Stupid 4024 freq divider question (shaft encoder resolution)

[Randy MacKenna]

Hi,
I have a motor with a diffential optical encoder (A and B outputs),
which is configured as a 1000 count-per-revolution device. I really
want this thing to be a 250 CPR device...so rather than replace the
encoder, I figure there must be an easy way to divide the pulses coming
out of the encoder.

It's been years since I've played around with this stuff...so is it
as simple as piping the A and B outputs of the encoder into two
separate 4024 ICs, and then picking off Q2 of each, as the new output?

Also, the encoder I have says it is a "line driver" output. Does
that mean it is 5V, but can sink more current than typical TTL? If so,
I think I'm okay if I just put my 4024s within a few inches of the
motor control, correct?

Last stupid question: If the output signal from one channel of the
encoder looks like this:

____|-|_________|-|__|-|__|-|________________|-|____|-|___|-|____|-|_____

Does the output of Q2 on the 4024 look like this?:

__________________________|-|____________________________________|-|_____

In other words, is Q2 only transmitting every 4th pulse? That's
exactly what I am looking for, since I want to be able to tell the
motor control software that the shaft is in the spot that it would be
in, if the encoder were 250 CPR instead of 1000 CPR.

Thanks for your patience & help.

-Randy

 

[Larry Brasfield]

Hi,
I have a motor with a diffential optical encoder (A and B outputs),
which is configured as a 1000 count-per-revolution device. I really
want this thing to be a 250 CPR device...so rather than replace the
encoder, I figure there must be an easy way to divide the pulses coming
out of the encoder.

I doubt that the encoder produces pulses. More likely, it
is a quadrature encoder which produces one of four output
states according to the shaft position modulo 4.

It's been years since I've played around with this stuff...so is it
as simple as piping the A and B outputs of the encoder into two
separate 4024 ICs, and then picking off Q2 of each, as the new output?

If you intend only to use the encoder as a tachometer,
it could be that simple. But the fact that you mention
use of both outputs leads me to believe you want to
read shaft position, not just velocity. In that case, you
have a more complex task ahead of you. You need a
state machine to convert the input 2-bit grey code into
a divided-by-4 output 2-bit grey code. In other words,
you will produce another quadrature encoding that has
1/4 as many output transitions but preserves position
within the limitation imposed by the lower resolution.

Also, the encoder I have says it is a "line driver" output. Does
that mean it is 5V, but can sink more current than typical TTL? If so,
I think I'm okay if I just put my 4024s within a few inches of the
motor control, correct?

Last stupid question: If the output signal from one channel of the
encoder looks like this:

____|-|_________|-|__|-|__|-|________________|-|____|-|___|-|____|-|_____

Does the output of Q2 on the 4024 look like this?:

__________________________|-|____________________________________|-|_____

In other words, is Q2 only transmitting every 4th pulse?

No, the counter will increase its count output
at every input pulse, (specifically at the leading
edge of the input pulse). This means, if you use
Q2 as the output, it will toggle at every other
input pulse.

That's
exactly what I am looking for, since I want to be able to tell the
motor control software that the shaft is in the spot that it would be
in, if the encoder were 250 CPR instead of 1000 CPR.

If, with that language, you reveal that you are trying
to control position rather than speed, then a simple
divider simply will not work.

Thanks for your patience & help.

Good luck.

 

[Randy MacKenna]

Okay, thanks...I thought that the A / B channels of the encoder were
more or less independent, and it was up to the motor driver hardware to
integrate them. I figured I could just put a divider on each channel,
and 'trick' the motor driver into seeing every 4th pulse on both the A
and B channel.

I just looked at the datasheet for the encoder... the A and B channels
sure look like square wave pulses to me. The B channel seems to lag
the A channel by about 25% of the pulse width. The spec says "90
degree quadrature".

Not meaning to be a pest, but is there any hope here for a simple
circuit, or do I really have to replace this encoder?

-Randy

 

[Larry Brasfield]

Okay, thanks...I thought that the A / B channels of the encoder were
more or less independent, and it was up to the motor driver hardware to
integrate them. I figured I could just put a divider on each channel,
and 'trick' the motor driver into seeing every 4th pulse on both the A
and B channel.

