great job man, im currently doing a large gap devive myself and im struggling with the logic side so i was wondering what sort of power driver you have for the coils are you using a h bridge? also if you by chance have a schematic available for your set up to use as reference?
Yes I am using two H bridges, one for each diagonal of the control coils. Each bridge is implemented by one TDA8932 class D audio amplifier modified for DC input coupling. As for the schematic, most of it is in the video: 0:52 the block diagram 2:50 the sensors and signal conditioning. You need two, one for each axis. 3:33 the power amplifier TDA8932. You need two, one for each axis. Do not forget the modification for DC input coupling. The output of each TDA8932 feeds a diagonal of the control coils. 3:42 the offset indicator. Optional but very useful. One for each axis. It helps positioning the top before letting it float. 3:47 the physical layout. Not covered in the previous schematics are: a reverse polarity protection diode 6A10 and two regulators 7809 feeding a 7805. A single 7805 is enough but it must have a heatsink if operating over 15V. These 3 components (plus a few capacitors) are at the bottom of the physical layout. Nikos
@@nikos44317 for the attenuation of the amplifier input signals from the hall sensors could you let me know the values of the resistors and capacitors and whether or not they are in series or in paralell if you have the values at hand of corse. much appreciated for the reply and explaination.
Nice build and design there! I have seen some yt videos where people are able to float the magnet with only one big electromagnet(wounded on an iron core). However that is extremely hard( i have myself tried but could not mimic) and then finally i came up on these types of designs(with 4 electromagnets(lol i could not understand why you have to use normal magnets at the base besides the 4 electromagnets)). Although nice build there. I will try to build this one rather now(looks muchhh muchh effective). Btw could you pls tell why do you need those extra magnets at the base?
Hello. The “extra” magnets at the base provide the required lifting force to the top. On the other hand, the purpose of the coils is to stabilize the top and not letting it slip aside or topple over. If the base is well levelled, the lifting force of the magnets is vertical and if the sensors are well centered, very little power is required for stability. It is certainly possible to build a device without permanent magnets at the base by using electromagnets, but the required power would be very important and - most probably - active cooling (noisy?) would be required for continuous operation. Nikos
Yes of course. I have successfully used ferrite ring magnets for magnetic levitation. You may watch th-cam.com/video/7oCw1vQuh2Y/w-d-xo.html (operation) and th-cam.com/video/Zah9ORLddAg/w-d-xo.html (building). This is a spin stabilized levitation device (no active control but able to spin indefinitely), but for what concerns the lifting force I do not expect any surprises. Nikos
@@nikos44317 thanksss! Btw what are your views regarding magnetic levitation projects that only use one electromagnet with a hall effect at the bottom. Have you ever given that a try?
Well, I have experimented with the following configuration: First, a horizontal crown of permanent magnets. Second, a central magnet. Third, an electromagnet over the central magnet. Fourth, a Hall sensor on top of the central electromagnet (not at the bottom!). This configuration provides horizontal stability and limited rotational (about any horizontal axis) stability. It is unstable regarding vertical translation - just one degree of freedom that must be stabilized! The controller may be very simple, just a MOSFET or a very simple PWM circuit. My initial results showed some disadvantages: very sensitive adjustment, very short levitation gap (between the sensor and the hovering top) and very low payload. I concluded that this could not satisfy my requirements (several centimeters levitation height and significant payload), so I abandoned the idea. Unfortunately I do not have a video of these attempts. I wish you success! Nikos
Amazing stuff! I've always wanted to make something like this, but I don't quite have the know-how. I also find it hard to find good sources. Do you have any tips for where and what to read? I'm very curious as to how this works, so it would be great if you could answer some (possibly stupid) questions: 1) Seems like this is a fully analogue device; have you tried to make a similar device with a micro controller instead, or would the sampling rate be too low? 2) What keeps the floating magnet from flipping over? The inertia of the disk? 3) How did you tune the PD controller? Thank you, and again, this is really cool!
