In that case motor must rotate on higher rpm, problem will be also inertia ratio wich cause lower accuracy in positioning, by the way great job and thanks for uploading good videos with explanation
Really nice potato measurements! I was too expecting this this to work a lot better but hope that it will with powerful enough power supply. You have become one of my favorite youtube makers because your concepts are really good, they are not too specific for only one purpose and I see these might have some actual use.
This is great content. I'm an engineering student and like to watch your videos and then try to apply the things I'm learning in school to what you have done. Thank you for producing this!
Looks like you should check Org-mode, where you can write document and shell scripts in the document. And you can start the scripts/shell commands with simply Ctrl-C Ctrl-C. and from that Org-mode document, you also can generate PDF-document and HTML documents.
@@Mr30friends : Until you mentioned it, I was assuming we were talking about the motor, encoder, a couple of belts & pulleys, and the 3D printed parts. There is, of course, the electronics side of the equation, which I had conveniently ignored! I don't know the cost of the complete package, but I will be interested to learn that if possible, later on. In my mind, I was relating this to the cost of an "off-the-shelf" closed loop system... If this is not cheaper, then it loses it's significance. We may be breaking a bigger bank than I realised. It's still a neat project tho'! :o)
@@PiefacePete46 The odrive costs 120$, plus 40$ for the encoder, plus 80$ to 100$ for the motor, plus the huge special bearings which probably cost around 30 to 50$. Plus something like 40$ for the power supply. Add to that the 3d printed parts, axels and pulleys. So around 350$ to recreate the exact setup from the video. But of course there are some parts that could be used for more than one joint. For example the odrive can control 2 motors. And a power supply could probably also take care of more than 1 joints. I think he plans to make a 6 joint robot with this kind of joint. So the cost will reach quite high. 3 Odrives 360$, 6 encoders 240$, 6 big bearings 200$(?), 6 motors 450$, lets add another 100$ for power supplies cause I have no idea how many it will need. Lets also add 150$ more for any structural stuff. So if my rough estimates are somehow close to the truth, the robot could cost from 1500$ to lets say 1800$. That's quite a big project.
@@Mr30friends : My bank is definitely broken! :o( His stepper-driven arm is what caught my attention, and I had not taken on board the size difference between that and this current project. I would be curious about cost comparisons between the original arm, and an arm of identical size and power that uses a closed loop scaled down from the one shown here. Closed loop is appealing, but seems to move cost and complexity beyond all but the most determined builder, and that's not me!
@@PiefacePete46 Yeah. For that price, I would expect less backlash in the system... I wouldnt exactly call this the pinacle of budget robotics... At least until chinese factories step in, take the open source odrive and its encoder that cost about 120$+40$ currently. And somehow transform them into 20$+5$ products. Coming soon on aliexpress probably. (or not) Also that would not be theft. The odrive maker made everything open source.
14:33 some sort of brake system would probably be ideal so that it could lock it in place, and when the arm goes to move it releases the brake. The disc brake system on some bicycles would work very well, have a servo to control it. This would also allow you to control how fast the arm would drop under load without having to use the motor to be the sole thing to provide resistance on slowing or stopping the arm under load. And you wouldn't really have to change too much of the current design to do it either.
Very impressive! Really interested how the actuator will perform with full power. It would perhaps be a good idea to introduce some active cooling to the motor when doing that, since the motor magnets are quite susceptible to higher temperatures.
Stall torque is NOT going to be power supply limited, because the applied voltage is many times lower than the supply voltage under those conditions. For an electric motor, PHASE current is proportional to torque, and PHASE voltage is proportional to rotational speed. At stall or low speed, the motor is not turning (either at all, or just very slowly) and so the phase voltage applied by the drive is only that required to drive the phase current though the DC resistance of that phase. Because motors are mainly reactive rather than resistive, that voltage will be small, likely in the order of a few volts. As such, the drive is acting as a step down regulator, and so the phase current available is many times greater than the supply current. For example, if we ignore losses for a moment, your supply is 24v and 17A (24 x 17 = 408W) , if the phase voltage at stall is say 2.4V, (10 x lower) then the motor phase current can be 10 x higher, ie 170A !! (2.4 x 170 = 408W) !! Even with loses, which will be in the order of 10% at stall, you will still have plenty of torque at low rotational speeds! What you need to do is to work out if your Odrive is limiting the phase current and hence the torque, as chances are it is the "weak link" in the chain, so to speak......
@@pavelzhdanov Increasing the ratio will of course generate a higher stall torque. It cannot increase the systems total power, and will reduce the maximum output speed of the actuator. At high ratios the rotational inertia of the drive motor starts to dominate as that inertia is referred to the load by the same reduction ratio. If it were me, i'd characterise the motor first one it's own, for torque vs current, and then "do the math" as it were to see if my overall system performance is in fact suitable to drive the load i wanted to drive. One easy way would be to 3d print a "drum" of a known diameter, attached that drum to the motor drive output, mount the motor, and run string around the drum, hanging down to a known mass. With the Odrive you can slowly increase the phase current until the motor juuuust lifts the mass, and then you can calculate the torque the motor makes, and hence it's KT (Torque per amp). The belt drive system and bearings are of course not lossless, so i'd knock off something like 15% to account for those loses when the reduction system is in place.
Based on other comments as well, the whole system with such motors would not work finally even with higher ratio. However interesting to see the result
Excellent! I am one of your patreons and I've got to tell you that you're a very good investment: I learn something new from every video you post. Great work and it looks like you're having fun too, so do I, this is one of the best reality tv series, lol
Great work! Torque at stall means that all of those amps go into heating, without active cooling it will be very easy to damage the motor if you push 50A through it for longer than a second or two. You might want to consider the motor's motor constant and the maximum amount of heat that you can dissipate from the motor. The motor you are using is designed to provide a large amount of work at a relatively high RPM, for this application a flatter motor with a higher motor constant may be preferable. There should be some hobby motors that are more suited for the task at a similar price if you are interested in changing your design at all.
