CG5-GT, rebuild and onstep coversion

[FrenchyArnaud]

Hi guys, 
not really a question today but rather sharing the current situation.

As some of you may know, my trusty CG5-GT has died some time ago; the dec axis has seized, which put so much stress on the motor that the gear box literally exploded, shearing teeth of some cogs in the process. Irreparable - or so I thought. See here : https://stargazerslounge.com/topic/408523-rip-cg5 

After studying for a very long time all options to build a GEM from scratch, it occured to me that if I could fix the axis I also could convert the CG5 to onstep, which would be a very good first step into onstep before trying to also build the mechanics. 

So I did... just that. After hours and hours taking the mount apart and "un-seizing" the dec axis then spending even more hours polishing the parts by hand, regreasing etc until it was butter smooth, I now have a fully reassembled mount and adjusted the worms with no detectable backlash. 

 

 

The CG5GT natively uses encoded servos with a 50:1 gearbox driving a 144:1 worm on both axes. 
Now to drive each worm it has a belt drive 12th>48th (4:1 ratio) powered by a nema 17, 400 steps 0.9deg, 48Ncm, 1.68A, (ref JK42HM48-1684-16A)

Onstep is on a mega 2580 + ramps1.4+TMC2209+RTC. The power source is a dedicated stabilized 12V3A from mains. 

The motor bracket in dec is screwed in the saddle, the motor bracket in RA is simply superglued (by chance it sits very nicely where it belongs, like the mount was designed for it!) 

I am now just waiting for the belts of the correct length to be delivered so that I can begin to test and adjust in software. Right now it is set at 400steps * 64usteps * 4:1 * 144:1, giving a max rpm comfortably at 96rpm, and a tracking precision at 0.09arsec and a max slewing speed of 2.5deg/sec. Of course, I will need to test a lot and surely to address unforeseen problems, especially on tracking speed and guiding, but up to now I am very optimistic.

I was very intimidated by OnStep but right now, I am seduced! 

[FrenchyArnaud]

After weeks of redesign, delicate work on the mechanical cores, stripping all the original electronics and refitting DIY custom ones,  my trusty CG5GT, like the phoenix, raises from its ashes! 

 

 

First test run tonight, hoping the damn thing will start at all 🤣

Also, it's technically no longer a CG5GT but a very upgraded,  onstep-based mount that only re-uses the axis of the CG5 in this total rebuild, so it surely needs a new name 🧐

[Swoop1]

The FrenchyFrankenMount?

[FrenchyArnaud]

Very first test - not even balanced and with a 28arcmin error in PA! Over the moon. With everything aligned and tuned as it should, the RMS will defo be well in the region of 0.5.

[knobby]

23 hours ago, Swoop1 said:

The FrenchyFrankenMount?

Frenchy-Q-Cinq 🙂

 

[knobby]

1 hour ago, FrenchyArnaud said:

Very first test - not even balanced and with a 28arcmin error in PA! Over the moon. With everything aligned and tuned as it should, the RMS will defo be well in the region of 0.5.

Thats looking very promising, nice job.

[bosun21]

 Phoenix would be an apt name for your mount since it came back from the dead.

[FrenchyArnaud]

IGNORE - DELETED

Edited June 17, 2023 by FrenchyArnaud
PICTURES NOT SHOWING

[FrenchyArnaud]

As requested here is a summary of the conversion of my celestron CG5-GT to OnStep. Its dec axis had seized, causing the catastrophic failure of the gearbox. It's my understanding, but don't quote me on this, that the “polaris” mounts are quasi-clones.

After exploring different options, I decided to try and convert it to onstep. It has been a lot of work but well worth it and the final result is actually a massive upgrade on the original celestron system.

I must immediately point out that this mount is used exclusively on a pier for astrophotography. The pier has its own electrical supply (12V 8A) and a ramp of relays. The mount is piloted entirely by NINA, via USB connexion, from a dedicated laptop indoors (except for the relays that are commanded par a DIY python GUI). So the conversion does not involve any handset and it does not use the other options of OnStep either (focuser, rotator, wireless etc) , this is purely designed to enable goto, plate solving / centering then fine guiding, via USB.

Consequently, all the values and drive ratios were chosen for accuracy at the expense of speed because in my case that is the whole point. Your priorities might be different and the values you need different too.

Detail of the motor on the dec axis. The small pulley on the white drilled metal plate is not part of onstep (it's my arduino focuser) 
The bracket is simply screwed in the saddle. 
The black dots are because during development it was too difficult to judge if the pulleys were actually turning and at what speed. 

Detail of the motor on the RA axis. This bracket is simply superglued in place, it sits nicely in the grooves where the original system was. 

NB : there is no way to adjust the tension of the belts. They must be chosen of the right length to achieve proper tension. 

Global view : the electronics are in the grey electric box. 

And for fun, the silly mug of very self satisfied me! 

 

I will not go in details into the repairs to the dec axis; I will just mention the tutorial by Astronomy Boy here : http://www.astronomyboy.com/cg5/ , that is a big help in undertaking the servicing or repair of this mount. The repair of the mount itself costed only the price of the grease – I went with white lithium. I can't price up the amount of elbow grease used though, but that was quite a few litres. The total work from taking down to putting back up was about 50 hours but 2/3 of that was to fix and polish the seized axis.

So the first step for me was to totally strip the mount from motors and electronics (discard them), and totally service (polish + re-grease + re-adjust the worms) until I was confident I had a solid, sound mechanical base.

Then I fitted the electronics previously assembled/tested on the bench.

The OnStep system is based on Arduino Mega2560 + Ramps1.4.

I included a RealTimeClock because it seems the Mega carries a clock that is not good enough; besides, the RTC module allows the system to keep time between power cycle, streamlining the boot of the pier (no need to set the time and location or sync the mount to NINA's values.)

