They can be used for good and evil.
I need a solution thats cheap and good
need to replace the top cover by diy aluminum one
that way the case will no longer fail
also cnc thread
pic related: my fingers
>>881039
what solution?
measure the distances between the centers and the outer bolt holes
mill out clearance for gears
mill out bearing for right, sleeve bearing for center, stub for left
>>881039
>the case will no longer fail
How are the cases failing? It's possible that your application places stresses on the servo shaft that the case was not designed to handle. So the solution might be to keep the plastic case and add a bushing, bearing or flexible coupling on the outside of the servo.
>>881663
I never did any metal work.
There are so many things out there for milling, like CNC, doing it by hand, molding.
I just don't know how to make my prototype servo. Trying to think this through as much as possible before I start working on it, ordering materials and stuff.
>>881687
This man did a pretty good writeup on the good and the bad with this servo. I thought it would be very interesting to engineer a way to use the good parts of these to make better servos.
End goal is to use them for a humanoid robot (16" tall) that can fight. Maybe in a few years be able to compete at Robo-One, or some other event.
"3. Weak case
The MG995 seems to have been made with the least possible amount of plastic. The most common failure I have found is the plastic lip around the center shaft in the gearbox."
(from the article: Deficiencies of the TowerPro MG995 )
>>882315
How are you breaking these servos in your robot? Are you not supporting them on both sides? Perhaps you should use dynamixels.
>>882334
Sure if you buy me 17 mx24 Dynamixels I sure will.
One of the issues is to find how to make a second hinge point. Then the case strengthening would be priority #2. Replacing the pcb with one that takes serial communication and heat sensor for the motor, as well as a better potentiometer is priority #3.
I'm having a hard time getting started because I'm not a mechanical guy, I'm more EE inclined.
servo thread huh>?
just ripped this fucker out of the ol' slutbot
750w 3phase,brake, 12bit serial encoder, commin atcha
guy from the servo place came to pick her up today. they better not fuck this up.
if its more than $2000 for a repair/referb im gonna be so fucking pissed off and it will be the last time i ask for help/accept help from anyone
with the servo out, shes limp and never in the mood. imma stick it in that planetary gear hole
also, if you buy toy servos you will get toy results
you dont need to use all dyna servos, but they are the best toy servo
i understand these types of servos are easy to use and cheap, and have a nice working package as well , but real ass servos will give real ass results.
maxxon? maxon or faulhaber make tiny ass servos with encoders.
I see you want
>cheap
>good
if you really want cheap and good, buy a maxon or faulhaber micro servo on ebay and save a shit ton.
>>882387
i dont want to buy though
why would I do that?
i do this for the challenge and the pain in the ass
that it will cause me
also i want to get more xp in engineering products
looks like first thing will be to order a 17dof robot frame
so that i can start actuating the legs and stress test the servos
i already received 5 or 6 of these mg995
>>882360
how the fuck are you breaking these servos? Show a picture or something for crying out loud
>>One of the issues is to find how to make a second hinge point
3d print some servo brackets. Like this:
http://www.thingiverse.com/thing:356398
Second, why the fuck are you going to CNC stuff? That's gonna be more expensive than just buying servos.
Third why do you need serial communication and a heat sensor? What does that really buy you? Why swap out the potentiometer? At that point why not just buy a new servo? AX-12s aren't that expensive, like 30 bucks a pop if you buy them in bulk.
I would really like to hear your reasons for soldering in a whole new PCB. There are good reasons to do this, it just seems you are doing so for the wrong reasons. The Chinese are starting to make servos like this for the right reasons. They will be out by next year.
>>882398
>>also i want to get more xp in engineering products
move to shenzen, go to school, or get a job.
>>looks like first thing will be to order a 17dof robot frame
if you want to get xp in engineering products why are you ordering a 17dof robot frame?
OP look, with robots it's really easy to spend a lot of money and not get anything done. Get your stuff working in simulation before you spend a dime.
Here's V-REP a FREE robot simulator:
http://www.coppeliarobotics.com/
Go download it( go here and get the proEDU one: http://www.coppeliarobotics.com/downloads.html) and get some of the humanoids to do what you want first before you spend any money.
>>882384
How much do servos like that usually cost to refurb?