Do you know what's the possible cause behind the slight jittering at 12:54? Is is the controller? I feel like it overshoots a little every time the goal is changed. Maybe over damping can help?
This controller have PID control and it is possible to change all parameters for this control. I hope I just need adjust this parameter and it will solve this problem. This is something I should try. Thank you!
Very nice! I agree that the torque is lower than it should be, but I don't think it's the power supply at fault. Can you try to run the encoder offset calibration AFTER index search?
Great Job!!, I like your videos. It would be great if you can continue to develop this type of works and share it. When I can I would like to support you from Patreon. am interested in learning about robotics and when I saw your work, I was impressed.
Thank you for this inspiring comment! I will continue with this development, and I do plan to share it, when I will be happy with design and when it will work properly. Thank you!
You need to tune your PI-values on the odrive. The motor overshoots quite a bit and corrects back. Of course, every time you change the mechanical load, the PI-values need to be tuned again...
great job, thanks . The Arm material would be ligther, I would recommend carbon fibre, and it does not have to be so fast (to be fast , you will loose your torque at the same power) By the way I am excited to see the next, finally doing reverse kinematics equations and calculations
The problem is not so much with the power supply, as in the non-tuned PIDs in the position. This is seen in the video. Without a load, the motor does not reach the position. Good luck!
@@Skyentific Yes, of course, but not configured PIDs will not allow the controller to create the necessary currents in the motor windings with any power supply.
doesn't it have way too high RPM/way too low reduction??? I think you mentioned 1:15 redction which would mean it would run at 5700/15=380RPM. Depends on the kind of robot youre tryna build but I think the speed is quite (too) high!
Re Power Supply: I’ve heard some people get away with a low wattage power supply by also adding a LiPo battery in parallel to help feed the high instantaneous currents (I’m pretty sure I read that somewhere in the Odrive forum). I’m not sure how it would go with longer duration holding torque currents, like your application requires, but it might be worth a shot if you have some high discharge rated batteries on hand before you go out and buy your new power supply.
I was curious about that and googled: from : hackaday.io/project/11583-odrive-high-performance-motor-control "Use of a high power density Li-Po battery means you can achieve >1kW peak power output with only a modest power supply."
LiPo battery in parallel. That sounds legit. I have been planning to drive a BT30 mill spindle by a BLDC. The show stopper is the power supply. Our 220V 13A power supply can only put out 2.5KW which is not quite powerful enough. Then by adding LiPo in parallel to the DC supply. May be I can use the more powerful 80x100mm BLDC which has a rating of around 5KW. Long time + continuous cuts is not my intention. My plan is to see if a home BT30 spindle can make dark blue chips out of milling steel. If the test is successful, may be some day ODRIVE will have 48V version for some serious business.
I checked on their site There is a latest 56V version for beta testing with limited cards made The 56V should be targeting 13.8volt lead acid cells. 4 bowling ball size battery can take a lot of abuse
Ran into your channel today, dropped a subscription. I'm still a robotics student so I could be wrong but do you not think the belts could be a major source of this torque deficiency? Running a two stage belt system at high tension seems like a likely source of torque loss and backlash? Also you could check your hypothesis about your power supply by using a multimeter on your power supply and when it is in use looking for concerning voltage drops, or just directly checking the current. I would expect more performance for 400W than what you're getting in this video. Anyway, I don't mean to be negative and I wish you the best with your channel and projects!
This looks fantastic! However, I've done some numbers for a while trying to do my own servo actuators, and the torque conversion needs to be about 1:100 for this kind of application. Otherwise the motor current will need to be too high, which is not practical for repeated motion. I wouldn't be surprised if this demands 30 amps for a few foot-pounds. We are all experts here lol.
For testing and mobility you might be better off with an RC lipo battery. An 6S 4000mAh 25C battery could provide 100A constant for example, and you can go higher in any of those values, both capacity and C rating
Very good video, thanks. The bang/click heard at 12:32 & 14:38 sounded more like belt slippage under torque, probably due to plastic deformation in the 3D printed plastic casing. The holding power (2kg) is almost certainly due to under-power as you say. Thanks again, excellent stuff, I'm jealous. :)
From your accent I presumed you are Russian, but your methodology and work ethics spoke otherwise. The calmness and order in your studio also didn't go with that presumption. Then I saw that you are located in Switzerland. i knew something was fishy! :D An awesome actuator btw
Belt is the issue here, not meant for high gear ratios, try to print some gears and try again, results should be better even with 3d printed gears, the distance is not an issue here as everything is quite close
Around 13 minutes the lever arm has some aftermoves/positioning, the encoder tries to correct to the desired position perhaps because of the high inertia. Can this be eliminated with full motor current aka more holding torque?
Also notice it. 12:30 12:34 12:44 ... all motions with lever). If it related to underkurrent, limiting current in program can solve it. Also I suspect high latency in closed loop control. So it slow react on command and slow correct position after movement.
I think this is because the PID parameters are not well adjusted for this configuration. My idea is to first have a good design of the robot joint and after I will adjust PID control.
I think what that dude James does is his algorithm instructs the Odrive controller to step down the velocity to about 90'% once the encoder reads that the motor is about 75% or so towards it's goal, and then gradually slows down as it gets to point B in the loop. Slowing that rotational "vector landing," I guess I'll call it that lol, but that may help. It seems that PID issue is maybe that the motor over shoots it's goal, and then creeps back awkwardly towards the goal in the loop cycle, as the encoder reads that the rotation has gone to far. It may even be something silly as maybe you have your baud rate set too low from your encoder to Odrive.
@@AngryRamboShow I have been doing research on Odrive system and as for my application which is CNC, residual movements are not ideal. I guess hybrid step servo motors are still a way to go.
@@3dprintwiz378 Stepper motors are good for CNC; they're like servos but can spin continuously and have encoders. My printer has 4 NEMA 17's, and I've seen here on RUclips people make CNC routers homemade, using steppers.
