What's new
What's new

This is a project - and it's about ENCODERS!

skround

Plastic
Joined
Feb 14, 2013
Location
Italy
Hi everybody! I'd kindly ask the moderator to move this to a different section if that's the case.

Myself and some fellow astronomy club members are building a 0.8m telescope for an observatory.
For tracking accuracy encoders are key to detect the position of the telescope mount - you can think of that as a 2-axis machine.

Now the accuracy to take nice pics of the sky -and to do some science with that- translates into a linear accuracy of an incremental encoder of 0.3micron or better.

There are 2 main round surfaces that rotate with the axis and they are/will be ground to -hopefully- less than 0.01mm run-out.
The idea is to attach linear scales to these rotating surfaces and utilize an encoder.
The challenge is to do that on a budget i.e. buying used components.

For example I just bought a used RGH22 – it was too cheap not to buy it and just having a closer look at that.
To start with silly questions: can a Renishaw encoder work with a different linear scale as specified?

Again on linear scales – can it be salvaged from machinery? Is it as ‘expensive’ as the encoder or it’s the cheap part of the equation? I would probably need in excess of 1m for each axis. On ebay it looks like you can find encoders for less than £100 but there aren’t many scales…

Btw the motor control sys is already there and has convenient phase A & B inputs.

Thanks in advance!
Michele
 

Attachments

  • RGH22.JPG
    RGH22.JPG
    41 KB · Views: 245
Old electronics are like ex wifes no one wants them. Buy new stuff... the live of electronics is about 10 years and the caps die...Phil
 
Last edited:
Sounds like you’re trying to translate angle to a linear scale, presumably by mounting a linear scale some distance from the pivot axis?
 
You can buy 0.1um linear scales new from China now for pretty cheap (no association with seller, though I recently bought a 0.5um unit from them): 0.1um /0.5um/10um EIA-422 Linear Glass Scale DRO with DB9 Plug for Lathe Milling | eBay

I believe the output is a quadrature signal, so probably not particularly difficult to read the signal with your own unit.

I can't vouch for the quality of the 0.1um units, and if you really care about 0.3um accuracy 0.1 um resolution may not be enough.

Since it's a high resolution linear measurement on a rotary axis, I'm guessing you only need small displacement measurements; this is a nulling sensor to measure relative rotary motion? In that case there are likely LVDTs that would be a good fit, and whose signals are easy to read.

Also, you may already know this, but actually getting useful submicron measurements is extremely difficult and requires careful engineering *and* manual proficiency (cleanliness, operational best practices, etc.). Going from engineering concept to actually clean data requires managing lots of sources of error simultaneously.
 
If you place your linear encoder gradient strip on a decently large radius, you can use a more common 0.5um linear encoder with excellent results.

We have used these to great effect for rotary axis positioning . . . https://www.lika.it/eng/products/linear-encoders/incremental-1/sme11

The read head has no difficulty reading the scale on a radius.

Depending on the drive, lots of signal options from quadrature incremental to a variety of absolute feedback protocols.
 
I have a few 40s radios that needed to have capacitors, usually electrolytics, reformed or replaced and a few later instruments that have some get leaky, but that is about all. I can't recall anything from the 70s on that had a failure except for cases where the capacitors were highly stressed. Some collectors automatically replace all, but I don't.

The biggest problem with old equipment is finding replacements for components that are out of print. In those cases, there usually are modern equivalents such as the TTL 7400 series ICs that have been superceded by MOS devices that do the same thing.

Bill
 
Confused on rotary and linear? Do not understand the mounting being used for a telescope.
This is the tape scale system?

Sounds like you’re trying to translate angle to a linear scale, presumably by mounting a linear scale some distance from the pivot axis?

If you place your linear encoder gradient strip on a decently large radius, you can use a more common 0.5um linear encoder with excellent results.
.

First of all thanks to everybody for the prompt replies - I'm humbled to see such a passion!

It'll take me some time to read all of them and reply - bear with me.

Bob, Rabler, Motion Guru, what I need to detect is the angular position of an axis - so I need an angular information out of the sensing system. However, for the reasoning below I reckon it's more cost-effective -as DIY and on a budget- to implement a linear encoder with a scale that is attached to a radius.

The challenge here is the level of resolution.Target is to get 10 ticks – or pulses- each arcsecond which means 36,000 ticks each degree.
As a consequence a full revolution of an axis calls for a 36,000 x 360 = ca.12 million ticks.

There are rotary encoders that can go that high but they are top of the bunch and I suppose pretty rare and expensive (I guess 1,000/2,000£ easily or more).

On the other end, I can fit a linear scale to a 1200mm diameter ground surface - which was conceived from the design onset for such purpose- that is solid to the axis and read with a read-head.

Read-heads seems fairly attainable on ebay ( like less than 100$/£ for a RGH22/RGH24Series if you're lucky and patient).

