The existing system actually has two separate loops - velocity signals from tachometers in the motors, which feed back into the existing Bosch servo power amps, and position encoders which feed back into the existing Heidenhain control computer.
(By the way, the planned new 5x interpolator for the Heidenhain 11uApp analog encoders is here -
https://github.com/lukeweston/SinusoidalEncoder )
This is very similar to the machine we're working with:
http://www.machineability.com/Bridgeport_series_II.htmlOn the above page he's used the following products, which would also be suitable for what I want - if they have the 5i20 mainboard in stock, or any board that can interface into their 7i33 analog servo controller.
http://store.mesanet.com/index.php?route=product/product&product_id=66http://store.mesanet.com/index.php?route=product/product&product_id=74However their 5i20 board is out of stock, plus I don't think a PCI card is ideal for compatibility with most modern PC motherboards, if it needs future replacement, plus these Mesa boards are fairly expensive once you combine the analog output boards and the main motherboard. I think PCIe or USB would be better for a modern system. But I could handle PCI (or even parallel), it's just a matter of finding a spare older motherboard to do the job.
You're right in that I don't specifically care about a step-direction interface - anything that can ultimately interconnect back to the PC and talk to common software such as LinuxCNC/Mach3 would be ideal.
If no better options are available, the Kerr servo controller looks workable - if it is actually in stock and obtainable.
I'd like to minimise the amount of bespoke electronics construction for this project and plug-and-play off the shelf components where possible, but I also don't want costs to be too high.
If it ends up being essential, I've thrown together a quick schematic for a PWM input (5V TTL) to +/-10V converter, where 50%=0V. The schematic is attached, what do you think?