TMC 2100 and ramps - any pitfalls to avoid?

So I've got two of these TMC 2100 stepsticks:


Planning to try them out on my corexy, which currently has ramps/mega (I've got a re-arm in the post which is the next upgrade).

I've read this instructable about them [www.instructables.com]

I am aware that these boards are upside down having the solder pads for the heatsinks on the top compared to the original design, so I will be careful to insert them the right way around and remove the microstepping jumpers from the ramps board.

The instructable mentions bridging GND and CFG1 to get spreadcycle, 1/16th interpolated to 1/256 microstepping which is the mode I would like to use.

Anything I've missed? Anyone using these got any tips?

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I've got 4 coming in the post soon as well, so am following this with interest.

So installed them, after removing the jumpers on the ramps, the trim pots do need to face the other way from A4988's.

They seem to work out of the box in 1/16th mode. So I soldered a bridge from GND to CFG 1 as per wiki [reprap.org] to get spreadcycle mode


Its harder to tell if they are as silent as the TMC 2660's on the delta as the corexy is enclosed, there's a lot more belts and pulleys etc.. and so everything resonates that bit more. But they definitely are quieter, now I wish I had not been so cheap and ordered 4 as all I can hear now is my extruder motor and z motor which are still running A4988's. Two more on order now

Set vref initially to 0.800v but got a few missed steps, so beefed it up to 1.0v, next print so far so good.

These one's have nice big heatsinks and they are fairly warm so don't throw your electronics fan away just yet.

Edited 2 time(s). Last edit at 01/26/2017 07:26PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Good stuff DJ, it's always a bit of a gamble buying random stuff off ebay, so I'm glad to hear that it worked.

The main reason for me isn't sound, it's that 1/256 interpolation which _should_ eliminate the very fine 'tree rings' I'm getting on my prints when they are flat on the X axis (perpendicular to the Z tower). I've no idea how you'd test that on a coreXY, maybe you'd need to print a very slight curve, aligned mostly at 45 degrees to the X axis? That might get one of the drive motors to have a microstep every few mm, which is visible when using standard microstepping rates.

I recently changed from 1.8 degree steppers, with 8825 drivers, to 0.9 degree steppers and back to 4988 drivers. Same steps per mm (the 8825's were 1/32 steps, the 4988's are 1/16, but the new motors have more steps), but better quality. Those 8825's are rubbish. I can still see the individual microsteps in this setup though, when holding the part up to the light.

IMHO the solder bridge for CF1 is not necessary, as long as you have a Ramps. They come with a pulldown resistor on MS1.So leaving the jumpers open is the same.
Haven't checked other board schematics yet, but Rumba board should have one too.

A warning for future readers: Make sure you don't have a RAMPS with traces under the jumper pins!
Otherwise the TMC won't run in the best config. and the CF1 mod would even shortcut the 5V regulator of the Mega board.

One more thing to consider: My steppers all ran backwards. So reverse them in firmware or plug in the stepper cable the other way round.

Edited 2 time(s). Last edit at 01/27/2017 03:26AM by o_lampe.

Yes they run in reverse so I powered off and reversed my motor plugs.

Interesting what you say about the CF1 mod, I initially modified one of them and replaced it, so I could compare the noise level from one motor against the other. Not a straightforward comparison on a corexy as I had no 45/135 deg gcode moves ready to test it, but listening to the two motors during moves, one seemed much quieter than the other, so I modded the other one. However I can't rule out cognitive bias in this.

Also worth pointing out the vref is the RMS current so you probably need a higher vref than with A4988's for a given motor. I'd agree with this as my second print above at 1.0v skipped some steps and failed also. They come pre-set to 1.25v.

Interesting about the faulty/alternative ramps boards with the non-functional jumpers, I've had one of these and didn't notice until I wanted to change a stepper to 1/8th stepping and couldn't. So yes check continuity across the jumpers before installing TMC's on ramps, if they are joined you might have to cut some traces.