This is very interesting research. What kind of considerations are there for deciding where to put the encoder on a system like this? If the encoder is at the motor or if the encoder could be placed as close as possible to the output? Thanks again for your great videos.
A Robotic Arm may remain stationary but under load for extensive periods of time. How well would do you think an outrunner would handle this scenario? I made a very similar design a couple of years ago with 3 outrunners in a planetary setup (I wanted to use 3 in order to use them at half of their rated load), but never went through with it because when asking around while looking for motors with the appropriate specs, I kept receiving the same response: These motors are not designed to be used under stationary loads and even at half their rated power they will overheat and burn out. Do you have any thermal measurements? I want to give my old project another go
May I recommend LiPo batteries? They can offer you very high currents easily and may be a cheap option at least until you know the current that you need. Awesome work by the way!
Very nice! Couldn't complexity and price be reduced by using an inexpensive brushless esc and a feedback on the arm? You would have real feedback and not need any calibration after first setup.
When to motor is not moving you will just have your i^2 r lossess. Say the motor is 10 miliohms, at 53a that would be 530mv for 28watts. So your holding torque is fairly low power and should be within the range of the power supply. The motor controller acts as a buck converter with the motor inductance and multiplies the current from the psu to your motor command current. Either your motor is saturating with the current, or you don't have the odrive set to the right current. The mass of your arm also contributes to the torque required.
грейт видео, мэн. проблема скорее не в питании, а в выборе мотора. у этих хоббийных моторчиков всё очень плохо с моментом и удержанием ротора. они спроектированы таким образом что свой рассчётный киловатт-два-три они кушают на максимальных оборотах при хорошем охлаждении,так что делать на них серво привод не очень разумно. если обратишь внимание, то в промышленных сервах используется иной тип мотора, с большим моментом, но низкими оборотами (3000 обычно). если хочешь получить что-то с этого хоббийного мотора, то, как тут уже правильно заметили, стоит значительно повысить передаточное отношение. эти моторы любят обороты. можно добавить планетарную ступень перед ремнём - это повысит момент и точность позиционирования.
Fantastic projects, I ve seen all your videos in 2 days :) ... one question though, when u move into a position the motor always stops and goes back a little. Why does this happen? Can it be corrected? Is it overshoot?
bench power supplies like you need get very expensive very fast. You can gang up old server power supplies in series to get that voltage/current rating.
At stall, 1.2 kW of heat generation is a whole lot for that little guy-are you hoping to have that as a peak value? 30 Nm for such an actuator is very ambitious from my experience.
It's very good for education closed loop servo-motor. But problem what you will have, for real working model: small torque, poor braking, and poor keeping position. Probably use gearbox and braking sistem will solve those problems, but in real life more easy to use stepper moter, wich is more enough for this model. Удачи в экспериментах!
you are at the level of understanding it want to gain!! *i click all the stuff and comment, because i am not in the situation to bekome a patreon ... looking forward to ne new PSU and aksing myself if it will work or sholuld have a litte more than needed?! greetings stefan
Great video, and great channel... I dont think it is power supply, I would suspect regulator tuning, On 0RPM voltage on motor coils is low so if there is any capacitance between PS and Motor drive It should give nececerry 53A. It is easily checked, if you have 24V on PS output PS is ok. I suspect tuning because it is a bit slugish to reach final position in video, maby Integral time should be smaler (or integral constant bigger), Try lifting weigth realy slow 0.1mm/sec from floor, just to give PID time to acomodate to weigth...
I suggest that you don't use a higher wattage power supply as the solution (consider the amount of heat that would be generated and the total power requirements that would need to be considered if it was in a real world robotic application). Consider instead, using a planetary stage/s reduction for the BLDC motor. This will give you some standing torque that may help on the output stage. With the current speed of BLDC motor, you can use a 10:1 or less; downfall is speed/torque ratio as you know your output stage will be slower. My idea: Using this BLDC motorized systems in the current robot arm you have developed: This newly design BLDC motorized gearbox does not need to be that fast for your robot arm. You can then convert 3/4 of that speed to torque using a planetary gear system at the output shaft of BLDC motor. All the developed controllers and software would not have to be changed (Software tweaking may be required). I don't have any ideas if a planetary gear system was designed specifically for BLDC motors. However, I have seen planetary gear system that was designed for constant current DC motors adapted and mounted on steppers motors and place in 3D Cartesian printers and gantry type 3D printers. I think this idea will be a better solution to your torque requirements. My Two Cents I did a quick search and found that you may have to change the type of BLDC motor that you used. The type you have in this video is mostly used in RC models and I don't think they have planetary gear systems that fit that particular type. You may have to switch to NEMA 17 format type BLDC. I look forward to what you choose as a solution.
Also what is your intended application? I think you could get by with much more reduction, which will give you much more torque and less motor heating.
Are you intending to use this actuator for a robot build? Because if so, it will definitely not work in anything remotely precise or rigid, since the axis itself has several *degrees* of backlash and almost zero rigidity. I have a tiny 3kg payload KUKA robot, and it takes over 100Nm of torque to move not just axes 1 or 2, but even axis 3, without the brakes. And if the motors are on, I can only make the actual frame and transmission flex. Even the highest static overload with brakes not applied consumes under 150W for all 6 axes, and the robot is rated for over 1kW continuous.
@@MsSomeonenew It doesn't even have to be "obscenely expensive". The cheapest, shittiest worm drive not intended for robotics will have less backlash and more output holding torque than this.
