Telescope Drive Master (was: The Weather)

[Jim]

Yes, sorry, I bought a TDM. Please accept my abject apologies....

www.mda-telescoop.com

Gets the PE on the mount down to about 2 arcsec

[JohnP]

Jim - would be really interested to see how that works out... As I understand you have to remove the polar scope from your mount so you need a permanent setup with good polar alignment... Would be good to see before & after error graphs...

John

[Rocket Pooch]

Hi John,

I had a chat with Jim about this earlier today, oddly when I was at collage all those years ago I had to build a closed feedback loop circuit for keeping two sets of cogs in sync, very accurately.  I built this using a magnetic shaft encoderr to measure small % of degree of movement against a know suggested movement and adjust the next movement etc etc. 

This is exactly what these guys have done, you don't actually need to take the mount outside to measure it.  What is being measured is the pulse to the motor i.e. an assumed siderial rate and then mesurement of the actual movement on the RA shaft, you then calculate the differece and very very quickly (depending on the motors and software) adjust the next drive pulse.  So the limit will only be the ability to measure and adjust, quite simple really.

All the reports I have seen today seem to make an EQ6 track at sum 2 arc thingy peak to peak, with possible guiding adjustments to DEC.

Chris

[MarkS]

It's the first I've heard of this TDM.

I'm intrigued by how the rotation of the shaft can be measured to the required accuracy.  Let's say for the sake of argument that the accuracy is 1 arcsec.  

That means measuring the shaft position to 1/3600 of a degree.
Or, another way of looking at it, 1.3 million counts per full rotation (3600 x 360).
Or, given an encoder radius of 5cm the linear position must be read to an accuracy of 0.2 microns which is 25 times smaller than a typical pixel on a CCD.

I'm stunned that this is achievable.  But then their website:
(http://www.mda-telescoop.com/index.php?option=com_content&view=article&id=77&Itemid=109)
says they are working to 1/8 arcsec accuracy.  Now I'm totally gobsmacked.  Thats a linear accuracy of 0.025 microns at a 5cm radius.

Further internet research shows that devices exist that actually measure to one thousandth of a micron:
http://www.renishaw.com/en/incremental-optical-linear-and-rotary-encoders-now-offer-1-nm-resolution-and-ultra-low-positional-noise--11195

All amazing stuff!!

[JohnP]

Yes I think I understand. I think it uses a very accurate rotary shaft encoder to measure the small changes in the speed of the RA axis (either faster or slower etc.) due to imperfections in the mounts gearing etc. It knows exactly what the rate should be by comparing with an accurate quartz timer & then based on the difference between the two sends correction signals to ST4 port. It doesn't make corrections in Dec so you need either a very accurate polar alignment (drift) to start with or like you say you still need to use a autoguider. It does look good though & if results are correct produces performance comparable with mounts costing many thousands of pounds.... Jim you definitely need to start a new thread & keep us updated on your findings etc..

Cheers

[RobertM]

Great news Jim, Renshaw encoder type systems are supposed to work a treat so I'm really looking forward to the results. 

Robert
PS unless you got it more than six weeks ago I'm still blaming Mick for the bad weather.  The clock stars ticking at delivery time I think !

[MarkS]

All the reports I have seen today seem to make an EQ6 track at sum 2 arc thingy peak to peak, with possible guiding adjustments to DEC.

That's pretty good.  I typically get 0.2 pixels RMS guiding error at 300mm focal length i.e. 0.75 arcsec RMS which is more or less equivalent to 2 arcsec peak to peak.  But, as you say, the DEC still needs guiding - at least it does on my EQ6.  So I would still need a guider.  So if I need a guider anyway then it would be interesting to evaluate how much extra benefit the encoder feedback loop on the RA gives me.

Mark

[Rocket Pooch]

I'm intrigued by how the rotation of the shaft can be measured to the required accuracy.  Let's say for the sake of argument that the accuracy is 1 arcsec.  

They read the movement through a diffraction grating, the really accurate one use magnets and I have no idea how they work.

[Ian]

I'm intrigued by how the rotation of the shaft can be measured to the required accuracy.  Let's say for the sake of argument that the accuracy is 1 arcsec.  

