Jump to content


Photo

Encoder-based PE Correction on the cheap

  • Please log in to reply
306 replies to this topic

#226 vdb

vdb

    Messenger

  • -----
  • Posts: 426
  • Joined: 08 Dec 2009

Posted 07 May 2013 - 01:36 PM

Ah, so my advise was correct then, it was my suspicion from the beginning, the physical mounting of the encoder is also very important (not the only source of error but an important one)

#227 Starhawk

Starhawk

    Space Ranger

  • *****
  • Posts: 5559
  • Joined: 16 Sep 2008
  • Loc: Tucson, Arizona

Posted 07 May 2013 - 02:02 PM

Suggestion: If you can get a machinist to help, the answer might be to hand him the mount and just say, "This is what I am trying to do" and see what he says.

There are ways of causing this to true up using an expanding shaft. Imagine if you used the far end of the RA pass-through to cause the shaft to line up. Something like a tapered interference at the far end would cause the shaft to true up if the bottom end was tapered. So, what you would have, in essence, would be two tapers pulled into either end of the cylinder section with a long bolt holding them together.

Alternatively, you could just shim the shaft with something like aluminum air conditioning tape. You would want to have it at three equally spaced points around the shaft. That might get your current shaft working so you can make progress without having to wait for a major refit.

-Rich

#228 neilson

neilson

    Viking 1

  • -----
  • Posts: 538
  • Joined: 22 Aug 2010

Posted 07 May 2013 - 05:32 PM

Or some plumbers white teflon tape for pipe threads

neilson

#229 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 07 May 2013 - 09:59 PM

I've tried shimming with masking tape. Too thick. I would imagine the aluminum AC tape is even thicker.

I have some ultra-thin teflon sheet from my dead-end AP600 declination bushing experiment. I could try that..

#230 Mert

Mert

    Skylab

  • *****
  • Posts: 4222
  • Joined: 31 Aug 2005
  • Loc: Spain, Pamplona

Posted 13 May 2013 - 11:30 AM

Hi Orlando, have you already tried shimming the
encoder???

Great results so far, very impressive!

#231 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 13 May 2013 - 12:35 PM

Traveling, so no shimming. Have ordered some new shafts. Hopefully will receive them in a couple weeks.

#232 Mert

Mert

    Skylab

  • *****
  • Posts: 4222
  • Joined: 31 Aug 2005
  • Loc: Spain, Pamplona

Posted 13 May 2013 - 04:13 PM

I see, one more question though!

On the sketches you posted a while ago on the shaft you
would like to make, did you check if the end of your
RA-axis where the polar scope is screwed into, is trued
up???
If not, screwing your new axis flat against that RA-axis
end which wouldn't be trued up wouldn't make much of an
improvement I am afraid!
Maybe just rotating the RA-axis with a little feeler gauge
against the flat end will show??

I hope to read more on your project, this is very
interesting all together!!! :waytogo:

#233 mattflastro

mattflastro

    Vendor - Astrovideo Systems

  • -----
  • Posts: 622
  • Joined: 31 Jul 2009
  • Loc: Brevard County , FL

Posted 17 May 2013 - 11:24 PM

I see, one more question though!

On the sketches you posted a while ago on the shaft you
would like to make, did you check if the end of your
RA-axis where the polar scope is screwed into, is trued
up???
If not, screwing your new axis flat against that RA-axis
end which wouldn't be trued up wouldn't make much of an
improvement I am afraid!
Maybe just rotating the RA-axis with a little feeler gauge
against the flat end will show??

I hope to read more on your project, this is very
interesting all together!!! :waytogo:

you need an encoder mount similar to a push pull lens or mirror cell . No amount of shimming would make this mess true with the sub 1ppm accuracy you need . Scope makers know they couldn't build the telescope concentric and collimated and that's why there are collimation adjustments. Imagine you didn't have secondary mirror collimation adjustments and had to machine or shim the mirror holder until it was perfect. Adding collimation screws to your encoder isn't that big of a deal compared to all the IIRF , Bessel, etc FFTs you went thru . For a few pesos you could get some nice differential action submicron accuracy actuator screws like ther ones used in optical kinematic mounts:
Differential screws

The one from the link has a precision of 25microns PER REVOLUTION. That's 1 micron per 18degrees of rotation . Try that with shims or machining .

#234 Mert

Mert

    Skylab

  • *****
  • Posts: 4222
  • Joined: 31 Aug 2005
  • Loc: Spain, Pamplona

Posted 28 May 2013 - 12:59 PM

Hi Orlando,

By any chance have you had any chance to test something
more, since I'm very ( read VERY ) interested in your
project!
I hope to read more on your progress, keep up
the good work Orlando! :waytogo:

#235 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 29 May 2013 - 12:04 AM

Still waiting for my new shafts Mert.

I have been traveling for the past 2 weeks.

The new shafts are slowly making their way to me in Singapore from the US. I hopefully should have them by Saturday.