It seems you have small overshooting, you can tune vel_gain, pos_gain and vel_integrator_gain to reach better results. Good luck! (docs.odriverobotics.com/commands)
Геннадий, а почему вы не используете циклоидальные редукторы? ведь их тоже можно легко самому изготовить с помощью 3д принтера ruclips.net/video/B61fXYZmv38/видео.html
Дело не только в источнике питания. Ваш мотор - бездатчиковый. Реализовать с его помощью векторное управление достаточно тяжело, и я сомневаюсь, что ваш драйвер это умеет. Как следствие - снижение момента на старте. Можно реализовать векторное управление, используя линию Z энкодера, но для этого необходима калибровка при каждом холодном старте системы. Вообще, BLDC моторы с малым количеством полюсов - не лучшее решение для задач, где требуется развивать серьезный момент на низких оборотах.
Я делаю калибровку при каждом старте линии З энкодера. И плюс сделана калибровка относительно полюсов мотора и другими линиями энкодера. Так что это система не хуже мотора с датчиками. Даже лучше, так как мой энкодер лучшее разрешение имеет чем датчики на моторах. И этот драйвер был разработан для такой работы на низких оборотах.
@@Skyentific Я говорил не про драйвер, а про двигатель. Драйвер - это три полу-моста))) В любом случае, не готов с вами аргументированно спорить, не видя всей системы и алгоритмов. Если ваша система позволяет в момент старта знать положение полюсов относительно обмоток с точностью около 10 mil и в процессе движения - 100 mil, то реализация векторного управления не представляет труда. Почему я делаю вывод, что пока такой алгоритм управления не реализован: на видео видно (забавная тавтология), как привод "доезжает" в целевую точку, в том числе, со сменой направления движения.
@@dimstr8714 Тут другая проблема, это двигатель BLDC с трапецевидным ЭДС самоиндукции и питать его нужно прямоугольной ШИМ. О FOC (векторном) управлении можно забыть. Нужен PMSM двигатель для питания синусом.
@@kokotmkokot4926 Вы в плену инерции мышления, уважаемый. Да, ОПТИМАЛЬНО с точки зрения КПД подкидывать на обмотки трапецеидальные импульсы. И я так и делаю - в установившемся режиме. А вот в переходных режимах и, особенно, при старте под нагрузкой можно кратковременно закрыть глаза на дополнительный нагрев обмоток и рулить BLDC чистым или не очень синусом. Да, нужен или драйвер с запасом по мощности (для компенсации обратной ЭДС), или снаббер/дампер с той же целью. Но выигрыш при старте под нагрузкой существенный.
@@dimstr8714 вы всем незнакомым категорично утверждаете кто в чьем плену? вы с работой снабберной цепи вообще знакомы? судя по написанному - нет.. дискуссия закрыта, тролль
That's the wrong type of motor for such a project. This is very inefficient use of this motor. You'd need a very low KV motor with higher torque. Likely much smaller power will be plenty too and a lot higher gear ratio. Not sure what your intended use is but the way you built it makes no sense to me outside of experimenting with the concept with parts you already had.
Hmm, i think something in your calculations is wrong. When you have 0 rpm abd just static torque... You still have 0 watts + some losses. But a high current in the rotor windings. Maybe your current controller or position controller has too small of a kp or ti....
Nope! Stepper motors for toys. From this setup I already get 3N.m torque. The stepper motor with such torque weight from 1.5 to 2kg. And my setup is only 1kg. And with proper power supply it will get even higher torque. In general the brushless motors have much higher power on the unit of mass than the stepper motors.
@@Skyentific dont forget steppers are also brushless motors..just another kind of motor.anyway,in the real world,if you want accuracy,torque and holding torque,excluding servo motors,steppers are the way to go.check out my projects,i also started usind dc motor with encoder,using my arduino code and eventually went the stepper way to jave work correctly.
At 1.2KW, your motor is an absolute beast. a higher gear reduction may give you the needed torque without using as much electricity
100% agree with you. I even think that may be I should develop several actuators like this one with different gear ratios. Thank you for the idea!
In that case motor must rotate on higher rpm, problem will be also inertia ratio wich cause lower accuracy in positioning, by the way great job and thanks for uploading good videos with explanation
Really nice potato measurements! I was too expecting this this to work a lot better but hope that it will with powerful enough power supply. You have become one of my favorite youtube makers because your concepts are really good, they are not too specific for only one purpose and I see these might have some actual use.
This is great content. I'm an engineering student and like to watch your videos and then try to apply the things I'm learning in school to what you have done. Thank you for producing this!
Looks like you should check Org-mode, where you can write document and shell scripts in the document. And you can start the scripts/shell commands with simply Ctrl-C Ctrl-C.
and from that Org-mode document, you also can generate PDF-document and HTML documents.
Please, don't hesitate to leave your comment. This will help to promote the video!
It is great to see a closed loop actuator with enough power to be useful, without breaking the bank. Thank you.
But doesn't this setup cost around 350+ $?
@@Mr30friends : Until you mentioned it, I was assuming we were talking about the motor, encoder, a couple of belts & pulleys, and the 3D printed parts. There is, of course, the electronics side of the equation, which I had conveniently ignored! I don't know the cost of the complete package, but I will be interested to learn that if possible, later on.
In my mind, I was relating this to the cost of an "off-the-shelf" closed loop system... If this is not cheaper, then it loses it's significance. We may be breaking a bigger bank than I realised. It's still a neat project tho'! :o)
@@PiefacePete46
The odrive costs 120$, plus 40$ for the encoder, plus 80$ to 100$ for the motor, plus the huge special bearings which probably cost around 30 to 50$. Plus something like 40$ for the power supply.
Add to that the 3d printed parts, axels and pulleys. So around 350$ to recreate the exact setup from the video.
But of course there are some parts that could be used for more than one joint.
For example the odrive can control 2 motors. And a power supply could probably also take care of more than 1 joints.
I think he plans to make a 6 joint robot with this kind of joint. So the cost will reach quite high. 3 Odrives 360$, 6 encoders 240$, 6 big bearings 200$(?), 6 motors 450$, lets add another 100$ for power supplies cause I have no idea how many it will need. Lets also add 150$ more for any structural stuff. So if my rough estimates are somehow close to the truth, the robot could cost from 1500$ to lets say 1800$.