Someone explain to me how to interpret the maxon servo specs. They list Mn as a handful of millinewtonmeters of torque for 200w motors and that makes no sense to me. How do I work out the holding torque for a motor on their website? Say this is for a linear actuator for a man machine interface I am working on.
>>882568
The maxxon motors are some of the best, made in Switzerland or something, used in general for dentist torture tools.
They are also used in some of the better Dynamixel servos (which seem to be the best robot servos, even they are not used on the robots in Robo-One for some weird reason).
It would be great to be able to get more info on them. Stall torque, bulk price, etc...
>>882537
>how the fuck are you breaking these servos
I'm not but the article I cited explains it, refer to :
>>882315
>>882537
>3d print some servo brackets
Nice find, I'll look into this.
>why the fuck are you going to CNC stuff
Because I don't trust my hand to make such small things and I want to make a lot of pieces so that I can make a lot of robots and be able to break them a lot and quickly replace the parts.
This may be moot as I have never researched CNC, but it seemed like the way to mass produce metal parts.
>if you want to get xp in engineering products why are you ordering a 17dof robot frame?
Because I don't think making my own robot frame is interesting engineering-wise, and they cost only $75..
>I would really like to hear your reasons for soldering in a whole new PCB
Because I want to be able to say I designed a PCB with all the motor driving components. Bonus: be able to add serial communication and daisy chaining the servos to remove most of the wire bloat that we see on robots that don't have serial and daisy chained communication and power.
>AX-12s aren't that expensive, like 30 bucks a pop if you buy them in bulk.
These servos have plastic gears, which I don't want (pic-related).
It would cost $510 worth of servos for one single humanoid, so cost prohibitive.
They are the weakest servo from the entire product line (15 kg.cm, which is not too shabby, but still very weak for moving fast or fighting).
>>882639
continued:
I know the MG995 ($5, 13kg.cm stall torque, metal gears, can be bought in hundreds at a time) only has 13kg.cm (again, we are talking about stall torque, the point at which you don't want to be at, especially for extended duration) and this is why I'm looking into replacing everything around the gear train to give it more torque, including maybe a better motor.
>>882640
oops forgot the pic for the plastic gears on the AX12.
>>882639
yeah, I get that they are premium shit. but someone explain how to read the datasheets please. is 27nm maximum holding torque a lot for a servo that costs 700 swiss francs?
should I be looking at stepper motors? where can I find servos or steppers that have high holding torque or should i really be looking at gearing up the servo for a slower linear actuator to accommodate the 10000rpm servo?
I've seen the very simple motorized linear actuators that are just a rod connected to a wheel. and i like this. but I want precise rapid movement control for my study of prodding walnuts that are down shallow holes. unfortunately my prodding device needs to be budgeted for at least 300g and it would be best if i could have a multi axis arm so that I can prod walnuts from different angles.
>>882647
link to said datasheets and everything else?
>>882651
so if I wanted to use this, I could just use a very tiny gear on the servo and a very large gear so that the maximum rpm of 70 000 would give me more thrusting force? it's just that sometimes the aperature of the hole through which I am prodding sometimes shrinks and applies friction to the payload and so I would need to overcome the friction and still be able to consistently thrust at the programmed velocity.
>>882655
woops wrong photo
>>882655
>prodding
Are you asking about how to PID your drill's downward force/velocity?
>>882647
Power is a product of torque and speed. To get the rated power, you'll need to run the motor near the maximum speed, since the available torque at low speeds isn't that much higher.
>>882662
no I should be able to do this without hassle.
I am new to servos though so I don't really understand the capabilities of commercial models. just like I know that a steel bracket of a size that's big enough can hold a shelf, It's like I'm asking, what bracket is definitely too small and will not support the shelf.
I'm trying to work out how to select a motor that wont stall if say the arm comes up against a resisting force. lets say that the aperture can grip the wallnut prodder. and the force vector for that resistance can be up to 100 newtons
>>882671
I dont really understand how to interpret the datasheet to work out how much stall torque I can get and the slutbot/occasional murderbot uses servos which has been mentioned, can crush a man before he can blink yet can only sustain a 2.6kg payload. so i am just kind of confused by the tiny payload capabilities cited and the capacity for murder.
