[Zig]

I have a requirement for a higher pulse rate from existing and immatable encoder.

I have seen some gear on the web where the vendor offers a rate multiplier box which accepts quadrature signals from an encoder and outputs the signal at a higher rate.
The phase is maintained through the multiplier process.

To be more precise consider a 10 Khz output signal form a quadature encoder where signal a leads signal b.
This signal is fed into the multiplier and utcomes say 100 KHz signal where signala is still in quadrature with respect to signal b and it is still leading it.

I was wondering if there were any open source solutions to this problem out there in the wild?

[mike_Kilroy]

to increase resolution you need to start with a sin/cos quadrature encoder; you cannot increase the resolution of a 5v TTL square wave unit - no info there to do so. 'so before you put a lot of time into this search for the divider circuit, do you have 5v square wave output encoder or 1v sinewave output?

if u have sine encoder, then try googling "encoder divider" to find simple available chips/circuits to do it.

[Al_The_Man]

What kind of application is this and what rpm of the encoder does this represent?
Al.

[Zig]

If this were a rotary shaft encoder then there would be no problem; it would be readily replaceable.

This is a linear encoder.
The output from the encoder is used in side two loops: a velocity loopa nd a positional loop.

Positional loop has no problems.

Encoder pulses are fed into a tacho circuit which generates a voltage to be further used in the velocity loop.

Tacho circuit calls for a higher pulse rate to gennerate "meaningfull" feedback signal.


A sine resolver is not much more different to a quadrature logic level signal.
Both convey same informatio that is direction of travel and distance.

[mike_Kilroy]

so again, what is the output signal format of your scale? ttl 5v square waves? 1Vpp sine waves? 11uA sine waves? absolute something like SSI? Endat? Other?

[Zig]

Encoder output is logic level ( 5 V TTL) quadrature signal.

[mike_Kilroy]

Encoder output is logic level ( 5 V TTL) quadrature signal.

zig, that means there is NO extra data to multiply with. what u got is what u got. period.

[Al_The_Man]

Is this some kind of proprietary controller or what is the purpose of a position loop, which you say is adequate and a velocity loop?
The information so far is rather vague.
"Meaningful" feedback is usually obtained from the positional feedback which you say is adequate?
For a more qualified answer, more info on the present set up is required.
Al.

[Zig]

Mike,

Normally I would agree with You however in this instance the velocity loop is not going to protest if a single pulse from the encoder is "lost" due to some signal processing to obtain a higher frequency quadrature signal.

Most servo amps ( of the analogue kind ) explicitly call for a tacho feedback.
Servo amps I intend to use generate tacho signal from the encoder quadrature signal.
With the linear encoder I have no way of mechanically gearing the encoder to achieve higher pulse rate. No gearox, no leadscrew pitch.

The tacho circuit in the servo amp generates 1 volt per 22KHz of encoder signal.

In setting up the servo the inner velocity loop is tuned and then the positional loop is adjusted.

This means that while the velocity profile may have some error the positional error is zero.
If I wanted velocity error to be zero i would need to nest the acceleration feedback loop within the velocity loop.

So for intents and purposes I need to synhesise a velocity only pulse train which only needs to be direction correct and step "good enough".

There are some commercial products which divide the pulse train in the correct fashion but none that I have come across which multiply it.

In summary no special problem other than the fact we are talikng linear scales on top of a linear motor.

[Al_The_Man]

Mike,

Most servo amps ( of the analogue kind ) explicitly call for a tacho feedback.
Servo amps I intend to use generate tacho signal from the encoder quadrature signal.
.

I predominantly use analogue Servo amps and do not use any feedback to the drive of any kind, tach or encoder?
Both DC brushed and BLDC type drives.
The drives are configured for the Torque mode of operation.
Al.

[Zig]

Al,

I had tried current drive ( torque ) and voltage configurations.
My requirement is complicated by the fact that I am using linear motors with iron core and subsequent nonlinear thrust ( torque ) issues.
Veklocity feedback seems to minimise cogging nature of thrust.
Ploe distance is in the order of 30 mm whereas the scale resolves down to 5micron.

In any case you do use an encoder which provides positional feedback to the control loop.

[mike_Kilroy]

Mike,

Normally I would agree with You however in this instance the velocity loop is not going to protest if a single pulse from the encoder is "lost" due to some signal processing to obtain a higher frequency quadrature signal.

In summary no special problem other than the fact we are talikng linear scales on top of a linear motor.

Zig, you miss that THERE IS NOTHING TO MULTIPLY here..... yes, you can divide so that you make a new output string of pulse that say consist of 1 pulse every 10 of the original, but there is no physical way in the world to take 1 pulse and make a meaningful 10 out of it. none. period. physics. think of it more...... you have a pulse that is 1 pulse width wide...... how do you mutiply it to make it 10? I suppose if you really wanted to you make a circuit to say every time 1 pulse comes by make it into 10 at 10x skinnier width and then do it again for next pulse etc.... but think of it: there is no new data there so it is useless info. so how does this help you at slow speed or stop? if stopped you still need to wait for the next real pulse before you can make 10 more fake ones... what of slow speed? so say you have 1 pulsesec but you make it into 10 pulses..... so now you have 1pulse/sec of meaningful info but you spit out 10 at a frequency 10x higher until the next one..... that req has to be at the highest speed freq you will run, so now you get batches of say 10 pulses at 10khz, once every sec when a real pulse comes by - see what that will do to your velocity controller? blow its mind! it will try to run at 1okhz speed for short bursts every 1 sec. 9 of the 10 fake are just that - fake and carry no value or no new informatiojn for your velocity loop to use to allow more precise responce. the pulses have to have MEANING or they are of no value to alllow you to tune higher bandwidth. no one does what you ask because it physically gains no additionjal informationj so is useless.

