Last updated: 19 June 2009

Autostar - Thoughts on ETX drive training

Jump to the start of today's update.

Sent: Saturday, January 19, 2002 10:35:18
From: (Victor van Wulfen)



Much has been written on the importance of precise training of our little blue scopes' drives. It solves many problems like inaccurate GOTO's and 'rubber banding' most users experience early on in their ETX careers. First a short explaination of how training works inside the gearboxes.


Created by AccuSoft Corp.The gears that drive the ETX are small and made of plastic, very similar to what you'll find in a toy car. These little gearboxes allow considerable play. If one gear turns it doesn't immediately make the next one turn as well. It first has to take up some space in between the teeth of the two gears before they engage and the motion is passed. This happens between every one of these little gears before you will watch your scope slew. This phenomenon is what we call "backlash". What training does is telling the Autostar controller how long the drive motor in an axis has to spin to take up the slack in the entire geartrain. Once it has spinned the motors long enough it knows the scope will now slew in the direction you want it to move in. If your ETX is undertrained the Autostar will think the scope is moving when really the gears are still taking up the slack. If it is overtrained the scope will be slewing while Autostar thinks it is still taking up gear slack. Both of these errors make your GOTO's miserable or at least very inaccurate!

Created by AccuSoft Corp.

How does the Autostar know how much it moved? Well, in the drives there are little things called "encoders". Small toothed wheels which turn in front of a LED (Light Emitting Diode). Every time it senses the light is blocked by a tooth, it counts a tick. The amount of ticks it counts tell it how far the gears have spun.


So will training take care of rubber banding and bad GOTO's? Yes, if you do it correctly. Sure, if you centered the object you trained on every time your GOTO's will improve greatly. However, I think a lot of ETX'ers don't realize it is very easy to let Autostar take up the slack twice when training, although only once is required! Your GOTO's may be acceptable but I'm sure they can be even better!



This is what happens when you train your ETX's drives. This applies to both R.A. and DEC training, I'll use DEC training in the example:


Autostar asks you to center the reference object. You do so and press ENTER. Now the scope will slew down a certain amount and slightly back up. Next it tells you to slew up using the key on the keypad, until the reference object is again centered in the eyepiece. What it did is slewing your telescope to a position below the object, reversing the slewing direction (thereby taking up the slack in the gear) and have you tell it through the Autostar how long to spin the gears to re-center the object. What you really told it is how many encoder ticks it needs to count to be back at the reference object. It already knows how many ticks it slewed the scope down and back up, all it needs to know is how many ticks are needed to bring it back to the original position. It will do the math for you and determine the amount of slack to take up the next time you tell it to GOTO an object. That is what training does!



So what can go wrong, you ask?


If you centered your reference object by slewing your scope UP you have introduced a rather large margin of error in your training. Since the scope will slew down first in the DEC training process it will do the following:


-       Take up the gear slack (because you moved it UP, remember?);

-       Slew down;

-       Take up the gear slack again (because it has to reverse the direction of slewing);

-       Slew it slightly up.


Now you re-center the your training object and you're happy! Not really. The gears encountered slack twice! And you had the Autostar calculate how much slack to take up once!



How to counter this?


Before starting the training process, always center the reference object by slewing DOWN. The gears will then only have to take the slack up once and make your training more accurate then ever. When training the R.A. axis always center by slewing LEFT. Be careful here, the scope will move counterclockwise but the image in your eyepiece will move in the opposite direction. Knowing this, the worst thing you could possibly do when you start your training is centering the object by slewing up and to the right. It would then would take up the slack twice in both the R.A. and the DEC axis! In a nutshell, here's what to do:


-       Center the reference object by first pointing the scope higher than the object;

-       Slew down until the object is centered. Do not overshoot, you don't want to reverse the direction you slewed in. If you do overshoot, start over;

-       Scroll to "Train Drives" and pick declination;

-       Since the reference object is already centered you can press ENTER again;

-       The ETX will do it's thing, then you re-center with the keypad;

-       Press ENTER and the ETX will now slew up, slightly down (yes, to take up the slack) and have you re-center again;

-       When centered press ENTER and you're done training your DEC drive!


Repeat this process for R.A. training, using the techniques mentioned in this article.



