About¶

Why libdrone looks the way it does — and what drives every decision. A collection of principles from aerospace, mechanical engineering, biology, and common sense that shaped every design choice. Written for someone building their first drone who wants to understand not just what, but why. Read in any order — each principle stands on its own.

About¶

Why the drone looks the way it does — and what drives every decision? This document is not a technical specification. It is a collection of principles — from aerospace, mechanical engineering, biology, and common sense — that shaped every decision in libdrone. Written for someone building their first drone who wants to understand not just "what", but "why".

Read in any order. Each principle stands on its own.

Engineering Principles 101: Makermapper V2.4.6¶

Flight Performance¶

Thrust-To-Weight Ratio¶

The single most important number for any multirotor. At 1:1 the drone just hovers. Anything above 2:1 gives usable performance headroom. libdrone targets ~3:1 on a depleted battery.

Practical consequence: every gram saved on the structure translates directly into handling. The 30g saved by switching from hub+top plate to the X body sandwich is not just a number — it is a visible change in responsiveness.

The calculation is simple: Maximum thrust (g) / MTOW (g) = thrust-to-weight ratio 4 motors x ~800g thrust each = ~3200g total MTOW ~1050g (drone + battery) -> ratio ~3:1 ✓

Moment Of Inertia And Response¶

The lower the moment of inertia (mass closer to centre), the faster the drone responds to inputs. V2.4.6 has a CG ~8-12mm lower than V2.14 (battery lies flat on the body instead of on a tall stack). Shorter pendulum arm = faster roll and pitch.

Software consequence: PID D-term must be reduced 10-15% from V2.14 baseline at maiden. A lighter, lower-CG frame responds more aggressively to D.

Why Floating Motor Mounts¶

Motors vibrate. Vibration propagates through rigid material into the gyro. The gyro sees noise instead of motion. Software must filter -- filters add latency -- latency degrades handling.

Silicone o-rings and sleeves mechanically interrupt the vibration path before it reaches the gyro. The RPM filter then receives a cleaner input and can be configured less aggressively. Result: lower latency at equal or better signal quality.

This is an example of a general principle: solve the problem as close to its source as possible.

Structural Principles¶

Failure Hierarchy (Crash Energy Management)¶

A deliberately designed failure sequence is better than a random one. In libdrone V2.4.6:

shaft fractures -> tab stays -> T-slot intact -> X body intact -> electronics alive

The same principle is used in: * Automotive crumple zones (bonnet deforms, cabin survives) * Aircraft engine fuse pins (engine detaches, wing survives) * Electrical fuses (fuse blows, house survives)

The key: each element must be calibrated weaker than the one behind it. Shaft must fracture before tab. Tab must yield before T-slot. If the sequence is not intentional, crash energy finds its own path -- usually through the electronics.

Monocoque And Sandwich¶

A classic flat hub is a continuous plate: crash force from one arm propagates across the full surface into the other three arms and into the electronics without interruption.

The sandwich deliberately structures this continuity. Our X body has gaps between arm extensions -- force must pass through a narrow transition into the core zone. Energy dissipates along the transition length instead of propagating directly.

Aerospace rule for sandwich structures: Every hole weakens the structure globally, not just locally. Minimum wall around every hole = the hole's own diameter. libdrone maintains 3mm wall around all T-slots, rod channels, and bolt holes.

Pre-Tensioning¶

A pre-tensioned structure is stiffer than a loosely joined one at the same mass. Our CF rods are placed under compression at assembly via the pinch slit -- the body is held in tension.

Same principle: suspension cables, pre-stressed concrete beams, tennis rackets.

A pre-tensioned joint transmits load immediately with no slack -- no movement before activation. For the gyro this means: frame flex that would otherwise appear as noise simply cannot occur because the rods hold geometry under tension.

Zonal Stiffness Layering¶

The best structures are neither soft nor stiff -- they are zonally layered.

STIFF where you need precision: -> gyro mount, motor axis, X body core

COMPLIANT where you need damping: -> motor-to-frame interface (floating mount)

DELIBERATELY WEAK where you manage the crash: -> arm shaft as fuse

libdrone V2.4.6 implements all three zones. A brutally stiff race frame (10mm CF plate) ignores the first and third zones -- it works, but at the cost of weight and destructive crash potential.

Exact Constraint Design¶

Every part should be constrained by exactly as many joints as it needs. Fewer = freedom (part can move). More = over-constrained (internal stress).