Matt: am not too interested in the collimation-style adjustment. More opportunity for user error and quite laborious to adjust. And the TDM manages to do it without adjustments, so it should be possible.

My new shafts are 0.998" (instead of 0.992") for the 1" ID encoder shaft. Hopefully much less wobbling so no more gross periodic error.

#236 Edmond S.

Edmond S.

    Sputnik

  • -----
  • Posts: 42
  • Joined: 05 Jul 2010

Posted 29 May 2013 - 09:58 AM

Hi Orlando,

I have been following this post for a while. This is indeed a very interesting project.

Lately I start to investigate my CGEM performance and getting more intrigued by your solution. Keep us posted how it turns out.

another thing: is it possible to use PECPrep to figure out the fundamental PE and the 8/3 PE, create/generate the combined correction on the fly and send it to the mount ? Sort of like a PE from the computer, but takes 24min to generate profile.

Edmond S.

#237 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 29 May 2013 - 10:45 AM

Hi Edmond, no it is not possible.

You see the non-integer 8/3 cycle only repeats every 24 minutes. Since the PEC indexer on the CGEM repeats every worm turn, you don't know where you are in the 24-minute cycle.

Some of the Meade mounts have a 3-cycle indexer so they can figure out where they are.

Also the CGEM only has an 88-cell PEC. So even if it were possible to do 24 minutes, you would only have 88 cells, so that's only (88 / 24 minutes) = 3.67 corrections per minute (or 1 PEC correction every 16 seconds).

#238 Starhawk

Starhawk

    Space Ranger

  • *****
  • Posts: 5559
  • Joined: 16 Sep 2008
  • Loc: Tucson, Arizona

Posted 29 May 2013 - 06:56 PM

I'm still thinking about this as well.

-Rich

#239 Mert

Mert

    Skylab

  • *****
  • Posts: 4222
  • Joined: 31 Aug 2005
  • Loc: Spain, Pamplona

Posted 30 May 2013 - 08:00 AM

Hope to see an image of the new shafts Orlando, CNC-ing
is fun!!!

#240 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 09 June 2013 - 12:29 AM

small update. Installed the new shafts. Much more sturdy. Very tight fit, had to sand down the shafts a bit.

bad news: huge error still there.

I will be attaching an indexer to the shaft (a 128ppr $10 absolute encoder) and will try to apply PEC to the huge error. Inelegant solution, but I need to get this working (for academic reasons - I want to use it as a project for my astronomy postgraduate course).

Seems that an absolute encoder is a better choice, am still checking if the (5000ppr, 25-bit resolution) Heidenhain ECN425 is cost-effective. But it's instructive to note that this encoder only has a +/- 10" accuracy. I expect this to be true of any 5000ppr sine/cosine encoder. Makes me wonder (further) how the TDM does it, since they use the same darn encoder technology (ERN180/480).

#241 Starhawk

Starhawk

    Space Ranger

  • *****
  • Posts: 5559
  • Joined: 16 Sep 2008
  • Loc: Tucson, Arizona

Posted 09 June 2013 - 12:48 AM

Is the plate attached to the back of the mount still the thin sheet with CCA standoffs? That should be beefed up as well.

What does it do if you unlock the axis and swing through 360° with no OTA attached?

-Rich

#242 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 09 June 2013 - 01:13 AM

Yes there still is that thin sheet... but it's only there to prevent the encoder from rotating. All the physical support is provided by the extremely beefy shaft. I have rotated the mount 360* a couple times (to calculate the huge error). Have not seen how that sheet reacts...

#243 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 10 June 2013 - 11:19 PM

Update: the large periodic error is still there. I believe I've tracked it down to the torque mount on the encoder.

A torque mount of this type will invariably induce a tilt in the encoder. So a flange-mount encoder would be ideal.

The challenge is, there is a 3-month lead time if I order the Heidenhain flange mount encoder, and I'd have to re-fabricate the shaft coupling because Heidenhain doesn't make a 1" bore encoder. Having no lathe really bites at times like these..

#244 vdb

vdb

    Messenger

  • -----
  • Posts: 426
  • Joined: 08 Dec 2009

Posted 11 June 2013 - 12:03 AM

I did look at your first post with pictures, is this still the way the encoder is fixed? If you look at the TDM you see the adaption is really solid, so I don't think you need a new encoder but better adaption of the current encoder ...
Wild guess, would the CGEM TDM adaptor not be adaptable to your encoder?

#245 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 11 June 2013 - 12:39 AM

Unlikely, the TDM uses a specific Heidenhain encoder (the one with a 3-month wait list if I order - and pay - now).

Also the CGEM adapter for the TDM costs $400. I'm not willing to pay that.

#246 SWR

SWR

    Explorer 1

  • -----
  • Posts: 51
  • Joined: 23 Jul 2010

Posted 11 June 2013 - 02:45 AM

Here's a crazy idea for a low cost 2M PPR DIY digital encoder. :)

Use a Blu-ray disk as the encoder wheel and a PCB with three blu-ray pickups. Two pickups will generate the quadrature signal, and the last encoder will generate a coded index signal to give an absolute position after a few hundred counts in either direction.

