Hardware Timer for accurate pulsout?

[gtslabs]

I am trying to send a stepper motor step signal with a range of 0.1 steps per second to 2000 steps per second. I need less than 1% error. I have not been able to get accurate frequencies measured on my scope using the pause command. Using the Pulsout I get an accurate pulse but the rest of the loop adds error. I also tried looping just 2 pulsout statements with a dummy pin not used but I still had some error.
<code><pre><font size=2 face='Courier'>
top:
pulsout 4,5000 ' step signal
pulsout 5,5000 'nothing connected
goto top:
</font></pre></code>

I just ordered some 28x1 and 40x1 chips but waiting for those to arrive.

Could I use the hardware timer and a crystal oscillator to get accurate square waves at this range? Can PWMout go this low?

Or is there a better way to get the signal at this range? I would like to have 16bit resolution on the range.

 

[moxhamj]

Errors may still be &gt;1% with a picaxe. 2000 steps per second is way more than I have ever managed to drive a stepper, but maybe you are using special ones. You can get absolutely accurate stepper control with an xtal, divider, counter and AND/NAND logic gates. It will end up being 6+ chips though. What is the application?

 

[gtslabs]

Dr. Acula, thanks for your comments. I am building one of these <A href='http://www.soiltest.com/pdf/Soil/triaxial.pdf' Target=_Blank>External Web Link</a> for my own inhouse use. I dont need mine to that slow. I also have a need for 2000 steps per second or faster for a particular test we perform.

Mid band resistance may be a problem that fast so I am considing adding Viscous Inertial Dampers if needed. I may have to add a manual gear changer to get the speeds higher but then I would need a bigger stepper.

Do you know of any example crcuits using the components you suggested?

Thanks
Steve

 

[moxhamj]

Building it with discrete components means the speed can't be varied very easily, though 'divide by n' counters can be used. I suspect machine code on a pic would be the most practical option. Have you built these sorts of things before?

 

[gtslabs]

Not the circuit but I already have a proyotype working. I am using a surplus controller/driver frmo slo-syn <A href='http://www.electrosales.com/warner/Warpdrive/SS2000.html' Target=_Blank>External Web Link</a> but I am not getting accurate rates. This was a surplus item so I started with it. It costs over $1K so I wont be using it.

I used Liberty Basic to generate the interface to that controller. <A href='http://www.gts-labs.com/motor_control_picture.JPG' Target=_Blank>External Web Link</a>

My actual rate was 1/2 the desired rate at this slow speed. It was better when I ran it faster but not where I need it to be. Since I did not plan on using that controller anyway I did not investigate why.

I really dont want it controlled by a computer anyway. A keypad and LCD interface is what I am working on.

Why can the PWMout command work on a lower frequency? Could a picaxe with a very low Hz oscillator bring it into the range I need?

 

[gtslabs]

I wanted to share what I found that should work.
I belive this chip should work for my application<A href='http://www.analog.com/en/prod/0,,770_843_AD9834,00.html' Target=_Blank>External Web Link</a> This is a AD9834 chip.

An application:<A href='http://www.myplace.nu/avr/minidds/index.htm' Target=_Blank>External Web Link</a>

Steve

 

[premelec]

That chip certainly will put out - I have a 0-20MHz digital oscillator using this - [www.norcalqrp.org FCC-1&amp;2]- perhaps a bit overkill? A ring counter with weighted resistors can put out a nice stepped waveform easy to filter... p328 of my CMOS Cookbook shows this approach - also the famous 8038 function gnerator might be made to work... Whatever... good luck with it!