Guide to Robot Servos

Picking the Right Servos

There are a plethora of components to buy on the internet, including servos. For your robot project you could go around and comparison shop. You could pick the serovs that had the best specifications for the price. But this could be a big mistake. Servos are not always as good as their specs, and so it really helps to know what you are getting into before you buy a bunch of them. Some manufacturers are not very accurate in their specifications. Some servos are less reliable then others. Some servo specifications are misleading, particularly torque figures.

Servo Brands

The brands of servos that are dependable enough for Robotics and are also usually the ones that are widely available. These are Hi-Tec, Robotis/ Dynamixel, and Futaba. There are other good servos available, but it is harder to know what you are getting. Some RC Hobby stores have their own line of servos that are good. Hi-Tec and Futaba are probably the two largest manufacturers of servos and so are quite common. Robotis is probably the most popular manufacturer of servos exclusively for robotics.

It's a good idea to limit your servo search to these manufacturers because they are dependable and commonly found. All motors and gears on a robot are likely to fail at some point. You want to be able to replace the gears or the servo if they do fail. You can sometimes find deals on strange servos on the internet, but I recommend being careful about this. They can be oddly shaped, or difficult to calibrate.

Features and Specifications of Servos

Torque

Torque is the amount of force a servo motor generates at a given amount of voltage, assuming sufficient current. For walking robots torque is really important because it relates directly to how much weight your robot can carry. There are definately some trade-offs between speed and torque that I will discuss below, but as a rule of thumb you almost always can use more torque in a walking robot.

Torque is usually listed in ounces per inch (oz-in) or kilograms per centimeter (Kgf-cm). The torque of a spinning motor acts like a lever in the sense that the longer the lever arm, the less the force. So a servo with 10 oz-in of torque delivers 10 ounces of torque when connected to a 1 inch arm, 5 ounces when connected to 2 inch arm, and 1 ounce of force when connected to a ten inch arm. If you are thinking about a walking robot that lever arm length is the length of your legs. The longer you make your legs the less force they will be able to apply. But increasing the length of your legs increases the distance of your gait and hence speed. Increasing the length also usually increases the size of the obstacle you can clear.

There are a few different measurements of torque that can be used, such as dynamic torque and stall torque. Stall torque is the force that it takes to bring a servo to a dead stop at its normal power. Dynamic torque is the amount of force the motor is imparting while it is turning at some given speed. Stall torque is sometimes listed because it is usually the greater of the two measurements. But in practice it is not a good number to base your robot design around. You want to design the size and weight of your robot such that it can be carried by the normal torque of its motors, not the size and weight that will cause it to stall, or even get near to stalling. Because of obstacles, inclines, and momentum, the force on servos is not going to be constant. So if you are close to stalling, you will stall in certain common conditions.

An example of this is that the excellent Dynamixel AX-12+ servos are sometimes listed as having 15 Kgf-cm stall torque. That is a lot. But I have heard people point out that they cannot normally be expected to carry that weight. I believe the naming conventions of the Dynamixel servos give you a tip to what torque the engineers were expecting the servos to normally be expected to deliver. The AX-12 probably can handle about 12 Kgf-cm, the RX-28 about 28 Kgf-cm, the RX-10 about 10 Kgf-cm, and so on. I don't know this for a fact. But I point it out because you need to think about the torque the motor can really handle and not some theoretical peak figure.

Speed

Speed is very important for servos in robot legs. In general, the fewer legs you have the faster the legs have to move to maintain balance. Bipeds need fast and accurate servos for their moves. Hexapods don't have to worry about speed or precision so much because they should always have a triangle of three legs on the ground.

There is a potential design trade-off in taking a fast servo and making a mechanical linkage to reduce speed and improve torque. Some leg designs use a dog bone joint to convert a longer servo movement into a shorter, slower action with more force. This is not the preferred approach by me. It seems to me that you can achieve the same effect with a servo with a lower gearing ratio, since both are taking advantage of the same mechanical principles. Also, in my experience a dog bone joint fails more than a servo gear does. And besides speed, you loose a lot of flexibility. A servo can usually turn at least 120 degrees, sometimes more. A dog bone joint has perhaps 60 degrees of travel.

Weight

It is easy to underestimate the importance of servo weight, since batteries can weigh so much more than any individual servo. But while you may only have one to three batteries, you can easily have a walking robot with twelve or eighteen servos. So the relative weight of those servos can matter a lot.

An important thing to consider about servo weight is the weight of the servo plus whatever bracketing is required to mechanically attach the servo. Some servos come with most of the bracketing built in, particularly the AX-12+. With hobby servos you will have to add a bracket, and perhaps a metal hub and axel to get the mechanism you are looking for. The weight of these things usually isn't much, but again, you have to figure that you will have many of them so you need to account for it.