>>882671
stall torque on geared motors is per distance from the center of the last gear's shaft
if you need 100Newtons, or 10 KG, you'll need a servo that gives a stall torque of twice as much to be safe
the servo will need to be 20kg.cm
the ".cm" part means you can have that torque if the arm is 1 centimeter long
if the arm is 2 cm long, you'll need a 40 kg.cm torque
I don't know the relationship between the final torque at the last gear and any of the specs of the motor behind the gear. I am really interested in learning this relationship (obviously for that servo project for the humanoid, knowing what motors to get to replace the weak one would be a game changer).
I know the gear ratio can be calculated even though I've never done it and am not quite sure about the specifics on how to do it, but that's all I know.
Some guy last week was telling me to use a servo rated at stall torque that is at least three times higher than normal torque use, but who knows if that's correct.
>>882675
Oh the gear ratios seem pretty straight forward now. It's just linear. I could get a max speed of 60rpm with 34nm which is quite a respectable amount of force but I'd need a gear reduction of 1000:1 so I don't think this would be practical nor could I really imagine a gearbox that could manage this without many large gears. And as the mass of the gearbox increases, acceleration would suffer.
It would be possible with three stages of 1:10 gears bit that's still pretty fucking huge so I might need to look at higher torque servos.
>>882639
It will probably cost more than $510 to do CNC and make entirely new PCBs. For low production runs CNC is expensive.
>> plastic gears
Doesn't matter as long as it provides the torque you need
>>weakest servo
isn't the servo you are trying to modify only rated for 10 kg/cm?
>>882640
At that point why are you even modifying servos? Why not just build your own from scratch?
>>882687
>At that point why are you even modifying servos? Why not just build your own from scratch?
I wouldn't mind doing that.
I was thinking modifying these would be a good start into making my own as a next goal. I can do the PCB myself without any real challenge, for the rest there are many unknowns, so this is research/testing phase.
I think I'll go print some servo brackets tonight at the local hack shop to see if that's a good alternative to adding a second hinge point.
I'm wondering whether or not printing the servo case would be good for high torque servos. If some well thought out critical pieces (drivetrain parts) are picked for cheapness and goodness and fitted into the plastic case.
>>882714
You are going to spend more money making pcbs and cases than you were if you were just to buy some good servos.
Besides there are new servos coming out in a couple of months that will probably change robo-one: https://www.kickstarter.com/projects/1128055363/7bot-a-powerful-desktop-robot-arm-for-future-inven
These look like they have torque sensing capability and are pretty powerful. These are the only conceivable reasons you should be modifying servos.
Serial communication doesn't buy you much, neither does temperature sensing, weak cases can be solved by not buying such a shitty servo in the first place.
>>882685
>> 1000:1 gearbox
Check out harmonic drives, cycloidal drives and other gearboxes used by robots. These can get insane gear ratios in a compact size.
If you are walnut guy, just controlling velocity will result in harm to your walnut. You need force control
>>881039
Have you tried wood. Or perhaps katanas or Iraqi sword replicas
OP here
printing a servo bracket
estimated 1 hour
20 min left
pic related
Seems decent.
Just don't have the actual 2nd hinge to put in the bracket.
Printing U_servos2.stl
Full robot joint.
>>882896
My first print ever, how to make all these better?
One thing I came up with on my own is to add material specifically for the servo bracket (topmost piece) because right now one hole is busted, also the bracket was not tall enough for the servo (by just a bit) so I need to make it a little taller.
>>882900
Go through the calibration guide on the reprap wiki. Do all the steps in order to tune your slicer settings. Your hotend is also too high.
Use the catalog and find /3dpg/ for more help.
>>882911
Sigh... a new book to open. Haha.
I was making this with a "hood rat" Ultimaker 1, maybe it's not too good.
I found a place with 4 new Makerbots. Gonna be able to print faster now and probably with better precision, if not I'll allocate some time to tuning the machines and actually learning the tricks to it.
I want to see if these plastic brackets are a proper replacement for the "industry" standard aluminum robot bracketry.
Also some smart dude last night on IRC found what appears to be a really good deal for strong robot servos that have 2 hinges and brackets included:
http://www.alibaba.com/product-detail/LEVIHOBBY-15kg-cm-Dual-shaft-Metal_2014830730.html
>>882657
Were you able to translate the motor specs to usable servo torque specs?
>>883002
in the end, yes. I worked out that motor could deliver over 300N of linear force given the proper gear reductions and a walnut prodding depth of 20cm. as I lay in bed unable to sleep last night, thinking about servos.
