COMMENT ⊗   VALID 00005 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002				Encoder Specifications for Hand-Eye Work
C00005 00003		For reasons of simplicity and economy, and because of the rapidly
C00008 00004	      	Here are some typical encoder specs.
C00009 00005		LITTON ENCODER INFO.
C00016 ENDMK
C⊗;
			Encoder Specifications for Hand-Eye Work
	
	This file outlines some  of  the  requirements  we  have  for
encoders  and their interfaces for some of the Hand-Eye project work.

	Up  till now we have been using Potentiometers as our general
purpose  resolving  devices.   For  increased  resolution,   improved
linearity,  and  better  stability, it appears desirable to switch to
optical encoders on many of our "joints".

	Two general types of encoders are available- "whole word" and
incremental".   Whole  word encoders have as 
many tracks as bits of 

output resolution and provide an n bit coded output of  the  position
of  the encoder. Incremental encoders have two output tracks, (plus a
zero reference mark sometimes).  They provide two output  signals  in
quadrature  which  are  electronically  decoded  to drive an external
up-down counter.   For  unidirectional  applications  some  of  these
incremental  encoders  merely  have  one  track and drive the counter
directly. Or else, a outside direction sensing switch changes the count direction
of the counter.



	For reasons of simplicity and economy, and because of the rapidly
decreasing cost of i.c.electronics, most general purpose encoders are
of the incremental type.  I am proposing to use these on most of the
new mechanical devices we design, and I am proposing to retrofit these
incremental encoders onto some of the already existing devices, starting
with the T.V. cameras.  

	The first device to be retrofitted will be the Sierra Camera because it
was designed to use encoders in the first place.  Then the Cohu.  Each camera will
use 2 encoders to start with.  They will be quadrature output devices with a
zero reference track.  4X count multiplication circuitry
will be necessary to multiply the count output.  Quadrature decoding will
also be required.  The count should be stored in hardware counters to eleiminate
the need to reset the computer count everytime the system dies (done by
servoing the camera past the zero reference mark). I suggest that 16 bit counters
be designed to accomodate all possible encoders.  These counters, and the related
logic must be stable and noise free to enable them to count error free for
periods of weeks or months.  Possibly a battery pack, standby unit can be used to 
provide power when the kludge bay power is off, otherwise the zeroing procedure will 
have too be repeated.
      	Here are some typical encoder specs.

	Number of lines- 1024 (gives 4096 counts per turn)
	Output-unamplified model- 
	Output-integral amplifier model-
	Max. expected count rate-50khz.

	LITTON ENCODER INFO.

This page  covers  some information  gleaned from  a  Mr. Rendler  of
Litton Industries, Encoder Div.

Model  71  and  Model  73.    Model  71  is  a  larger  version  with
phototransistor output.  Not differential.  Prices for 512  lines are
$135, 145,155 for one two or  three channels of amplified output.  No
cover is supplied. Model 73 is newer and smaller.  It fits into a 1.5
inch dia. package,  has a cover  (normally closed end, but  available
with thru shaft opening),  and differential transisitor output on all
channels.  Price  for 512  lines is $180,190,200  for 1,2,3  channels
output.  All square waves.  Both will take a
.5 inch dia shaft.

	What follows is some information on special encoders and also
on  encoder practice in  general.  They  charge about $400  to make a
master disk  and mask.   Typical  specifications  are the  following.
Concentricity of i.d. and  track + or- .005 to .010  Conc. of o.d. to
track  or i.d. + or- .005  to .010 Disk light  to dark width is 50-50
Mask light  to  dark width  is less  than 50-50,  dark  is >50%  (the
impression I get is that it is say like 40-60 or so. They will adjust
it to yeild a  roughly uniform sine wave  output. The standard  glass
thickness is about  .070-.080 for disks.   For low inertia  they make
the disks .020 thick.  The masks are .010 thick glass.  Litton grinds
their own glass, and will make it flat enough to suit the resolution.
They  center  the disk  on  the  hub  optically  to about  .0003  in.
concentricity  with a microscope.   For high resolutions  they use an
electrical method  which has  two sensors  180 degs  apart and  alows
alignmeent to within 1/8 cycle.   But not necessary even on 2048 line
disk.  Typical track width  is about .040 but it  can be wider.   The
mask is about  .5 in wide and houses  all the sources and  sensors as
close  together  as  possible.   They  generally  use a  differential
method, which uses  two sources and  sensors per  channel.  Only  one
track is used for the disk.  Quadrature is obtained from two pairs of
channels  90 degs. out of  phase.  These channels  each have two mask
grids 180  degs apart.   The zero  ref. track  is also  differential.
They  use a  single outer  track which  is all  clear,  as a  sort of
reference.  An inner  track has all  light and a  small line at  zero
(masked).  This gives  a sort of differential output.   Their typical
layout  is an inner zero  ref. track, then the  count track, then the
zero calib. track(the  one with no marks).   The differential  layout
adapts  for  greater temp.  variations,  also  allows for  ageing  in
sources, sensors, voltage variations, etc.  Gap.  As a rule of thumb,
the gap whould be  about equal to or less  than the width of  a line.
Their masks have each sensor  placed about .10 apart and I think they
use the .062  dia. devices.   This places them  close together  where
they have the least quadrature variation. Litton also has an aligning
tool good for up to 1000 lines.  You slip it on the shaft and its two
pins engage the head holes. This centers everything without  the need
to  do  any  special  electrical  measurements.  Oh  yes,  for  glass
thickness, let them select things for us, the std. thickness may be a
variable, its around .060-.090.  
	They mount  the disks to  the hubs  with a slightly  flexible
epoxy to keep the  bond from cracking with temperature changes, which
would occur if a rigid epoxy were used.
	That about sums up my conversation of Jan. 20.  We will order
one of  their model 73 encoders with 512  cycles, and zero ref. incl.
all electronics. This seems to be a  better deal.  By the way,   they
say that the  phototransistor output on all channels  is connected to
the  circuit  board by  wires  and we  can unsolder  these  wires and
connect them to our  outside electronics if desired.   The collectors
are aparantly common.  Light source is LEDs for each sensor.  
	Yes they still use the  electron beam master maker, at a cost
of about $400 per  disk and mask.   They will  also supply chrome  on
glass disks as standard.