That's quite a big project.
@@Mr30friends : My bank is definitely broken! :o(
His stepper-driven arm is what caught my attention, and I had not taken on board the size difference between that and this current project. I would be curious about cost comparisons between the original arm, and an arm of identical size and power that uses a closed loop scaled down from the one shown here. Closed loop is appealing, but seems to move cost and complexity beyond all but the most determined builder, and that's not me!
@@PiefacePete46 Yeah. For that price, I would expect less backlash in the system... I wouldnt exactly call this the pinacle of budget robotics...
At least until chinese factories step in, take the open source odrive and its encoder that cost about 120$+40$ currently. And somehow transform them into 20$+5$ products. Coming soon on aliexpress probably. (or not)
Also that would not be theft. The odrive maker made everything open source.
14:33 some sort of brake system would probably be ideal so that it could lock it in place, and when the arm goes to move it releases the brake. The disc brake system on some bicycles would work very well, have a servo to control it.
This would also allow you to control how fast the arm would drop under load without having to use the motor to be the sole thing to provide resistance on slowing or stopping the arm under load.
And you wouldn't really have to change too much of the current design to do it either.
Very impressive!
Really interested how the actuator will perform with full power. It would perhaps be a good idea to introduce some active cooling to the motor when doing that, since the motor magnets are quite susceptible to higher temperatures.
You are 100% right. I would do this. And in the next version I plan to install the cooling fan. Thank you for your comment!
Stall torque is NOT going to be power supply limited, because the applied voltage is many times lower than the supply voltage under those conditions.
For an electric motor, PHASE current is proportional to torque, and PHASE voltage is proportional to rotational speed. At stall or low speed, the motor is not turning (either at all, or just very slowly) and so the phase voltage applied by the drive is only that required to drive the phase current though the DC resistance of that phase. Because motors are mainly reactive rather than resistive, that voltage will be small, likely in the order of a few volts. As such, the drive is acting as a step down regulator, and so the phase current available is many times greater than the supply current. For example, if we ignore losses for a moment, your supply is 24v and 17A (24 x 17 = 408W) , if the phase voltage at stall is say 2.4V, (10 x lower) then the motor phase current can be 10 x higher, ie 170A !! (2.4 x 170 = 408W) !! Even with loses, which will be in the order of 10% at stall, you will still have plenty of torque at low rotational speeds!
What you need to do is to work out if your Odrive is limiting the phase current and hence the torque, as chances are it is the "weak link" in the chain, so to speak......
Will reduction ratio solve this problem?
@@pavelzhdanov Increasing the ratio will of course generate a higher stall torque. It cannot increase the systems total power, and will reduce the maximum output speed of the actuator. At high ratios the rotational inertia of the drive motor starts to dominate as that inertia is referred to the load by the same reduction ratio. If it were me, i'd characterise the motor first one it's own, for torque vs current, and then "do the math" as it were to see if my overall system performance is in fact suitable to drive the load i wanted to drive. One easy way would be to 3d print a "drum" of a known diameter, attached that drum to the motor drive output, mount the motor, and run string around the drum, hanging down to a known mass. With the Odrive you can slowly increase the phase current until the motor juuuust lifts the mass, and then you can calculate the torque the motor makes, and hence it's KT (Torque per amp). The belt drive system and bearings are of course not lossless, so i'd knock off something like 15% to account for those loses when the reduction system is in place.
Based on other comments as well, the whole system with such motors would not work finally even with higher ratio. However interesting to see the result
Literally, thank you a lot for sharing such videos. Keep up the good work!
Excellent! I am one of your patreons and I've got to tell you that you're a very good investment: I learn something new from every video you post. Great work and it looks like you're having fun too, so do I, this is one of the best reality tv series, lol
Great work!
Torque at stall means that all of those amps go into heating, without active cooling it will be very easy to damage the motor if you push 50A through it for longer than a second or two. You might want to consider the motor's motor constant and the maximum amount of heat that you can dissipate from the motor.
The motor you are using is designed to provide a large amount of work at a relatively high RPM, for this application a flatter motor with a higher motor constant may be preferable. There should be some hobby motors that are more suited for the task at a similar price if you are interested in changing your design at all.
Another great video. Its really impressive what you do. Keep going...
Simply amazing
Do you know what's the possible cause behind the slight jittering at 12:54? Is is the controller? I feel like it overshoots a little every time the goal is changed. Maybe over damping can help?
This controller have PID control and it is possible to change all parameters for this control. I hope I just need adjust this parameter and it will solve this problem. This is something I should try. Thank you!
Skyentific Thank you for doing all of these on RUclips! I am very impressed by your channel!
Super Excellent videos thank you, I am enjoying them a lot
Great Scott batman, an excellent reference!
Very nice! I agree that the torque is lower than it should be, but I don't think it's the power supply at fault. Can you try to run the encoder offset calibration AFTER index search?
More info here: things-in-motion.blogspot.com/2018/12/how-to-select-right-power-source-for.html
Great Job!!, I like your videos. It would be great if you can continue to develop this type of works and share it. When I can I would like to support you from Patreon. am interested in learning about robotics and when I saw your work, I was impressed.
Thank you for this inspiring comment! I will continue with this development, and I do plan to share it, when I will be happy with design and when it will work properly. Thank you!
Man like God of robotics!
Thank you!
You need to tune your PI-values on the odrive. The motor overshoots quite a bit and corrects back. Of course, every time you change the mechanical load, the PI-values need to be tuned again...
Excellent videos thank you, I am enjoying them a lot!
Thank you for your kind comment!
great job, thanks .
The Arm material would be ligther, I would recommend carbon fibre, and it does not have to be so fast (to be fast , you will loose your torque at the same power)
By the way I am excited to see the next, finally doing reverse kinematics equations and calculations
The problem is not so much with the power supply, as in the non-tuned PIDs in the position. This is seen in the video. Without a load, the motor does not reach the position. Good luck!
