Ola Fremming's homepage

Servo test results

Presented below are the test-results of the servo's tested so far, good and bad, old and new. Description of test method and how the results are presented can be found here.

NB : Servo torque is presented with the unit kg-cm, to convert to oz-in multiply by 13.86 !

During development of the system and testing of the servo's I had lying around I learnt a lot of interesting things, here are some :

By far most servo's are supplying power to the motor in pulses, these pulses match the update frequency of servos. The amount of power supplied during each pulse is controlled by the pulse-width as can be seen in the two graphs below. The first one is on a JR DS-8401with a 2[kgcm] load, the second is with more load on the same servo. As can be seen the max current in each pulse is very similar, the pulse-width is different. By counting the pulses per a certain time-frame the update frequency of the servo can be calculated, from the graph below : 31 Pulses in 102 [ms]  (31 / 0.102) indicates an update rate of 303[Hz]. The more pulses you count the more accurate this result becomes.

The current consumption during start of movements (as during speed testing) shows that the servo are doing it's very best to start immediately, there is no need for precision involved, just start as fast as possible .  When the servo is to stop the picture is quite different, the the task is not only to stop, but to stop at the correct place. Inside the servo there is a regulating system (in the literature for control-systems, a so called servo-system:-) )this performs the math (and control) needed to make the servo stop in the wanted position. For each current-pulse it's calculating the pulse-width needed to perform the requested positioning.  The graphs below shows how a JR DS-8401 start it's movement in 30-40 [ms] with long pulses, and stops in 40-50[ms] with pulses that are of different length. 

When making the test-rig I have made an effort to supply the servos with as stable voltage as I could.  To verify this the graph below shows the current consumed and the voltage deviation from 5.0[V] during start of movement.  The voltage drop is just above 20[mV] (that's 0.02V) during a 1.7[A] current pulse, which is far better than to be expected in a real life situation in our models.  Our batteries and cable resistance makes it far worse leading to slower and weaker servos than what my test results indicate (unless if you run a higher voltage initially, of course).