The motors are 400steps, used at 64 microsteps, driving a gt2 pulley+belt reduction (12teeth to 48teeth so 4:1) which in turn drives the original 144:1 worm gear of the CG5. This gives a total mechanical ratio of 576:1. This combination gives a theoretical tracking resolution is 0.09arcsec.

Both axes are designed exactly the same way, with exactly the same values.

The controller is an Arduino Mega 2560R3 clone distributed by KEYESTUDIO (the yellow one) - £17.99

The RAMPS board is a RAMPS1.4 by DollaTek. - £6.99

The RTC module is a DS3231 (RTC Module AT24C32). - ebay, £3.95

The motors are nema17, with TMC2209 drivers.

NB: it turns out these drivers are far from ideal because in this config you cannot get them to change microstepping after they are plugged on the RAMPS. It is seemingly possible but it requires some soldering / modifications that I did not have the confidence to undertake. I recommand that you dig into the wiki for a better choice if slewing speed is important for you.

The motors are 400 steps 0.9deg, 48Ncm, 1.68A, (ref JK42HM48-1684-16A). - ebay, 2x £17.99
They are fixed on the mount using right angle nema17 brackets : ebay, 2x £4.92

The drivers are “WANGCL” Motor Driver TMC2209 v2.0. - 2x £7.99

The pulleys are GT2 6mm, 12teeth 5mm bore : ebay, 2x £3.95 and GT2 6mm, 48 teeth 6mm bore : 2x £5.25.
There is finally one 70teeth/140mm GT2 belt and one 69teeth/138mm GT2 belt : ebay, 2x £2.95

The cables of the motors are sleeved with ETOPARS 6mm braided sleeve : amazon, £6.79 for 10metres.

The electronics are housed in an Junction Box (190mm x 140mm x 70mm Waterproof IP56) fixed to the pier : amazon, £9.10

The total final cost of the conversion/rebuild is: £123.02
NB : this is far less than the cost of just the encoded servo + gearbox if supplied by Celestron.

The relevant parts of my config.h looks like this :

#define MOUNT_TYPE GEM //

#define TIME_LOCATION_SOURCE DS3231 //
#define SLEW_RATE_BASE_DESIRED 2 //

#define SLEW_RATE_MEMORY ON //

#define MFLIP_SKIP_HOME ON //

#define MFLIP_PAUSE_HOME_MEMORY OFF //

#define MFLIP_AUTOMATIC_MEMORY OFF //

#define PIER_SIDE_SYNC_CHANGE_SIDES ON //

#define PIER_SIDE_PREFERRED_DEFAULT BEST //

#define STRICT_PARKING OFF //

#define STEP_WAVE_FORM PULSE //

Chosing PULSE proved IMPORTANT.

and for both axes :
#define AXISx_STEPS_PER_DEGREE 40960.0 //

#define AXISx_STEPS_PER_WORMROT 102400 //

#define AXISx_DRIVER_MODEL TMC2209 //

#define AXISx_DRIVER_MICROSTEPS 64 //

All the rest is left as default values.

The results!

In short, I am over the moon with that.

First let's begin with the 2 negative points. Let's be clear, they are not negative for me but you will want to be aware before you replicate my conversion.

1) Slow slewing speed : 0.26deg/sec. It takes 6 mins to go horizon to horizon. However, with runtime adjustment, PULSE waveform and SLEW_RATE_MEMORY activated, this goes permanently up to 0.51deg/sec, which is slow but honestly, not ridiculous. In real life test, each slew to target takes about 2 minutes.

2) Control: there is absolutely no way to activate, control and direct the mount without the laptop. This design is perfect, in my opinion, for a pier/observatory or any sort of permanent setup but would not do for a mobile/portable one.

The good points!

Let's cut to the chase, the one figure that counts :

RMS guiding error 0.60”, without any sort of tweaking for balance, PHD2 values, and with a 1 arcmin off PA. During the couple of hours of clear skies I had, the error actually hoovered in the high 0.40s to the low 0.50s, needless to say I was very impressed with that!

I am absolutely sure that, tweaking a bit and working on PA I'll be able to get the RMS under 0.50” consistently. And that's no minor feat considering the budget of the rebuilt and the fact that perfs like that would demand several £1000s with commercial gear.

Due to fresh servicing, worm adjustment, stepper motors + belts, the backlash is virtually undetectable.

Also, the mount is 100% silent. There is absolutely no way to tell, by ear, if it is on, slewing, or off.

The GoTo precision is amazing. During tests I was on target first time every time after a couple of stellarium syncs. Plate solving in astap (tolerance set to 1arcmin for framing in NINA) put me centered on target in 2 attempts consistently (slewing>solving>reslewing>centered)

Because Onstep parks ridiculously precisely and keeps high precision track of date + time, and because the mount is on a pier, this precision is GoTo and tracking will be not only repeatable but reliable.

Edited June 18, 2023 by FrenchyArnaud

[FrenchyArnaud]

To really close the thread, and give ideas to whever will use this for their own conversion, here is the very first picture taken with the setup; M101.

The imaging train is: canon 450D modded > Skywatcher Coma Corrector > skywatcher 130PDS (as sold, no tweaking done yet)
28x 180seconds (=84minutes) + calibration frames.  

Guiding (225mm x 60mm + asi120mm) degraded slowly during the night (balance? cable snag?) from 0.65 arsec to 1.40arsec. With my 1.83arcsec/px scale it was no issue. 

Not super impressive but considering there is not astro dark at all untill august and it is the very, very first attempt, I am pretty chuffed.

Cropped to 1/2 dimensions. 

[Alan White]

Arnaud, what a super job, you must be very pleased.

That mount is like a cat more than one life.