Maybe. But that would depend on a factor I mentioned
in my previous post and which still remains a mystery.

I just looked at the datasheet for the encoder... the A and B channels
sure look like square wave pulses to me. The B channel seems to lag
the A channel by about 25% of the pulse width. The spec says "90
degree quadrature".

Clearly, it is a conventional quadrature encoder.

Not meaning to be a pest, but is there any hope here for a simple
circuit, or do I really have to replace this encoder?

I cannot tell, yet.

You caught me in a good mood today, so I'll play
the 20 question games. To save some time, I'll pose
a few at a time:

1. Does your application have to position the motor?

2. Does the motor direction ever have to change?

 

[Michael Black]

Okay, thanks...I thought that the A / B channels of the encoder were
more or less independent, and it was up to the motor driver hardware to
integrate them. I figured I could just put a divider on each channel,
and 'trick' the motor driver into seeing every 4th pulse on both the A
and B channel.

I just looked at the datasheet for the encoder... the A and B channels
sure look like square wave pulses to me. The B channel seems to lag
the A channel by about 25% of the pulse width. The spec says "90
degree quadrature".

But that's the whole point of the encoder. If it just put out
single pulses, you'd have no clue which way the shaft was turning.
But having two outputs, you can extract direction. If X leads Y,
the shaft is turning clockwise. If Y leads X, the shaft is turning
counter-clockwise. The number of pulses is how far the shaft is
turned.

A classic example is setting a digitally tuned radio. The encoder
would feed an up/down counter, with one the pulses counted by
the clock input, and the direction telling the counter whether
it's adding or subtracting the counts. Turn clockwise, and it
increments the counter each time there is a click. Turn counterclockwise,
and it decrements the counter each time there is a click.

Michael

 

[Randy MacKenna]

1. Yes. 2. Yes.

Actually, "I" don't want to try to set position and rotation of the
motor -- I already have motor driver hardware and software that does
this -- I just want to make the motor advance faster.

Here's an example. The motor is currently at zero degrees. I want to
move and stop the motor at 360 degrees - one full rotation. Today, I'd
have to pulse the driver 4000 times to effect that amount of rotation
(1000 CPR encoder, quadrature = 4000 pulses per revolution). Instead,
I want to pulse the driver only 1000 times.

Therefore, my thinking is that if I can "throw away" all but every 4th
returned pulse out of both the A and the B channel, then the driving
hardware (the stuff with the real 'brains') will only have to be pulsed
1000 times to effect one full rotation. I will tell the driver that it
is connected to a 250 CPR encoder (I set this via software), thus when
it sees every fourth returned pulse from the A and B channel, it will
say "Ahh, my motor was just advanced by 1.44 degrees", an it will do
the math the correct way and "land" the motor at the right spot.

One other way to do this is to replace the optical disk in the encoder
with a different disk of lower resolution. That, in effect, is a
'mechanical' way of dividing the A and B outputs by 4. I can't buy a
replacement disk for this encoder...thus my desire to solve this
digitally.

I found this, which I could go out and buy (I'd need 4 of them). I'd
like to avoid that cost, and at the same time solve this puzzle for my
own education.

http://www.usdigital.com/products/edivide/index.shtml

I guess that this thing could be complex enough to contain a state
machine and be doing the analysis of both A and B at the same time, and
synthesizing new A and B outputs, based on the division factor
specified by the DIP switches.

If you think that is indeed the case, then I am in way over my head,
and I just have to spring for the $160 to buy four of these units for
my four motors...

-Randy

 

[Rob Gaddi]

Okay, thanks...I thought that the A / B channels of the encoder were
more or less independent, and it was up to the motor driver hardware to
integrate them. I figured I could just put a divider on each channel,
and 'trick' the motor driver into seeing every 4th pulse on both the A
and B channel.

I just looked at the datasheet for the encoder... the A and B channels
sure look like square wave pulses to me. The B channel seems to lag
the A channel by about 25% of the pulse width. The spec says "90
degree quadrature".

Not meaning to be a pest, but is there any hope here for a simple
circuit, or do I really have to replace this encoder?

-Randy

Of course there's a simple circuit, a micro. Now all the complexity and
headache is in software instead.

