### Email Archive: emc-developers (read-only)

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-05-06 15:47 ``` On Sunday 06 May 2012 17:40:25 Yishin Li wrote: > On Sun, May 6, 2012 at 10:38 PM, dave wrote: > > > On Sun, 6 May 2012 19:53:58 +0800 > > Yishin Li wrote: > > > > > On Tue, May 1, 2012 at 3:34 AM, dave wrote: > > > > > > > On Mon, 30 Apr 2012 13:41:12 -0400 > > > > EBo wrote: > > > > > > > > Sorry all of your text went away when I hit reply. > > > > I just posted this in a reply to Andy. > > > > > > > > caris.mech.ubc.ca/Publications/sonja.pdf > > > > > > > > But alway remember: Good, Fast, Cheap ... pick any two. ;-) > > > > > > > > > > > I tried Sonja's approach for final deceleration and approaching > > > target. However, I can not get the same accel/vel curves while > > > approaching target. > > > > > > Has anyone tried Sonja's method? > > > > > > -Yishin > > > > Hi, > > Any idea what is limiting the accel, etc? Of course, the real question > > is how smooth is it and then can you speed it up? Thanks for trying it. > > > > > Instead of follow the whole Sonja's algorithm sets, I just tried to limit > the final accel with sin template. > (3.2' (modified 3.2)): a(t) = 0.5 * acc * sin(PI * t /dtmax - PI/2) - 0.5 * > acc > (3.4) dtmax = 0.5 * PI * acc / jmax > I think I made mistakes for integrating 3.2' for vel(t) and pos(t). > > I would like to learn how to use octave to integrate 3.2'. > I'm newbie to octave. Any guide would be appreciate. > > -Yishin Octave is primary for numerical math. Try maxima, a CAS (computer algebra system). http://maxima.sourceforge.net/ Joachim > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers > ```

[Emc-developers] S-curve andy pugh <bodgesoc@gm...>
 Re: [Emc-developers] S-curve From: EBo - 2012-04-30 17:41 ```Andy, If you get a full jerk-limited motion planner let me know -- I'll pass it to a friend who needed such a beast for retrofitting some old equiptment... EBo -- On Mon, 30 Apr 2012 18:09:45 +0100, andy pugh wrote: > I am trying to write a super-simple jerk-limited motion planner for a > very simple application. > Looking for equations of motion I found: > http://sprott.physics.wisc.edu/pubs/paper229.pdf > Which seems to suggest a risk of chaotic motion once the equations > become third-order. > However, I think the chaos in my motion planner at the moment comes > from less esoteric sources, such as me over-simplifying. ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-04-30 18:34 ```On Monday 30 April 2012 19:09:45 andy pugh wrote: > I am trying to write a super-simple jerk-limited motion planner for a > very simple application. > Looking for equations of motion I found: > http://sprott.physics.wisc.edu/pubs/paper229.pdf Nice paper. It refreshes my knowledge. > Which seems to suggest a risk of chaotic motion once the equations > become third-order. I assume, you are working in 3D (xyz), so deterministic chaos is possible! > However, I think the chaos in my motion planner at the moment comes > from less esoteric sources, such as me over-simplifying. To measure your chaos I point you to: Grassberger, Procaccia: "Estimation of Kolmogorov entropy from a chaotic signal" Phy. Rev. A 28,4 Page 2591, 1983 additional to reference 19. But maybe you let us know which equation of motion you are using? Joachim ```

 Re: [Emc-developers] S-curve From: andy pugh - 2012-04-30 18:51 ```On 30 April 2012 19:34, Joachim Franek wrote: > But maybe you let us know which equation of motion > you are using? At the moment, a completely incorrect one. -- atp The idea that there is no such thing as objective truth is, quite simply, wrong. ```

 Re: [Emc-developers] S-curve From: dave - 2012-04-30 19:31 ```On Mon, 30 Apr 2012 18:09:45 +0100 andy pugh wrote: > I am trying to write a super-simple jerk-limited motion planner for a > very simple application. > Looking for equations of motion I found: > http://sprott.physics.wisc.edu/pubs/paper229.pdf > Which seems to suggest a risk of chaotic motion once the equations > become third-order. > However, I think the chaos in my motion planner at the moment comes > from less esoteric sources, such as me over-simplifying. I'm not certain that any ... motion is simple but take a look at this one. sine accel/decel plus jerk limited. caris.mech.ubc.ca/Publications/sonja.pdf HTH Dave > ```