PID would not change the maximum torque.
@@Skyentific Yes, of course, but not configured PIDs will not allow the controller to create the necessary currents in the motor windings with any power supply.
doesn't it have way too high RPM/way too low reduction??? I think you mentioned 1:15 redction which would mean it would run at 5700/15=380RPM. Depends on the kind of robot youre tryna build but I think the speed is quite (too) high!
Re Power Supply: I’ve heard some people get away with a low wattage power supply by also adding a LiPo battery in parallel to help feed the high instantaneous currents (I’m pretty sure I read that somewhere in the Odrive forum). I’m not sure how it would go with longer duration holding torque currents, like your application requires, but it might be worth a shot if you have some high discharge rated batteries on hand before you go out and buy your new power supply.
I was curious about that and googled:
from : hackaday.io/project/11583-odrive-high-performance-motor-control
"Use of a high power density Li-Po battery means you can achieve >1kW peak power output with only a modest power supply."
LiPo battery in parallel. That sounds legit.
I have been planning to drive a BT30 mill spindle by a BLDC. The show stopper is the power supply.
Our 220V 13A power supply can only put out 2.5KW which is not quite powerful enough.
Then by adding LiPo in parallel to the DC supply. May be I can use the more powerful 80x100mm BLDC which
has a rating of around 5KW.
Long time + continuous cuts is not my intention. My plan is to see if a home BT30 spindle can make dark blue
chips out of milling steel.
If the test is successful, may be some day ODRIVE will have 48V version for some serious business.
@@lpjunction Sounds very similar to what I have planned. Let me know if you're successful! PS odrive have 48v version on their website ;)
I checked on their site
There is a latest 56V version for beta testing with limited cards made
The 56V should be targeting 13.8volt lead acid cells.
4 bowling ball size battery can take a lot of abuse
Ran into your channel today, dropped a subscription.
I'm still a robotics student so I could be wrong but do you not think the belts could be a major source of this torque deficiency? Running a two stage belt system at high tension seems like a likely source of torque loss and backlash?
Also you could check your hypothesis about your power supply by using a multimeter on your power supply and when it is in use looking for concerning voltage drops, or just directly checking the current. I would expect more performance for 400W than what you're getting in this video.
Anyway, I don't mean to be negative and I wish you the best with your channel and projects!
@ Oran Deutsch Could you please explain why a two stage timing belt causes torque loss and backlash? Thank you.
This looks fantastic! However, I've done some numbers for a while trying to do my own servo actuators, and the torque conversion needs to be about 1:100 for this kind of application. Otherwise the motor current will need to be too high, which is not practical for repeated motion.
I wouldn't be surprised if this demands 30 amps for a few foot-pounds.
We are all experts here lol.
cool:))) Добавь, значительно, передаточное число, минимизируй люфт, настрой разгон-торможение и огонь!
I will! Thank you!
For testing and mobility you might be better off with an RC lipo battery. An 6S 4000mAh 25C battery could provide 100A constant for example, and you can go higher in any of those values, both capacity and C rating
Very nice work!
Thank you!
Very good video, thanks.
The bang/click heard at 12:32 & 14:38 sounded more like belt slippage under torque, probably due to plastic deformation in the 3D printed plastic casing. The holding power (2kg) is almost certainly due to under-power as you say.
Thanks again, excellent stuff, I'm jealous. :)
This is probably one of the better O-Drive demonstrations I've seen so far. I'm curious about the holding power of this setup.
From your accent I presumed you are Russian, but your methodology and work ethics spoke otherwise. The calmness and order in your studio also didn't go with that presumption. Then I saw that you are located in Switzerland. i knew something was fishy! :D
An awesome actuator btw
Belt is the issue here, not meant for high gear ratios, try to print some gears and try again, results should be better even with 3d printed gears, the distance is not an issue here as everything is quite close
1190 w ! Powerfull and cool
Around 13 minutes the lever arm has some aftermoves/positioning, the encoder tries to correct to the desired position perhaps because of the high inertia. Can this be eliminated with full motor current aka more holding torque?
Also notice it. 12:30 12:34 12:44 ... all motions with lever). If it related to underkurrent, limiting current in program can solve it. Also I suspect high latency in closed loop control. So it slow react on command and slow correct position after movement.
super great
I see some residual movements after stopping. Is that normal?
I think this is because the PID parameters are not well adjusted for this configuration. My idea is to first have a good design of the robot joint and after I will adjust PID control.
I think what that dude James does is his algorithm instructs the Odrive controller to step down the velocity to about 90'% once the encoder reads that the motor is about 75% or so towards it's goal, and then gradually slows down as it gets to point B in the loop. Slowing that rotational "vector landing," I guess I'll call it that lol, but that may help. It seems that PID issue is maybe that the motor over shoots it's goal, and then creeps back awkwardly towards the goal in the loop cycle, as the encoder reads that the rotation has gone to far. It may even be something silly as maybe you have your baud rate set too low from your encoder to Odrive.
@@AngryRamboShow I have been doing research on Odrive system and as for my application which is CNC, residual movements are not ideal. I guess hybrid step servo motors are still a way to go.
@@3dprintwiz378 Stepper motors are good for CNC; they're like servos but can spin continuously and have encoders. My printer has 4 NEMA 17's, and I've seen here on RUclips people make CNC routers homemade, using steppers.
@@3dprintwiz378 Properly tuned parameters solves the issue. I'd not shy away from ODrive due to what you've observed.
This is very interesting research. What kind of considerations are there for deciding where to put the encoder on a system like this? If the encoder is at the motor or if the encoder could be placed as close as possible to the output? Thanks again for your great videos.
A Robotic Arm may remain stationary but under load for extensive periods of time. How well would do you think an outrunner would handle this scenario? I made a very similar design a couple of years ago with 3 outrunners in a planetary setup (I wanted to use 3 in order to use them at half of their rated load), but never went through with it because when asking around while looking for motors with the appropriate specs, I kept receiving the same response: These motors are not designed to be used under stationary loads and even at half their rated power they will overheat and burn out. Do you have any thermal measurements? I want to give my old project another go
Excellent.