That said, how fast of a motor are you talking about? Are these 1000
and/or 250 pulses going to be coming once an hour? Minute? Second?
Millisecond?

 

[Tim Wescott]

Hi,
I have a motor with a diffential optical encoder (A and B outputs),
which is configured as a 1000 count-per-revolution device. I really
want this thing to be a 250 CPR device...so rather than replace the
encoder, I figure there must be an easy way to divide the pulses coming
out of the encoder.

It's been years since I've played around with this stuff...so is it
as simple as piping the A and B outputs of the encoder into two
separate 4024 ICs, and then picking off Q2 of each, as the new output?

Also, the encoder I have says it is a "line driver" output. Does
that mean it is 5V, but can sink more current than typical TTL? If so,
I think I'm okay if I just put my 4024s within a few inches of the
motor control, correct?

Last stupid question: If the output signal from one channel of the
encoder looks like this:

____|-|_________|-|__|-|__|-|________________|-|____|-|___|-|____|-|_____

Does the output of Q2 on the 4024 look like this?:

__________________________|-|____________________________________|-|_____

In other words, is Q2 only transmitting every 4th pulse? That's
exactly what I am looking for, since I want to be able to tell the
motor control software that the shaft is in the spot that it would be
in, if the encoder were 250 CPR instead of 1000 CPR.

Thanks for your patience & help.

-Randy

To sum up:

Randy, you have a quadrature encoder, but every time someone mentions
"quadrature" you seem to read it as "". Just dividing the two channels
will _not_ get you where you want to go, you need direction information
as well.

Go do a search on "quadrature encoder" (NOT "encoder"!). Learn how they
work. Then you'll know over half of your answer. With a PLD, a PIC, or
a few 7400 series parts you can do this -- but first you have to
understand how a quadrature encoder works.

 

[Larry Brasfield]

[In answer to:
1. Does your application have to position the motor?
2. Does the motor direction ever have to change?
]

1. Yes. 2. Yes.

Actually, "I" don't want to try to set position and rotation of the
motor -- I already have motor driver hardware and software that does
this -- I just want to make the motor advance faster.

That's why I referred to "your application".

Here's an example. The motor is currently at zero degrees. I want to
move and stop the motor at 360 degrees - one full rotation.

The controller will have to recover correct position
information from the encoder subsystem, then. For
example, many positioning systems will overshoot
the setpoint a little bit, then correct for that. And to
stay in one position, they often have to hunt around
the setpoint, jogging back and forth a little bit.

Your simple divider simply cannot retain position
correctly in the face of direction reversals.

Today, I'd
have to pulse the driver 4000 times to effect that amount of rotation
(1000 CPR encoder, quadrature = 4000 pulses per revolution). Instead,
I want to pulse the driver only 1000 times.

That's good thinking, as far as it goes. Be aware,
however, that by decreasing the position sensing
resolution, the hunting is likely to be worsened.
(This may not happen for some controllers and
some motor/hardware configurations.)

Therefore, my thinking is that if I can "throw away" all but every 4th
returned pulse out of both the A and the B channel, then the driving
hardware (the stuff with the real 'brains') will only have to be pulsed
1000 times to effect one full rotation. I will tell the driver that it
is connected to a 250 CPR encoder (I set this via software), thus when
it sees every fourth returned pulse from the A and B channel, it will
say "Ahh, my motor was just advanced by 1.44 degrees", an it will do
the math the correct way and "land" the motor at the right spot.

The problem is that, unless you know the direction
is always the same, you cannot just throw away
3 of every 4 position transitions. Some of them
may be ones that represented a new direction.
Without a direction sensitive discard, and some
state to track where the shaft is within the hidden
set of positions, the correct ones cannot be "thrown
away".

One other way to do this is to replace the optical disk in the encoder
with a different disk of lower resolution. That, in effect, is a
'mechanical' way of dividing the A and B outputs by 4. I can't buy a
replacement disk for this encoder...thus my desire to solve this
digitally.