They read the movement through a diffraction grating, the really accurate one use magnets and I have no idea how they work.

I would think they use Hall effect sensors http://en.wikipedia.org/wiki/Hall_effect

[Jim]

I see I've started something here!   Just had a look at the encoder website and here's the description of how it works

"The encoder uses image scanning.  This functions by means of projected-lightsignal generation: two graduations with equal grating periods are moved relative to each other—the scale and the scanning reticle. The carrier material of the scanning reticle is transparent, whereas the graduation on the measuring standard maybe applied to a transparent or reflective
surface. When parallel light passes through a grating, light and dark surfaces are projected at a certain distance. An index grating with the same grating period is
located here. When the two gratings move relative to each other, the incident light is modulated. If the gaps in the gratings are aligned, light passes through. If the lines of one grating coincide with the gaps of the other, no light passes through. Photovoltaic cells convert these variations in light intensity into nearly sinusoidal electrical signals. Practical mounting tolerances for encoders with the imaging scanning principle are achieved with grating periods of 10 μm and larger"



There's a lot of debate as to whether these things are worth it, the encoders are pretty expensive and they are not going to turn my mount into a AP or Tak. However I think it's just another tool that adds to your mounts accuracy. Good autoguiding can also achieve the same result and does not need as good a polar alignment but I think where this thing might shine is allowing unguided 300sec exposures. I seem to waste an inordinate amount of time getting guiding to work so seems worth a gamble. In any event I don't think I've lost anything (apart from money!) as I'd stopped using the polar scope anyway after doing my back in one session, and started using alignmaster which gives me a pretty good alignment in about 5-10 mins. Both my Owl and M95 pics were taken using this method. Anyway I've yet to use this in anger but have done a PE chart below showing the first 4 mins with the unit off and then with it on.

[MarkS]

Which encoder site did that come from?

Whilst it's a good explanation of the previous generation of devices, the Renishaw can give a resolution down to 1nm  i.e. approx 2 wavelengths of light.  I'm guessing they must be using some kind of holographic grating with a laser reader.

[Jim]

TDM uses a Heidenhain ERN-480 encoder. See this document on Google Docs.

(Edit: Shorten visible part of link -- Rick)

[MarkS]

Thanks for the link to the doc.

It quotes rotational resolution of around 12 arcsec for that encoder. 
But TDM is quoting rotational resolution of 1/8 arcsec which is 100x better.

There's a missing piece of the puzzle somewhere ...

[Jim]

Hi Mark, that's a good point but clearly it works so as you say must be something else going on. I found a useful document that explains a bit more about how optical encoders work; don't pretend to understand it all, you'll probably make better sense of it than me, but look at pages 25-30. They use a formulae

"The resolution of an encoder depends on the number of pulses it can provide per revolution
(PPR = pulse per revolution or lines per revolution). The higher the resolution, the greater the
number of holes on the disc and hence the encoder will be more expensive. If we consider
a 4096 PPR quadrature encoder, 4096 pulses per revolution multiplied by 2 channels (A+B) multiplied by 2 edges per pulse gives 16,384 counts per revolution (CPR), the shaft resolution
is simply 360/16384 = 0.02197266° or 1.32 arc minutes (1' 19")"


http://www.eurotechsa.com.ar/data/download/newsletter_docs/002/position_guide_lowres.pdf

[MarkS]

I've found some other online discussions regarding what is going on.

Jerry Hubbell explains it well in this thread:
http://www.cloudynights.com/ubbthreads/showflat.php/Cat/1,2,3,4,5,8/Number/4890522/Main/4885225

The key point is that the output from this particular encoder is not a series of pulses but is an analog sinusoidal wave - 2 channels, I think. These sine waves are sampled using A/D converters.  The net effect of this is to boost the accuracy from 5000 cycles per revolution (this is the number of lines on their optical grating) up to the 10,000,000 counts/rev required for sub arcsecond accuracy.

It's a very clever way of pushing the encoder well beyond it's design parameters.  It won't be as good as using an (expensive) alternative that has this accuracy built in from the outset (i.e. by having a finer resolution on the optical grating - more lines per revolution) but it does appear to give good results in practice.

I'm very interested to see how you get on with it. 

Mark