May I recommend LiPo batteries? They can offer you very high currents easily and may be a cheap option at least until you know the current that you need. Awesome work by the way!
Totally agree!!! This is very good option. But I don't have any. I will try to find one.
@@Skyentific
May be a car battery can act as a current booster too.
Amazing work! Be careful though, this robot might really kill you!
Thank you! Yeah, this becomes powerful!
I love the way you used potatoes to measure torque, I'm a fan of detergent bottles myself xD
nice video but why not use a much higher reduction like 60 it was moving way too fast than necessary
agreed, but I was thinking like 100
gran trabajo
I've always wondering how you pass wires in an arm with more joints without them getting twisted eventually.
okbrb o ring with wipers? Or coiled wires that can stretch with software limited to a few turns?
Very nice! Couldn't complexity and price be reduced by using an inexpensive brushless esc and a feedback on the arm? You would have real feedback and not need any calibration after first setup.
If you measure the current being used when the motor is at standstill, you should be able to infer the force being used to keep the arm still.
When to motor is not moving you will just have your i^2 r lossess. Say the motor is 10 miliohms, at 53a that would be 530mv for 28watts. So your holding torque is fairly low power and should be within the range of the power supply. The motor controller acts as a buck converter with the motor inductance and multiplies the current from the psu to your motor command current.
Either your motor is saturating with the current, or you don't have the odrive set to the right current.
The mass of your arm also contributes to the torque required.
грейт видео, мэн.
проблема скорее не в питании, а в выборе мотора. у этих хоббийных моторчиков всё очень плохо с моментом и удержанием ротора.
они спроектированы таким образом что свой рассчётный киловатт-два-три они кушают на максимальных оборотах при хорошем охлаждении,так что делать на них серво привод не очень разумно.
если обратишь внимание, то в промышленных сервах используется иной тип мотора, с большим моментом, но низкими оборотами (3000 обычно).
если хочешь получить что-то с этого хоббийного мотора, то, как тут уже правильно заметили, стоит значительно повысить передаточное отношение.
эти моторы любят обороты.
можно добавить планетарную ступень перед ремнём - это повысит момент и точность позиционирования.
А если редуктор 1:100 сделать? Решит вопрос? Или там возникает ещё один вопрос с инерции?
Is it just me, or is there half a second of latency before the motor responds to the commands?
I want to learn robotics arm from your video,can you provide link for starter like me
Fantastic projects, I ve seen all your videos in 2 days :) ... one question though, when u move into a position the motor always stops and goes back a little. Why does this happen? Can it be corrected? Is it overshoot?
bench power supplies like you need get very expensive very fast. You can gang up old server power supplies in series to get that voltage/current rating.
Can u give the details about the power supply u r going to use in next video I.e. 53 A power supply
Awesome !!
Thank you.
There is some jerkiness at the end of the move (even unloaded), is the encoder correct for position?
I think I just need to adjust PID parameters in the ODrive. I just left the default values.
I noticed that as well
At stall, 1.2 kW of heat generation is a whole lot for that little guy-are you hoping to have that as a peak value? 30 Nm for such an actuator is very ambitious from my experience.
Whoops, 15 Nm. That is more reasonable, but still quite a lot.
Nice and impressive work! What motor model did you use?
I think it's the Turnigy Aerdrive SK3 or something like that...
What 3D printer and filament are you using? Thank you for these videos.
It's very good for education closed loop servo-motor. But problem what you will have, for real working model: small torque, poor braking, and poor keeping position. Probably use gearbox and braking sistem will solve those problems, but in real life more easy to use stepper moter, wich is more enough for this model.
Удачи в экспериментах!
you are at the level of understanding it want to gain!!
*i click all the stuff and comment, because i am not in the situation to bekome a patreon ...
looking forward to ne new PSU and aksing myself if it will work or sholuld have a litte more than needed?!
greetings
stefan
Great video, and great channel...
I dont think it is power supply, I would suspect regulator tuning,
On 0RPM voltage on motor coils is low so if there is any capacitance between PS and Motor drive
It should give nececerry 53A. It is easily checked, if you have 24V on PS output PS is ok.
I suspect tuning because it is a bit slugish to reach final position in video,
maby Integral time should be smaler (or integral constant bigger),
Try lifting weigth realy slow 0.1mm/sec from floor, just to give PID time to acomodate to weigth...
I suggest that you don't use a higher wattage power supply as the solution (consider the amount of heat that would be generated and the total power requirements that would need to be considered if it was in a real world robotic application). Consider instead, using a planetary stage/s reduction for the BLDC motor. This will give you some standing torque that may help on the output stage. With the current speed of BLDC motor, you can use a 10:1 or less; downfall is speed/torque ratio as you know your output stage will be slower.
My idea:
Using this BLDC motorized systems in the current robot arm you have developed:
This newly design BLDC motorized gearbox does not need to be that fast for your robot arm. You can then convert 3/4 of that speed to torque using a planetary gear system at the output shaft of BLDC motor. All the developed controllers and software would not have to be changed (Software tweaking may be required). I don't have any ideas if a planetary gear system was designed specifically for BLDC motors. However, I have seen planetary gear system that was designed for constant current DC motors adapted and mounted on steppers motors and place in 3D Cartesian printers and gantry type 3D printers. I think this idea will be a better solution to your torque requirements.
My Two Cents
I did a quick search and found that you may have to change the type of BLDC motor that you used. The type you have in this video is mostly used in RC models and I don't think they have planetary gear systems that fit that particular type. You may have to switch to NEMA 17 format type BLDC. I look forward to what you choose as a solution.
Make a video about 6384 motor with gear
You will likely need forced air cooling at that much power to keep the motor happy.
I agree. I will add cooling for the next test.
Also what is your intended application? I think you could get by with much more reduction, which will give you much more torque and less motor heating.
Where you from?
Born in Russia. But I live in Switzerland. You have very interesting youtube channel!
what filament are you using?
I'd like to see other robots rated in rope-potatoes :)
Are you intending to use this actuator for a robot build?
Because if so, it will definitely not work in anything remotely precise or rigid, since the axis itself has several *degrees* of backlash and almost zero rigidity. I have a tiny 3kg payload KUKA robot, and it takes over 100Nm of torque to move not just axes 1 or 2, but even axis 3, without the brakes. And if the motors are on, I can only make the actual frame and transmission flex. Even the highest static overload with brakes not applied consumes under 150W for all 6 axes, and the robot is rated for over 1kW continuous.
You are telling me home made plastic actuators aren't the same as obscenely expensive industrial gear? The wonders will never cease...
@@MsSomeonenew It doesn't even have to be "obscenely expensive". The cheapest, shittiest worm drive not intended for robotics will have less backlash and more output holding torque than this.
Most accurate test of strength, lifting potatoes
Oh, vi iz Anglii!
Из Швейцарии.
encoder details plz
why not using LIPO batteries?
Because I don't have them. But probably the LIPO batteries is the easiest solution.
Или я чего-то не понимаю, или крутящий момент должен измеряться от "пол шестого" до "без пятнадцати девять". Или я всё же чего-то не понимаю?
It seems you have small overshooting, you can tune vel_gain, pos_gain and vel_integrator_gain to reach better results. Good luck! (docs.odriverobotics.com/commands)
Yes! Thank you for the link.
Is ther stl file
Геннадий, а почему вы не используете циклоидальные редукторы? ведь их тоже можно легко самому изготовить с помощью 3д принтера ruclips.net/video/B61fXYZmv38/видео.html
Дело не только в источнике питания. Ваш мотор - бездатчиковый. Реализовать с его помощью векторное управление достаточно тяжело, и я сомневаюсь, что ваш драйвер это умеет. Как следствие - снижение момента на старте. Можно реализовать векторное управление, используя линию Z энкодера, но для этого необходима калибровка при каждом холодном старте системы. Вообще, BLDC моторы с малым количеством полюсов - не лучшее решение для задач, где требуется развивать серьезный момент на низких оборотах.
Я делаю калибровку при каждом старте линии З энкодера. И плюс сделана калибровка относительно полюсов мотора и другими линиями энкодера. Так что это система не хуже мотора с датчиками. Даже лучше, так как мой энкодер лучшее разрешение имеет чем датчики на моторах. И этот драйвер был разработан для такой работы на низких оборотах.
@@Skyentific Я говорил не про драйвер, а про двигатель. Драйвер - это три полу-моста))) В любом случае, не готов с вами аргументированно спорить, не видя всей системы и алгоритмов. Если ваша система позволяет в момент старта знать положение полюсов относительно обмоток с точностью около 10 mil и в процессе движения - 100 mil, то реализация векторного управления не представляет труда.
Почему я делаю вывод, что пока такой алгоритм управления не реализован: на видео видно (забавная тавтология), как привод "доезжает" в целевую точку, в том числе, со сменой направления движения.
@@dimstr8714 Тут другая проблема, это двигатель BLDC с трапецевидным ЭДС самоиндукции и питать его нужно прямоугольной ШИМ. О FOC (векторном) управлении можно забыть. Нужен PMSM двигатель для питания синусом.
@@kokotmkokot4926 Вы в плену инерции мышления, уважаемый. Да, ОПТИМАЛЬНО с точки зрения КПД подкидывать на обмотки трапецеидальные импульсы. И я так и делаю - в установившемся режиме. А вот в переходных режимах и, особенно, при старте под нагрузкой можно кратковременно закрыть глаза на дополнительный нагрев обмоток и рулить BLDC чистым или не очень синусом. Да, нужен или драйвер с запасом по мощности (для компенсации обратной ЭДС), или снаббер/дампер с той же целью. Но выигрыш при старте под нагрузкой существенный.
@@dimstr8714 вы всем незнакомым категорично утверждаете кто в чьем плену? вы с работой снабберной цепи вообще знакомы? судя по написанному - нет.. дискуссия закрыта, тролль
That's the wrong type of motor for such a project. This is very inefficient use of this motor. You'd need a very low KV motor with higher torque. Likely much smaller power will be plenty too and a lot higher gear ratio. Not sure what your intended use is but the way you built it makes no sense to me outside of experimenting with the concept with parts you already had.
вы такой сексуальный в шортиках😃
Еще бы! Жена выбирала! :)
Skyentific а я сижу и не понимаю, что-то английский сильно русским духом пахнет!
LOL was that a calibrated kilo of potato's
12:36 :D
doesn't seems to have holding torque
Hmm, i think something in your calculations is wrong. When you have 0 rpm abd just static torque... You still have 0 watts + some losses. But a high current in the rotor windings. Maybe your current controller or position controller has too small of a kp or ti....
Минус что двигатель перегревается при удержания
Вы русский?
Ты чё русский?
Need metal, heed more hard.
great work an all,but its kinda pointless...just use a proper stepper motor and get the job done...this thing is only for toys
Nope! Stepper motors for toys. From this setup I already get 3N.m torque. The stepper motor with such torque weight from 1.5 to 2kg. And my setup is only 1kg. And with proper power supply it will get even higher torque. In general the brushless motors have much higher power on the unit of mass than the stepper motors.
@@Skyentific dont forget steppers are also brushless motors..just another kind of motor.anyway,in the real world,if you want accuracy,torque and holding torque,excluding servo motors,steppers are the way to go.check out my projects,i also started usind dc motor with encoder,using my arduino code and eventually went the stepper way to jave work correctly.
Ужасный акцент.
Это фича канала. :) К сожалению в русских школах учили именно такой акцент. Но я стараюсь исправиться.
Так Вы русский но в видеороликах объясняет на английском?