Answers per question: Do you have any tips for where and what to read? I found the idea of a bridge configuration for the control coils here (see figure 4): mozgochiny.ru/electronics-2/samodelnyiy-levitron/ It is in Russian but google translate does a decent job. The bridge provides optimal positions for the sensors (see th-cam.com/video/unxhKSoGacM/w-d-xo.html at 2:06). 1. Seems like this is a fully analogue device; have you tried to make a similar device with a micro controller instead, or would the sampling rate be too low? Yes, it is fully analog (excepting perhaps the class D power amplifiers). Yes I tried an AVR at 16 MHz using the embedded 10bit ADC. To make it simple I used two AVRs - one per axis. Sampling rate (15625 samples per second) was OK. The problem was the quantization noise that was excessively amplified by the numerical differentiation. D control is a necessity because the controller must correct very fast the top position (before the deflecting force becomes too big) with suitable current in the control coils, but current needs time to build up because of the inductance of these coils! 2. What keeps the floating magnet from flipping over? The inertia of the disk? Before building anything, I did extensive numerical simulations and I found that at the vertical equilibrium point, the top is rotationally stable about a horizontal axis. A small rotation causes a restoring torque. Problems arise when an active horizontal translational stability system is in operation: A (small) rotation about a horizontal axis changes the magnetic field at the hall sensors, the controller generates a (non required) restoring force and an unavoidable parasitic torque. If this torque is in phase with the initial rotation and larger (absolutely) than the normal restoring torque (which is almost always the case), it leads to oscillation than results in flipping over. In other terms, horizontal translation and rotation about a horizontal axis are coupled through the controller! This coupling may be reduced by making the two natural oscillation frequencies very different. It is difficult to change the horizontal translational natural frequency by tuning the controller (control is easily lost), so I changed the rotational natural frequency of the top by increasing its corresponding moment of inertia. This can be done by stacking several magnets to make the top (as you can see in many videos), or by selecting a wide top (as I did in this video - diameter of 60mm). But one must be careful: The optimal position of the sensors is a function of the top diameter! 3. How did you tune the PD controller? Thank you, and again, this is really cool! Numerical simulation is the tool! I was an IT professional and I like from time to time to write in C++. I hope this helps.
@@nikos44317 Thank you for such a detailed answer! Great source as well, I'll read through it thoroughly. Interesting to see you've done numerical simulations. I was planning to do some numerical analysis beforehand as well, as I have some experience with numerical simulation and FEM analysis. May I ask how you did the numerical implementation? Did you do all simulations in C++ based on some FEM model of the dynamics of the electro-magnetic system?
Actually it is simpler than it looks. Magnets and control coils were simulated as simple current loops. The top is not very close to the base and this crude approximation is accurate enough. The current loops are decomposed in very small circle arcs (i.e. 1000 arcs per loop) and arcs are approximated by elementary linear segments. In order to calculate forces and torques, every element of the top is combined with every element of the base magnets (and the control coils) and elementary forces are calculated by Ampère’s force law (see th-cam.com/video/eTMG4BCsgYU/w-d-xo.html at 1:21). These elementary vectors of forces are combined to give a total force and torque acting on the top. To calculate the magnetic field induced by the top to the sensors, the same decomposition of the top into elementary linear segments is used, and the Biot Savart law (see th-cam.com/video/eTMG4BCsgYU/w-d-xo.html at 0:55) is used to calculate the vertical component of the magnetic field (Bz) at the position of each sensor. For the linear movement of the top, force divided by mass gives acceleration, acceleration is integrated to velocity and velocity to position. For rotational movement, torque divided by moment of inertia gives angular acceleration, angular acceleration is integrated to angular speed and angular speed to attitude (rotations are considered to be very small, so no solid body modelling is used). The simulation step can be 1ms and the integration uses a variant of the Runge Kutta method. A few hundred steps show if control is effective or not.