 Re: [Emc-developers] S-curve From: dave - 2012-04-30 19:34 ```On Mon, 30 Apr 2012 13:41:12 -0400 EBo wrote: Sorry all of your text went away when I hit reply. I just posted this in a reply to Andy. caris.mech.ubc.ca/Publications/sonja.pdf But alway remember: Good, Fast, Cheap ... pick any two. ;-) Dave > ```

 Re: [Emc-developers] S-curve From: EBo - 2012-04-30 20:01 ```For years I've wanted to take a stab at this using a geometric analogy via Farin's Blossom notation... If I only had an infinite amount of time and resources... EBo -- On Mon, 30 Apr 2012 12:34:06 -0700, dave wrote: > On Mon, 30 Apr 2012 13:41:12 -0400 > EBo wrote: > > Sorry all of your text went away when I hit reply. > I just posted this in a reply to Andy. > > caris.mech.ubc.ca/Publications/sonja.pdf > > But alway remember: Good, Fast, Cheap ... pick any two. ;-) ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-04-30 20:21 ```http://hal.archives-ouvertes.fr/docs/00/67/91/18/PDF/2012_beudaert_lavernhe_tournier_Feedrate_interpolation_with_axis_jerk_constraints_on_5_axis_NURBS_and_G1_tool_path.pdf Joachim ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-04-30 20:24 ```http://www.et.byu.edu/~ered/ME537/Notes/Ch5.pdf Joachim ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-04-30 20:40 ```Harmonic jerk: http://www.simtech.a-star.edu.sg/Research/TechnicalReports/tech-reportV8N1/STR_V8_N1_07_MCH.pdf Joachim ```

 Re: [Emc-developers] S-curve From: andy pugh - 2012-04-30 20:43 ```On 30 April 2012 21:24, Joachim Franek wrote: > http://www.et.byu.edu/~ered/ME537/Notes/Ch5.pdf My current plan is to use the equations from here in a simplistic way. (I found it at the weekend). For my simple exact-stop and constant-speed requirements, I am planning on an approach that says, on every servo cycle: If I was to decrement acceleration by the max jerk from the current velocity and acceleration, how long would it take for v = 0? v(t) = v0 + at -j.t^2/2 => t = -v0 + sqrt(a^2 + 2.v0.j) / 2.v0 If the distance travelled in t + one thread > dtg then start to jerk-down the accel. Otherwise, jerk it up unless a > amax or v > vmax. Note that my requirement is not for any sort of coordinated motion -- atp The idea that there is no such thing as objective truth is, quite simply, wrong. ```

 Re: [Emc-developers] S-curve From: EBo - 2012-04-30 20:50 ```Cool! It will take me awhile to read and digest this. When I was thinking about this I was looking at the osculating circle (or higher order osculant) and surface fairing where one of the dimensions is time. I did not take it to the point where I was able to prove, in a mathematical sense, that the higher order smoothing was indeed equivalent to minimizing jerk. I would love to see this done, but I doubt that I will ever have the time/inclination to pick that one up... EBo -- On Mon, 30 Apr 2012 22:21:00 +0200, Joachim Franek wrote: > > http://hal.archives-ouvertes.fr/docs/00/67/91/18/PDF/2012_beudaert_lavernhe_tournier_Feedrate_interpolation_with_axis_jerk_constraints_on_5_axis_NURBS_and_G1_tool_path.pdf > > Joachim > ```

 Re: [Emc-developers] S-curve From: EBo - 2012-04-30 20:54 ```Joachim, How do you know these references off the top of your head, or are you just taking the time to do some interesting internet searches ;-) EBo -- On Mon, 30 Apr 2012 22:37:31 +0200, Joachim Franek wrote: > Harmonic jerk: > > http://www.simtech.a-star.edu.sg/Research/TechnicalReports/tech-reportV8N1/STR_V8_N1_07_MCH.pdf > > Joachim ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-04-30 20:59 ```just found: http://www.cadanda.com/CAD_8_4__583-592.pdf Joachim ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-04-30 21:10 ```On Monday 30 April 2012 22:54:17 EBo wrote: > Joachim, > > How do you know these references off the top of your head, no. > or are you > just taking the time to do some interesting internet searches ;-) Exact! > EBo -- ```

