;
+            const currentRawAvg = (info.power * 1000) / info.hashRate_1h;
+            
+            // Wenn der Wert zum ersten Mal berechnet wird, nimm den aktuellen Wert
+            if (this.smoothedEfficiencyAverage === 0) {
+                this.smoothedEfficiencyAverage = currentRawAvg;
+            } else {
+                // FILTER: 99% alter Wert + 1% neuer Wert. 
+                this.smoothedEfficiencyAverage = (this.smoothedEfficiencyAverage * 0.99) + (currentRawAvg * 0.01);
+            }
+            
+            // Weise den geglätteten Wert der Anzeige zu
+            this.efficiencyAverage = this.smoothedEfficiencyAverage;
          } else {
             this.efficiencyAverage = 0;
+            this.smoothedEfficiencyAverage = 0;
          }
 
         this.expectedEfficiency = this.calculateEfficiency(info, 'expectedHashrate');```

[mutatrum]

Maybe it would make more sense to apply the same 1m/10m/1h statistics for power as well, so we can have efficiencies over those ranges as well? Testing with EMAs gave me a lot of trouble when changing power/frequency settings and are slow to settle after a boot.

[chillingbee]

If you can implement that, by all means!
That’s a good idea, too!
Personally, I think the current setup is just right, that’s exactly how I understand an average—but it needs to cover a period of at least 30 minutes, if not longer.
Anything shorter doesn't really count as an average to me.
But that’s just my opinion!