I'm sure that a simple PLD, with no more than 8
registers, could do this job provided you can live
with a slight loss of transition time resolution. The
loss can be kept negligible for ordinary positioning
applications. It could be clocked continuously and
produce the right output for any input speed up to a
maximum of one input transition per clock. Let me
know if this solution looks attractive and I will take
a little more time to describe it in more detail.

I found this, which I could go out and buy (I'd need 4 of them). I'd
like to avoid that cost, and at the same time solve this puzzle for my
own education.

http://www.usdigital.com/products/edivide/index.shtml

I guess that this thing could be complex enough to contain a state
machine and be doing the analysis of both A and B at the same time, and
synthesizing new A and B outputs, based on the division factor
specified by the DIP switches.

If you think that is indeed the case, then I am in way over my head,

I think you are in deeper than your first realized.
But the problem is not really that difficult.

and I just have to spring for the $160 to buy four of these units for
my four motors...

I would do the $0.80 PLD, in a heartbeat.

 

[Jim Thompson]

To sum up:

Randy, you have a quadrature encoder, but every time someone mentions
"quadrature" you seem to read it as "". Just dividing the two channels
will _not_ get you where you want to go, you need direction information
as well.

Go do a search on "quadrature encoder" (NOT "encoder"!). Learn how they
work. Then you'll know over half of your answer. With a PLD, a PIC, or
a few 7400 series parts you can do this -- but first you have to
understand how a quadrature encoder works.

Somewhere in my bag of tricks I have a synchronous divide-by-two that
might do what the OP wants. I'll search for it.

...Jim Thompson

 

[Randy MacKenna]

Tim, I wasn't ignoring 'quadrature' out of arrogance...I didn't
understand what the implication was. I will research, as you
suggested, and better educate myself about this technology.

As I stated, I am looking to save money (I'm a hobbiest, trying to
build a woodworking CNC router) -- and at the same time, I'm trying to
learn as much about every aspect of my project as possible.

I do appreciate you guys taking the time to help educate me.

Thanks,
Randy

 

[petrus bitbyter]

1. Yes. 2. Yes.

Actually, "I" don't want to try to set position and rotation of the
motor -- I already have motor driver hardware and software that does
this -- I just want to make the motor advance faster.

Here's an example. The motor is currently at zero degrees. I want to
move and stop the motor at 360 degrees - one full rotation. Today, I'd
have to pulse the driver 4000 times to effect that amount of rotation
(1000 CPR encoder, quadrature = 4000 pulses per revolution). Instead,
I want to pulse the driver only 1000 times.

Therefore, my thinking is that if I can "throw away" all but every 4th
returned pulse out of both the A and the B channel, then the driving
hardware (the stuff with the real 'brains') will only have to be pulsed
1000 times to effect one full rotation. I will tell the driver that it
is connected to a 250 CPR encoder (I set this via software), thus when
it sees every fourth returned pulse from the A and B channel, it will
say "Ahh, my motor was just advanced by 1.44 degrees", an it will do
the math the correct way and "land" the motor at the right spot.

One other way to do this is to replace the optical disk in the encoder
with a different disk of lower resolution. That, in effect, is a
'mechanical' way of dividing the A and B outputs by 4. I can't buy a
replacement disk for this encoder...thus my desire to solve this
digitally.

I found this, which I could go out and buy (I'd need 4 of them). I'd
like to avoid that cost, and at the same time solve this puzzle for my
own education.

http://www.usdigital.com/products/edivide/index.shtml

I guess that this thing could be complex enough to contain a state
machine and be doing the analysis of both A and B at the same time, and
synthesizing new A and B outputs, based on the division factor
specified by the DIP switches.

If you think that is indeed the case, then I am in way over my head,
and I just have to spring for the $160 to buy four of these units for
my four motors...

-Randy

Randy,

Assuming you have nice digital input signals, lets call them A en B, and you
want to make new ones, lets call them X and Y, you can try the following
quick and dirty circuit:

A four bits positive edge triggered up/down counter is the heart of the
circuit. Connect A to the clock input and B to the up/down input. The ripple
output of the counter can be used to toggle a flipflop giving you the X
signal.

