Re: [CAD_CAM_EDM_DRO] Re: EMC Following Error

Brian,

I'm not really looking for "for impossible base resolutions and near light
speed at the same time". I got to 32000 steps per inch by using what I had
available (2000 cpr encoders, 8 tpi lead screws and 2:1 reduction from motor
to screw). I have already started on new pulleys to get rid of the 2:1 and
thereby halve the steps per inch. I have also reduced the velocity and
acceleration parameters and this has also reduced the steps per inch and got
rid of the EMC following error messages.

I like your suggestion of using some sort of hardware assistance but have to
convince myself that I can actually get this machine to cut metal in a
useful way before I try to improve on it (and spend more money along the
way!).

My experiments today with a ball point pen in the spindle and a sheet of
paper on the table showed me that, even as it stands, I can draw clock
wheels and dials at reasonable cutting speeds. No real rapid movement but I
just have to be patient!

I think if I change the pulleys to 1:1, drop the encoder resolution a bit
and/or swap the 8 tpi ACME leadscrews for 5 tpi ballscrews I will get down
to ~10000 steps per inch and have something that moves quickly enough.

Regards,

Dave.

David Micklethwaite

> The key to achieving higher speeds and smooth operation is to
> relieve the motion processor from spending much of its time twiddling
> bits on a slow legacy parallel port. The parallel port control is
> nice because it is simple and available. Many of the folks here seem
> to be advanced beyond the simple and looking for impossible base
> resolutions and near light speed at the same time. Even with great
> improvements in processing the basic system design for the parallel
> port access will limit bit rates.
>
> This is where a velocity based system such as a true servo system
> works best like Mr. Elsons for EMC or several of the other hardware
> assisted solutions.
>
> http://www.pico-systems.com/univstep.html
>
> Take 2 examples with a .2 lead screw geared 2:1 and running a G201
> for example. This setup will take 20,000 micro steps per inch. To get
> 180 Inch per minute rapids you would need to output 60,000 pulses per
> second.
>
> For step pule generation this means for every second I have to
> service 60,000 interrupts. Each one I stop, respond to the interrupt,
> save context, flush the pipeline, branch to service, set the bits,
> wait for pulse width, clear the bits etc.. And during all this I
> expect to read a 100,000-line file, service the mouse and display
> full color graphics and compute constant velocity contours with
> offsets and tool compensation.
>
> In a velocity based system I set the loop interval to reasonable
> value say 1000 per second. Now rather than waste all my effort
> bouncing bits I just output target velocity profiles. When moving at
> mach 1 I don't really need the several micron following error and
> when it slows down for detail work the update interval is fast enough
> to create precision. In this system the overhead is far less and more
> consistent.
>
> You can do a rate multiplier but to work well the system has to have
> a good phase lock and knowledge of what max rate is. Then you take
> the ratty pulse train, buffer it, smooth it back out while filling in
> the intermediate holes. What you have done is take a velocity
> profile, integrate it into steps, send it out as pulses,
> differentiate it back to velocity, and send a new pulse train out
> with smaller steps.
>
> The real solution is to adapt the driver software for closed loop
> velocity control vs. open loop position for the demanding
> applications. This is what Jon has done with his board.
>
> To buffer the step pulses the minimum I would guess is to slow the
> step software to a reliable maximum where it outputs good steady
> pulse trains. The buffer then reads step pulses and delays them until
> the next pulse or a time out to compute the intended rate. To
> preserve the sync you should clock all the streams the same as well
> as buffer the direction lines. Or simply buy a G210..
>
> The problem being any rate multiplier will lower your resolution. If
> you need both the fine resolution and speed you can't get there so
> you are back to the velocity approach. If you just need the micro
> step smoothness then a multiplying driver is the answer.
>
>
> Brian
> BSP