Re: [CAD_CAM_EDM_DRO] Re: Black Box - UMCI spec for API posted

Steve Stallings wrote:

> Alan,
>
> The credit for the API document goes to Bob Shaver who works
> with me. He is also behind the software and getting our algorithm
> ideas into the Xilinx chips. My role is mostly hardware and design
> for manufacturing, and of course understanding the machine tool
> related aspects.
>
> A DDS is a rather simple concept, just a value that gets repetitively
> added to a register. The overflow of the register becomes the step
> output.

DDS usually is associated with generation of Sine waves through a table
indexed by the DDS counter. If you are broadening the term to include
generation of pulses, then the thing I have prototyped is also a DDS, but
a specialized one. It has some extra logic in it to make sure that setup
and hold times between direction changes and step pulses are always
observed. It also stretches the step pulse to a fixed width (300 ns to 25.5
uS).

> We have been working with DDS's and precision frequency control for
> over a decade. Nowadays we end up describing them in Verilog which
> gets compiled into a Xilinx pattern.

Well, I still find a schematic more concise, but some things like Gray
code to binary conversion is best done in a HDL. Xilinx FPGAs are just
incredible, I'm still getting a feel for how much you can do in one, and
how much faster it will be than a bunch of MSI 74xx chips.

> Things do get quite a bit more
> complicated when you throw in hardware management of acceleration
> and we are still fiddling with that part. Likewise, circular motion
> is a whole different problem and at present we intended to rely on
> the application program or the API to convert to short linear moves
> that get sent to the hardware.

Since EMC (and other CNC programs) already do such a good job of this,
I didn't see the reason to move that work out of the CPU. With the 'DDS'
only in the hardware, the system looks much like a servo system to the
CNC control. The only difference is that in my 'DDS', the velocity value is

non-linear. But, that is no problem, the CPU can easily compute the correct

value to put in the velocity register to get linear velocity response out of
the
'DDS'.

Jon