If I could find linear scales are inexpensive then bingo! but I seem not to find any of them – the optical ones with 20/40microns gratings, at least.

I'll soon attach an image to try and explain the configuration of the system.

Cheers, Michele
 
So if I understand this right...
You already have a motor control system for this machine, you just need accurate feedback for positioning control. That means you're measuring.

With any kind of measurement, you have precision and repeatability. You're trying to resolve mechanical motion into numeric signals that match.

Precision is the result of the number of increments within a given range of motion. Repeatability is being able to place it in the same numeric, and get the same physical position. This comes down to simple division, and eliminating slop.

I suspect that your motor drives already do substantial division, in order to get the position. That being the case, having an encoder coupled to the drive motor gives you the ready capacity to count turns, and (assuming no lost pulses) determine a fraction of a turn, so you're measuring. The only concern past that, is that you're assuming no lost turns AND... no mechanical slop... the numeric result would be a pretty good 'estimation' of position...

Would that be sufficient to get started?

After that, having some other feedback source that would provide reference to confirm position in either direction, and (with a quality sensor) perhaps be able to 'count' the variations in change with change in direction, it would have the ability to measure it's own slop, and thus, upon change of direction, 'improve' it's positional estimation by subtracting or adding out half the deviation offset, and that would put you predictably within less than the 'slop' range on each motion event...

Or another way, which could be used to improve accuracy of a 'primary' system: Use optical encoding which is much, much farther from the center of your rotating axis. Use the sky.

if your imagery is going into a computer... use the image as your reference... detecting color and intensity of elements, and look for key landmarks (err... skymarks?), compare them to a 'basic' map of the sky, and then determine the physical azimuth and elevation based on that image, compare that to the absolute positions indicated by your platform's position feedback, then recalculate for the desired position, determine the general position needed, then command step-direction-speed towards that target. Once the general area is in window, re-shoot the sky, repeat the calculation, and make another accuratizing run to place it just as you want.
 
Hi everybody! I'd kindly ask the moderator to move this to a different section if that's the case.

Myself and some fellow astronomy club members are building a 0.8m telescope for an observatory.
For tracking accuracy encoders are key to detect the position of the telescope mount - you can think of that as a 2-axis machine.

Now the accuracy to take nice pics of the sky -and to do some science with that- translates into a linear accuracy of an incremental encoder of 0.3micron or better.

There are 2 main round surfaces that rotate with the axis and they are/will be ground to -hopefully- less than 0.01mm run-out.
The idea is to attach linear scales to these rotating surfaces and utilize an encoder.
The challenge is to do that on a budget i.e. buying used components.

For example I just bought a used RGH22 – it was too cheap not to buy it and just having a closer look at that.
To start with silly questions: can a Renishaw encoder work with a different linear scale as specified?

Again on linear scales – can it be salvaged from machinery? Is it as ‘expensive’ as the encoder or it’s the cheap part of the equation? I would probably need in excess of 1m for each axis. On ebay it looks like you can find encoders for less than £100 but there aren’t many scales…

Btw the motor control sys is already there and has convenient phase A & B inputs.

Thanks in advance!
Michele
We have built a tracking system to 1arc second, repeatable to 2arc seconds. The biggest challenge is ringing/vibes, but it sounds like you may have the mass to deal with that. We used a drive system that was geared high enough that we could use standard encoders on the servo motors.

The other big issue is backlash, but if you are only tracking, you are fine.

Using a linear tape system will work.

Have you chosen a controller yet?

Sent from my SM-G975U using Tapatalk
 
1) You probably don't care about accuracy, just resolution.
2) For long distances (1000mm?), plan on everything acting as if it were made of rubber. Things will sag by 5um (or more) easily.
3) Really nice idea on previously mentioned; use the encoder value every so often, and use motor/gear resolution to move small bits. Sky land marks sound even better, since that is what you really are indexing to.
 
Hi Dave, I have to admit that for somebody that is not into telescopes you are quite on point!

So if I understand this right...
You already have a motor control system for this machine, you just need accurate feedback for positioning control. That means you're measuring.
CORRECT


I suspect that your motor drives already do substantial division, in order to get the position.
CORRECT – 2000CPR with a motor to telescope axis ratio of 8000:1 gives a resolution of 16M.
The problem is that it’s far from being accurate enough for the task. The foe is slop, backlash, gear/belts cyclic error in the gear train – you can reduce that although eliminating it’s practically not feasible.
The task of the system is to ‘Track’ i.e. continuously ‘Aiming’ at a moving target. That is needed to take pictures where the camera is exposing for a time that can span from tens of seconds to several minutes – you know, photons are scarce from out there.
See an example of good vs bad tracking:
good vs bad tracking.JPG

Or another way, which could be used to improve accuracy of a 'primary' system: Use optical encoding which is much, much farther from the center of your rotating axis. Use the sky.
EXCELLENT POINT – in Astronomy that’s called ‘Guiding’ and involves sampling a star at a high frame rate with a dedicated imaging sensor, calculate the centroid and feed-back the motor controller with the deviation to be compensated.
So you can guess that the motor controller is quite sophisticated as it is ‘negotiating’ the input from the motor encoder, axis encoder and the guiding system.
 
