A trampoline project

[fotoopa]

I have started a new project with trampolines. Four modules are foreseen. Each trampoline is set up in a rotating position. The whole thing is also set up on a turntable. The hinge point of the trampoline lies exactly in the trampoline surface. Because of this, there is no shifting of the center point when setting a new angle.
Each drive is done with an XM motor and includes a quadrature encoder. A home point is also provided. For the encoder disks I use neodym magnets D2x3mm and hall detector chips. The encoder disc contains 8 magnets which results in 32 pulses per revolution. The basic rotary table contains a 58-tooth gear and worm reduction. A full revolution has 32x58= 1856 pulses. The upper trampoline table is rotated with a T80 gear segment and is also driven with an XM motor and worm reduction. This gives 32x80=2560 pulses per tour.
A few pictures:


HD on Flickr: https://www.flickr.com/photos/fotoopa_hs/51595797719


HD on Flickr: https://www.flickr.com/photos/fotoopa_hs/51595797549


HD on Flickr: https://www.flickr.com/photos/fotoopa_hs/51595797034

[url=https://flic.kr/p/2mBiFfJ]
HD on Flickr: https://www.flickr.com/photos/fotoopa_hs/51595354218/in


HD on Flickr: https://www.flickr.com/photos/fotoopa_hs/51594317477
Instead of the trampoline, a mirror can also be inserted. This is then very suitable for experiments with light beams.

An extra stock of XM motors was needed for this.

HD on Flickr: https://www.flickr.com/photos/fotoopa_hs/51595132776

I am now in the process of preparing all the modules. This requires a lot of work. The 8 XM motors with quadrature decoders and home pulse all need to be connected. I am using my DE0-nano-soc module for this purpose. This module is linked via an I2C connection to the TXT controller. The Robopro software program is used to control everything. The motors are controlled with the TB6612 modules on the SPI bus. In total I can connect 12 motors together with all their fast encoder signals.

Within a few weeks I hope to be able to provide images of the results with you. The idea is to jump from one trampoline to another. At the end they have to come in a green funnel to bring them up again. The order how the balls jump is set with the motors. Whether it goes for double or triple salto I have to wait and see. The next step will then be tests with lights and the mirrors.

Frans.

[rubem]

Hi Frans,

Looks nice as always, your detailed photos with labels are always a pleasure to look at. I look forward to the finished model (or models)!

Best regards

Rubem

[fotoopa]

Thank you Rubem!

I am now performing tests to determine the proper drop height of the balls and the position of the trampolines. The rebound of the balls is never really even, which makes it difficult to move from one trampoline to several others. I have now made an dual additional adjustable guide that can be set in X and Y direction to drop the balls. This allows me to more accurately determine the point of impact on the first trampoline.

With the current settings, I can already jump to the green output funnel via 2 trampolines. The biggest problem is that the balls never really rebound at the same angle. The different height settings have a strong influence. I also modified the input green funnel with a more precise internal diameter so that the balls have less slack during the fall. This makes the fall direction more stable. My end goal is to be able to reach an exit goal via 3 trampolines. A total of 9 output green funnels are set up around the trampolines. From there, I have to get a via Flex profiles and chain back to the top. There should be a ball detection added to that so I can control the flow when the ball is allowed to fall down. Some additional parts are made via 3D printing.

Frans.

[fotoopa]

The first trampoline module is assembled and tested. This module contains 2 XM motors each with a quadrature encoder that has 32 pulses per worm revolution. The base rotation consists of a gear of 58 Teeth. This makes 58x32 = 1856 pulses per full revolution available. This corresponds to a rotation angle of 360/1856= 0.194 degrees per pulse.
The trampoline has a gear segment of 80T. This makes 80x32= 2560 pulses per tour available. The angle setting is 360/2560= 0.14 degrees per pulse.

Quadrature encoders never give errors. These counters run under hardware of the FPGA. Even if the software fails, the counters keep their correct state. Even if you would manually rotate them yourself. At powerup, a calibration must first be done via a home point. At this point the counter is set to zero. This detection is done with a small neodym magnet and a hall detector chip. As long as the home point is active the counter is kept at zero.

The picture shows the adjustment to set the balls exactly in the middle of the first trampoline. There is an X-Y table on top that can be moved manually. With a weight on the cord, you can see exactly where the ball will fall. The frame for this is made of aluminum to be very stable. Any vibration when the balls fall must be avoided.

