Clem's Stepper Motor Puzzle!

Of course It is always slightly different. When I was testing it for a first time (outside view of camera) it was about 960 real presses and about 3300 total rising edges.

the practical testing seems to confirm your suspicion!
going backwards would mean the pulse just happened to be exactly lining up with a interpolation pulse.

Going backwards was interesting.

I am a little surprised there weren't more guesses given the nice prize.

As I indicated I wasn't going to weigh in, but my preliminary analysis would have guessed you would probably get a different number of button pushes each time you tried to do a full rotation test.

Hello Element14 community.

Today I completed my practical part of this puzzle, and I would like to post my solution. In recent two weeks I did theoretical analysis and two practical experiments. The first of them failed. Here I will describe my journey with TMC2130 and stepper motors. It is my first experience with stepper motors. I already used basic DC motors, but stepper motor I used for a first time. I split my solution to several parts as described in following list:

• Theoretical analysis
• Building my own module with TMC2130
• Buying and experimenting with Trinamic module

Theoretical analysis

At first, I of course start to find answer in datasheet. But at the beginning I need to find datasheet. Finding datasheet of TMC2130 chip was easy but I had troubles with datasheet of the module which has some important details about connection of the chip.

In schematics it is mentioned as Pololu module but unluckily I did not find any Pololu module containing TMC2130. I opened large view of the photo, copied its URL address, downloaded it and in local viewer I over-zoomed it. Then rotated 180 deg for getting text in right direction instead of upside down.

As you can see on the module is something which look like URL address. I originally did typo and missed double t in URL. Then tried to open it and website without double t redirected me to some dangerous site and adblock prevented me to open it:

So, I gave it but later I found the correct page. Instead, I went for the chip datasheet in meantime and noticed that Trinamic site in Related products contains eval board referred as SilentStepStick which have similar form factor like module on Clem’s photo.

Trinamic SilentStepStick is available in multiple variants with several different chips. Variant with TMC2130 have no datasheet but there is schematics which replaces datasheet.

At this moment I had schematics of the module, schematics of the circuit and TMC2130 datasheet which should be everything needed for theoretical analysis.

Floating pins analysis

The first thing which interested me was letting config pin floating. I remember rule of thumb that no pin of any chip should be left floating, and this violates this rule. So, at first, I was searching what floating config means and found that in case of TMC2130 it is valid situation. Most relevant information about this I found in section number 24:

Then I searched for meaning of all 7 configuration pins. CFG4 and CFG5 are on the module directly connected and there is no (easy) way to change them. I basically skipped them. CFG6 is enable/disable pin:

Remaining important configuration options are CFG0, CFG1, CFG2, CFG3. Note that there is numbering shift. MS1, 2, and 3 are numbered starting at 1 while CFG names are numbered starting at 0. Pay attention that MS2 != CFG2 and instead MS2 == CFG1. Later I found that MSx naming convention comes from different stepper motor driver chip and designer of TMC2130 made it compatible.

CFG0 is not very attractive and important for this puzzle. It configures timing of driver operations:

Most important for this puzzle are pins CFG1 and CFG2 which configures mode.

In this table is hidden answer to the puzzle question. With both pins let in open state each step consists of 16 micro steps which are internally interpolated to 256 micro steps by the chip. Mode of operation is StealthChop which is the more advanced mode of operation of this chip and there are also mentioned some other register configurations which generally do not impact answer to the puzzle question.

Because I am stepper motor newbie, I had to learn what the microstep means. I found answer in several other articles on internet including some explanation of internal details. Basically, I understood it is technology which splits one step of stepper motor to smaller steps by properly applying right current to the motor thus letting motor in position between two full steps. From practical point of view, it means that with 16 micro step configurations for making one full step I need to toggle STEP pin 16 times. This means that for 200 steps (full rotation of my stepper motor) I need to toggle it 200 * 16 = 3200 times.

The last configuration option CFG3 is important but do not change previously deduced answer. It configures way the TMC2130 use for measuring current flowing through motor and allowing limiting it. Changing these settings requires hardware modifications, so I let this pin open for making it working with setup which is used on module (current limit can be adjusted by potentiometer which is connected to AIN).

Button bouncing

The previous answer saying that we need to press button 3200 times for doing one full rotation is only theoretical and in practise actual number of button presses depends on some other factors. The biggest issue which I realized in theoretical phase was button bouncing. When you press or release button it may bounce for a short period of time after the button “status” changed. This duration depends on button. Usually, it is up to 10 ms long. In this time button status oscillate and can generate several pulses. Even before I started playing with TMC2130 I connected logic analyser to my button and found that every press generates 1 to 8 rising and falling edges. Of course, this comes from experiment, and you should not suppose that you never get 9 edges. Then I wrote program using STM32G031 MCU with I2C OLED display which show number of detected edges and also debounced number of presses. After about hundreds of presses, I found that number of detected edges is about 3 to 4 higher than the (debounced) actual one.