@@nikos44317 Yeah, that is simpler than I expected! Thank you so much for the detailed answer! I'll see if I can set something up myself based on that. Looking forward to see some more of your amazing projects!
Hi Nikos, Thanks for this - very interesting stuff! How practical do you think it would be to scale this up? I am thinking of using high power SCR's and running at say 20A. It might be better to use an all electromagnet design in this case as the magnets would need to be huge. Then it would be just a question of designing the coils and configuring a suitable floating platform. Any ideas?
Yes, it can be scaled up, but, if you go for electromagnets at 20A, you will need thick wire (at least 4mm2). If you wish to wind (say) 1000 turns, the cross section of the winding will be about 70 * 70 mm which is also very large. Windings will have considerable dissipation and active cooling might be required. Lastly, saturation of the iron cores will diminish the core benefits. The model for the simulation will have to be nonlinear and this will be an additional complexity. In the positive side, you will not need separate control coils: you will have just to modulate the main lifting current of groups of electromagnets. Besides SCRs, I would also consider power MOSFETs in PWM operation. I wish you luck and keep us informed!
How much power do you need to carry say that pile of sticky notes for a minute? Have you seen the video using pyrolized graphite and permanent magnets?
I feel sad when you show something very knowledgeable and good but you don't get attention because here people just want comedy and time pass in any topic.. often happens with me..
I see that the last time I run such a simulation was on November 2019. Give me one weekend (or two...) and I believe that I will come back with a positive answer!
Amazing stuff mate. Thanks for sharing.
Glad you enjoyed it!
great job man, im currently doing a large gap devive myself and im struggling with the logic side so i was wondering what sort of power driver you have for the coils are you using a h bridge? also if you by chance have a schematic available for your set up to use as reference?
Yes I am using two H bridges, one for each diagonal of the control coils. Each bridge is implemented by one TDA8932 class D audio amplifier modified for DC input coupling.
As for the schematic, most of it is in the video:
0:52 the block diagram
2:50 the sensors and signal conditioning. You need two, one for each axis.
3:33 the power amplifier TDA8932. You need two, one for each axis. Do not forget the modification for DC input coupling. The output of each TDA8932 feeds a diagonal of the control coils.
3:42 the offset indicator. Optional but very useful. One for each axis. It helps positioning the top before letting it float.
3:47 the physical layout.
Not covered in the previous schematics are: a reverse polarity protection diode 6A10 and two regulators 7809 feeding a 7805. A single 7805 is enough but it must have a heatsink if operating over 15V. These 3 components (plus a few capacitors) are at the bottom of the physical layout.
Nikos
@@nikos44317 for the attenuation of the amplifier input signals from the hall sensors could you let me know the values of the resistors and capacitors and whether or not they are in series or in paralell if you have the values at hand of corse. much appreciated for the reply and explaination.
iv just watched the video again and you have all the values outlined, thank you verry much again
Nice build and design there! I have seen some yt videos where people are able to float the magnet with only one big electromagnet(wounded on an iron core). However that is extremely hard( i have myself tried but could not mimic) and then finally i came up on these types of designs(with 4 electromagnets(lol i could not understand why you have to use normal magnets at the base besides the 4 electromagnets)). Although nice build there. I will try to build this one rather now(looks muchhh muchh effective). Btw could you pls tell why do you need those extra magnets at the base?
Hello. The “extra” magnets at the base provide the required lifting force to the top. On the other hand, the purpose of the coils is to stabilize the top and not letting it slip aside or topple over. If the base is well levelled, the lifting force of the magnets is vertical and if the sensors are well centered, very little power is required for stability. It is certainly possible to build a device without permanent magnets at the base by using electromagnets, but the required power would be very important and - most probably - active cooling (noisy?) would be required for continuous operation.