>>883012
>as I lay in bed unable to sleep last night, thinking about servos
are you me?
>>883012
300 N is enough to cause injury.
>>883002
Those motor specs tell you enough to properly model the motor. Ball parking, the continous torque speed curve should give a good idea of the torque that it could apply. Choose the desired speed and look up the torque! Easy.
>>883159
my wallnut proder is made of a soft elastomer and the walnuts are quite durable, being able to withstand more than 1000N of force. the force is dampened as well because the proder can deform and flex to absorb and distribute the force it applies.
>>883161
it's not really.
it doesnt even list continuous current ratings. 0/10 would not buy.
>slightly faster than 1rpm
>no current rating
>no load rating
>62g
>shitty hollow plastic everything
it should be fine for a toy or a remote control car. if you take a look at hobbyking and how they list the specs, they sell servos for RC stuff, so its all built to handle that application and all the fancy numbers dont really matter, people just need it to fit their current vehicle and make sure its the right one they need. so it's fine to only give those handful of data stuff. but if you are building your own robots, it's pretty important to know these things.
http://www.hobbyking.com/hobbyking/store/__71029__Turnigy_8482_S3317B_BB_Servo_2_5kg_14sec_19g_AR_Warehouse_.html
>>883234
>>it doesnt even list continuous current ratings. 0/10 would not buy.
You can calculate that from the thermal resistances, thermal time constants, max winding temperature.
I am referring to this >>882657 btw, if you are referring to one of OP's shitty servos I totally agree with you.
>>but if you are building your own robots, it's pretty important to know these things.
I honestly cannot agree more on this. Been trying to hack some servos to make a force controlled robot and it's a bitch not knowing the servo specs.
>>883232
while your walnut might be able to withstand 1000 N of force, your hole might not be able too. You might consider sticking a rod shaped elastomeric prodding device on a novint falcon. You can them for less than 100 dollars on ebay. It's basically a force controlled delta robot. It probably has acceptable specs for prodding walnuts
>>883257
>if you are referring to one of OP's shitty servos I totally agree with you.
I was referring to the one this guy posted in the link. >>883002
it just has the voltage and time taken to rotate 60o. which basically gives you nothing to work with. no power, current, temperature, no if you do this it will explode specs.
>novint falcon
I can't find them online. the official page lists them for for 250USD though.
I took a look around trying to find force specs for the falcon and I couldnt really fund much. I found this table made by this guy, "Professor William Harwin Director of Research for the School of Systems Engineering at the University of Reading, where his research interests encompass cybernetics and the interfaces between humans and smart machines as typified by haptic devices, and medical and rehabilitation robots."
so I figure he knows his stuff. the problem with adapting the haptic devices seems to be that they're limited by the mechanical constraints of the human wrist. did you know that gymnasts routines can't last for more than about a minute? it's because even though the the muscles can be strong enough, even with the wrists heavily strapped, they will fail if used any longer than this. even the best, most hardenned athletes in the world can't last this long. https://www.youtube.com/watch?v=2407_uPHpIk
the servo >>882657 has a holding torque when geared up that far to be considered ridiculously strong (1580mNm stall torque x 1:1000 gear ratio = 1580Nm or 15800N or 1580kg with a 20cm stroke depth[10cm from axis].) of course this is the absolute maximum force though and assuming no mechanical losses in the system. even the experimental have pretty weak holding torque at about 1kg, but it's just enough to stop your fingers from moving the joystick, not to really hold.
>>883278
cont
so I drew this diagram to sort of explain my rough plan
>>882537
>I would really like to hear your reasons for soldering in a whole new PCB
To add serial communication to the servo. That way you can tell the micro the exact position of the servo at any time.
I just received the robot frame 17DOF SainSmart. Gonna see what the mg995 can do. Some will probably break early. I also ordered 2 mg959 rated at 30kg and aluminum case. Will put them at the two spots the robot has the most difficulty to move.
gotta build the servo driver board now.
Gonna start with an atmega328p and some voltage regulators, a small battery probably a 2S 7.4V 1300mAh, because the servos are rated for 6V so I won't lose too much energy in heat from 7.4 to 6 at the voltage regulator.
Thoughts?
Leg is ready, I'm missing those servo horns still so waiting for them. They are coming straight from China..sigh
It's okay I'll need time to design the servo driver board anyway.