Edited 1 time(s). Last edit at 01/27/2017 05:10AM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I've been trying those TMC 2100s for a while. I've got them from a vendor on Aliexpress. I've also bought some original ones. I use them with the jumpers removed and I don't short anything else. They are very very silent compared to 4988s however I find them to be more unstable, so I occasionally change back to 4988s. I've added a 80 mm fan now, so maybe they'll perform better.

Just a question: You've mentioned that they were working on 1/16th mode out of the box. How did you understand that? Is there a way to tell which stepping mode the stepper is working on ?

And another question: Are you on 12 volts electronics? If you are, are you hearing a high pitched noise from the steppers when they are activated? Interestingly, my original TMC2100s make that noise but my cloned ones don't, so I use the clones instead of the originals, but that makes me question how identical the clone is.

Edited 1 time(s). Last edit at 01/27/2017 07:37AM by drmaestro.

I knew they were on 1/16 rather than any other true microstepping mode as my carriages were moving the distance commanded. I could not tell if they were interpolating to any high microstepping. As I said above I thought they were quieter after the mod, but it was not a very scientific test. The wiki suggests with CF1 and CF2 open they would be on 1/16 to 1/256 interpolated in stealthchop mode, where I've been led to believe we want them on spreadcycle for better torque, but I was unaware that CF1 was grounded by the pull down resistor.

I am using 12v on that machine, I couldn't really hear a high pitch noise but my audible range is rather poor in the upper end after too many years in noisy nightclubs.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

So set to 1.25v and printing well(0.9deg 48mm, 1.68A nema 17's), make a nasty grumble on fast travel moves, maybe I'm hitting some sort of resonance, but no skipped steps.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Watch out for the Watterot version of the TMC 2100- that has the trim pot underneath, and a hole in the board so you can adjust it.

They helpfully put "TOP" on one side of the board, but I foolishly ignored that to get the trim pot on top... oops.

Depending on the version of TMC2100 (clone/Waterott/other), beware there are slightly different configs. Some clones also have CFG1/2/3 labelled incorrectly, same but different.

Principally two modes of operation. Stealthchop and Spreadcycle. Pro's and cons with both. There are even deeper config options available with both modes, based on very small soldered jumpers on the underside.

They will be prine to hum/whine on a 12v system particularly in spreadcycle mode. Hotend and Bed heater power switching also has potential to impact their performance if running off same supply.

Trimpots should be closest/facing toward the power and USB socket on a RAMPs board, which is the opposite of 4988.

Trinamic drivers are good (better than 4988 out of the box) but not necessarily a silver bullet. Tuned/modded 4988's and 8825's can also be made to run very well. However proper setup and config information is beyond a simple post and so one should do some (quite a bit of) reading and assessment to derive your optimal setup. Warning: You may well damage one or two (or more) during the process so good to buy some spares.

One can get 4988 drivers to run with near comparable quality, if you check and set decay modes correctly. Most of the drivers commonly used in 3D printing, are designed for general stepper motor control (Pololu etc) and not specifically 3D printing which requires steppers with absolute motion control integrity. ie no micro-step jumping for best results.

HTH
J

Edited 1 time(s). Last edit at 01/28/2017 02:50AM by Mutley3D.

---

Flex3Drive.com

Quote
Mutley3D
Most of the drivers commonly used in 3D printing, are designed for general stepper motor control (Pololu etc) and not specifically 3D printing which requires steppers with absolute motion control integrity. ie no micro-step jumping for best results.
J

Does that mean we still don't have the ideal stepper drivers yet?

Quote
drmaestro

Quote
Mutley3D
Most of the drivers commonly used in 3D printing, are designed for general stepper motor control (Pololu etc) and not specifically 3D printing which requires steppers with absolute motion control integrity. ie no micro-step jumping for best results.
J

Does that mean we still don't have the ideal stepper drivers yet?

Ideal would be servo based, but would it be practical? Our architecture' is predominantly stepper based. Some stepper drivers are better out of the box than others as mentioned, but all of them are quite configurable beyond the stock settings/circuitry. Waterott/Trinamic based are a significant step forward. There are a couple others out in the wild using alternative driver chips that also work well. Processor based drivers are another leap.