A tab in a T-slot is over-constrained in the height axis if the PCCF layers are not exactly 15.0mm. A floating tab pocket (shallow groove instead of a fixed face at the junction) compensates for +/-0.3mm manufacturing tolerance -- the correct answer from exact constraint design.

Gaps Between Arms¶

Modern CF drones have a deliberate gap in the body between arms. Reasons: 1. Interrupts destructive force transfer from one arm into the others 2. Allows airflow through the body for ESC and FC cooling 3. Reduces weight without losing functional stiffness

Our X body implements this -- 60x60mm core with arm extensions separated by gaps. T-slots pass from extension into core through a narrow transition that acts as a local force transfer damper.

Material Principles¶

Why Three Materials On One Drone¶

Every material does some things well and some things poorly. The key is to assign the right material to the right function:

PCCF (Prusa PC-CF): + extremely stiff, low thermal expansion, dimensionally stable - brittle, poor impact tolerance, difficult to machine -> X body structural layers, wherever dimensional precision is needed

PETG

+ tough, absorbs impact energy, easy to adjust with Dremel
- lower stiffness, creep under sustained load above 60 deg C
-> arm shafts (deliberate fuse), arm tabs, GPS bracket
-> bottom X body layer (impact face -- first contact on landing)

ASA

+ UV stable, good thermal resistance, good surface finish
- more expensive, requires enclosure to print well
-> bumpers (externally exposed surface)

Petg Bottom Layer As Impact Face¶

PCCF is stiffer but more brittle -- first contact with a hard surface on impact favours tough material, not stiff. PETG at the bottom absorbs impact and deforms in a controlled way. PCCF above takes over the structural role.

Analogy: a shoe sole is rubber (tough), not steel (stiff).

Interference Fit For Passive Clamping¶

At the CF rod crossing zone in the core we use #RodDiaChannelCore = 2.1mm instead of the standard #RodDiaChannel = 2.2mm. The rod must be pressed into the material -- the material is held in radial compression around the rod passively, without additional geometry or bosses.

Applies only to the PETG bottom layer: PETG deforms elastically and holds the rod. PCCF layers keep 2.2mm -- PCCF is stiffer and more brittle; interference fit could cause micro-cracking around the channel on assembly.

Fallback: if 2.1mm PETG grips the rod too tightly, 30 seconds with a Dremel gives you 2.2mm. You do not want to do this to PCCF.

The Weakest Link Principle (And Why You Want To Know Where It Is)¶

A chain is only as strong as its weakest link. In drone design this is not a problem -- it is a tool. If you know where the weakest link is, you know where the drone will fail in a crash. If you do not, the crash surprises you.

The arm shaft is deliberately the weakest structural element. Printed vertically (layers perpendicular to bending load = maximum strength in that direction), but still PETG and still a 26x15mm cross-section over 100mm. It fails predictably, at a predictable location, and it is the cheapest part on the drone. A well-designed weakest link.

Electronics And Signals¶

Why Bidirectional Dshot¶

Standard DSHOT runs one way: FC -> ESC (command). Bidirectional DSHOT adds a return channel: ESC -> FC (actual motor RPM).

The FC now knows exactly how fast each motor is spinning at every moment. The RPM filter can notch out exactly the motor and propeller frequencies instead of estimating them. Result: significantly cleaner gyro signal, lower latency, better handling -- with no hardware change whatsoever.

This is an example of the principle: feedback always improves a system.

Tvs Diode -- Overvoltage Protection¶

When a battery is disconnected in flight (or on a hard impact when the connector bounces), motors act as generators and push back-EMF into the circuit. On 6S this spike can exceed 30V -- enough to damage MOSFETs in the ESC or capacitors on the FC.

The SMBJ28A sits between VBAT and GND. Under normal conditions it does nothing. When voltage exceeds its clamping voltage (~28V) it shunts the excess to GND in microseconds. Cost: pennies. Protection: entire stack.

Principle: a cheap fuse protects an expensive system.

Conformal Coating -- Why Before The First Flight¶

Electronics are designed for a dry environment. A skatepark has dew, rain, wet concrete, condensation when moving from a warm car into cold air.

Conformal coating is a thin layer of silicone or acrylic that covers every solder joint, every component, every trace on the PCB. Electrically transparent, mechanically protective.

Once applied = permanent protection. You cannot remove it cleanly without damaging the board. Therefore: all soldering completed before coating, all connectors masked.