Price

This seems like an obvious consideration, but the price I am thinking of is the cost of the whole servo system including brackets, hinges, and controllers. Hobby servos in particular need additional mounting hardware in order for them to be mounted reliably on a robot. Some of the cheaper hobby servos may have plastic hubs which you may want to replace with metal hubs, which will come at a cost.

HiTec Servos

I use HiTec servos a lot. They are common, inexpensive, and well made for hobby servos.

These are good quality and inexpensive hobby servos. I use HS-422 in fluffy. They are not very powerful, and so I replaced HS-422s on the back leg to help carry the weight in the back. They are fine servo, but probably not adequate torque for all applications.

The HS-485 is a very good inexpensive servo. They are very similar to the HS-422/425, but with more torque. In Fluffy I exchanged HS-485s for more expensive HS-645s that had gotten a worn gear. They worked great. For the cost these are terrific servos.

This servo has been around for a while, I think because it is popular. It is small, strong, and fairly cheap. What is not to like? If you are making a lynxmotion robot this is a good choice. I have had these servos strip their gears more than once. I think it is because they are among the more powerful servos you can get in such a small package. The small package means smaller gears. You can get replacement gears for them.

This is a quarter scale servo with good torque. It fits into Lynxmotion servo brackets and has durable metal gears. It is a little heavy, but it is powerful and durable. I have seen discussions of whether servos really perform up to their specifications, with these I think they really live up to their specs. I have used this servo with absolutely no problems on Sterylite6000.

A quarter scale servo with amazing torque for the price. Unfortunately they do not fit into Lynxmotion brackets. I use them for the leg lifting motors in Sterylite6000, in RobotZone brackets. That is fabulously strong.

The HSR-1425 is a servo that rotates continuously-- in other words it turns around like a conventional motor. You control it by pulse width so that you can adjust the speed and direction.

The HSR-1425 is great value that I highly recommend. For $16 you can buy a geared motor that is easy to mount and easy to attach a wheel to. It is not only reversing, but it also can adjust its speed. It can be controlled by most microcontrollers with no additional circuitry. It isn't as strong as some gear motors and you can't run as much power through it as a motor with an H-Bridge or a an electronic speed controller. But for quick light robot they work great.

I use HSR-1425s in the Peanut Tin of Terror. I considered using gear motors. Good gear motors cost $25 a pop, and the cheapest are probably $10 each. Once you have a gear motor you have to control it with an H-Bridge or electronic speed controller. The H-Bridge is cheaper, but it doesn't give you any speed control. An ESC can easily cost you $40 and up. So I could have spent $80 for a two wheel drive system ($20x 2 for motors, $40 speed control) but instead spent $32 on two HSR-1425s. You can definately get a lot more power or speed with other solutions, but the HSRs are cheap and they just work.

Dynamixel

Dynamixel servos are made for robots and seem to offer some clear advantages over hobby servos. Because they use serial communication they require fewer wires to the microprocessor, and less computational babysitting. They also can report back their position.

I have read that dynamixel servos have great holding torque but otherwise don't always live up to their torque ratings. Their are enough really good walking robots using Dynamixel servos that I am comfortable saying they are a good choice for a robot.

They do require very fast communication, so not all microprocessors are going to work well with them. In particular, I think you could have a problem using a BASIC Stamp and some Arduinios. Check ahead of time if your microprocessor can handle the communication speed required.

When you look at the features vs. the cost, these are terrific servos. When you consider the cost of brackets (which are usually included with these) they really do not cost a lot more than good hobby servos like the HS-645. But they have a lot more features, such as position reporting and heat detection.

Position reporting lets your controller read the position of the servos. This means you can pose a robot and have the controller remember the pose, which is really useful. It also means you can detect when the servo isn't going where you planned, such as when it overshoots a particular position.

These just came out and seem like a stronger version of the AX-12, with a faster coreless motor. They seem to be about twice the cost of AX-12s, without a doubling of performance. It doesn't mean they are not worth it, it just shows what a great deal the AX-12s are.

I only know of this servo from the robots that use it. But they are great robots made by talented builders. If you really want to spend a lot on servos, this is probably a very good choice. Google Giger, or the RNSL hexapod.

Where to Buy Servos on the Internet

Lynxmotion is a good source for robot kits, robot brackets, and robot servos. They tend to only sell servos that are appropriate for robots.

Servocity has a great selection of servos, R/C and mechanical bits. Not all of them are good choices for robotics but the selection is good.

Crustcrawler sell good quality high end robotics components. This is a good place to get AX-12s and brackets for them.

Trossen is a good site for a wide variety of robotic components.