They key to maxing out the performance of a given driver however is to read the data sheets closely. As usual, the variability between machines means what works for one may not work for others.

---

Flex3Drive.com

Following on from a discussion here [forums.reprap.org] about these drivers it is worth noting the current needs to be set just right, and that the right level - which should correspond to 70-80% of your motor's maximum current handling will result in warmer motors and much warmer drivers than you might be used to running A4988's. Too high and you're hitting thermal cutout issues. So there is a sweet spot.

From what I read on reprap wiki the vref equates to the current(max) - so 1.3v is 1.3A max

This image shows the issue. At 1.0v (I was running my A4988's at about 0.8v) with 0.9deg nema 17, 1.68A motors skipping is occurring and it is clear at which point I've nudged them up to 1.3v.


Motors now run at 45 deg C so feel fairly hot to touch.

Edited 1 time(s). Last edit at 02/08/2017 03:10PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Are you absolutely sure that they are running in spreadcycle mode, rather than stealthchop?

I'm pretty sure but I'll recheck it. I've got no jumpers under them and a bridge from GND to CF1.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

I believe their Vref has to be raised because of the 256 steps interpolation. Every new step sent from the controller starts with a current ramp-up instead of bang-bang. And the ramp-up current has to be sufficient to move the rotor. After the ramp, the current is much higher than necessary causing the stepper to heat up.

My TMC drivers have config pads available on the underside. ( Like Mutley3D mentioned )
IMHO they are all in "open" state. Maybe it's worth to set them up differently?

I've flipped through the TMC2100 datasheet to find more information about the cfg0 and cfg3-6 pins and it seems the best setting for spreadcylce would be:

CF0 = GND ; low chopper timing
CF1, CF2= stepper config as seen in the Wiki-table
CF3 = open ; use external current and sense resistor
CF4 = GND ; low switching hysteresis
CF5 = VCC(5V); medium chopper blank time
CF6= used as enable pin ( no idle current reduction possible without tristate driver )

The datasheet is to large to attach...

Edited 1 time(s). Last edit at 02/09/2017 03:33AM by o_lampe.

Thanks for that O_lampe, my drivers arrived today so will be using those settings as a starting point.

How did those settings work out?

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

There was another data sheet that I saw that had clearer presentation leaning toward an application/deployment note however for my life, I cannot find the darn thing. As soon as I do, I will post it here.

I also found these drivers to work far better at 24v. If you do upgrade from 12v to 24v, make sure you do upgrade at least your hotend cooling fan.

HTH

Edited 1 time(s). Last edit at 02/11/2017 09:47AM by Mutley3D.

---

Flex3Drive.com

My Rumba board is 24V ready, so I'll give it a try soon. Matches the 0.9° steppers better, so it's a win-win for me.
I haven't tried the new found cfg-settings yet. My Prusa is busy to print parts for the D-Bot. Never change a running system, they say...
It's a pity even Watterott, the German reseller isn't able to provide a better documentation. Seems this Chinese-disease is highly epidemic...

Edited 1 time(s). Last edit at 02/11/2017 11:11AM by o_lampe.

I simply removed all jumpers, plugged them in without making changes or soldering wires or underside jumpers. Reversed motor plugs. Off I went at 12v.

They ran better than 4988's or 8825's, but gave an alien high pitched beating sound when running especially when the heatbed and hotend were running on PID. Had them for some while like that but everything was running fine and good prints. Then more recently I noticed an issue on my machine which was eventually (and painfully) found to be the the PSU slowly letting go.

Slapped in a 24v PSU, dropped max currents a tad for PID in firmware and all good. Drivers were now silent. Just check your vrefs, and motors or drivers shouldn't run hot, at least mine dont. One step back, two forward.