[Al_The_Man]

Is there a possibility that the drives also have DC tach input? (The AMC drives I use do).
If so there may be a possibility to use something such as a LM2917 IC, Freq. to analogue convertor, this could be scaled to represent a suitable value for your feedback?
Al.

[mike_Kilroy]

Al, I imagine his controller thing does this internally already to take derivative of position pulses to make tach sig for the velocity loop; so unless he gets m ore pulses by replacing the scale with one that has say 10x more pulses per unit movement then making his own tach should result in the same amount of information as the present controller has, no?

Or if he makes a change of time between pulses converter instead of V/freq converter.... that will give more information from the time between pulses - if it is at least moving. We did an experiment with this kind of converter once where we had a 1024ppr encoder on a motor and using this instead of crude v/freq converter simulated the sun rising with a glued mirror on motor shaft rotating at 1 rev per day (0.0007rpm) and laser pointing at mirror then off 50' to wall with 1024 ppr circles (3.8" apart).... a bit jittery but you could watch it sorta smoothly moving at this speed up the wall between pulses of feedback data. 1/2 hour later and it was obviously higher.... some drives do this conversion but most don't I think.

[Al_The_Man]

Al, I imagine his controller thing does this internally already to take derivative of position pulses to make tach sig for the velocity loop; so unless he gets m ore pulses by replacing the scale with one that has say 10x more pulses per unit movement then making his own tach should result in the same amount of information as the present controller has, no?
.

I was thinking in terms of a outboard designed Freq to Analogue where in the case of the IC similar to the one mentioned, the output can be scaled, IOW 10Khz could be made to represent a desired DC tach voltage in order to scale the existing rate.
Al.

[Zig]

Al,

Quite correct.

I am trying to press AMC BE15A8s into service.
Sadly they do not have analogue tacho input.

I have had great joy with BE12A6 in a similar situation using linear scale as positional feedback and rotary shaft encoders with much higher effective line count on motorshafts driving the axis.

Mike if it is possible to succesffuly divide down a train of pulses and retain some semblance of information that is direction and speed then it is also possible to multiply the same information ( direction and speed) up into higher pulse rate.

The keyto this is that the tacho circuit ( i expect ) is not pulse duration sensitive since most probably internally it shapes the incoming pulse streams ( inphase and quadrature) and then converts the signal to analogue value.

The phase discriminator being a digital beast does not necesarily need to indicate phase magnitude but only /lead/lag relationship between the thwo signals.

This as I have said relies on the above expectation and i am about to build a prototype unit to try out the idea.

The circuit is capable of receiving the quadrature signals, extracting phase relationship between incoming siignals generating a pulse train at higher rate and and outputting a quadrature signal at 16x the rate. The output qudrature signal phase is monitored and corrected to match input phase.

Al this with a handfull of CMOS logic gates ( so far ).

Basic blocks .. phase discriminator... edge triggered one shot... quadrature generator with selectable phasing .... phase discriminator .. and input /output phase comparison logic and correction flag


Hope to have good outcome with this approach.

[mike_Kilroy]

let us know how you do.

you probably have seen pictures of your quadrature signals but in case not, here is a sample. You will see there is only FOUR (4) not 16 possible pulses you can get per cycle; that is why it is called quadrature. anyway, I know of no controller that does not allow this x4 already. so again, I think you will gain nothing new compared to your present controller.

if you want to tell us the model and name of the positioner with velocity loop inside we can probably tell you what you already have so you know what you are comparing to?

[mike_Kilroy]

I have seen some gear on the web where the vendor offers a rate multiplier box which accepts quadrature signals from an encoder and outputs the signal at a higher rate.
The phase is maintained through the multiplier process.

To be more precise consider a 10 Khz output signal form a quadature encoder where signal a leads signal b.
This signal is fed into the multiplier and utcomes say 100 KHz signal where signala is still in quadrature with respect to signal b and it is still leading it.

do you recall who advertised such a thing for a quadrature TTL encoder input? 10khz in, 100khz out. I would be mighty interested to see how they do the impossible.

[mike_Kilroy]

I am using linear motors with iron core and subsequent nonlinear thrust ( torque ) issues.
Veklocity feedback seems to minimise cogging nature of thrust.
Ploe distance is in the order of 30 mm whereas the scale resolves down to 5micron.

what brand & model of linear motor is it? I never heard of a linear motor with nonlinear thrust? I would like to learn about it.

[Al_The_Man]

. You will see there is only FOUR (4) not 16 possible pulses you can get per cycle; that is why it is called quadrature.

Mike, Actually that is a popular misconception, the term quadrature comes from the 90° phase difference between the A & B pulse, whether the pulses are multiplied by X1 X2 or X4 is up to the design of the receiving unit.
Even if used in the basic pulse count, it is still quadrature count.
Al.