 Re: [Emc-developers] S-curve From: EBo - 2012-05-01 00:34 ```On Mon, 30 Apr 2012 23:00:58 +0200, Joachim Franek wrote: > On Monday 30 April 2012 22:54:17 EBo wrote: >> Joachim, >> >> How do you know these references off the top of your head, > > no. > >> or are you >> just taking the time to do some interesting internet searches ;-) > > Exact! Some of those are cool finds. Are you interested in playing with that, or just helping the conversation along? If you do want to play with this stuff there might be a couple of peole who can help you along... EBo -- ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-05-01 08:32 Attachments: ch5_convex_01.odf      ch5_cancave_02.odf ```On Monday 30 April 2012 22:42:45 andy pugh wrote: > For my simple exact-stop and constant-speed requirements, I am > planning on an approach that says, on every servo cycle: > If I was to decrement acceleration by the max jerk from the current > velocity and acceleration, how long would it take for v = 0? > v(t) = v0 + at -j.t^2/2 => t = -v0 + sqrt(a^2 + 2.v0.j) / 2.v0 This is equation 5.12 and 5.14?. Attached are libreoffice math files. Joachim ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-05-01 10:34 ```Harmonic jerk 2: http://www.aspe.net/publications/Spring_2001/01Sp%20Extended%20Abstracts/Arevalo.PDF Joachim ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-05-01 12:28 ```Harmonic jerk 3: http://cdn.intechopen.com/pdfs/4267/InTech-On_algorithms_for_planning_s_curve_motion_profiles.pdf Harmonic compares to 5th order: see fig. 10. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-01 17:41 ```Look at http://www.roymech.co.uk/Useful_Tables/Maths/fourier/Maths_Fourier_transforms.html and compare the fourier transform of: 1. top hat (sinc) 2. triangle (sinc^2) 3. gauss (gauss) trapezoidal is inbetween top hat and triangle. for accaleration function 1. top hat: jerc is unboundet 2. triangle and trapezoidal: jerc is constant 3. gauss: jerc is gauss _Conclusion:_ gaussian accaleration gives the _lowest_ frequency components! Unfortunately gaussian jerk is impossible. You have to wait infinite time to start the motion. Which function looks like a gaussian? The sinusoidal type in STR_V8_N1_07_MCH.pdf: Fig.1 (a) or 4267.pdf: page 8, upper trace is a candidate! Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Kenneth Lerman - 2012-05-01 18:11 ```My understanding of the goal is that it is to have a minimum time subject to certain constraints. The current constraints in LinuxCNC are to have certain acceleration and velocity limits. It has been suggested that it would also be appropriate to limit the jerk. Doing so could only make the time longer. For a variety of reasons (smaller error, less vibration, ...), some commercial machines add such constraints. I suggest that if enhance LinuxCNC to add the jerk constraints, that will be sufficient to "compete" with commercial solutions. =============== I've noticed that most of the papers I've looked at seem to discuss the problem of moving a single axis. Presumably, changing this to an N-axis solution is straight forward. LinuxCNC, though, handles both linear and circular interpolation. I've never looked closely enough at the design to understand how circular interpolation would change the problem. Regards, Ken On 5/1/2012 1:41 PM, Joachim Franek wrote: > Look at > http://www.roymech.co.uk/Useful_Tables/Maths/fourier/Maths_Fourier_transforms.html > and compare the fourier transform of: > 1. top hat (sinc) > 2. triangle (sinc^2) > 3. gauss (gauss) > > trapezoidal is inbetween top hat and triangle. > > for accaleration function > 1. top hat: jerc is unboundet > 2. triangle and trapezoidal: jerc is constant > 3. gauss: jerc is gauss > > _Conclusion:_ > gaussian accaleration gives the _lowest_ frequency components! > > Unfortunately gaussian jerk is impossible. You have to > wait infinite time to start the motion. > > Which function looks like a gaussian? > The sinusoidal type in > STR_V8_N1_07_MCH.pdf: Fig.1 (a) > or > 4267.pdf: page 8, upper trace > is a candidate! > > Joachim > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: dave - 2012-05-01 18:54 ```On Tue, 01 May 2012 14:11:15 -0400 Kenneth Lerman wrote: > My understanding of the goal is that it is to have a minimum time > subject to certain constraints. > > The current constraints in LinuxCNC are to have certain acceleration > and velocity limits. It has been suggested that it would also be > appropriate to limit the jerk. Doing so could only make the time > longer. For a variety of reasons (smaller error, less > vibration, ...), some commercial machines add such constraints. > > I suggest that if enhance LinuxCNC to add the jerk constraints, that > will be sufficient to "compete" with commercial solutions. > =============== > > I've noticed that most of the papers I've looked at seem to discuss > the problem of moving a single axis. Presumably, changing this to an > N-axis solution is straight forward. LinuxCNC, though, handles both > linear and circular interpolation. I've never looked closely enough > at the design to understand how circular interpolation would change > the problem. > > Regards, > > Ken > > On 5/1/2012 1:41 PM, Joachim Franek wrote: > > Look at > > http://www.roymech.co.uk/Useful_Tables/Maths/fourier/Maths_Fourier_transforms.html > > and compare the fourier transform of: > > 1. top hat (sinc) > > 2. triangle (sinc^2) > > 3. gauss (gauss) > > > > trapezoidal is inbetween top hat and triangle. > > > > for accaleration function > > 1. top hat: jerc is unboundet > > 2. triangle and trapezoidal: jerc is constant > > 3. gauss: jerc is gauss > > > > _Conclusion:_ > > gaussian accaleration gives the _lowest_ frequency components! > > > > Unfortunately gaussian jerk is impossible. You have to > > wait infinite time to start the motion. > > > > Which function looks like a gaussian? > > The sinusoidal type in > > STR_V8_N1_07_MCH.pdf: Fig.1 (a) > > or > > 4267.pdf: page 8, upper trace > > is a candidate! > > > > Joachim Ken, Joachim, et al My understanding is: decent acceleration within the mechanical constraints of the machine. gauss ( e^-k) in very simplified form as opposed to sine. I'm guessing that any smooth function will allow one to increase acceleration to achieve good performance without jerk, therefore it must be jerk limited. Smoothness is very desirable in trying to get good curves or surfaces. One takes their pick and pays their money. ;-) For many years now we've had the computing power to do control much more sophisticated than G1, G2 and G3. An italian group extended emc with nurbs several years ago and to my knowledge it doesn't get used much. With a really good 3D nurbs within emc there is the possibility of doing most surfacing in nurbs. However, this awaits an easy to understand in implement CAM. Sorry for the digression; just couldn't resist. Dave > > > > ------------------------------------------------------------------------------ > > Live Security Virtual Conference > > Exclusive live event will cover all the ways today's security and > > threat landscape has changed and how IT managers can respond. > > Discussions will include endpoint security, mobile security and the > > latest in malware threats. > > http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > > _______________________________________________ Emc-developers > > mailing list Emc-developers@... > > https://lists.sourceforge.net/lists/listinfo/emc-developers > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. > Discussions will include endpoint security, mobile security and the > latest in malware threats. > http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ Emc-developers > mailing list Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-01 19:41 ```On Tuesday 01 May 2012 20:11:15 Kenneth Lerman wrote: > My understanding of the goal is that it is to have a minimum time > subject to certain constraints. > > The current constraints in LinuxCNC are to have certain acceleration and > velocity limits. It has been suggested that it would also be appropriate > to limit the jerk. Doing so could only make the time longer. This is not allway true. It is true for stiff mechanics. The lowest resonant freuquency is _much_ higher compared to the exitation frequencies. In this case it makes no sense to reduce the acceleration. It is wrong for example for a gantry setup (2 y-axes). http://wiki.linuxcnc.org/cgi-bin/wiki.pl?FiveAxisGantry The traverse acts like a spring (bends) and the z-axis as a mass: -> a low frequency mechanical vibration is possible. With an unlimited jerc a movement of the 2 y-axes induces a frequency spectrum. If the 5th harmonic stimulates this mechanical resonance and the resulting vibration is not tolerable it is recommanded to reduce the amplitude of the high harmonics. With a jerk=f(t) it is possible to increase the acceleration because the higher frequency components are low in amplitude and do not stimulate machanical resonances. But the higher frequency components with low amplitude are always present. Good mechanics has some damping! > For a > variety of reasons (smaller error, less