Gps Antenna And Cf Material¶

Carbon fibre is electrically conductive. GPS signals (1.5 GHz) cannot penetrate CF material -- an antenna above a CF plate has no direct line of sight to satellites.

Therefore: the GPS patch antenna must have an unobstructed sky view. In V2.4.6 the M10Q sits at the top of the GPS/camera bracket, above the camera, with a clear 180-degree view. The VTX antenna must not block the GPS view from above.

Same principle: a WiFi router behind a TV or inside a metal box loses signal.

Systems Principles¶

Single Source Of Truth¶

Every number exists in one place. reference/LD_-_Variables_v246.md contains all parametric variables -- #RodDia, #ArmShaftLength, #SandwichHeight, etc. All other documents reference these variables.

If you change #ArmShaftLength from 100mm to 105mm, you change it in one place. If the same number lived in Spec, Cookbook, WBS, and PRUSA -- you change it in four places and forget the third. Then you have inconsistent documentation and crashes caused by wrong numbers.

Principle: data redundancy is the enemy of consistency.

Coupon Before Production¶

Before printing 3 PCCF layers (~4.5 hours of printing) we print a 50mm sample with a T-slot and one tab. The test takes 2 minutes. If the fit is wrong, adjust the tolerance and reprint -- in 20 minutes, not in 4.5 hours.

Coupon testing is standard in aerospace and automotive manufacturing. For a hobby builder it feels like a mental barrier ("I want to print the real part"), but the economics are clear: a cheap test saves expensive material and time.

Our coupons: Coupon 8 -- T-lock fit (CRITICAL before X body production) Coupon 8b -- Rod interference fit 2.1mm in PETG

Maiden As A Data Point, Not A Celebration¶

The first flight is the most important data collection event in the entire project. Blackbox records the gyro signal, RPM filters, motor outputs. Gyroflow shows stabilisation quality. Post-flight visual inspection checks T-lock engagement, screws, o-rings.

Maiden does not answer whether the drone flies -- it answers how it flies and where the weak points are. Every deviation from expectation is information, not failure.

"No upfront redesign needed. Build it, fly it, observe it."

Moscow Prioritisation¶

Every project item has a priority: Must -- without this the project does not exist Should -- significantly improves the outcome Could -- nice to have if time and budget allow Won't -- consciously deferred, not forgotten

"Won't" is as important as "Must". An explicit decision not to do something prevents scope creep -- the tendency of a project to grow continuously by "just one more detail". Frame 2 is Won't in V2.4.6. Not forgotten -- deferred.

Operational Principles¶

The Compliance / Stiffness Balance¶

Every drone sits on a spectrum between two extremes:

SOFT <> RIGID absorbs energy transmits energy harder to tune PIDs precise gyro response field-repairable swap economy less destructive crash more destructive crash

foam -> floating -> our V2.4.6 -> race frame -> 10mm CF mount mounts sandwich 4mm CF slab

Race pilots go to the far right (4-10mm CF plate, ~40-60g). V2.4.6 is deliberately in the middle -- stiff enough for accurate gyro response, compliant enough for zonal crash management.

Where exactly on the spectrum does V2.4.6 sit? Blackbox after maiden tells us. If noise floor is significantly higher than V2.14 -> too far left -> add stiffness. If first crash was destructive -> too far right -> consider a flex neck.

Winter Protocol¶

LiPo batteries lose capacity non-linearly below 0 deg C. A battery at +5 deg C may behave as if 30-40% of its rated capacity is unavailable. PETG arm shafts are more brittle at sub-zero temperatures.

Rule: do not fly below 0 deg C in V2.4.6 baseline. Store batteries at room temperature; bring to the flying site in the last 5 minutes before flight.

Prop Balance¶

An unbalanced propeller vibrates at its rotation frequency. Vibration travels through the motor into the floating mount (which attenuates but does not eliminate it) and onward into the gyro. A magnetic balancer catches propeller asymmetry before the first flight.

Time: 3 minutes per propeller. Result: cleaner gyro signal, lower Blackbox noise floor, longer bearing life.

Inspiration From Other Fields¶

Biomechanics: Bones, Tendons, Cartilage¶

Our architecture unintentionally mirrors the biological model:

Bone = PCCF (stiff, brittle, carries compression and bending) Tendon = CF rod (stiff in tension, lightweight, connects distant points) Cartilage = silicone o-ring (damping, load distribution)

From biology: the bone-to-tendon transition is always gradual -- stiff material transitions through a zone of intermediate stiffness to soft. A sharp boundary concentrates stress and causes tearing. Our tab junction is potentially too abrupt -- a more gradual geometry would distribute load better. Worth exploring in a future revision.

