Onstep and celesctron CGEM


Good day

I begin this topic to ask if you think if replacing the motors (and the whole electronics) in a CGEM mount, for stepper motors and an onstep system would enhance my mount performance. Has you may know, the CGEM is known for the low quality of its gearboxes, which add non periodic errors. I have used PECPrep to analyze my mount and the main sources of error are at the gearbox frequencies so I think that changing the motors and electronic in this mount would give me some improvement.

I have planned to use NEMA 17, 0.9 degrees, probably from orientalmotors, TMC2130 as drivers and a Teensy 3.6 as microcontroller (I already have one).

Howard Dutton

The worm wheel's have 180 teeth on this I think?  That makes directly driving the worms with 400 step motors ill-advised in my opinion (too little torque and tracking resolution isn't so great either)... especially if you're shooting for the very highest levels of performance.

Khalid Baheyeldin

You are obviously targeting astrophotography, otherwise you would not have bothered with PEC.

Before you jump to OnStep as a solution, check your mount's internals. Sometimes due to poor manufacturing, there is a lot of metal pellets that get into the gears. This guy was doing a hypertuning fir his CGEM, and found out about it. The pictures are frightening.

Someone else tried to convert a CGEM to OnStep, as per this CN thread , and as Howard said, directly driving the 180 worm wheel will give you a resolution of 0.56 arc second, and there may not be enough torque.

Not sure if you can insert pulleys and belts inside the housing. There does not seem to be a lot of room there.

This leaves geared steppers as the the last resort, such as this 19:1 planetary, but that means you will now have non-periodic error from the gearbox, so you are back to the original problem.


Thanks  for the fast reply. I forgot to mention that I would use a 4:1 or 3:1 reduction, I need to check how much space I have to work.

Does anyone know which is the reduction on the eq6 mounts? If I use the same I should be able to drive the mount with the motors I have on mind.

Khalid Baheyeldin

On Tue, Feb 12, 2019 at 04:51 PM, <librepensador_1591@...> wrote:
I forgot to mention that I would use a 4:1 or 3:1 reduction, I need to check how much space I have to work.

Does anyone know which is the reduction on the eq6 mounts? If I use the same I should be able to drive the mount with the motors I have on mind.
Start by filling in the values in the spreadsheet, and also reading the construction page.

If you manage to get a 4:1 reduction inside the mount, then you can get good resolution.

The spreadsheet will have 400, 32, 4, 180, and that will give you 0.14"/step, or 400, 16, 4, 180 for 0.28"/step.
Either should work for astrophotography.

The issues become: a) do you have room, and b) will this reduction introduce non periodic error?

There are other considerations, such as slewing speed, and that is dependent on the MaxRate parameter, which in turn is depending on the board you will be using.

Francois Jr. Bouchard

Hi guys,

I am in the process of finishing the complete retrofit of a Celestron CGEM to Onstep.

So far, the results are showing significant improvements. My goal was to validate if the CGEM could potentially have better tracking and guiding performance be replacing motors/gearbox, spur gears, electronics & firmware. To give you an idea of the PHD2 Autoguiding results:
Original Hypertuned CGEM = Peak to peak guiding graph of -4 to +4 arc-second with an RMS of near 2 arc-second.
With the replaced parts =   Peak to peak guiding graph of -1 to +1 arc-second with an RMS of near 0.7 arc-second.
"No more elongated stars"
Here is what I used:
Oriental Motor Steppers (PKP244MD08A-L) - Nema 17 / 400 steps
Timing belt  and 4:1 pulleys to replace the spur gears in the CGEM
Onstep (currently finishing the assembly of MiniPCB 2: Teensy 3.2 + TMC2130)

No modifications were made to the CGEM assembly. I had to fabricate a plate for the motors to replace to stock plates. Also bought a USB endoscope in order to be able to make the belts/pulleys alignment once the motors were in the mount.

Using 400 steps steppers ith the 1:180 worm gear and the 1:4 pulleys, I get a 0.281 tracking resolution at 16 micro-steps.


Khalid Baheyeldin

Thanks for sharing Francois.

Interesting thing about the USB endoscope. Must be very hard to attach the pulleys and belts.

