USB-to-step/direction (was Interface. Gauging interest.)

Hi, everyone. I've been lurking here for a while, and I've been
following the discussion with great interest.

What perfect timing on this topic!! :-)

Just last night, I got a prototype working for a USB-->HobbyCNC board
interface. For those not familiar with the 4AUPC HobbyCNC board (is
there anybody like that on this list?), it's a kit that controls 4
unipolar steppers, and it expects to connect via a parallel port, using
STEP and DIRECTION.

What I have working so far is that I am able to control a single stepper
(the X port), with acceleration/decelleration at the beginning and end
of the move, from a USB port. And, I don't have a parallel port on this
machine! :-)

I'm not connected to a mill yet, but I'm getting a top speed of about
360RPM on the stepper, which will end up being about 36ipm with my 10TPI
leadscrew.

More work (mostly software) is clearly necessary, but I'm fairly
confident I can get all the way there (4 motors controlled at the same
time). I don't seem to be losing any steps, but certainly more testing
is needed here, too, just to be sure.

Here's a summary of what I've done so far:

1) FTDI FT2232 chip, with dual USB FIFO's, running at "9600 baud" (in
quotes). I actually used the DLP module from dlpdesign, to save some
wiring time on the FT2232 side.

These FTDI chips are some of the ones that are being used for
USB-to-parallel-port interfaces. I chose this one specifically because
it supports a high speed parallel mode, called "Bit Bang". Most of the
other ones don't, AFAIK.

2) So, I'm running in "Asynchronous Bit Bang" mode, which basically
means that I give it a whole buffer of bytes, and the USB driver takes
care of streaming them out at high speed, one at a time. It actually
transfers them in batches, for efficiency over the USB bus. I have the
lowest bit connected to STEP, so I have to store a zero, then a one, in
my buffer, for each step I want.

Note that this chip has an internal 16X multiplier on the baud rate.
So, "9600 baud" here really means that the port spits out one byte every
1/(9600*16) seconds (6.5 microseconds), because of an internal 16X
multiplier.

My timing control on the STEP line then has a resolution of about 6.5 us.

3) I am driving a small, cheap Jameco stepper, using a 12V 25A switching
supply I had lying around. (My mill project is designed for cheap, not
for speed.) This stepper has 1.8 degrees per step.

4) I am using 16X microstepping mode on the HobbyCNC board. So, one
revolution actually requires 200 * 16 steps. Motion seems smooth and
repeatable so far.

5) The 2232 chip actually has 2 USB channels. And, each side can
optionally be put into a serial port mode. So, while I was at it, I
used the other 2232 channel in serial mode to drive a Scott Edwards 2
line LCD display, and a beeper. Just for fun!

6) I'm driving it from a fairly fast machine, an Athlon 3200, because
that's what my desktop machine is. I haven't tried it on slower
machines. But, the overhead (as measured by the Windows Task Manager)
seems like it's on the order of 3% of the CPU, worst case, so I would
guess that almost any machine could drive this chip.

7) I don't have a scope handy, so I don't know if there are any gaps in
the stream, due to USB overhead. In any case, the motor doesn't seem to
mind.

I suspect that I may have violated some assumptions here. For example,
I (perhaps naively) did not try to duplicate a 50.25us waveform. I just
tried to create an interface where I could stream out zeroes and ones at
a very high, but *regular* rate. This USB chip seems to do it.

Note: this means that the actual step pulses will contain jitter -- a
13us period followed by a 19.5us period, for example. I am counting on
the microstepping plus the inertia of the motor to smooth this out, by
the time it gets to the leadscrew.

Will this end up working perfectly? Well, lots more testing is required
to determine that. But, I did want to share with you all that the
concept seems like it *could* work, based on a real prototype! :-)

Mike

Mariss Freimanis wrote:

>>ballendo wrote:
>>
>>What we truly need is a usb EQUIVALENT of the parallel port.
>
>
> That's just not possible for CNC programs that output step and
> direction signals from a parallel port. Step pulses have timing in
> the microsecond range. Example:
>
> 20kHz step pulse freq = 50uS period
> 19.9kHz step pulse freq = 50.25uS period
>
> The difference is only a 1/4 microsecond. USB serializes data which
> strips this timing information; the reconstituted parallel data would
> bear no resemblence to the sent original.
>
> Mariss