vibration, ...), some commercial > machines add such constraints. > > I suggest that if enhance LinuxCNC to add the jerk constraints, that > will be sufficient to "compete" with commercial solutions. This depends on the mechanical setup. > =============== > > I've noticed that most of the papers I've looked at seem to discuss the > problem of moving a single axis. Presumably, changing this to an N-axis > solution is straight forward. LinuxCNC, though, handles both linear and > circular interpolation. I've never looked closely enough at the design > to understand how circular interpolation would change the problem. > > Regards, > > Ken > Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Viesturs Lācis - 2012-05-01 19:51 ```2012/5/1 Joachim Franek : > >> For a >> variety of reasons (smaller error, less vibration, ...), some commercial >> machines add such constraints. >> >> I suggest that if enhance LinuxCNC to add the jerk constraints, that >> will be sufficient to "compete" with commercial solutions. > > This depends on the mechanical setup. If I understand this discussion correctly, then jerk limit can be set to very high number, so the machine would behave very similarly as if jerk was not limited at all. Viesturs ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-01 20:01 ```On Tuesday 01 May 2012 21:50:32 Viesturs Lācis wrote: > 2012/5/1 Joachim Franek : > > If I understand this discussion correctly, then jerk limit can be set > to very high number, so the machine would behave very similarly as if > jerk was not limited at all. > > Viesturs Look for example to 4267.pdf, page 8. If you reduce t1 towards zero (t1-t0) ->0, (t3-t2)->0 etc. you get a top hat as accelaration. This trigonometric model allows continously changing from nearly unlimited jerc to harmonic behaviour. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Alexey Starikovskiy - 2012-05-01 20:21 ```It is all good looking in theory, but how do you transfer this information to the servo drive? As I understand, now we have all calculations of trajectory done in servo thread with repeat rate equal to ~1kHz. This means that we calculate new data points for drive (either acceleration or velocity) only once per 1ms. It stays constant for this time period, and at best we have some ladder approximation of the A or V. How do you feet gauss or sine function for jerk into this setup? Alex On Wed, May 2, 2012 at 12:01 AM, Joachim Franek wrote: > On Tuesday 01 May 2012 21:50:32 Viesturs Lācis wrote: >> 2012/5/1 Joachim Franek : >> >> If I understand this discussion correctly, then jerk limit can be set >> to very high number, so the machine would behave very similarly as if >> jerk was not limited at all. >> >> Viesturs > > Look for example to 4267.pdf, page 8. > If you reduce t1 towards zero > (t1-t0) ->0, (t3-t2)->0 etc. > you get a top hat as accelaration. > This trigonometric model allows > continously changing from > nearly unlimited jerc to harmonic behaviour. > > Joachim > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-02 06:32 ```On Tuesday 01 May 2012 22:21:50 Alexey Starikovskiy wrote: > It is all good looking in theory, but how do you transfer this > information to the servo drive? > As I understand, now we have all calculations of trajectory done in > servo thread with repeat > rate equal to ~1kHz. This means that we calculate new data points for > drive (either acceleration > or velocity) only once per 1ms. It stays constant for this time > period, and at best we have some > ladder approximation of the A or V. How do you feet gauss or sine > function for jerk into this setup? From http://www.aspe.net/publications/Spring_2001/01Sp%20Extended%20Abstracts/Arevalo.PDF page 4, uppe left: "In a i486–DX4 type processor at 66 Mhz (with FPU integrated), running an ISR (interrupt service routine) that feeds a servo controller at a rate of 250 Hz, under a multitasking OS (Linux), there is not an appreciable difference for the computational load of calculating any of the profiles reviewed. " If this works 2001 I assume it is possible to get it working now. Or we have to switch to i486 processors. At this time I am not familiar enough with the existing code to make usefull recommendations. