Re: I-LPT & step/dir servo amps

Carlos Guillermo wrote:

> From: "Carlos Guillermo" <Carlos@...>
>
> Jon -
>
> Yeah, as soon as I sent that message, I remembered a key issue. I've read
> that the encoder resolution on a servo system should be at least 5-10 times
> the desired positional resolution, because of the losses due to of
> dithering, hunting, etc. (I think I read that in Tech 80's "Designing with
> Motion Handbook").

Yes, that's true. Fine encoder resolution also permits motion to be smooth
even at extremely slow feeds.

> That roughly agrees with your .000025" encoder
> resolution, which I would guess is using 5 TPI screws, 2:1 reduction, and a
> 1000 cpr encoder wheel (with quadrature, of course).

No, the .000025" resolution is 10 TPI screw and a 1000 cycle/rev encoder
directly driven. I have very strong objections to putting belt drive between
the encoder and the screw. The belt can have pitch variation, and the sprockets
can have eccentricity. Both of these can cause really large cyclical errors.

> What if the step/dir frequency was limited to the desired positional
> resolution multiplied by the max feedrate, and the step/dir servo amp

Ughh! This whole idea of having step pulses controlling a servo is just
making me sick! What a waste of good servo motors and encoders!
I just don't believe that most of these systems have the timing resolution
to move very smoothly over a wide range of velocities.

Let's work out an example. Let's say we want resolution of .00005", and
velocity up to 120 IPM, that is a step rate of 2 IPS x 20000 step/In =
40000 step/sec, or 25 uSec per step. Now, let's say we need to go at
119 IPM, that's 1.983 IPS x 20000 Step/In = 39666.66 Steps/sec or
25.21008 uSec per step. Hmmm, that's a small variation of about 210
nS in the step rate. But, what if we needed 119.9 IPM? That is 25.02085
uS, or about 21 nS. Can your program of choice deliver steps to 3
(or more) motors with 20 nS resolution? Obviously, not. Can it do
210 nS? I suspect, still not, without custom timing circuitry to generate
the step pulses. But, most of the programs output step pulses from the
parallel port, directly from software. Most of these programs have a
fixed interrupt clock, and either generate a pulse, or don't, every tick
of that clock. I would suspect that this clock can't be much over 10 KHz
for software step pulse generation. So, right there, your highest rate
would be 10,000 steps/sec, and the next lower rate would be 5,000!
That is a VERY coarse jump in step rate.

With a 12-bit msgnitude plus sign velocity DAC, as used in the Servo-to-Go
card, you get 4096 equal intervals of velocity. If you set max (4095) to
be 120 IPM, then you get approximately 34 quanta on the DAC for each
IPM. So, the servo routines have a choice between 120.0, 119.97, 119.94
etc. This allows very smooth adjustments of velocity to keep the actual
position close to the desired position at all times, even near max velocity.

I really think this is where stepper systems fall down, and it is already in
deep trouble before the signals have come out the parallel port. Namely,
information has been lost, and there is no way to bring it back once it is
lost!

Jon