In total there are already 4 complete trampoline modules ready and tested. Soon these will also be set up. The first central trampoline can then send the balls to one of the 3 other trampolines. Each other trampoline can in turn send the balls to 3 different funnel modules. With proper programming, 9 outputs can thus be used. Whether it will be sufficiently stable, I still need to test further.

In the background you can see that I have made a second SPI motor module. So now I have a total of 12 motors available with 48 input signals. All motors can be controlled simultaneously and each hardware counter can process up to 15 Khz pulses. The LCD display of the FPGA box can show all signals on the screen in real time at any time. Also servo values and analog values are always readable through different pages. This works independently of a connected TXT controller.

Translated with www.DeepL.com/Translator (free version)


HD picture on Flickr: https://www.flickr.com/photos/fotoopa_hs/51645364082

Frans.

[fotoopa]

First test program to control and adjust already 3 trampolines. The settings to go via 2 trampolines to an output funnel with the balls is not easy. With the right adjustments I managed to get most of the shots correct. Now I can make the position optimization of the funnels. Especially driving sharp corners in the right direction is not easy. Once the values are determined I can use them in a final program. Setting the turn table and trampoline angles is very precise. The calibration is also perfect. Because of the quadrature decoders there are no cumulative position errors.


HD picture in Flickr: https://www.flickr.com/photos/fotoopa_hs/51648817312

With Robopro, creating the programs is very easy and very fast. The DE0-nano-soc FPGA module controlled by the TXT Controller via I2C is very reliable and fast. My Fischertechnik TXT Controller is used here only as a graphical user interface.

Frans

[fotoopa]

Complete setup of 4 trampoline modules and 9 Funnel capture units. The input of the balls is still done manually at the top left. Via an XS motor, a ball is passed by the program. At the exit where the balls come down there is a hall detector placed. This signal can be used to finish further action of the program. The signal can also be useful to take high-speed pictures during the falling of the ball. After all, you have a correct starting point.

Each trampoline module can rotate in 2 directions via 2 XM motors. This allows you to change the direction how the ball moves during the fall. The first task is to always move over 2 trampoline modules to one of the 3 funnel outputs. This goes like this:

T1 to T4 to funnel 1 to 3
T1 to T3 to funnel 4 to 6
T1 to T2 to funnel 7 to 9

The first tests give a pretty good result. The hit rate is quite high. Most balls are going to fall exactly where they were planned. Setting up the trampolines is very fast. The trampoline could be changed to a new position even between fall movements. This is part of further testing.

There are still some details to come. So there will be a LED indicator per funnel so that in addition to the Neo LED display, you also get an extra indication of where the ball should go.


HD picture on Flickr: https://www.flickr.com/photos/fotoopa_hs/51659115235

Now I first write the full TXT program. With the already finished test program I can determine all position values. Consequently, there is still some work to be done to work out a number of sequences. My old IR remote control is used to select a funnel output. On that old DVD IR control the numbers are on. In total I have 43 push buttons available. The Neo led display then becomes the output reading. The control goes over the TXT I2C to the FPGA module where the IR decoder is also on. All IR values can always be read out on the LCD of the FPGA as well as on the PC via the TXT and Robopro.

Translated with www.DeepL.com/Translator (free version)

Frans.

[rubem]

Hi Frans,

Looks great! it's always a pleasure to follow your posts and see your progress. Will you post a video at some point?

<OT>I'd love to see what others model makers are doing too. I wish there were more posts at this subforum, after all making models is what fischertechnik is all about!</OT>

Best regards to all,

Rubem

[fotoopa]

Thanks Rubem!

Looks great! it's always a pleasure to follow your posts and see your progress. Will you post a video at some point?

Yes a video I would love to make. Currently the test program is ready to set all trampolines. The next task is to measure all the positions of the trampolines for the 9 different output funnels. That is a time consuming task. I have to work with average settings. Balls don't always reflect exactly the same way. A 100% success rate will not work.
I have now also added a FT led per funnel. These leds are controlled from my SPI interface of the servo module. There I had 24 servo outputs but I have changed the division to 12 servos and 12 digital outputs. The signal levels are the same on the connector. Those are controlled from a 74HCT595 and their high level is 5V. With this 5V I have more than enough to light up an FT led even brightly.

Also my IR remote box is now available for the TXT. I send a signal to that every time a new IR code is received. The TXT via the Robopro program can then perform a task according to the code. The codes are readable on the screen. This makes it easy to create a conversion table for the 43 existing keys. This already works very well on the test program. Through these commands I can also control the neo led matrix display of 8x16 leds in color. The selection of a new output funnel in manual mode is also displayed there. During the input another color is also taken. By the poweron of the TXT it is also indicated whether there is already a calibration done. If not, this is also indicated on the display and you must perform it first.
As long as the FPGA remains powered up, the calibration is valid even if you stop the TXT. This is because all counters are hardware processed in the FPGA.

