The OpenServo

[RobotJay]

Hi guys!

I've been lurking here for about a year now, and I finally decided to step into the community see what you guys think: Because of information I found on this site, I was able to come across the OpenServo project. Its a project intended to create cheap digital servos by replacing the PCB found in standard analog servos. The project has HUGE potential because of the robustness of the circuitry. The circuits use I2C/TWI instead of your typical PWM control, which allows the servos to be daisy-chained. This could SERIOUSLY cut down on the number of wires used in your bot as well as the number of outputs required from your controller. Go to their website and take a look:
http://www.openservo.com

Can you imagine? A digital, daisy-chainable servo, with power consumption feedback? For under $30? That's amazing IMHO. Let me know what you guys think, or if you've already heard of the project.

[Joe]

I've been lurking here for about a year now, and I finally decided to step into the community see what you guys think: Because of information I found on this site, I was able to come across the OpenServo project. Its a project intended to create cheap digital servos by replacing the PCB found in standard analog servos.

It doesn't really do that, though, because most of the cost of the servos seems to be in the mechanics, not the electronics. At the same time, we've got hungry young companies like Bioloid offering 16 kg cm servos for $45 that are already digital. Can you put together an OpenServo with torque anywhere close to 16 kg cm for $45? I've asked this question before, and so far, it seems that nobody can, because even analog hobby servos with that much torque (or close to it) tend to cost a lot more.

On the other hand, JR Robotics recently came out with some new analog servos, including one with 9.5 kg cm of torque (or so) for $30, which is a darn good deal. So, if you could get your OpenServo board down to $15, you could get a 10 kg cm digital servo for $45... but personally I'd still rather have the 16 kg cm for the same price. But there might be some advantages to the JR/OpenServo combo, like maybe being able to fit into Lynxmotion brackets. (Although Bioloid brackets are substantially cheaper, and probably stronger and lighter too, so this would be a draw only if you already had a lot of Lynxmotion brackets lying around.)

Can you imagine? A digital, daisy-chainable servo, with power consumption feedback? For under $30?

What such servo could you build for under $30? How much torque would it have? Please show me. I want to get excited about the OpenServo project, I really do, because I'm a cheapskate and I dig open source. But for high-torque servos of the sort needed to build Robo-One bots, I just don't see how it can come even close to competing with commercial offerings such as Bioloid.

Best,
— Joe

P.S. Yes, we've already heard of the project.

[RobotJay]

Joe,

What about those of us who already have analog hardware? Or those of us that want to upgrade our low functioning digital hardware (like the Kondo ICS servos)? Sure, the Bioloid is nice, but even $45 a unit is expensive. Chech this out:

http://www.peakmodel.com/index.php?main_page=product_info&products_id=208

$12 per servo. At single unit price. From just this distributor. More searching might yield even lower pricing in low to medium quantity...

I am selling the OpenServo boards for $18 each at 16 units. (The minimum number for a low functioning humanoid.) $30 per unit is still expensive, but not as expensive as Bioloid's.


Also, not all humanoids require massive amounts of torque from their servos. Only Robo-One style robots do. What if you wanted to build a team of smaller, lighter, soccer bots? They don't need the torque to push over their competitors. Robot Labs' "Plen" robot seems to get enough functionality out of 60 oz-in servos... I think you could easily get away with micro servos like this one:

http://www.peakmodel.com/index.php?main_page=product_info&cPath=47_79&products_id=211

$7.20 for 50 oz-in is not too shabby. Again this is just single unit pricing from this particular distributor.

Either way, the actual cost of digital servos are FAR less than the major makers want you to believe. Are you really going to pay that much extra for something that is already open knowledge?

[Joe]

Sure, the Bioloid is nice, but even $45 a unit is expensive. Chech this out:

http://www.peakmodel.com/index.php?main_page=product_info&products_id=208

$12 per servo.

Wow, $12 for 13 kg cm is a really good price. Almost too good to be true... I recently looked into some other off-brand servos that had a similar price/power ratio, but people who tried them told me they tended to self-destruct rather quickly and would end up wasting more money than they saved. But that may not be the case with these; could be that Tower Pro is simply underselling to try to break into the market. It's also possible that some of the problems of these cheap servos could be corrected with better electronics (i.e. OpenServo).

I am selling the OpenServo boards for $18 each at 16 units. (The minimum number for a low functioning humanoid.) $30 per unit is still expensive, but not as expensive as Bioloid's.