Nikos
@@nikos44317 oh right! Btw can i use one large ferrite core magnet rather then those neodymium smaller ones?cause i have seen some people do this
Yes of course. I have successfully used ferrite ring magnets for magnetic levitation. You may watch th-cam.com/video/7oCw1vQuh2Y/w-d-xo.html (operation) and th-cam.com/video/Zah9ORLddAg/w-d-xo.html (building). This is a spin stabilized levitation device (no active control but able to spin indefinitely), but for what concerns the lifting force I do not expect any surprises.
Nikos
@@nikos44317 thanksss! Btw what are your views regarding magnetic levitation projects that only use one electromagnet with a hall effect at the bottom. Have you ever given that a try?
Well, I have experimented with the following configuration: First, a horizontal crown of permanent magnets. Second, a central magnet. Third, an electromagnet over the central magnet. Fourth, a Hall sensor on top of the central electromagnet (not at the bottom!). This configuration provides horizontal stability and limited rotational (about any horizontal axis) stability. It is unstable regarding vertical translation - just one degree of freedom that must be stabilized! The controller may be very simple, just a MOSFET or a very simple PWM circuit. My initial results showed some disadvantages: very sensitive adjustment, very short levitation gap (between the sensor and the hovering top) and very low payload. I concluded that this could not satisfy my requirements (several centimeters levitation height and significant payload), so I abandoned the idea. Unfortunately I do not have a video of these attempts. I wish you success!
Nikos
Amazing 👍🏻
Thanks for the visit!
Amazing stuff! I've always wanted to make something like this, but I don't quite have the know-how. I also find it hard to find good sources. Do you have any tips for where and what to read?
I'm very curious as to how this works, so it would be great if you could answer some (possibly stupid) questions:
1) Seems like this is a fully analogue device; have you tried to make a similar device with a micro controller instead, or would the sampling rate be too low?
2) What keeps the floating magnet from flipping over? The inertia of the disk?
3) How did you tune the PD controller?
Thank you, and again, this is really cool!
Answers per question:
Do you have any tips for where and what to read?
I found the idea of a bridge configuration for the control coils here (see figure 4): mozgochiny.ru/electronics-2/samodelnyiy-levitron/ It is in Russian but google translate does a decent job. The bridge provides optimal positions for the sensors (see th-cam.com/video/unxhKSoGacM/w-d-xo.html at 2:06).
1. Seems like this is a fully analogue device; have you tried to make a similar device with a micro controller instead, or would the sampling rate be too low?
Yes, it is fully analog (excepting perhaps the class D power amplifiers). Yes I tried an AVR at 16 MHz using the embedded 10bit ADC. To make it simple I used two AVRs - one per axis. Sampling rate (15625 samples per second) was OK. The problem was the quantization noise that was excessively amplified by the numerical differentiation. D control is a necessity because the controller must correct very fast the top position (before the deflecting force becomes too big) with suitable current in the control coils, but current needs time to build up because of the inductance of these coils!
2. What keeps the floating magnet from flipping over? The inertia of the disk?
Before building anything, I did extensive numerical simulations and I found that at the vertical equilibrium point, the top is rotationally stable about a horizontal axis. A small rotation causes a restoring torque. Problems arise when an active horizontal translational stability system is in operation: A (small) rotation about a horizontal axis changes the magnetic field at the hall sensors, the controller generates a (non required) restoring force and an unavoidable parasitic torque. If this torque is in phase with the initial rotation and larger (absolutely) than the normal restoring torque (which is almost always the case), it leads to oscillation than results in flipping over. In other terms, horizontal translation and rotation about a horizontal axis are coupled through the controller! This coupling may be reduced by making the two natural oscillation frequencies very different. It is difficult to change the horizontal translational natural frequency by tuning the controller (control is easily lost), so I changed the rotational natural frequency of the top by increasing its corresponding moment of inertia. This can be done by stacking several magnets to make the top (as you can see in many videos), or by selecting a wide top (as I did in this video - diameter of 60mm). But one must be careful: The optimal position of the sensors is a function of the top diameter!