I tuned by real world outcome, turned vref low, did a few homing actions, and G1 X100 F10000 then repeated F15000 and same for Y axis turning the vref up until the motion was reliable. then ran a few prints. Had to turn up vref a fraction more on my X axis as it skipped on a print. Been sweet ever since.

---

Flex3Drive.com

Well interesting-ish update to this. The machine in which I had the two TMC2100 drivers installed in, for X and Y axes is now using Re-ARM instead of arduino mega, and running smoothieware. The motors are now very silent (no driver hacks apart from the GND to CFG1 which is not needed on well made ramps boards). And the print quality is stunning compared to when I had arduino/marlin running it.

So massive thumbs up for 32bit board and TMC driver combo.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Are you running both boards on the same voltage? I guess the Re-Arm runs on 24V instead of 12V? Do you know if the control signals are 3.3V or 5V?
I'll try my Rumba board ( 8bit ) with both voltages soon to evaluate the reason for the silence. My best bet is, the current control chopping frequency is 4x higher at 24V than with 12V. Which makes it non audible for us, but the cats in the neighborhood will get a headache

PS: I had the same experience with the Duet 0.6, when I dumped the Arduino/Ramps on my Delta.

Edited 1 time(s). Last edit at 02/20/2017 02:49AM by o_lampe.

Well that's the thing if I had gone to 24v I would expect a change, I am using 0.9 deg motors for x, y, z. But all I did was pull the mega off the ramps and slap the Re-ARM in. Then there was quite a bit of learning how to make smoothie work.

It has 5v tolerant inputs and I'm told it outputs 5v for some functions that need them, everything works.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Installed mine tonight.

Started at 0.5V reference voltage, but it was too low, so went up to 0.7 volts. Seems OK there so far. I didn't put the jumper in to get into spreadcycle mode, it's in the default stealthchop mode.

I'm glad to say that the tree rings that I've had forever, are now gone. The printer is now so much quieter that I can hear the balls clicking in the linear rails when the carriages move.

Coming down from the homing position to the plate (at the start of a print) seems to hit some sort of resonance, not sure what it is, but I'll adjust that speed to get around it.


The Titan didn't make any difference to the tree rings, but it was on the list of updates for this week.

These two prints were done using the exact same Gcode file. The main difference here is the stepper drivers.





The diagonal lines in the last image are also present in the first image, but the tree rings drown them out. I think they are due to using toothed belt on a flat bearing. They are extremely faint.

Edited 2 time(s). Last edit at 02/21/2017 10:44AM by nebbian.

Yeah I hit some kind of resonance on travel moves around (guessing) 80mm/s. Oddly with the rearm the motors run quieter and cooler for the same driver current - can't work that out but not complaining.

My reading of it is that you get 1/16 to 1/256th interpolation spreadcycle with no ramps jumpers... As the CFG 1 is grounded by default, but I haven't put a meter on it and confirmed this. The higher spec tmc's have spi comms for config, like on duetwifi.

For me tmcs made a little difference, rearm made a huge difference. Might get one for my micro delta, its going away from the cheap and simple ethos, on that machine but it prints pretty well with ramps/a4988, as its small and easy to calibrate well, might print amazing with rearm/tmc. I need that 0.2mm nozzle might get a 0.4 and peen it over and re drill it.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Hello,

I received my tmc's yesterday but I'm still pretty confused. I want them to run at spreadcycle mode.

I know I have to remove the 3 jumpers from the ramps but the "removing pins" and "soldering bridges" is what confuses me.
How exactly I do it?
It's really necesary to solder the bridge?

Thank you!

Edited 1 time(s). Last edit at 02/21/2017 03:58PM by trefhi.

I'm told no. Remove the jumpers. Use a multimeter to see if the CFG 1 pin in your driver socket is grounded. If it is then in theory you just plug them in. You can test the jumper pins to make sure you don't have a faulty ramps with traces joining them (this is only a problem if you do the CFG 1 to GND bridge) and then use them. I haven't tried any other modes on these drivers but have tried other microstepping settings on tmc2660 on duetwifi, see duet forum for more info.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions