Post by thread:Need a smooth needle motion!

part 0 1342 bytes <META content=text/html;charset=iso-8859-1 http-equiv=Content-Type> <META content='"MSHTML 4.72.3612.1706"' name=GENERATOR> </HEAD> <BODY bgColor=#ffffff> <DIV><FONT color=#000000 size=2>Comrades;</FONT></DIV> <DIV><FONT color=#000000 size=2></FONT> </DIV> <DIV><FONT color=#000000 size=2>I have a project with a stepper controlled needle indicator.  I have a simple 8 bit input.  I have a maximum needle speed.  The problem is that I have a jerky needle motion.  I have an IIR filter in the software, and you can see the needle jerk in chunks that resemble binary distances.  (IE the needle surges less and less as it homes in).  I need good snappy response, but I need the needle to move smoothly.  Are there any filter gurus out there that can identify my problem?</FONT></DIV> <DIV><FONT color=#000000 size=2></FONT> </DIV> <DIV><FONT color=#000000 size=2>Thanks</FONT></DIV> <DIV><FONT color=#000000 size=2>Chris Eddy</FONT></DIV> <DIV><FONT color=#000000 size=2>Pioneer Microsystems, Inc.</FONT></DIV></BODY></HTML> </x-html>

At 08:49 17/08/99 +1000, you wrote: {Quote hidden}>> Chris Eddy wrote: > >> I need good snappy response, but I need the needle to move >> smoothly. Are there any filter gurus out there that can identify my >> problem? >> > >Maybe you need to work out the amount of steps needed for a given move >and then step in a linear mode to that point. If the input value >changes, then recalculate to the new position. I guess you could get >fancier and do a logarithmic move so that the needle moves fast >initially and slows down as it approaches the end point. > > >-- >Best regards > >Tony > >http://www.picnpoke.com >Email spam_OUTsalesTakeThisOuTpicnpoke.com > > Yep this is the sort of thing one needs to do if it is to look like an analogue counterpart. You say in chunks, sounds like the time constant of the IIR is too short (Not enough taps) And that you are reponding at a rate that may be way too fast. We all need to know a little more before we attempt to look at the problems (Yes I think that there is more than one and I don't think that it is the filter as yet, although it may be fixable in this area). Firstly how well does it work when it has to move from a large value to a small value and vice versa, secondly how does it move when subjected to small veraiations ie. 3 or so bits worth (The IIR filter should do its job here) Then how often do you sample, do you leave the ADC in continuios sample mode while the IIR is active, or do you use a sample and hold type thing? Please tell us more. I don't know much, but may be able to help. Dennis

One way to achieve a smooth needle movement is to continuously modify the velocity based on the distance the needle has to move. The larger the distance, the shorter the time interval between steps. You would constantly compare the present position with the desired position. The absolute value of the difference would be used to govern the current step rate. The sign of the difference would be used to determine the direction of the step. In pseudo-code the procedure would be something like this: Initialization: Home stepper motor to 0 position Clear CurrentValue to 0 Set Direction to UP Set TargetValue to 0 Set Difference to 0 StepperMeter: (Main Routine) Call GetTargetValue. <TargetValue now has 8 bit value>. Call FindDifference. <Difference and Direction are now loaded> Call OneStep. <Step ONCE UP/DOWN and update CurrentValue> Call FindDifference. <New Difference and Direction now loaded> Call DelayDifference. <Delay based on remaining difference> Goto StepperMeter. <Loop> GetTargetValue: (Subroutine) Check availability of new value. IF nothing new to be acquired THEN Return. ENDIF Acquire TargetValue via ADC/SerialPort, or via other means. Convert TargetValue into 8 bit form if not already received in this form. <8 bit unsigned binary.> Return with 8 bit value in TargetValue. FindDifference: (Subroutine) Compare TargetValue and CurrentValue IF TargetValue=CurrentValue THEN set Difference to 0 set Direction to UP Return. ENDIF IF TargetValue>CurrentValue THEN set Difference to TargetValue-CurrentValue set Direction to UP Return ENDIF IF CurrentValue>TargetValue THEN set Difference to CurrentValue-TargetValue set Direction to DOWN Return ENDIF OneStep: (Subroutine) IF Difference=0 THEN Return. ENDIF IF Direction=UP THEN step up once decrement Difference Return ENDIF IF Direction=DOWN THEN step down once decrement Difference Return ENDIF DelayDifference: (Subroutine) IF Difference=0 THEN Return. ENDIF Compute number of extra delay units: Units=255-Difference IF Units=0 THEN Set Units to 1 ENDIF Call StepTime (minimum step time) DelayLoop: Call IttyBittyTime Decrement Units IF Units>0 THEN Goto DelayLoop ENDIF Return StepTime: (Subroutine) Delay for 5 milliseconds (or whatever min step time is) Return IttyBittyTime: (Subroutine) Delay for 100 microseconds (experiment with this value) Return ************** Fr. Tom McGahee

The short answer: Nope. The Long answer - what you need to do is NOT to filter the signal to your stepper motor, but use microstepping. The best place to look for this info is Jones on Stepper Motors http://www.cs.uiowa.edu/~jones/step/ This is a complete treatise on the subject of stepper motors and microstepping. There are a lot of others. I am involved in a project now using stepper motors to control a pan system for video cameras. Same issue - at slow speeds they are jerky. I am looking into microstepping the camera motor to eliminate the jerking motion. {Original Message removed}

I don't think it is related to switching of the stepper. The motor moves smoothly through individual movement segments. The issue is how to match the filtering of an input to the rapid motion of a needle. With a digital display, we simply do not notice the jerky nature of a filter function as it homes in on a value. The needle does, though. So the question is one of filtering such that it is smooth, and yet not so slow that it is not, well, slow. I see great ideas so far, and am about to try Tom's ideas. Thans Chris Eddy Russel said: >Then, if you are using a stepper motor, there is microstepping - instead >of just applying steps you PWM between adjacent steps and vary the PWM >ratio smoothly. > >This can work very nicely.

{Quote hidden}> I don't think it is related to switching of the stepper. The motor moves > smoothly through individual movement segments. The issue is how to match > the filtering of an input to the rapid motion of a needle. With a digital > display, we simply do not notice the jerky nature of a filter function as it > homes in on a value. The needle does, though. So the question is one of > filtering such that it is smooth, and yet not so slow that it is not, well, > slow. > > I see great ideas so far, and am about to try Tom's ideas. > > Thans > Chris Eddy > > Russel said: > > >Then, if you are using a stepper motor, there is microstepping - instead > >of just applying steps you PWM between adjacent steps and vary the PWM > >ratio smoothly. > > > >This can work very nicely. Just to add another possibility..... Steppers are not inherently continuous devices, contiguous yes. It might be a better approach to to actually use a continuous type device... I know moving coil meters are a pain in the butt ... BUT (humour) this is because they lack feedback. I have started to play around with junked harddrives that use moving coil armatures to position the heads with digital/optical positional feedback.... and are they smooth. This investigation arose because of a requirement for different meter characteristics dependent on mode of operation, fast response, peak detect,averaging, rms, peak hold, min hold etc and a digital circuit vastly simplified the circuit. The problem with these mechanisms is I haven't seen them anywhere than in harddrives. _____________________________ Lance Allen Technical Officer Uni of Auckland Psych Dept New Zealand _____________________________