Number & Korvax
Number Number
Korvax Korvax
Number Number
I was just crunching the latest telemetry from autonomous drones, and I spotted a pattern that could shave off flight time—want to dig into the numbers together?
Korvax Korvax
Sure, send me the raw telemetry logs and I’ll run the regression on the flight profiles right away. The first thing I’ll check is the lift‑to‑drag ratio over the last 48 hours, then we can tweak the prop‑efficiency constants. Any anomalies you already spotted?
Number Number
Here’s the raw telemetry dump for the last 48 hours, timestamped every 30 seconds. Each row lists the flight ID, timestamp, altitude, speed, engine temp, battery voltage, and sensor checksum. I’ve flagged any rows with a checksum mismatch or a sudden temperature spike (above 10 % of the rolling average) with a red highlight in the attached spreadsheet. The most notable anomaly is flight‑ID “X7‑23” – a 3.5 % drop in lift‑to‑drag ratio around 14:32 UTC, which could be due to a prop spin‑inertia issue. Let me know what you think.
Korvax Korvax
Got the dump. The red‑highlighted rows are a good start, but I’m seeing the checksum errors cluster right after the 3.5 % lift‑to‑drag dip on X7‑23. That suggests a sensor‑sync problem when the prop stalls. I’ll run a quick cross‑correlation between engine temp and battery voltage to see if the temp spike is an independent event or just a side effect. In the meantime, can we get the raw PWM logs from the prop controller? That’ll tell us if there’s a sudden change in thrust that caused the drag increase. Also, let’s baseline the lift‑to‑drag curve at 12 m/s versus 14 m/s; the difference is about 2 % across the fleet, so we may need to adjust the pitch map. Once I have the PWM data, I can propose a firmware tweak to smooth the torque curve and avoid that spin‑inertia hiccup.
Number Number
Sure thing, I’ve pulled the raw PWM logs for the prop controller—each sample is timestamped to the nearest 10 ms and includes throttle, duty cycle, and motor RPM. I’ve also added a column for the calculated torque using the thrust model. I’ll forward the file in the next message. In the meantime, let me know if you need the data in a different format or if you want me to pre‑filter the spikes for you. Once you run the cross‑correlation on engine temp and battery voltage, we can quantify how much of the temperature rise is driven by the power draw versus a sensor lag. Looking forward to your firmware tweak proposal.
Korvax Korvax
Got the PWM file – the 10 ms granularity is fine. I’ll import it into the MATLAB script I built for the torque model, then run a high‑pass filter at 20 Hz to eliminate the low‑frequency drift before computing the correlation. I’ll also flag any duty‑cycle jumps that exceed 5 % of the preceding 100 ms window; that’s the typical threshold for a spin‑inertia bump. Once I have the correlation matrix, I’ll isolate the portion where the temperature spike begins – that will tell us if the rise is a direct response to the increased power draw or just a lagged sensor reading. For the firmware tweak, I’m thinking of a simple PID‑based torque ramp: a deadband of 10 % and a smoothing constant of 0.2 should keep the RPM from jerking while still letting the drone hit the target lift. I’ll send the updated script and the proposed torque ramp parameters in the next message.
Number Number
Sounds solid, I’ll line up the PWM data with the lift‑to‑drag table and keep an eye on the 5 % duty‑cycle jumps. The PID ramp parameters look reasonable—just make sure the deadband doesn’t clip any of the finer thrust adjustments we rely on at low altitudes. I’ll run a quick sanity check on the updated script once you send it and we’ll compare notes on the correlation matrix.