Tensegrity (Buckminster Fuller)¶

Structures where compression members (rods) float within a network of tension members (cables) without direct contact. Extremely lightweight, extremely stiff.

Our drone is not tensegrity -- but the pre-tensioned CF rods in the sandwich come close. The key insight: route loads into tension, not compression. PCCF is significantly stronger in tension than in bending. Any geometry that carries load in tension instead of bending is more efficient.

Aerospace: Fuse Pins¶

Aircraft engines are held to the wing by deliberately weakened shear pins (fuse pins). On impact (bird strike, hard landing overload) the pins shear -- the engine detaches, the wing survives intact.

Our arm shaft is the direct equivalent. Deliberately dimensioned as the weakest element in the crash energy path. Field-replaceable in 5 minutes.

Electromagnetic Compatibility (Emc)¶

Every Wire Is An Antenna¶

Every wire carrying changing current radiates an electromagnetic field. The field strength is proportional to: current \times length \times rate of change.

In a drone the worst radiators are: * ESC phase wires: high current, high-frequency PWM switching (~48kHz) * Battery leads: high current spikes at every throttle change * Buck converter output: switching ripple (~180kHz)

These fields couple into the gyro (vibration measurement), compass (heading), and receiver (control link) — corrupting the signals these components depend on.

Twisted Pairs Cancel Fields¶

Two wires carrying equal and opposite currents twisted together cancel their magnetic fields at any external point. The twist forces the fields to alternate direction every half-twist — net external field approaches zero.

Implementation in libdrone: * Motor phase wires (3 per motor): twist all three together before routing through the arm cable groove. ~1 twist per 15mm. * Battery leads (+/−): twist together from XT60 to ESC pads. * Cost: zero. Benefit: dramatic reduction in gyro and compass noise.

Star Grounding¶

Multiple ground connections between components create closed loops. A closed loop is an antenna — any changing magnetic field through the loop induces a current, which creates noise on the ground plane, which appears as noise on every signal referenced to that ground.

Star ground eliminates loops by making all grounds meet at one point:

ESC GND pad (master ground) ├── Battery GND lead ├── FC GND (short direct wire) ├── VTX GND → buck converter GND ├── Primary capacitor GND (directly on pad) └── GPS GND → via FC (not separate wire)

NEVER

✗ Separate GND wire from VTX directly to battery
✗ GND wire running parallel to signal wire
✗ Multiple ground paths creating closed loops

Capacitor Placement — Distance Kills Effectiveness¶

A decoupling capacitor absorbs voltage spikes on the power rail before they propagate as noise. Its effectiveness drops sharply with distance from the source of the spike — a cap 50mm from the ESC pads is significantly less effective than one at 5mm.

Two-stage filtering: * 1000µF electrolytic directly on ESC VBAT/GND pads: handles slow large spikes * 100µF MLCC ceramic on FC 5V pads: handles fast small spikes the electrolytic cannot respond to quickly enough (electrolytics have high ESL at high frequency)

Both caps must be soldered directly to pads — no pigtail wire between cap and pad. The wire inductance defeats the capacitor at the frequencies that matter.

Power/Signal Separation¶

Parallel power and signal wires form a transmission line — energy couples from the power wire into the signal wire capacitively and inductively. Minimum 15mm separation wherever parallel runs are unavoidable.

Routing discipline: * Signal wires (UART, GPS, receiver, MIPI): one side of the frame * Power wires (ESC phase, battery leads): opposite side * Never route power wires toward GPS bracket — compass is at nose, maximum distance from power wiring is the design intent

Ferrite Beads — Targeted High-Frequency Damping¶

A ferrite bead on a wire acts as a frequency-selective resistor — it presents low impedance at DC and low frequencies (passes power normally) and high impedance at high frequencies (damps noise). Unlike a capacitor it does not resonate and does not affect DC efficiency.

Application: 3–4 clip-on ferrite clamps (TDK 3.5mm, clip-on type) stacked on the VTX power wire at the buck converter output. Clip-on type preferred — no rethreading required, repositionable after build. Damps the XL4015 switching frequency (~180kHz) and harmonics before they reach the VTX. Note: clip-on ferrites have slightly lower impedance than solid cores due to the clam-shell air gap — stack 3-4 to compensate.

"A drone is a collection of compromises.¶

A good drone is a collection of deliberate compromises."

Revision History¶