Re: [CAD_CAM_EDM_DRO] Re: Axis Stall Problems

Wayne C. Gramlich wrote:

>--- In CAD_CAM_EDM_DRO@yahoogroups.com, Jon Elson <elson@p...> wrote:
>
>
>>Wayne C. Gramlich wrote:
>>
>>
>>
>>>--- In CAD_CAM_EDM_DRO@yahoogroups.com, "Paul Kelly" <tech@c...> wrote:
>>>
>>>
>>>
>>>
>>>>I run my 297+298 driver (circuit pretty much straight from
>>>>the datasheet) at 18kHz.
>>>>
>>>>Thinking about your setup (I've built drives very similar to
>>>>this and had similar problems) I reckon 20kHz sample rate is
>>>>about a factor of 10 too slow.
>>>>
>>>>A typical PWM drive starts with a square wave at the PWM
>>>>frequency. Turns on the drive transistor for a minimum period
>>>>(ie ignores the current spike from switching transients),
>>>>then uses a comparator to turn off the drive when the
>>>>current reaches a threshold. All this is happening VERY
>>>>fast. I can remember having to change an op amp I was using
>>>>to amplify a current sense voltage because it wasn't fast
>>>>enough.. If you are switching at say 10KHz and sampling at
>>>>20kHz then you effectively only have two current settings.
>>>>(it's a bit more complex I know..)
>>>>
>>>>
>>>>
>>>>
>>>I understand what you are saying, but the measured current
>>>does indeed average around 1 Amp.
>>>
>>>
>>>
>>But, are the individual pulses all around the same width, or is it
>>making a very long average of dozens of narrow and wide pulses
>>averaging out to 1 A? I think Paul is on to something here.
>>I have several drives that I've worked with that run at about a
>>20 KHz rate, but they are all capable of varying the duty cycle in
>>microsecond increments. if you are really only taking one sample
>>every 50 us, then the current rise and fall rates might be very rough.
>>This may work fine at low speeds, but when the speeds pick up,
>>the PIC may not respond to coil current relative to the new coil
>>polarity needed in a timely manner. Obviously, when the step rate
>>reaches 5000 steps/sec or so, then things will really break down,
>>as there are only four sample per step. (Or, maybe, things will
>>actually run well there!)
>>
>>
>
>It is clear to me now that I am not sampling the current
>sense voltage any where near fast enough. What is happening
>is that I am getting is chop waveform that is pretty much
>a 10KHz square wave with occasional half width pulse stretches
>at either the low portion of the cycle or the high portion.
>This is not anywhere near the waveform that the L297
>produces, which is a relatively constant frequency of
>variable width pulses.
>
>Alas, the PIC16 series of processors is not gifted with
>faster A/D's, so I think I will have to abandon hope of
>fast rapids with my current hardware design. I top out
>at about 200-225 full steps per second with the current
>hardware (which is way less than I was hoping for.)
>Given that this is just a hobby for me, I'll use it
>event with the really slow speeds.
>
>

One possible way to fix this, and still get your microstepping eventually,
is to put a comparator on the current signal, and generate the reference
for the comparator from a DAC on the PIC (if it has one). I don't know
if the
PIC can even sample this fast enough, but it should be able to get it
down to 5 us or something like that with tight code.

Jon