The polar scope could be partially blocked by the top motor, but there may be enough 'aperture' for it to be functional. Or you may be using another method and the polar scope is not of use to you.

I added it to the show case on the wiki.


Looks great we have a old cgem laying around. I'm very interested in the plate design you made and where and what type of pulleys and belt you used. Looks like it is going to be a very clean build.

Francois Jr. Bouchard


The CGEM had to be disassembled in order to install the pulley to the worm gear. I had experience with the mount due to the fact that I had hypertuned it myself.

Endoscope was use only to align the motor pulley with the one attached to the worm gear and to make sure the belt was in the right spot. The small mirror attachment was a savior.

You are right, the polar scope (if you use one; I don't) is no longer functional.

Francois Jr. Bouchard

Hi Matthew,

Here is the Solidworks files for the plates.

I ordered the pulleys and belts from

2 ea. MP (MXL) Synchroflex® Timing Belts
Width in mm - 6
Belt Length in mm - 182.88
Tension Cords - Steel
Material - Standard PU

2 ea.15 Tooth MXL Pulley (AL15MXL025FB)
  • Modify Bore? - 5mm
  • Grubscrew Holes? - 2off M3
  • Keyway? - Not Required
2 ea.60 Tooth MXL Pulley (AL60MXL025)
  • Modify Bore? - 12mm
  • Grubscrew Holes? - 2off M4
  • Keyway? - Not Required
Hope this help.



Awesome thank you.

Chris Vaughan

Thanks for this it looks great. After I did my EQ5 I was thinking about onstepping my cgem but didn't think I could fit a pulley on the worm shaft. Thought the space was too small and that it would be to difficult to fit a belt from it to a stepper


What you are doing is exactly what I want to do with my mount. And I have to say that the information you shared is going to be extremly useful, so thanks for it.

Chris Vaughan

Have you got STL file versions. Like to see how you did the stepper mounts. I can't open the sldprt files

George Cushing

Used free trial solidwks viewer to get these.

Base Plate

Dec Bracket

RA Bracket

Chris Vaughan

Thanks for the glovius link

George Cushing

If I had gone a little deeper into my google, I'd have noted that SolidWorks has their own free viewer.

Chris Vaughan


Any chance you can put up STL files of the plates for the CGEM. I have tried to use your SLDPRT files but cannot get them to the correct scale

Thanks, Chris

Chris Vaughan

I've finally got back into trying this CGEM mod. I've currently got
- 400step NEMA17 motors (
- 4:1 gear reduction using the mxl pulleys (as described by Francois above).
- using a miniPCB, have tried fystec TMC2130s and TMC2209s spreadcycle
- 32 microsteps and the 160 worm wheel I get 0.141sec per step.

Stripped down it seems to be working OK, but the most I can get is about 1.8 deg.per.sec slewing before it starts stalling !
I'd like faster slewing. The easiest thing for me to do is to reduce the gear ratio, say to 2:1 - reducing resolution to 0.281 sec.per.step, but then can hopefully get 3.5 deg.per.sec slews !!. Or maybe 3:1.

My question is - what sort of resolution do I really need for imaging.  Would reducing the resolution to 0.281 sec still be OK ??
I'm mostly using an 8" F5 1000mm focal length newt with an ASI071 on this mount - so an imaging scale of just under 1arcsec.per.pixel (

Thanks, Chris

Howard Dutton

As always feel free to experiment, but it is difficult to imagine why a 4:1 reduction isn't more optimal for reaching higher slew speeds...

The motors are high current/low inductance type so should be able to reach useful RPM's without dropping much torque.  In-fact I looked at a similar motor with a datasheet that includes torque vs. speed curves and indeed it maintains >50% of its torque to 400 RPM (3.3 deg/s speeds) so on paper cutting the ratio in half will lessen torque (relative to what is required) at both tracking and goto speeds.

I'm wondering about what voltage you're running at and if mid band resonance is causing a problem.  Hopefully you're running from a 24VDC power supply... and it has been my experience that the fastest slews happen at 30-50% of the rated current, this is because mid-band resonance often causes reduced torque availability at moderate speeds.  And for the most part the speeds we use are moderate.  I assume you know to use spreadCycle mode for slews (not _VQUIET.)