Re: G320s, New and Improved!

> Is the multiplier problem only in the G210 or is there is similar
> problem with the multiplier in the G340 shipped in the last few

weeks?


Both the step and the sevo drive use the same multiplier board, so
the problem is common both drives. The problem has been solved and
entails adding one transistor and one resistor to the existing boards.

I try very hard not to make any mistakes in my designs, but being
human, some slip past in spite of my best efforts. This is one of
those cases. I belive it is incumbent on me to let everyone know
whenever one does and to make right any previously sold products with
a design error. It is the only honorable thing to do.

For those that are interested, the following is the cause and cure
for this problem. It traces the genesis of how this happened and the
steps necessary to fix it:

Good afternoon! It has been an interesting 24 hours from finding a
problem until finding a fix. Here's the problem:

The G901 is a phase locked loop. Being a PLL, a loop is involved, and
a loop means response time. The original breadboard was wired for
a "times 10" since that is what is technically most difficult. The
loop design had two conflicting requirements; a fast response is need
for rapid rate changes while a slow response insures the 10
synthesized pulses are evenly distributed in time. I chose a 1 mS
response time so the loop would respond to a 1,000 step per second
instantaneous change in speed, figuring no step motor can physically
respond that fast.

The rub is, at "times 1" setting this only works out to 100 full
steps per second.

The G901 works a little bit like a G320, the steps sent to the G901
go into a counter, the steps generated by the voltage controlled
oscillator (VCO) on the G901 are sent to another counter (after being
divided by 1, 2, 5 or 10). An arithmetic adder finds the difference
between the two, converts thru a 4-bit D to A into an analog voltage
which is then applied to the VCO. This is just like the position
error on the G320. When the loop response is to slow, more steps
received accumulate than steps sent until the +/- 8 step difference
is exceeded and an overflow occurs (missed steps).

Here's the solution:

I added a circuit that looks a the "position error" voltage and feeds-
forward compensates (gooses) the VCO voltage when the error exceeds 6
steps. This rapidly speeds up the VCO frequency and takes up the
error to less than 6 steps, where the normal loop response can take
over and reduce the error to zero (normal operation).

This solution requires one NPN transistor and one resistor. I have it
running right now and as a consequence it will not miss a step if
even 100,000 steps per second are abruptly applied to the input. No
sane person would ever run a motor like that. Next is to see what is
entailed to rework my existing stock of G901s, I will of course trash
all of my existing unstuffed boards, rework the board layout and
order new ones ASAP. I should have working examples Monday that I
would like to send you to try out. Will keep you posted and sorry
about this screw-up, it has me feeling very embarrassed.

Mariss