The Y signal is the problem. You need to decode 0011 of the counter(two
inverters and an AND.) This signal has to clock a D-type flipflop with B XOR
X on its data-input, giving the Y on the output.

petrus bitbyter

 

[Randy MacKenna]

The problem is that, unless you know the direction

is always the same, you cannot just throw away
3 of every 4 position transitions. Some of them
may be ones that represented a new direction.
Without a direction sensitive discard, and some
state to track where the shaft is within the hidden
set of positions, the correct ones cannot be "thrown
away".

Ahhh...now I get it...thanks!

I'm sure that a simple PLD, with no more than 8
registers, could do this job provided you can live
with a slight loss of transition time resolution. The
loss can be kept negligible for ordinary positioning
applications. It could be clocked continuously and
produce the right output for any input speed up to a
maximum of one input transition per clock. Let me
know if this solution looks attractive and I will take
a little more time to describe it in more detail.

That would be very nice. I suppose that if I go down this route, I'd
have to buy some sort of FPGA programmer (like Digikey sells?) and
learn how to code the solution. A long time ago, I was a grad student
at Syracuse, writing VHDL and doing other digital logic coursework.
Life took a different turn, so I've long forgotten most of it. I do
remember how much fun it was setting up the system clock, mapping out
state transistions, and reducing the problem to a circuit...so if
there's an outside chance I can actually figure this out, that'd be
nice.

All I need to know is the path I need to take, e.g.:

1. learn how a quadrature encoder really works
2. learn how to program a PLD
3. figure out the inputs/outputs and state transistions
4. design the circuit
5. code the PLD

I understand that there's a loss of resolution here (which buys the
added speed)...but we are talking about way below the spec of the
woodworking machine I'm trying to build. The servo driver is tunable,
so using an o-scope, I ought to be able to get the motor to be
critically damped. All I need to do is to decide to go down one of
these paths:

a. sell the encoders I have and buy new ones
b. add the $40 divider box to each encoder circuit
c. sell the drivers I have and buy new ones that have pulse
multipliers onboard
d. explore building my own encoder frequency division circuit.

I'd like to explore 'd' a little bit more, before giving up...if 'd' is
a pipe-dream (for me), then my next viable option is 'c'.

-Randy

 

[John Fields]

In other words, is Q2 only transmitting every 4th pulse? That's
exactly what I am looking for, since I want to be able to tell the
motor control software that the shaft is in the spot that it would be
in, if the encoder were 250 CPR instead of 1000 CPR.

 

[Mac]

Ahhh...now I get it...thanks!



That would be very nice. I suppose that if I go down this route, I'd
have to buy some sort of FPGA programmer (like Digikey sells?) and
learn how to code the solution. A long time ago, I was a grad student
at Syracuse, writing VHDL and doing other digital logic coursework.
Life took a different turn, so I've long forgotten most of it. I do
remember how much fun it was setting up the system clock, mapping out
state transistions, and reducing the problem to a circuit...so if
there's an outside chance I can actually figure this out, that'd be
nice.

All I need to know is the path I need to take, e.g.:

1. learn how a quadrature encoder really works
2. learn how to program a PLD
3. figure out the inputs/outputs and state transistions
4. design the circuit
5. code the PLD

I understand that there's a loss of resolution here (which buys the
added speed)...but we are talking about way below the spec of the
woodworking machine I'm trying to build. The servo driver is tunable,
so using an o-scope, I ought to be able to get the motor to be
critically damped. All I need to do is to decide to go down one of
these paths:

a. sell the encoders I have and buy new ones
b. add the $40 divider box to each encoder circuit
c. sell the drivers I have and buy new ones that have pulse
multipliers onboard
d. explore building my own encoder frequency division circuit.

I'd like to explore 'd' a little bit more, before giving up...if 'd' is
a pipe-dream (for me), then my next viable option is 'c'.

-Randy

The quadrature encoder is not that hard to understand. Depending on which
way the shaft is rotating, either A leads, or B leads. If A leads, let's
say that is clockwise. Then in order to count clockwise position
increments, you connect A to the up/down input of a counter, and connect B
to the clock input of the counter. So if A is high on the positive edge of
B, the counter counts up, because you are going clockwise. But if A is low
on the rising edge of B, then the counter counts down, because you are
moving counter clockwise. Does that make sense?

So if you have a clocked PLD, you could build your circuit, I think.