We have built a tracking system to 1arc second, repeatable to 2arc seconds. The biggest challenge is ringing/vibes, but it sounds like you may have the mass to deal with that. We used a drive system that was geared high enough that we could use standard encoders on the servo motors.

The other big issue is backlash, but if you are only tracking, you are fine.

Using a linear tape system will work.

Have you chosen a controller yet?

Sent from my SM-G975U using Tapatalk

I'd be glad to see that tracking system - 1arcsec is great. Do you have any pic or any link?

Tracking in a telescope is meant for astrophotography and it needs to pinpoint a star at high magnification for up to minutes. Backlash in the geartrain is killing the tracking capability.

The controller is purpose-built by a shop of amateur astronomers in Oregon for driving telescope mounts.
Home
 
Okay, well... tracking... the problem you're really going to be festered with, is mechanical... particularly, backlash, which on a mechanism exposed to any thermal change, will be inescapable.

The problem is entirely triganometric- you could have such a slight backlash and deflection that at the end of the telescope it's barely detectable, but several bizillion miles away, that angle's change is incredible...

...not to mention that no matter how 'tightly' fitted your mechanical support and motion equipment IS... you're still sitting on a techtonic plate on a spinning ball, and trying to catch 4.3 terahertz waves as they're being 'bent' by atmosphere which is changing WHILE your spinning ball is 'revising' your angle on that path.

Perhaps you'd be better to set it on a multi-ton slab of polished granite, rotating on a bearing pin and plate with air-bearings, teh granite surrounded by four huge pontoons floating in a bath of mercury, with the whole works carrying gyroscopic stabilizers... and I STILL don't know how you'd control it's position with the sky...

But it's a good question...
 
1) You probably don't care about accuracy, just resolution.
2) For long distances (1000mm?), plan on everything acting as if it were made of rubber. Things will sag by 5um (or more) easily.
3) Really nice idea on previously mentioned; use the encoder value every so often, and use motor/gear resolution to move small bits. Sky land marks sound even better, since that is what you really are indexing to.

Hi Greif,

#1 Both are vital for tracking although resolution is the primary information - if the accuracy error is not overly erratic it doesn't have "time" to show up in a 10min picture. What I can do without is repeatability - it's not really interesting for an application like this given that is less than, say x10 resolution.

#2 the Altitude axis for example just needs 90degree to be encoded. So 1000mm tape is enough. I like your mindset of thinking like it's rubber to solve problems upfront. For this project I ran FEA - modal and deformation analysis

#3 Regretfully it won't work - there's way too much backlash and other errors in an 8000:1 drivetrain to trust the motor encoders. True direct axis position is the only way to go. That's how all the pro and semi-pro telescope are controlled
 
Myself and some fellow astronomy club members are building a 0.8m telescope for an observatory.
For tracking accuracy encoders are key to detect the position of the telescope mount - you can think of that as a 2-axis machine.

Now the accuracy to take nice pics of the sky -and to do some science with that- translates into a linear accuracy of an incremental encoder of 0.3micron or better.

There are 2 main round surfaces that rotate with the axis and they are/will be ground to -hopefully- less than 0.01mm run-out.

The idea is to attach linear scales to these rotating surfaces and utilize an encoder.
The challenge is to do that on a budget i.e. buying used components.

I'm not picturing the mechanical setup here. Is this telescope an Equatorial or an Alt-Az mount? What angular accuracy is needed? How good is the astronomical seeing at the telescope location?

If it's one arc second (which is very very good for looking through the atmosphere - think mountaintops), the rotary encoder would need to be at least 20 bits. Such things exist, but are not cheap. One can gear-drive the rotary encoder so it can turn multiple times per full circle (360 degrees), allowing a coarser encoder to suffice. But then one must design a backlash-free mechanical system. Again, this can be done, but ... You get the point.

As for using linear scales over a limited angular range, that too can work, but most likely you will need some mechanical gearing to make this work, and rotary encoders may well be simpler and cheaper to use.

So, let's start with a mechanical sketch of the proposed telescope and its pedestal system. Plus some idea of jusat how accurate the angle measurements must be, and how this will be aligned or calibrated against Down and True North.
 
Astronomical? If so, have you considered trying to use your telescope image with "star tracker" software to derive the inertial-space orientation of the telescope?

I'm far from an expert on this subject, but if I remember right, the Jena "Astro 15" startracker used for spacecraft navigation uses a 50mm focal length objective lens and a square- grid photo sensor to determine pointing within an a second in two perpendicular directions, and rotation around the pointing direction within 10 arcseconds.
 








 
Back
Top