The encoder disk will be generated by programming the three tracks on a Blu-ray disk. There's a description on Wikipedia under "Optical disk" where you can see the resolution of the Blu-ray disk. The shortest pulse is 130nm which will give a quadrature resolution of 65nm.

The inside tracks are 46mm diameter, which will give a resolution of 2.22M counts per revolution corresponding to a 0,58" resolution.

If it can be aligned with sufficient accuracy, the trimmed down Blu-ray disk could be mounted under the orange ring instead of your thin aluminum plate. The PCB can be mounted on the RA axis using a hollow adapter that will allow the feed through of a wire to the guide port routed internally in the CGEM.

The pin on the guide port that is not connected, should be wired to 5V inside the CGEM. This way you have power supply and guide signals in the same wire.

The small microcontroller on the encoder PCB should implement the guide pulses as open-collector so that a guide camera can be connected in parallel.

If the alignment issues can be worked out, I think this will be a very low cost solution. :D

#247 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 11 June 2013 - 02:51 AM

That's the thing. The alignment issue is the biggest issue, I'm starting to realize...

I've tried to allow the torque mount to have a degree of freedom parallel to the encoder shaft (so it just slides along rather than tilt the encoder).

This doesn't work. I've noticed that the TDM uses the through-hole version of the Heidenhain encoder, but the difference between the Heidenhain and the Baumer I'm using, is that the H. has compression collars on both ends of the bore, whereas the Baumer only has a compression collar on one end.

I imagine the two compression collars help align the shaft more accurately.

#248 SWR

SWR

    Explorer 1

  • -----
  • Posts: 51
  • Joined: 23 Jul 2010

Posted 11 June 2013 - 04:16 AM

The advantage of using the mount as alignment is that you have no bearings that have to align. The bearings on the RA axis are already aligned. The only fiddely thing would be to push the small piece of Blu-ray disk under the orange ring around, until it is perfectly aligned with the RA axis. The PCB with the Blu-Ray pickups is automatically aligned with the RA-axis because it is mounted on this axis. Alignment of the PCB adapter is not important. It will just change the diameter of the pickup circle.

I suppose that this can be aligned by having several similar tracks side by side on the Blu-ray disk. Which one it catches is not important. This way the only alignment is the centering of the Blu-ray disk. I admit that this can be difficult, but at least it's only one thing that must be adjusted.

Perhabs an accurate hole with the right diameter can be made in the disk using a lathe? Otherwise it must be aligned manually.

#249 orlyandico

orlyandico

    Fly Me to the Moon

  • *****
  • Posts: 5512
  • Joined: 10 Aug 2009
  • Loc: Singapore

Posted 11 June 2013 - 09:15 AM

I think you underestimate how accurate the centering should be..

In my case, the encoder has a 1.000" bore. My original shafts were 0.992" (a difference of 0.008") and the resulting encoder swash and errors were huge. The replacement shafts are 0.998" (a difference of 0.002") and there still is significant error.

Also if the disk is not exactly parallel to the pickups (i.e. it is not perfectly perpendicular to the RA shaft) this will also result in encoder swash and errors.

There is a very good treatment of the numerous errors that afflict encoders here -

http://resources.ren...y-of-angle-e...

in particular check out the error induced by bearing wander, encoder swash, and eccentricity (all these errors are of the same magnitude):

Angular measurement error (arc seconds)
= bearing wander (µm) x 412.5/D
where D is the diameter of the encoder scale in mm

So if you have a 100mm disk, a bearing wander of 10 micrometers would result in 41 arc-seconds of error. So the 2 million ticks (~ 0.6 arc-second) is empty resolution, and is completely swamped by bearing wander/eccentricity/encoder swash.

And 10 micrometers is 0.0003937 inches (0.01mm). I don't think a blue ray disk could even be dimensionally stable enough to meet that spec..

#250 SWR

SWR

    Explorer 1

  • -----
  • Posts: 51
  • Joined: 23 Jul 2010

Posted 11 June 2013 - 10:30 AM

Thank you for the link. I will study this in detail. :)

Does it matter that the accuracy is 41" as long as the resolution is 0.6"? If the error is a slow changing error once per revolution, it can be removed in software by PEC training the encoder. If higher frequency cyclic errors from the gearbox and non repeatable errors from wind etc. can be removed, it should be possible to calibrate the slow moving cyclic encoder errors with a guide camera.

One idea that springs to my mind is this: One way to completely remove all alignment issues, would be to mount a blank disk and burn the pattern on the mount using the pickups when both disk and PCB has been mounted securely. The only requirement would be that the RA axis is spun at an accurate speed. Maybe a small/weak motor and a flywheel on an unloaded RA axis that is allowed to stabilize over several minutes could be used? The software could start by burning a single index pulse, and monitor the frequency of this pulse with great accuracy. This could then be used as an accurate timing reference.

Just some loose thoughts. :)






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics