Re: [CAD_CAM_EDM_DRO] Re: encoder to tach converter idea (was killer deal...)

skykotech wrote:

>Confused here Jon. The voltage from the tach would go from zero to
>several volts in a fraction of a second also. How do you close the
>loop with that?
>

An op amp does it quite nicely. You can build a completely continuous-time
system using an old stereo amp and some op-amps. The computer produces
a desired velocity command. That command and the tack voltage go into the
summing node of an amplifier, and the result is an error signal (difference
between actual and commanded velocity). This error progams another
loop, the current loop, to produce a calibrated torque to correct the error.
The velocity loop can have fantastic gain, 100,000 times or so and still
remain stable.

> Or are you saying that any DAC used would be too
>slow? Also, I don't see how a frequency to voltage converter would
>be discontinous...at 0 hz you get 0V...same with a tach. At 0 rpm on
>the motor I doubt your tach is putting out voltage, unless it is
>violating some laws of physics/electronics.
>
>

Yes, it does violate a la of physics, in a sense, because between encoder
counts, there is no information. So, it can't provide a linear response
to velocity. It can only provide a step-wise response at each encoder
count.
between encoder counts, it doesn't know whether the velocity is staying the
same, increasing or decreasing. So, it is essentially open-loop between the
encoder counts. The only way to fix this is delay, and that means more
sluggish overall response.

The DAC is not the problem, it is the lack of information between encoder
counts.

This all is important only at very low speeds. But, it is required for CNC
machines to sometimes move very slowly (as when boring a large hole)
and to stand as still as possible at times. If you have high gain and
periods
of open-loop, it will make the system massively unstable.

>*shrug* delta sigma dacs are 24 bits (16 million to 1) and have a
>response time of at least 30khz for the one I have in mind. Are you
>sure the lines from your tach are not picking up noise greater than
>7uV?
>

Yup, clearly they don't! The system would jitter and hunt if it were.
It doesn't
hunt. I have bar graphs on my servo amps, so I actually know what it is
doing all the time. The Z axis hunts due to the weight of the quill.

> I can't even measure 7uV very accurately on my lab bench with
>my $$$ oscilloscope.
>

Of course you can't. But, my servo amp really does. It is amazing, you can
try to move the machine, even with the positioning loop turned off, and you
still can't move it because the tach detects any attempt at movement and
the servo resists it. I do use an OP620 instrumentation amp, which is
incredibly
good at recovering small differential signals with common-mode noise.
If you
short the tach with a screwdriver, the amp faults within a fraction of a
second!

> Also, it may be feasable to forgoe using a DAC
>and just pwm an RC buffered by an op-amp...depends on the clock speed
>of the Xilinx part as to how many bits of resolution you can get vs
>response time. Something on the order of 16-18 bits and 1khz should
>be doable.
>
>Curious what the brush amps that use encoder input for velocity
>feedback are doing inside? Something similar?
>
>

Yes, but they have to lose performance, unless the encoder has very high
resolution.
The DAC is not the big concern, it is the change in velocity updates
with the
encoder.

Jon