True (though it doesn't quite match it in capability either). You should consider putting this package together and selling it as such, if it turns out to be a good combination. Yes, people are often willing to pay for things that they could, in principle and with sufficient time and skill, do for themselves!

Best,
— Joe

[RobotJay]

Joe,

I've only noticed one issue with the stock MG945's that I own: they "jitter" when idle. After I replaced the stock PCB with the OpenServo PCB, the jitters haven't come back.

I'm not sure why the servos are so cheap, but they work in my biped just fine. Most of the bad reviews I've seen are from airplane hobbyists, who demand MUCH more accuracy out of their servos than our (relatively) slow, ground based, bipeds. (Because things like a "jitter" on an airplane will end up crashing the plane and costing the hobbyist hundreds of dollars, whereas on a biped it will only make the robot shake a little.)

I WILL start selling the complete package, and in fact, I've already started selling the few sample servos I received from Tower Pro. I have 17 left and I am selling them for $39 each. Unfortunately, the Bioloid servos are only $5 more for 3 more kg-cm's, but I have done the math and I can easily sell the TowerPro OpenServos for $29 each, in quantity 100. At quantity 500 I can get the price down to $22. You can imagine where this is going. All it will take is more interest so I can guarantee I won't go broke and not make my money back...

If you're interested AT ALL, though (even if you have no plans to buy a servo) the OpenServo PCB can easily be modified to control ANY motor/potentiometer combo. You really should join their forums and post your thoughts.

[Ray]

great!

like your web page and your idea!
I have been studying digital servo control for a time,
I hope you can having more algorithm on the digital control especially when the loading is always changing non-linearly. Like some advanced control such as adaptive control, recursive control ...etc which is not so easily implemented in embedded controller because of the masive computation. (even though DSP can be used)

Just brought my MG995 (very very cheap!) and modify the feedback resistor so that overall angle can attain 180degree.

But, completely replace the driver and controller is much more challenging !

[limor]

Hi RobotJay,

Your fabrication of the OpenServo boards is a critical first step on a holly mission. Please don't let Joe put you down

Indeed AX12 servos are high torque and currently the only servo box designed for humanoid research. But different robot builders have different tastes and not everyone that wants to have advanced control features will necessarily want Robotis.

As I mentioned on the OpenServo forum, many KHR-1 and Robonova owners may want to upgrade their robots and experiment with more exciting control techniques that the OpenServo boards allow through position&current feedback and speed control. This gives rise to an opportunity to upgrade the main controler board with better processing power and in general give a whole new life to the robot.

I have an old KHR-1 that I'd be happy to make its servos go OS. But so far i haven't figured out how to extract the board and motor from the box.

I'd like to propose that I send you one of the servos and you document on this forum how to pimp it into an OS servo. I will then buy from you at least 17 OpenServo boards, complete the conversion of my KHR-1 into a KHR-1OS and replace the RCB-1 with a Robostix/Gumstix controler.

This will also probably encourage others on this forum to follow suit.

[RobotJay]

Limor,

If you sent me the servo, I would be more than happy to install the OpenServo v2.1 board, and document the proceedure. E-mail me at RobotJay@gmail.com, and I'll send you my address.

-Jay

[Joe]

Your fabrication of the OpenServo boards is a critical first step on a holy mission. Please don't let Joe put you down

Whoops — does that seem to be what I was doing? I honestly didn't mean it that way. I think OpenServo is a cool project and I'm glad it's being pursued. I've just gotten a little disheartened about producing or finding high-torque digital servos at a more reasonable price... but it's entirely possible that I've given up too soon.

I'd like to propose that I send you one of the servos and you document on this forum how to pimp it into an OS servo. I will then buy from you at least 17 OpenServo boards, complete the conversion of my KHR-1 into a KHR-1OS and replace the RCB-1 with a Robostix/Gumstix controler.

That sounds like a cool project, and I agree, if you can pull it off it will really encourage others in the community to consider doing the same.

I sincerely apologize if it seemed I was disparaging anyone's efforts.

Best,
— Joe

[limor]

regarding the 4 wire daisy chain connection, I looked on the openservo forum and don't quite see a concensus around this matter.

This solution looks nice although it would be nicer to have just a single 4-wire daisy-chain with 4-pin connectors (are 8 needed?)


it also requires the use of a dremel to modify the case...

what do you recommend is the easiest way to go about ..ehm.. daisy-chaining ?

