A prime
component in GOTO mounts and digital setting circles is the optical
encoder. There are two types - incremental and absolute. You will need
incremental encoders. In my case I have two unknown incremental
encoders attached to the back of the motors I plan to use. A better
option is to separate the encoders from the motors. Encoders work by
shining light - quite often infra red - through a disk with slots in it
with a light sensor on the other side. When the slot is between the
light source and the sensor the encoder puts out a logical high. When
the light is blocked by the disk, it puts out a logical 0. So
as the disk turns the output moves from 0 to 1 and back again in a
square wave.
You can then measure the speed of rotation or the amount of rotation by
counting pulses. If you count the ones or zeros - also called level triggered -
then you are doing x1 decoding. If you count the transitions from 0 to
1 and 1 to 0 - also called edge triggered - you are doing x2 decoding.
The problem with this is that while you can measure rotation by
counting the pulses, you don't know which direction it is going in. In
addition your precision is limited by how accurately you can create the
disk and the limitations of light leakage. So they add a second set of
sensors and light sources slightly offset from the others. This gives
you two square-waves on what are called channel A and channel B.
If A leads B in pattern then it is going in one direction and
if B leads A it is going in the other direction. In addition the second
channel increases the highest possible precision to x4.
Finally in some cases they are worried that the electronics counting
these things will get out of step . They add a third set of light
sources and sensors. These are generally toward the center, solid area
of the disk. There is in this case a single hole. This produces a pulse
once per revolution and is called the Z channel or the index channel.
My encoders have the hole in the disk but there don't seem to be any
wires to support the z channel.
Identifying the direction
You tell the direction of rotation by comparing the current values on
the two channels with the previous state.
|
|
Now |
|
|
01 |
11 |
10 |
00 |
| Previous |
01 |
S |
1 |
E |
-1 |
| 11 |
-1 |
S |
1 |
E |
| 10 |
E |
-1 |
S |
1 |
| 00 |
1 |
E |
-1 |
S |
In this table, S stands for the same or no change in state, E stands
for error ie it is theoretically impossible to get here, and the 1 and
-1 are the two directions.
One way of programming this is to create a table and use the current
value and the previous state as a 4 bit offset into the table.
Encoders and the Picaxe
Originally I had intended to decode the encoders in software on the
picaxe. This was until I read with interest the admiration of the hot
picaxe programmers on the picaxe forum for someone who had managed to
decode around 300Hz. I estimated that my encoders would run at about
13KHz. I began looking for a hardware solution.
Estimating decoding frequency.
You can estimate the frequency by simply multiplying it out. I plan to
run my motors on 12v and according to their datasheet they will run at
2000rpm at 12v. I counted the pulses from the encoders and got 80 per
revolution. I know this is probably inaccurate so say it is 100.
Using x4 decoding gives me 400 events per revolution that I
need to decode. So the frequency is:
f= 2000 x 400 / 60
= 13,333Hz
This is all fairly rough so I will need to allow plenty of margin for
error.
Hardware Decoders
You can build a decoder out of hardware. I found a logic diagram and looked at what it would take to build it. When I had identified 20 to 30 discrete chips for two encoders, I gave up. BTW the
logic design comes from
fpga4fun.
There are three sets of ICs that will do this job for you. There are no
doubt others but I found only three. The UK firm
Genapta
make some ICs that will do everything including the counting.
They appeared to be the most expensive and hard to get. LSI in the US
also make
encoder
interface
ICs. However they are designed to interface with either
counter ICs or microcontrollers. So they would need more chips than the
other two options. Having said that even with the extra chips they are
probably the cheapest option. They also seem to be very common. The
third is a
set
of chips made by Avago that also do all the work including
counting. These are expensive but not as expensive as the Genapta
offering. They are also in common use among the robotics fraternity.
Depending on where you are will depend on what is best for you. I could
neither afford nor get the Genapta chips. I had trouble getting the LSI
chips and found only one source for the Avago chips locally.
Considering my
skill level and my wounds from the uLCD experience I paid the $AUS32
for an Avago HCTL-2032. This will be the most expensive chip in the
whole unit.