Following outcome means that every press move motor about three to four times in average and thus required number of buttons presses reduces significantly from 3200 to somewhere between 800 and 1066.

Practical experiments later shown that there are more issues with button bounces. When the glitches are too fast, the motor may skip the step or even worse it may do step in reverse direction. This can move number of required button presses in both positive and negative directions. Datasheet contains detailed section about timing requirements (in standalone configuration which we use dedge=0):

Also note that there is filter which removes very fast glitches.

After theoretical analysis I went to practical one.

Designing my own TMC2130 module

For saving money I decided to request free same of TMC2130 and building my own module for evaluating it. Trinamic (now ADI) confirmed my sample order, but DHL needed some papers and later I had to pay VAT for it, but it was still cheaper than buying it on my own expense and paying full shipping (and it was very fast of couser). I ordered TQFP variant of the chip for easier soldering. QFN and TQFP are exactly the same parts in terms of internal logic but because of larger pins TQFP variant can handle 1.4A motor coil current instead of 1.2A limit of QFN part. After few days I received it:

There was no time to make proper PCB so I used generic TQFP to pinhead adapter from AliExpress which I bought several years ago and some pieces remained in my cabinet:

At first, I soldered chip and some SMD parts to it. I used SMD parts for passives which are connected to the place between two sibling pins. As a base for soldering I used schematics of the SilentStepStick module which I mentioned at beginning.

Then I soldered first THT passive and some pinheads. Here are pinheads for connecting motor soldered:

In schematics there were CSA shunt resistor. I had only chip SMD resistor. I soldered it rotated by 90deg and used wire for connecting second (flying) terminal to non-sibling pin.

I did several mistakes when placing parts, so I had to rework it. The biggest issue was that when reworking I destroyed one pin and had to connect capacitor directly the pin of chip. But luckily it was the first pin on the side, so it was easy to soler. At the end my module look as follows:

And some wires and caps are also at bottom side:

With soldering and reworking I spend about 3 hours, then I connected it to the circuit and tested it. On the photo below you can see (from left to right) STM32G031 disco with display counting button presses, logic analyser, breadboard with buttons, pull-up, and power LED. The rightest board on the breadboard is my own DC/DC module converting 12V input to 6V for motors. Above breadboard is motor and below is the most important part: my TMC2130 module which I described above.

And unluckily it did not work.

So, I spend additional 10 hours with debugging and reworking the module again. But unluckily it did not work. My outcomes from experiments and debugging were following:

• Connection did not burn, all parts survived, so it was not so bad.
• TMC was alive. Using logic analyser I have seen that generating diagnostic interrupt work.
• TMC received all my step rising edges and after every 64^th edge it drove interrupt signal as expected.
• Motor did not move
• Motor was not held by TMC2130
• On motor pins there were only short glitches when step was regularly driven otherwise motor pins were held high

Because TNC2130 respond to my step edges by at least diagnostic interrupt I think it was alive. My theory is that experiment failed because of over-current detection. For sensing currents there is CSA and shunts with very low value, but my DIY approach most probably added resistance on wires which is most probably significantly higher that it should be and then when TMC2130 trigger motor it immediately detects overcurrent due to large voltage drop on higher resistance path and stop motor due to overcurrent.

In previous theoretical analysis you have seen that there are several options for sensing current. I tried rework board for using different approach (reworks required for example bypass shunt, remove voltage divider and changing value of remaining resistor in original voltage divider) but all my attempts were without success.

Experimenting with bought module

Previous experiment was mostly waste of time. It did not work. I saved some money but lost lot of time. But I do not give up and decided to go more common (and slightly more expensive) way. Because I was interested in stepper motors and want to learn it, I bought TMC2130 SilentStepperStick module and go in standard (expected) way. After few days I received it (with many other components from wishlist which are unrelated to this experiment):

As you can see my module also have watterott domain written on it so I think it is exactly the same module as Clem have. Pin headers were bundled with module but I had to solder them manually. Note that it is not clear which side of the module is top at the first look. Usually, chips are on top side of breakout boards but in this case chip is at bottom side.

The first mistake was that I of course forget to solder bridge for selecting standalone mode (instead of SPI mode). So, I had to select very thin soldering tip and solder it later with paying attention to nearby pinhead. But I succeeded this time.