Let's say you have three inputs, A, B, and CLK. You have two outputs,
A_OUT, and B_OUT. You rig up A and B to the four-bit counter, just as
I outlined in the first paragraph. Then you build a clocked circuit
which watches the count. When it sees the count transition from 3 to
0, the watching circuit sends out phased pulses on A_OUT and B_OUT such
that A_OUT leads B_OUT. On the other hand, when the count transitions from
a count of 0 to a count of 3, the watching circuit sends out phased
pulses such that B_OUT leads A_OUT.

You just have to run the clock fast enough so that you don't miss any
transitions on the counter. And you have to make sure that your output
pulses meet whatever timing the subsequent circuitry expects.

There might be an even slicker way to do this to make it behave more like
a real quadrature encoder, but it might not be necessary.

Good luck, and don't give up!

--Mac

 

[Rich Grise]

Okay, thanks...I thought that the A / B channels of the encoder were
more or less independent, and it was up to the motor driver hardware to
integrate them. I figured I could just put a divider on each channel,
and 'trick' the motor driver into seeing every 4th pulse on both the A
and B channel.

I just looked at the datasheet for the encoder... the A and B channels
sure look like square wave pulses to me. The B channel seems to lag
the A channel by about 25% of the pulse width. The spec says "90
degree quadrature".

Not meaning to be a pest, but is there any hope here for a simple
circuit, or do I really have to replace this encoder?

It's trivial. Decode all 1000 positions, and shift right two bits.

Cheers!
Rich

 

[Rich Grise]

1. Yes. 2. Yes.

Actually, "I" don't want to try to set position and rotation of the
motor -- I already have motor driver hardware and software that does
this -- I just want to make the motor advance faster.

Here's an example. The motor is currently at zero degrees. I want to
move and stop the motor at 360 degrees - one full rotation. Today, I'd
have to pulse the driver 4000 times to effect that amount of rotation
(1000 CPR encoder, quadrature = 4000 pulses per revolution). Instead,
I want to pulse the driver only 1000 times.

What if you just give it bigger pulses, such that it moves 4 counts instead
of just the one?

Good Luck!
Rich

 

[Randy MacKenna]

Thanks Rich...yes, that would work (and would be the easiest solution
of all!). It wouldn't be bigger pulses, but rather multiple pulses on
the input side of all this hardware.

The problem is that this whole thing is being driven by the parallel
port (yuck!) of a PC, with the software that talks to the parallel port
being limited to a maximum pulse rate of about 45kHz.

Next, my motor through a timing belt and gear (for torque), and again
through a rotating screw & nut assembly (for linear motion) is limited
to a crawling pace of 50 inches per minute. For my application, I'd
like to move at 200 IPM.

Thus my dilemma

-Randy

 

[Tim Wescott]

Tim, I wasn't ignoring 'quadrature' out of arrogance...I didn't
understand what the implication was. I will research, as you
suggested, and better educate myself about this technology.

As I stated, I am looking to save money (I'm a hobbiest, trying to
build a woodworking CNC router) -- and at the same time, I'm trying to
learn as much about every aspect of my project as possible.

I do appreciate you guys taking the time to help educate me.

Thanks,
Randy

I didn't assume it was arrogance -- sometimes a term is superfluous, or
gives a nuance that might get you that last 2% of the understanding when
you only need 90%, so ignoring it is a good thing. As an example, when
I talk to my lawyer I often interpret as many as half of his words as ""
to retain my sanity.

In this case however it means everything, which is why I was stressing it.

 

[Randy MacKenna]

Gosh, yes...this is starting to make sense. I think the hard part is
going to be the state transition where you have to generate A_OUT
leading B_OUT (or vise-versa), and do it exactly 90 degrees
phase-shifted. It'll have to be that way, since the motor driver is
expecting to see a nice, clean quadrature-style signal.

The clock rates on all of this are measured in the kHz, not mHz, so no
worries about speed in this circuit.

I'm burned out tonight (long week at work), and I'm now working on the
fiscal books for my daughter's school (I'm the volunteer bookkeeper

So, I'll be checking back here on Monday...and not giving up!

Thanks to everyone, have a great weekend.

-Randy