[RobotJay]

Limor,

I've been looking into this for some time now, and I think the best solution is to use a combination of the 2x4 header, and a 4-wire lead soldered to the same spot as the header.

I've found a good 4 position connector from Tyco electronics.

In the picture of the OpenServo on the main page, you can see the 2x4 header I am talking about. The 4 pins on the bottom are only used for doing the initial bootstrapping, and in most instances, will never be used again. Therefore, the 4 pins on top (the 4 you can see) are where the 4-wire lead would be soldered.

In general, I've ruled out the ability to use ribbon, because it seems impossible to find ribbon in a wire gage thick enough to support more than 2 servos without melting, which totally negates the need to use daisy chaining in the first place.

It is, in fact, necessary to dremel a slightly bigger slot for the 2x4 header to stick out. One COULD solder each OpenServo to the next in line without even using the headers at all (and therefore just using the hole that is already present), but that makes the connections more permanent. This permanent connection might be a good solution for Robo-1 bots, because of all the vibration and falling.

Ultimately, the OpenServo is very flexible , and can be used in a wide variety of robotics applications that require digital control. And thusly, the connector needs may vary. I still think my idea of using the Tyco connector AND 2x4 header is the most versatile solution.

[limor]

Those Tyco headers are $4 each on Digikey. What are the part number used for the 2x4 pin male header and the 4-wire cable?

How about using only 3 wires and implementing an asynchronous protocol like the Bioloid bus?. The big advantage is that you can use the existing cable and just need to solder a second short cable with a female Futaba/Hitec header that can be bought for peanuts. Robotis bus has GND, PWR, TTL-data. USART of the Atmega gives 1mbps data speed and is quite sufficient for doing closed loop control at 200hz or more for 20 servos. Otherwise if only a GPIO line is exposed, it can be used with some bit-banging UART implementation instead of the TWI stack that requires 2 lines.

[RobotJay]

Limor,

Not sure where you saw the pricing you posted above, but check these DigiKey Part numbers:

2x4 Header: S2011E-12-ND $1.75 for a strip of 2x24, just snap them apart.

Tyco Housing: A28360-ND $0.42 each at single unit quantity.

Connector that fits inside the Housing: A28331-ND $0.151 each, 10 unit min order.

Total cost for connectors: $1.24 per servo.

Sticking it to the man by making your own digi servos: Priceless.

You can use any 4 wires that are 20-24 AWG thick. The reason for the 4 wires is because the I2C bus requires 4 wires. Why do we use I2C over the TTL or RS485 protocol in the Bioloid servos? Simple: The Microcontroller AVR on the OpenServo already has built in I2C support. In the spirit of cranking out a working version as fast as possible, we went with what was easiest. It is entirely possible to implement a 3-wire interface for the on-board MCU, but that would require a board-level redesign. Possibly a good candidate for OpenServo v3.0? Either way, I don't have much experience with Dynamixel's servos. Perhaps you can help us understand what they are doing, and maybe we can become inspired by some of their ideas. You should post your information about them at the OpenServo site.

I'm still working hard to get the first sub-$100, 19 DOF, biped to market, and I think I can do it by using the OpenServo. If the major servo makers want to over-charge for their technology, fine. I'll open source their ass and crash the market.

[Joe]

I'm still working hard to get the first sub-$100, 19 DOF, biped to market, and I think I can do it by using the OpenServo. If the major servo makers want to over-charge for their technology, fine. I'll open source their ass and crash the market.

You could overshoot your goal by a factor of three and still undersell all the competitors by another factor of three! It seems impossible to me, but I'll keep my fingers crossed. Good luck and keep us posted!

Best,
— Joe

[RobotJay]

And the "Nay-Sayer of the Year" Award goes to... *drum roll*

Joe! (just joking, please don't get mad)

It's not impossible. Far from it. Trust me. The only expensive part is building the tooling for injection molded plastic parts. After that, I have already worked out deals to produce the mechanical and electric components for under $50 per biped.

Of course, a $100 biped would not be as large or strong as current $1000 bipeds. But like I said before, not everyone involved in bipeds is interested in Robo-1. (And possibly there can be smaller weight classes?) Soccer bots can benefit from being small and fast. And fielding a 5 bot team would be much less daunting at $100 per bot. But either way, you'll see. Expect prototypes by summer.