OpenServo.com

[linuxguy]

Have successful Open Servo modifications been done with Hitec HS-645MG servos?

What is the most powerful motor that could be shoe horned into a standard servo case and be expected to perform reliably?

I want to use Open Servos for my BiPod (you need the latest version of Adobe's Acrobat Reader to view this seven page 3D PDF file with seven views of the robot) hybrid biped, but I am not sure the 133 oz/in. torque of the HS-645MG is going to be enough, at least not at a few of the real heavy lifting joints such as the ankle and hip vertical joints. There may be other joints of the 4DOF legs that might need more than 133 oz/in. torque. The HS-5645MG digital servo provides 168 oz/in. torque, which isn't that much better than the HS-645, considering the difference in price. Of course, the HS-5645 is a digital servo and the HS-645 is analog, which may account for the rest of the difference in prices.

I am pretty sure I would be better off using Open Servos for the BiPod, but am not convinced modifying a standard servos is necessarily the best way to go either. If I used stock servos, they would be a servo like the HS-645MG and I would pay somebody to build and program my Open Servo boards, and modify the servos. I'd rather see as much of my money as possible go to support a great Open Source project like Open Servo.

So, what is everyone's idea on this? I have the 3D CAD models completed for the BiPod, and the brackets I have are assembled according to the models. I just need to finish the left leg, build the feet, and decide where I want to put sensors.

The BiPod has many places for various kinds of sensors, including a compass at the back of the pan/tilt sensor turret, at least one 3-Axis accelerometers in each leg, one or two Ultrasonic/IR combos on each foot, and a TPA01 thermal array on the sensor turret. I designed the BiPod to be a biped with many places for sensors, since most bipeds I have seen didn't seem to have places for many sensors, or at least no good places for them. I have even allowed attachment points for two arms if I decide the BiPod can handle and use them effectively.

I just need about $75.00 more worth of SES brackets and I will be ready to start buying servos for the BiPod, which means probably in May.

8-Dale
_________________
No, Mr. Jobs, the BiPod is a ROBOT. It does not play music OR interface with iTunes.
The Dynaplex Network - Robotics, Open Source, Linux, and Technology Forums

[Cliff]

linuxguy,

I'm pretty sure that Mike has been doing testing with the HS-645MG and OpenServo, since he beat me up on my board dimensions for OSvX2, wanting to make sure my board would fit the HS-645MG. Mike took some pictures of the HS-645MG insides that can be found here .

Cliff

[mpthompson]

Although I have one HS-645MG that was given to me last year, I actually don't have it converted to an OpenServo yet. Therefore, I don't have direct experience with using the HS-645MG hardware.

In my biped I'm still looking to use the HS-475HB servo which is less powerful. If worse comes to worse, I'll replace some of the load bearing servos with something like the HS-645MG.

Because the servo investment for a biped is fairly substantial (for either regular or OpenServos) I would strongly suggest getting one servo and running some tests to evaluate it for your needs. It's not too hard to rig up a test stand so you can run a servo through it's paces under various loading conditions. Some up front testing could save you some grief down the line. Particularly with something experimental as the OpenServo is at this point.

-Mike

[linuxguy]

[SOI Sentinel]

I think the HS-5645MG digital servo will net you the same result as an OpenServo 645MG. I'm quite positive the only difference between the two is the driver board, as I really doubt the similar model numbers are simply a coincidence. Digital torque control can provide that extra torque, especially if you're using digital FETs instead of analog transistor control techniques.

[linuxguy]

[tillin9]

I ordered two boards from Jay to test the following servos:

Hitech 805, 5955, 5645, and MPI-35.

I probably will do the 805 and 5645 first. I will probably remove the OpenServo board from the 5645 and pop it into the MPI-35 next. Only if it makes a big difference in torque will I play with OpenServo conversion of my 5955s (at least until the other projects are done).

I have at least one of all but the MPI-35 (sub-micro) servos on my desk, they haven't come yet, ready to go. I'll post full results once I get my boards.

