Improving hobby-grade Servos

I happened across this last evening: **LINK**

https://hackaday.com/2021/03/19/robot-arm-achieves-amazing-accuracy-with-just-servos/

and, having very little experience with servos … I was impressed !!

.

Grateful for any observations from the cognoscenti

.

MichaelG.

Skipping forward … we get this:

.

.

MichaelG.

Edited By Michael Gilligan on 27/08/2023 08:27:59

As with the previous thread on the pico the big breakthrough will come with a DRO system independent of the motion system, which means self-compensation for mechanical error (just like with us)

My hobby router with nema17s will easily do 0.1mm again and again and again... until it gets wear and tear

The stage after self compensation is giving it the ability to "see" the object with a 3D scanner

.

Sorry, Ady … I don’t see the similarity

Could you please elucidate

MichaelG.

It's 4 or 5 axis but has the same issues

The big achievement is with the software doing multi-axis control, I think mach 4 does more than 3 axes as standard but there doesn't seem to be a lot of choice around for a hobbyist atm

Edited By Ady1 on 27/08/2023 12:50:12

I must be very dim, today:

My pico thread was about the fact that a dirt-cheap and very useful-looking board can be programmed using the Arduino IDE

This one is about a clever lad who upgraded a servo

MichaelG.

The big achievement is with the software doing multi-axis control, I think mach 4 does more than 3 axes as standard but there doesn't seem to be a lot of choice around for a hobbyist atm

(crossed post)

I would be far more interested in his software approach because it's yielding pretty impressive results

Edited By Ady1 on 27/08/2023 12:54:13

Very interesting Michael especially the calibration process - I must download the Arduino code and see what it does about the intertia of the motor.

I have been using a brushless geared motor like this

to drive the leadscrew on my little Taig lathe. It's 12v, 7watts, 60rpm rated at 15kg.cm of torque and about £25 from Ali express. The gearbox is 131.5:1 reduction. It has forward and reverse control, speed control via PWM input, and outputs 6 pulses per rev of the motor (787.5 pulses/rev of the output shaft) on a signal line - all at 5v logic levels.

It has worked very well, and my little Arduino controller used the pulses to measure leadscrew speed. Some time ago I decided to try and count the motor pulses as a way of determining carriage position with a view to implementing an auto stop. It worked OK, but I could not get the stop at a chosen zero. It consistently overshot (by up to 0.15mm) but repeatably so, and I assumed I had a code error.

It took a long time before the realisation dawned that there was enough inertial energy in the little rotor (@ 7000+ rpm) to drive the load of the carriage - through the gearbox and leadscrew for several ( up to 10) revolutions of the motor once the power was cut.

By computing and allowing for the overrun based on the speed of the motor, I can now get it to stop the carriage repeatably within 0.01mm.

We can pretty much all of us bodgers put widgets and wires together to do stuff with varying levels of competency and experience but once you get beyond the 3-axis zone you suddenly find your paths for a control solution rapidly narrow into very few choices

4 -axis CAM and the control systems to use them are thin on the ground for a hobbyist

Suddenly the choices go down and the $$'s start going rapidly upwards

Edited By Ady1 on 27/08/2023 13:13:10

I've had another squirrel and it's amazing work on multiple levels but there's no information on what he did for the CAM work and the G-code

It's a 6-axis job which really is next level stuff but I have no clue how he actually made it move about, unless it was maybe perhaps a python routine

All the stl files and freecad files appear to be build files for the project, 3D printing etc

So it looks like it could be a robot that can do fabby stuff, but it will only ever be doing one fabby thing unless you have serious programming skills

Edited By Ady1 on 27/08/2023 18:24:46

I guess that’s why we were not quite on the same page this morning, Ady

I hadn’t taken any real interest in the programming of the robot arm … just in the upgrade of the actual servo.

MichaelG.

.

https://bitbucket.org/adamb3_14/servoproject/src/master/README.md

Edited By Michael Gilligan on 27/08/2023 19:31:22

Not a prob. We all thrash stuff out from different angles in here which makes this place so interesting

The build in itself is an impressive feat, and then there's all that extra technical amazingness

(edit: I was really hoping someone had created an easy multi axis software solution for hobbyists)

Edited By Ady1 on 27/08/2023 19:49:08

Here’s a treat for you, Ady

It’s a few years old, and in a somewhat higher price bracket … but anyway:

.

.

MichaelG.

.

P.S. __ I doubt if the narrator really meant “nano millimetres “

Edited By Michael Gilligan on 27/08/2023 22:00:13

Not sure if this is what you're asking, but the 32bit ports of GRBL will support up to 6 axes (definitely for the first two, and probably for the third. (Open loop still, though.)

GRBL_ESP32

FluidNC

GRBL_HAL

GRBL_ESP32, although no longer actively supported is *very* compatible with 8 bit GRBL and its senders. Both it and FluidNC can be compiled in the arduino IDE.

FluidNC supports software configuration, so you can make machine changes without having to recompile. It will play with many of the Gcode senders, but not my favourite one (Openbuilds Control).

GRBL_HAL needs to be compiled in the appropriate IDE - I haven't used it as I can't be bothered to install yet another IDE, and the web based configurator has limited options.

GRBL_ESP32 works well for me on a 3 axis machine with the ability to run a second spinde (laser + rotary) and a 4th axis without having to change any firmware or wiring.

CAM is the big hurdle for 4+axes.

Edit: I should say *cheap* CAM is the big hurdle...

Edited By Andy_G on 27/08/2023 22:18:37

Why not use something that has been improved on for years and being used daily for real machines?

Linuxcnc will allow you to do true closed loop control.

I am even doing it through the the printer port (obviously speed limited) but very accurate using PID.

Using basic pwm amplifiers - My lathe currently uses $5us drives from amazon.

It has 9 axis of coordinated motion..

You can start as simple as this for closed loop... (replace the amc drive with the $5us drives from amazon.)

Plus you can buy inexpensive external interfaces that do hardware encoder counting, high speed step/dir and high speed pwm (among other things)

Andy, I wish you hadn't posted this! Now I'm thinking about upgrading my cnc to use esp32!

Sorry about that!

(P.S. you also get wifi control...)

Just beware that the ESP32 uses 3.3V logic so (usually) needs a buffer/level shifter to reliably drive stepper motor drivers. (Some people get away without.) Cheap level shifters don't work - DAMHIKT.

This was my approach, but you may prefer a ready made interface:

Edited By Andy_G on 28/08/2023 15:42:30