3. How did you tune the PD controller? Thank you, and again, this is really cool!
Numerical simulation is the tool! I was an IT professional and I like from time to time to write in C++.
I hope this helps.
@@nikos44317 Thank you for such a detailed answer! Great source as well, I'll read through it thoroughly. Interesting to see you've done numerical simulations. I was planning to do some numerical analysis beforehand as well, as I have some experience with numerical simulation and FEM analysis. May I ask how you did the numerical implementation? Did you do all simulations in C++ based on some FEM model of the dynamics of the electro-magnetic system?
Actually it is simpler than it looks. Magnets and control coils were simulated as simple current loops. The top is not very close to the base and this crude approximation is accurate enough. The current loops are decomposed in very small circle arcs (i.e. 1000 arcs per loop) and arcs are approximated by elementary linear segments. In order to calculate forces and torques, every element of the top is combined with every element of the base magnets (and the control coils) and elementary forces are calculated by Ampère’s force law (see th-cam.com/video/eTMG4BCsgYU/w-d-xo.html at 1:21). These elementary vectors of forces are combined to give a total force and torque acting on the top. To calculate the magnetic field induced by the top to the sensors, the same decomposition of the top into elementary linear segments is used, and the Biot Savart law (see th-cam.com/video/eTMG4BCsgYU/w-d-xo.html at 0:55) is used to calculate the vertical component of the magnetic field (Bz) at the position of each sensor. For the linear movement of the top, force divided by mass gives acceleration, acceleration is integrated to velocity and velocity to position. For rotational movement, torque divided by moment of inertia gives angular acceleration, angular acceleration is integrated to angular speed and angular speed to attitude (rotations are considered to be very small, so no solid body modelling is used). The simulation step can be 1ms and the integration uses a variant of the Runge Kutta method. A few hundred steps show if control is effective or not.
@@nikos44317 Yeah, that is simpler than I expected! Thank you so much for the detailed answer! I'll see if I can set something up myself based on that. Looking forward to see some more of your amazing projects!
Helll. How may get in touch with you?
You can send an e-mail to nikosaravantinos@icloud.com
Hi Nikos, Thanks for this - very interesting stuff! How practical do you think it would be to scale this up? I am thinking of using high power SCR's and running at say 20A. It might be better to use an all electromagnet design in this case as the magnets would need to be huge. Then it would be just a question of designing the coils and configuring a suitable floating platform. Any ideas?
Yes, it can be scaled up, but, if you go for electromagnets at 20A, you will need thick wire (at least 4mm2). If you wish to wind (say) 1000 turns, the cross section of the winding will be about 70 * 70 mm which is also very large. Windings will have considerable dissipation and active cooling might be required. Lastly, saturation of the iron cores will diminish the core benefits. The model for the simulation will have to be nonlinear and this will be an additional complexity. In the positive side, you will not need separate control coils: you will have just to modulate the main lifting current of groups of electromagnets. Besides SCRs, I would also consider power MOSFETs in PWM operation. I wish you luck and keep us informed!
How much power do you need to carry say that pile of sticky notes for a minute? Have you seen the video using pyrolized graphite and permanent magnets?
I feel sad when you show something very knowledgeable and good but you don't get attention because here people just want comedy and time pass in any topic.. often happens with me..
You are correct, but this a fact of life... I watched a few of your videos and I liked them. I will revisit.
Hey Nikos, Awesome work! Do you have the code available for the simulation that you used to confirm the controller before implementation?
I see that the last time I run such a simulation was on November 2019. Give me one weekend (or two...) and I believe that I will come back with a positive answer!
@@nikos44317 Awesome! :D
I have found the code and I can send it with an example run if you provide an e-mail address!
@@nikos44317 fryseboks123@gmail.com
your the best! Ill link you if I manage to do something awesome with it :)
Mail sent!