EDIT: The idea is that if a servo is electronics limited (i.e. really cheap FETs), an OpenServo board will provide a big improvement. This is more for the 805/5955 since they pull enough stall current that even good FETs can start to be limiting. My guess is that the 5955 and 5645 boards use the same FETs, though once I disassemble mine, I will let everyone know for sure what Hitec is doing.

[linuxguy]

[tillin9]

As per my 805 thread in the Lynx forum, the Hitec analogue servos seem to not use FETs, but bipolar transistors in their H-bridge. I initially assumed they were FETs and had marked difficulty find spec. sheets, though given their model other forum members were able to eventually find them. This more or less means that they burn a lot of wattage, some even suggesting they consume more than the motor (though I doubt this). Since the current handling properties of a servo's drive transistors are thermally limited, inefficient control transistors can cause lower current to goto the motor for the same potential/ thermal envelope as more efficient ones. As torque is proportional to motor current via the drive train's essentially constant proportionality, inefficient transistors can directly mean lower torque.

While it would be nice for this theory to be backed up by some experimental evidence (i.e. an OpenServo 645 has the same or higher torque than a 5645) it seems very logical. The jump from 133 oz/in. to 168 oz/in. with similar mechanical hardware also makes me hopeful for a similar percent increase in my 805 from 343 oz/in. to around 430 oz/in.

Please keep us informed about how your 645 conversion goes.

[linuxguy]

[SOI Sentinel]

*looks around* Geeze, you two, pick a forum!

As for my comment (in the other forum) about the bridge burning more current than the motor, it can be true. Most analog H bridges (not necesarilly RC servo designs) use an error signal difference from analog integration of the pulse width versus a similar signal from the pot (the PWM is used for noise resistance for the control signal in RC). This is used to turn on the bridges in an analog current amplifier (active) region. You'll see the voltage drop from the diode function, but at low levels (small motions) you WILL see more power loss in the transistors than the low ohm motor. Large motions exit the active region of the BJT and reach saturation, but you'll still lose 0.7Vxcurrent in power per transistor.

MOSFETs do operate similarly in end result, but instead of having a diode drop, they have a resistance. They also can burn a lot of power in their active region, but act as a switchable gate with low resistance in their saturation region. They also are not current amplifiers, but act as capacitors for their control.

You can turn on a MOSFET fully with a low current supply, but it will take time, and you can maintain their on state after that with very little additional power output. BJTs need their max instantaneous base current met to max out their power control ability, burning more power over time.

[tillin9]

Sorry about the dual forumness.

I definately get your idea about the power use, and am the first to admit my EE knowledge is no where near complete enough to talk definatively, but doesn't the nature of the gain limit the power loss in the bipolar? I certianly believe that this can be made much better by use of a FET, and worse since this isn't just an on/off saturation but an application of PWM. However, I really doubt we're using more power in the transistors than in the motor. I mean, first off that would just be really dumb. The question becomes what amount is being wasted by using cheaper transistors, is it 10%, 20%? I can believe those numbers. I also get your voltage drop argument, but don't think the resistance in the motor is low enough when compared to the junction. Again...the real answers will come when I do some testing.

As far as torque tests. I will try to do something similar to Jim's tests here: http://www.lynxmotion.net/viewtopic.php?t=1710&postdays=0&postorder=asc&start=45
However, I don't have the same SES U rod and mounting hardware. I could buy some, but I think I can make do with some more general lab hardware designed originally for mounting optics (essentially a larg piece of aluminum with a regular set of holes drilled and tapped). The idea is to put in some rods for mounting the force meter, though getting an accurate one is also proving difficult. If all else fails I do have some spring scales here I can just calibrate. Ideally I'd have some Mega style brackets (and may order some from evolution for this purpose), but some rods and chem. set style clamps will probably let me mount the servo okay. Then run it at different voltages and record the force and do some simple calculations if I can't space the end of the meter 1"" from the axis of the servo. Yes, there are some details in making sure the pull is only in one plane and make sure the force meter is free to rotate and so forth, but Jim's example already shows how to do this. Hope this helps. I will post picture of my setup on my thread on the Lynx forum when I actually have this. Unfortunately I'm having a bit of trouble getting my BS2 I2C setup the way I want (it's been awhile), so this may take a few days.