Then I connected module to the circuit. It is significantly simplified in comparison with previous circuit. I also bought and changed second stepper motor for testing that my previous failure was not cause by motor instead of my module :D

For new motor I used higher voltage about 7.2 volts, and it work.

On the picture you can see that I had connected logic analyser to the circuit. I have connected “signals” of one motor coil and the button.

When the motor is not moving motor signals looks as follows. Note that this is view form logic analyser but in real signals are analog. I do not have oscilloscope so this the most advanced view which I can provide.

All pulses are same length. Motor is held in fixed position.

After applying step signal (for tests I used the attached microcontroller instead of buttons) signal changed. On following screenshot from logic analyser, you can see one step pulse (the last line).

As you can see the first signal consist of short pulses of mostly the same length, but the second signal has pulses with changing length (in real it is sine I think, but logic analyser can’t show it). We can see that at beginning there were long pulses which were later shortening and extending again. In next step cycle signals switched roles and the same pattern was visible on first signal.

I think these signals are cause by interpolating micro steps.

So now I have working circuit, so I started making experiments.

Experiment with grounded CFG1 and CFG2

At first, I tried experiment with grounded CFG1 and CFG2. I use this mode:

It is the simplest mode making one full step per step cycle. I have motor with 200 steps per rotation, so I should click the button 200 times. But according to theory research it should be less because of button bounces. Let’s see:

Edit media

Dimensions x Small Medium Large Custom

Subject (required) Brief Description Tags (separated by comma) Video visibility in search results Visible Hidden

Parent content

Comment on Clem's Stepper Motor Puzzle!

Poster

Upload Preview

Ok

Cancel

As you can see motor several steps skipped, and sometimes it turned back. This behaviour I mentioned in theoretical section. At the end counter ended on 220 real presses and 393 rising edges including bounces.

Experiment with open CFG1 and CFG2

Now let’s try configuration asked by this puzzle. It is this configuration:

Let’s see on video:

Edit media

Dimensions x Small Medium Large Custom

Subject (required) Brief Description Tags (separated by comma) Video visibility in search results Visible Hidden

Parent content

Comment on Clem's Stepper Motor Puzzle!

Poster

Upload Preview

Ok

Cancel

As you can see steps were significantly smaller and some of them were efficiently invisible. In theoretical part I estimated button presses count required for full rotation in range 800 to 1066 when bounces taken in account. At the end counter stopped at 1016 real presses and 3562 detected rising edges (which includes button bounces). Expected value of rising edges was 3200 but as we have seen some steps were not very clear, some were skipped. In this mode motor very rarely moved back. These issues most probably caused deviation of missing 362 edges. But generally, values are very near expectation.

The last experiment which I will show is with STEP signal driven by MCU. This can prove theoretical values 200 for grounded CFG1 and CFG2 and 3200 for open CFG1 and CFG2. In these experiments I reconfigured MCU pins from input to open-drain output and drove signal low. When driving TMC2130 by MCU there are no glitches so one turn should exactly match 200 and 3200 steps.

Here is video for CFG1 and CFG2 grounded (SpreadCycle mode with one full step per step cycle):

Edit media

Dimensions x Small Medium Large Custom

Subject (required) Brief Description Tags (separated by comma) Video visibility in search results Visible Hidden

Parent content

Comment on Clem's Stepper Motor Puzzle!

Poster

Upload Preview

Ok

Cancel

Here is video for CFG1 and CFG2 open (StealthChop mode with 16 micro steps and interpolation):

Edit media

Dimensions x Small Medium Large Custom

Subject (required) Brief Description Tags (separated by comma) Video visibility in search results Visible Hidden

Parent content

Comment on Clem's Stepper Motor Puzzle!

Poster

Upload Preview

Ok

Cancel

As you can see StealthChop mode work very well under normal conditions and movement is very smooth.

Summary

This is all from my datasheet analysis and experiment. The answer is that for full rotation with all CFG0, CFG1, CFG2 and CF3 open It is needed to generate 3200 rising edges but when used with mechanical button there is no exact answer because of random spurious events and practically using my button it is about 1000 presses.

For me it was nice two weeks spend with TMC2130 datasheet reading, planning, soldering, measuring, and learning stepper motors in generals. I learnt that stepper motor drivers are hot while I was burning my finger when touching module. I learnt many other useful information about stepper motors and their drivers. I consider last two weeks as well spend.

Thank you for reading my answer (especial thank you for your patience if you watched full video with over 1000 button presses :D) and have a nice day.

yes, I also like to do things practically, but a look in the datasheet is always a good idea especially when the practical application does not behave just as expected.