servo benchmark
goal: find robust servo models for a Minus-Type robot, in the hopes of extending time between servo failures and repairs. A per piece price of up to 20$ may be acceptable (N20 Pololu motors are also in that range). Minus-Type uses Micro Servos, also called 9g or sg90.
tested are the following servo models:
1) DIYMore SG90, plastic gear, 1$: recommended by Vigibot. are very quick which makes for a comically fast head nodding that people find hilarious. used on most of my robots (not WalterJakob's pan-tilt), personal failure rate is pretty high.
2) DIYMore MG90S, metal gear, 3$: had servos in the past that were DOA or didn't work on the lower battery voltage, unfortunately can't remember the exact reason.
3) TiankongRC TS90A, plastic gear, 270°, 3$
4) Doooman Hobby DM-S0090MD, metal gear, 270°, 6$
5) DFRobot DS-S006L, plastic gear, 9$: has a clutch to protect internal gears.
6) FiTec FS90MG, metal gear, 6$
7) Spektrum A330, plastic gear, 18$
8) Futaba S3114, plastic gear, 20$: "sub-micro" servo doesn't fit the current 3D printer files, requires small adjustment.
9) HiTec HS-55+, plastic gear, 20$
10) HiTec HS-5055MG, metal gear, 30$
11) HiTec HS-65MG, metal gear, 30$
↓ See conclusions or table "benchmark table.xlsx" for servo failure times.
test tries to simulate real world usage, common scenarios:
• to work with Vbat voltage 3.3V - 4.2V
• usually low weight/momentum movements (camera is pretty lightweight)
• people may accidentally hit the robot
• robot may drive into walls or objects with a servo absorbing the hit
• servo may want to move into an object and be limited by it
stress test setup:
pre-test #0: power from 3.3V and check if servo is even functional (most servo specifications state a minimum operating voltage of 4V). stop movement with hand, to check if it stops immediately / gets stuck.
main test #1: power from 3.3V, move back and forth (90° - 0°) automatically (eg 1 cycle every 6s to keep temperature low), no weight attached. from time to time force into obstacle (180° position can't be reached) to simulate real world interaction. let run every night and check how many minutes they last before they stop working.
notes:
27.5.22: ran test #1 for 2400min (actual movement: 400min), 1 cycle (to 0° and back to 90°) every 6s, no failures.
-> replaced with 10A 3.3V power supply. increased speed from 1 cycle every 6s to 1 cycle every 2s. now movement from 3 servos overlap. can still be doubled. temperature IR meter: room = 27°C, servo avg = 35°C, servo max = 37°C.
28.5.22: ran test #1 for 480min (actual movement: 240min), 1 cycle every 2s no failures.
-> increased speed to 1 cycle every 1s. movement from 6 servos overlap, maximum possible. temperature IR meter: servo avg = 42°C, servo max = 46°C
29.5.22: ran test #1 for 800min (actual movement: 800min), 1 cycle every 1s,
-> divided runtime value by intensity to "actual movement" value.
VID_20220606_211516.mp4 13,93 Mo
18 téléchargement(s)
Conclusion for Botkins / Minus-Type Robot:
6 servos had good behavior and more than acceptable runtime until failure.
Metal Gears:
• DIYMore MG90S: Okay
• FiTec FS90MG: Okay
• Doooman Hobby DM-S0090MD: Okay (Pan 270°: cable fatigue!)
Plastic Gears:
• DIYMore SG90: Okay
• DFRobot DS-S006L: Okay (auto-revert would need some testing)
• TiankongRC TS90A: Okay (Pan 270°: cable fatigue!)
not recommend servos:
• Spektrum A330: No (failed early)
• Futaba S3114: No (.stl doesn't fit)
• HiTec HS-55+: No (Slow)
• HiTec HS-5055MG: No (3.3V unsupported)
• HiTec HS-65MG: No (Slow)
See table "benchmark table.xlsx" for servo failure times.
benchmark table.xlsx 10,69 Ko
26 téléchargement(s)
Modifié par firened, 25 juin 2022 - 06:54 .