I still need to program the 9th motor as well. This XS motor takes care of the input of the balls.
In total there are needed:
9 motors. All motors with quadrature encoder input and home point.
26 digital inputs.
9 led outputs --> 5.7 mA @5V FT led wihte.
1 neo led matrix 8x16 control 128 RGB leds.

Update:
Codes old DVD IR remote crontrol.

HD on Flickr: https://www.flickr.com/photos/fotoopa_hs/51663849947

Frans

[fotoopa]

I have some delay to make a video of the trampolines. The reason is that I have to add 2 extra motors at the top of the input module. The balls need to fall slightly off the center of the first trampoline to take a stable direction. I need to be able to adjust the X and Y direction of the falling balls. The motors will be 2 XS motors with quadrature encoders. This will bring the total number of motors to 11 units. 10 of them are equipped with quadrature encoders, one with just an end switch.
I also made a new quadrature encoder that replaces a regular FT-32064 block. This will make the setup very compact. The setup can be used for all motors: XM motor, XS motor and mini motor.


HD picture on Flickr: https://www.flickr.com/photos/fotoopa_hs/51690697133

The encoder disc contains 6 small magnets D2x3mm. Via quadrature this gives 24 pulses per revolution. The Halls chips can be used either on 5V or 9V. Quadrature signal pulses may be up to 15 Khz for each motor. The DE0-nano-soc module can handle those easily.

Frans.

[juh]

Hi Frans,
as always it's impressive and interesting to read about the progress of your projects! I'm interested in your neat encoder block, as I planned to do sth. similar sometime in the future. Why are there six wires, shouldn't four suffice, as GND and V+ can be shared between the two sensors? And have you considered a design where the wheel is enclosed within the building block to make it more compact?

Best
Jan

[fotoopa]

Why are there six wires, shouldn't four suffice, as GND and V+ can be shared between the two sensors? And have you considered a design where the wheel is enclosed within the building block to make it more compact?

Yes 4 wires are sufficient. There is a practical reason why I use 6 wires anyway. All my input connectors have a 3 pin connection, GND, VCC and Signal. The different input connectors should not really be connected sequentially. In the past I also used only 4 wires. With 2x3 wires, connecting to the box goes better. Also technically you have the advantage that the signal per hall as a twisted pair is more resistant to unwanted interference pulses at longer connections.

Another important point is the soldering of the hall connections. By using 3 wires per hall pin, you only need to solder one wire with solder holes on 2.54mm grid. Soldering an extra wire to connect the GND and VCC at the same place is too difficult for me. It is very difficult for me to solder something small. After 40 years of diabetes both my eyes are too bad. I can't make a super small pcb anymore. I still solder a small smd 0.1 uF capacitor between each power supply per hall.

Making the encoder even smaller is not a priority for me. The block with the halls in it is now 15mm wide. I am obliged to use hall types with wires. With a magnet of D2x3mm the halls must be at a distance of 8.5/4*3= 6.375mm. The magnets are then at 8.5mm to obtain a perfect duty cycle of 50% with a 90 degree shift of the pulses. The block with the halls has a center height of 12.5mm. The maximum disk diameter for the halls is therefore limited to 25mm (practically 24.5mm). Actually there can be placed 7 magnets which would give 28 pulses. I am still testing this because the distances would be a bit smaller. The deviation of the pulse ratio is normally given at 45/55. After all, the encoder disc sits on the end of a standard 30mm shaft: FT-35063. I prefer to use the worm FT-35072. It is more stable and has less backlash.

With smd pcb and smd hall chips it could be smaller. The distances of the magnets and halls would then have to be determined again. Whether it would also work with smaller magnets I do not know. In neodymium D1mm they don't have long versions, only D1x1mm.

Frans.

[juh]

Thank you Frans, that was very insightful.

[fotoopa]

I'm going to close this topic now. Robots should continue to help me for the new projects from now on. Intention is to make FT and Lego work together. At least the software is similar. Linking both groups will be fine with the help of extra 3D printed blocks. If there are results I will make a new topic.


HD on Flickr: https://www.flickr.com/photos/fotoopa_hs/51709594212


HD on Flickr:
https://www.flickr.com/photos/fotoopa_hs/51710378381
https://www.flickr.com/photos/fotoopa_hs/51709594257
https://www.flickr.com/photos/fotoopa_hs/51711056334

Frans.