I think you might find that your glass scales go to the servo drive, which uses, somewhere in its logic a step and direction control. The reason I say this is, I have several boards that take step and direction from the PC and have encoder inputs, and analog outputs to the servo drive. 

Hi,
there are servo drives that can accept a glass scale input in addition to the normal rotary encoder....but they are not common and command a premium over servos that have a rotary encoder
alone. For example Delta A2 series servos:

https://www.fasttobuy.com/ecmac10807rsasda20721f-delta-220v-750w-239nm-3000rmin-80mm-dmcnet-ac-servo-motor-drive-kits-with-3m-cable_p27766.html

This servo will use the rotary encoder for velocity and torque loops but can close the position loop on a secondary encoder like a glass scale.

In earlier times with analogue servos it was more common to use linear scales to close the position loop, however those servos typically had a tacho generator
on the servo shaft that was feed back to the servo amplifier to stabilise the velocity and torque loops. Such servos are now rare and fading away into history.

There are motion controllers like the Hicon and CSMIO/A that can close the position loop using a linear scale, but note they produce an analog output voltage for the transconductance amp,
not Step/Dir.

There are servo drives like the Gecko G320 for instance that accept Step/Dir input, with encoder feedback, but to my knowledge must be fed with a rotary encoder, rather than a linear encoder.
Any dither associated with the difference between a tightly coupled rotary encoder and a rather more loosely coupled linear scale will absolutely screw the torque loop bandwidth, and
once that's happened all bets are off.

Craig

Any servo drive that accepts incremental quadrature input from rotary encoders can be run from an incremental quadrature linear scale, however the scales have lower effective resolution and are not coupled as tightly mechanically to the motor. Mechanical backlash will cause problems. Typically you will have to tune them to operate more slowly with lower P gain, more I gain, and if available a dead band in order to be stable. Anilam did this on some of their older controls.

Hi,

Typically you will have to tune them to operate more slowly with lower P gain, more I gain, and if available a dead band in order to be stable. Anilam did this on some of their older controls.

This is equivalent to reducing the bandwidth of the torque loop.

A modern AC servo has a torque loop bandwidth of about 5kHz, resulting in a velocity bandwidth of about 1kHz and a consequent position loop bandwidth of 200Hz.
The critical point is the you need the highest possible torque loop bandwidth, because that in turn results in the best possible velocity an position loop bandwidths,
with the position bandwidth being what we are interested in, where the 'rubber meets the road' so to speak.

Any backlash or non-linearity of the linear scale severely curtails the high frequency accuracy of the torque loop, say down from 5kHz to 1kHz. If I'm not mistaken those
Anilam systems had tacho-generators on the servo shaft? This was to stabilise the velocity loop and thereby retain sufficient bandwidth for the position loop.
So the torque loop would be say 1kHz, the stabilised velocity loop 500Hz for a position loop of 100Hz. A 100Hz position loop was considered pretty damned good
for many years, and would still be entirely adequate on any hobby machine today.

Today with the likes of the A2 servo you can have high bandwidth torque and velocity loops by virtue of the high lineraity and resolution of the rotary encoder BUT
also the ability to 'load sense' with a linear scale for example. Best of both worlds. All the top servo manufacturers seem to have models that have the dual sensing feature,
but they all command a premium.

I buy Delta B2's, a 750W example costs me $438USD plus shipping. a 750W A2 (basic L version) from the same supplier is $682, so a $240 premium.

I am still only scratching the surface of what the B2 offers without paying for even more that I cant use!

Craig