Other thoughts on training


-       Train the drives in an "as you use them" attitude. I always use polar mode. For training I put my ETX in the home position and rotate the tube 180° in right ascension. Try to use a high object like a lamp post or chimney to train on. I prefer to set my declination to 0°, other techniques should work as well;

-       Use the highest magnification you can, preferably with a crosshair or reticle eyepiece. If you don't have one, "center" the object near the edge of the FOV and attempt to return to that position as accurately as you can;

-       Use the slowest speed possible ( [1] ) to make it easier for the Autostar to count encoder ticks. Be sure to stop slewing when the object is exactly centered again;

-       Remember to not use your R.A. and DEC percentages: I set them to 00% (zero) before training to take out any interference it may cause. Re-enter your values after training;

-       Before and after you train always (re-)calibrate the motors, this never hurts;

-       When you finish your training take a minute to test it on some landmarks. If you trained carefully you'll immediately see results;

-       Train with your accessories (dew cap/clip, piggyback adapter, etc.) attached. It'll probably only make a minor –if any- difference but we're looking for excellent scope performance, aren't we !!

Start of today's update
Subject:	Your thoughts please ...
Sent:	Thursday, June 18, 2009 11:57:03
From:	John Batiste
I have reviewed the ETX DRive Training in this posting from your Web
Site:  [this page]

The article is intriguing, but there is another algorithm that may come
into play.

Assume that the backlash is identical whether you are turning the worm
gears clockwise or counter clockwise.

If this is true, the Meade documented drive training is correct in that
when the software/firmware caused azimuth movement from center
clockwise, and then a little bit counterclockwise, you have been set up
to complete the movement to center the target.  You are now in a
position to have the software/firmware move counterclockwise and a
little bit clockwise.  The user's completion of movement to center the
target now presents a total of backlash plus required centering travel,
i.e. if the mechanical backlash is identical in both directions, the
data is available to derive backlash.  Same would be true for

I realize that I run the risk of black box analysis without the benefit
of input from the device designers.

Any thought on this?




From:	richard seymour (
Actually, i'm going to answer by drawing upon Andrew Johansen's super-detailed
analysis of how Drive Training works in the 497 Autostar.

The initial "how i slewed to reach the target to start" does not matter.
(at least in v43Eg of the firmware)
Meade's algorithm does not look at the encoder at the -start- of the
Train Drive sequence.
You center the target.
They slew out for a fixed *time* (not distance)
They slew back for a smaller fixed *time*.
They ask you to center.  You tap enter when done.
Only *now* do they take their first encoder position reading.
Then it slews out for a fixed *time* in the same direction you were moving to center.
Therefore no backlash happens on this "outbound" leg.
Then they slew back a smaller fixed *time*.
They ask you to center.  You tap enter when done.
Now they take their second encoder position reading.
The backlash measurement is simply the difference between the first and second reading.
(scaled by the gear ratio to arcseconds).

Before Andrew's analysis, i had thought that they took -three- readings
and averaged the difference.  They don't.  They only take readings
"surrounding" the second slew out-and-back, and that's it.
So Yes, they assume the backlash is the same in both directions
(i know it's not on the Alt axis of my ETX90, due to worn bearings and gravity)

have fun


Thank you very much.  "Black Box" analysis can be fun.

This discussion leads then to another question.

If the worm gears are loaded at the target currently in view, and then
you GOTO another target that keeps the work gears loaded, backlash
compensation would not be required.  The new question: Is the firmware
smart enough to know whether or not backlash compensation is required?

An inerteresting question only, since the scope normally puts the GOTO
target either in the eyepiece field of view or in the view finder and,
that's good enough for me..
Your thoughts on this?


The Autostar only engages the anti-backlash if it has to -reverse-
a motor.  If the motors always spin in the same direction,
there is no anti-backlash compensation applied.

You can test this by setting the Alt Percentage fairly high (50% or more),
then set the slew speed to "1" (guide) or "2"
Now use the slew buttons to move Alt up and down.
If you tap "up" "up" "up" you will not hear the anti-backlash
If you tap "up" "up" "down", you should hear or see the anti-backlash kick in
at the start of the "down".

have fun

Go back to the Autostar Information page.

Go back to the ETX Home Page.

Copyright ©1999-2002, 2009 Michael L. Weasner /
Submittals Copyright © 2002, 2009 by the Submitter