Chopper methodology was Re: stepper vs. servo

It is what's called an over-constrained feedback loop. Over-
constrained loops are unstable and are susceptible to a phenomena
called pulse doubling when the current decay slope exceeds the
current rise slope. This is guaranteed to happen with a step motor
coil because the current rise slope is (Vs - Vm) / L while the decay
slope is (Vs + Vm) / L where Vs = supply voltage, Vm = motor voltage
and L is the motor coil inductance.

You can graph it out to see this phenomena on square-ruled paper.
Have the increase slope equal +1, have the decay equal -2. Put the
clock "ticks" every two squares. The graph will quickly settle
at "on" for 2 clock cycles, "off" for 1 cycle no matter your initial
conditions. This is not good.

The cure is to add feed-forward compensation. This is a positive
slope signal from the oscillator that when summed with the current
feedback waveform insures the current rise slope is greater than the
decay slope.

Mariss





--- In CAD_CAM_EDM_DRO@yahoogroups.com, "ballendo" <ballendo@y...>
wrote:

> Mariss,
>
> Thank you for the reply. As I read what you wrote, it sounded
> familiar; first I thought of the Hard sync of two synthesizer

VCO's

> (music), then thought of the ST chips, which use constant-ON time...
>
> Here is a quote from the L297 spec sheet (under the heading circuit
> operation):
>
> "A common on-chip oscillator drives the dual chopper. It supplies
> pulses at the chopper rate which set the two flip flops FF1 and

FF2.

> When the current in a winding reaches the programmed peak value the
> voltage across the sense resistor (connected to one of the sense
> inputs SENS1 or SENS2) equals Vref and the corresponding comparator
> resets its flip flop, interrupting the drive current until the next
> oscillator pulse arrives. The peak current for both windings is
> programmed by a voltage divider on the Vref input."
>
> This sounds exactly like what you just decribed, but I know htat
> these chips spit, squeal and hiss too!
>
> What am I missing?
>
> Thank you in advance,
>
> Ballendo
>
> --- In CAD_CAM_EDM_DRO@yahoogroups.com, "Mariss Freimanis"
> <mariss92705@y...> wrote:
> > Constant off-time choppers hiss, squeal and "spit" because you

have

> > two free running variable frequency oscillators. A typical

chopper

> > (20 microsecond off-time) may run at 25kHz at min current,

dropping

> > to 15kHz at max current.
> >
> > There is an unavoidable tendency for these oscillators to phase-
> lock
> > (run at the same frequency). The winding currents diverge while
> phase
> > locked so the lock has to break after a number of cycles. This
> > repeated locking and breaking generates frequencies in the

audible

> > range, (lock for 9 cycles, break on the 10th generates a loud

2kHz

> > squeal for a chopper freq of 20kHz).
> >
> > I get around this by making the current servos synchronous, and

by

> > definition, always phase locked. A 20kHz clock rising edge causes
> > both coils to increase in current. Once current reaches the

desired

> > value, the current is caused to decay until the next clock rising
> > edge. Both coils switch at the same constant 20kHz, no audible

sub

> > harmonics are generated and the drive is silent.
> >
> > Mariss
> >
> >
> > --- In CAD_CAM_EDM_DRO@yahoogroups.com, "ballendo"

<ballendo@y...>

> > wrote:
> > > Mariss,
> > >
> > > Glad to see your input on this. Perhaps the Allegro designers
> also
> > > consider 98% of a design close enough to 100% <G>...
> > >
> > > What do you use instead of "constant-off" time? I seem to

recall

> > you
> > > sayingsomething about this a year or two ago?
> > >
> > > ST uses constant-ON in many of their products; and they also

hiss

> > and
> > > spit<G>. I'm assuming it's the "constant" that's the problem???
> > >
> > > Thank you in advance,
> > >
> > > Ballendo
> > >
> > > --- In CAD_CAM_EDM_DRO@yahoogroups.com, "Mariss Freimanis"
> > > <mariss92705@y...> wrote:
> > > > I think that is an interesting and confused assertion by
> Allegro.
> > > The
> > > > motor error of 3% probably refers to the non-accumulative

step

> > > error
> > > > of a premium motor. This is a cyclic error, not a step to

step

> > one.
> > > > Step-to-step errors are about an order of magnitude smaller,
> well
> > > > under 0.5%.
> > > >
> > > > Calling 98% "close enough" to 100% is not good enough; it is

a

> > > > perceptible error in motor smoothness anyone can spot in a

side-

> > by-
> > > > side comparison.
> > > >
> > > > Allegro also makes other, larger mistakes that affect motor
> > > > smoothness. This is their insistance on switching between
> > > > recirculating and non-recirculating modes while stepping.

That

> > > > introduces large offset errors at the full step locations

that

> > > > completely swamp the potential microstep accuracy.
> > > >
> > > > My other complaint is their continued use of a "constant off-
> > time"
> > > > chopper method for the current servo loops. This produces the
> > usual
> > > > assortment of hissing, squealing and grunting sounds.
> > > >
> > > > A well designed (read low distortion) drive used with an
> accurate
> > > > motor (the new square ones) should completely resonance free.
> > That
> > > > means no perceptible vibration at all as you slowly sweep the
> > zero
> > > to
> > > > 2 revs/sec speed range. Allegro doesn't even come close; they
> > > should
> > > > re-think their switching topology. Otherwise they're nice

chips.

> > > >
> > > > Mariss
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > --- In CAD_CAM_EDM_DRO@yahoogroups.com, "turbulatordude"
> > > > <davemucha@j...> wrote:
> > > > >
> > > > > Interestingly, Allegro's paper notes that at the 1st and

16th

> > > step,
> > > > > the current is so close to full that the value (98.089%) is
> > > smaller
> > > > > than the motor error (3%) so they ignore the very end of

the

> > > > ranges.
> > > > > I'd quote the pages, but Allegros site is not responding.

If

> > > > anyone
> > > > > wants the reference, please let me know.
> > > > >
> > > > > But that opens the question of how others can get 256
> > microsteps
> > > or
> > > > > more. And THAT really is a more general motion control
> > questions
> > > > as
> > > > > even though we tend to think our interests are 98% of the
> > > universe,
> > > > > CNC machining applications of motion control are probably
> more
> > in
> > > > the
> > > > > 2% range. I'll bet Hewlett Packard (one supplier) sells

more

> > > Laser
> > > > > Printers in one year using motion control than all the home
> > brew
> > > > CNC
> > > > > stuff ever built.
> > > > >
> > > > > Dave