Servos for Robotics

 

A few gems of info I found along the way whilst evaluating different servos for robotics 

Return to the home page of Robots Australia

Robots Australia Menu

Robots Australia Home 

Quadruped4 Robot

Biped Concept Discussion

Earlier Work

Robots Australia Blog on Blogspot

My Workshop 

Servos for Robotics 

Dynamixel AX-12 

Useful Links 

Contact Robots Australia

 

During the development of the quadruped robot, as it evolved through various incarnations I evaluated several different servos before I eventually settled on the Robotis Dynamixel AX-12 as the unit to use in the Quadruped4 design. The info I found out could be useful to others looking to use servos in robotic designs, though the list of servos and the amount of testing is certainly far from including all the many different servo options available.

Manufacturer's Specs

The following list represents a few options I looked at when first considering a robot based on hobby class R/C model servos (for the Quadruped2).  

Mfr

Servo

Claimed Torque (kg.cm)

Speed (s/60ْ)

Weight

Price (AUD)

GWS

GWS S04BB

13

0.20

110g

$20

Hitec

HSR-5995TG

24

0.15

62g

$143

Hitec

HSR-5985MG

12.4

0.13

62g

$100

Futaba

S5050

19

0.20

128g

$117

Futaba

S9351

13.8

0.13

57g

$143

Futaba

RS403PR

13.8

0.13

63g

$168

TowerPro

MG995

11

0.20

55g

$22

TowerPro

SG-5010

11 (tested 3)

0.11

39g

$35

Kondo

KRS-2346

20

0.16

57g

$254

 Out of this list I decided that the GWS brand servo offered the best price/performance ratio, though they are very bulky and heavy compared to some of the other options. The prices in the list represented the best price at which I could source each device at the time. I actually tested the torque of the GWS unit, and was pleasantly surprised to find it met the manufacturer's claims on torque. I also tested the two cheap Chinese TowerPro servos, and found they fell rather short. The MG995's I tested seemed to have badly tuned control system, and hunted about the target angle and so was rejected for that reason, and the SG-5010 was clearly mis-specified and had very low torque capability. I later decided that such a grand endeavour as building a complex robot shouldn't really be sold short by putting price first, and so the favourite became the Hitec HSR-5995TG as representing pretty much the best power/weight ratio available.

Hitec HSR-5995TG Information

The HSR-5995TG is an impressive servo, boasting high torque, titanium gear train, small size and a very light weight (62g). When I selected this unit for the Quadruped3 robot design I bought two of them to test and evaluate.

Overall they came up very well, with usable torque beyond 10kg.cm, and snappy response without overshoot or target position hunting.

When I purchased the two evaluation servos, I was under the impression that they supported Hitec's newly developed HMI protocol (Hitec Multi-Protocol Interface). This was developed by Hitec to add the capability of position feedback to their robotics targeted servos. The protocol extends the normal R/C servo PWM scheme (1 to 2ms pulse width with 20ms period) by detecting certain much shorter pulse widths as commands to return position (angle) feedback and perform some other functions.

With the HSR-5995TG being a robotics servo, I naturally assumed it would have support for this protocol! Alas, no, my testing using the PWM servo robotics controller (see Earlier Work) with this servo showed no response to the 50us pulse which should request angle feedback, and a later email to Hitec support confirmed that no, the 5995 is not capable of supporting the HMI bidirectional protocol. According to Hitec only the 8498 robot servo supports the HMI protocol at this time.

The other mechanism to sense actual servo position I provided for in the PWM Robotics Controller was to modify each servo by supplying an additional wire directly from the wiper of the internal pot out to a new connector. I replaced the standard three pin servo connector with a four pin variety, with position feedback being the fourth pin. My concerns regarding doing this were that the high impedance pot position sense wire would act as an antenna and introduce noise into the servo's control loop, however this turned out to not really be a concern, and I couldn't detect any problems during the testing I carried out.