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Viesturs Lācis - 2012-05-02 06:51 ```2012/5/1 Alexey Starikovskiy : > It is all good looking in theory, but how do you transfer this > information to the servo drive? If I understand correctly the question, then that is "mission impossible 5" to implement jerk limitation in the feedback loop, closed by servo drive, because AFAIK almost all of them are running closed source firmware. Jerk limitation is available only in feedback loop, closed by LinuxCNC, and I think that it is sufficient - imagine the torque loop closed in servo drive and velocity loop closed in LinuxCNC - requested velocity from LinuxCNC will be jerk-limited, thus will have s-shape curve instead of trapezodial. It does not matter if servo drive has jerk limit or no, it is already there in requested velocity. Viesturs ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-02 07:29 ```On Tuesday 01 May 2012 21:40:50 Joachim Franek wrote: > It is wrong for example for a gantry setup (2 y-axes). > http://wiki.linuxcnc.org/cgi-bin/wiki.pl?FiveAxisGantry > The traverse acts like a spring (bends) and the z-axis as a mass: > -> a low frequency mechanical vibration is possible. > With an unlimited jerc a movement of the 2 y-axes induces > a frequency spectrum. If the 5th harmonic stimulates I must correct me: sinc is not harmonic. > this mechanical resonance and the resulting vibration is not tolerable > it is recommanded to reduce the amplitude of the high harmonics. > With a jerk=f(t) it is possible to increase the acceleration because > the higher frequency components are low in amplitude and > do not stimulate machanical resonances. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Alexey Starikovskiy - 2012-05-02 08:10 ```No, the question was quite simple. In your example of velocity-controlled drive, it gets command input once per ~1ms, thus you can't really give it any smooth curve as velocity, as there is no way to reliably smooth it between those points. You might have low-pass filter in some MESA card, you might have low-pass filter in the drive itself, and it may be any combination of them. What will be the typical (ac/de)celeration time? Is it much larger than 1ms and we could assume all the above is irrelevant? On Wed, May 2, 2012 at 10:50 AM, Viesturs Lācis wrote: > 2012/5/1 Alexey Starikovskiy : >> It is all good looking in theory, but how do you transfer this >> information to the servo drive? > > If I understand correctly the question, then that is "mission > impossible 5" to implement jerk limitation in the feedback loop, > closed by servo drive, because AFAIK almost all of them are running > closed source firmware. > Jerk limitation is available only in feedback loop, closed by > LinuxCNC, and I think that it is sufficient - imagine the torque loop > closed in servo drive and velocity loop closed in LinuxCNC - requested > velocity from LinuxCNC will be jerk-limited, thus will have s-shape > curve instead of trapezodial. It does not matter if servo drive has > jerk limit or no, it is already there in requested velocity. > > Viesturs > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: Viesturs Lācis - 2012-05-02 08:45 ```2012/5/2 Alexey Starikovskiy : > No, the question was quite simple. In your example of > velocity-controlled drive, it gets command > input once per ~1ms, thus you can't really give it any smooth curve as > velocity, as there is no way > to reliably smooth it between those points. You might have low-pass > filter in some MESA card, > you might have low-pass filter in the drive itself, and it may be any > combination of them. > What will be the typical (ac/de)celeration time? Is it much larger > than 1ms and we could assume > all the above is irrelevant? > Well, then I think that this is general question, not related to s-curve velocity profile. Practice shows that 1 ms is short enough that vast majority of motors with their inertia and load cannot achieve any significant acceleration during this one servo thread period. 1 ms is short enough that most machines with velocities below 10-15 m/min have smooth velocity profiles without additional filtering. I think that successful PID loop with a 1ms update interval proves that it is sufficient for vast majority of machines. I guess that very strong linear servo motors with accelerations of 5G and more will require faster update rate, but they are so expensive that I doubt any LinuxCNC users will be able to afford them in any near future. Anywy the servo period length can be adjusted as well. I myself have set servo cycle at 1,5 kHz on D525MW based PC. I have read that other users have achieved even better numbers. Viesturs ```

 Re: [Emc-developers] S-curve: my summary From: Anders Wallin - 2012-05-02 09:56 ```One should prefer numbers, experiments, and real data over speculation! Here's a picture that illustrates these ideas: http://www.anderswallin.net/sandbox/trajectory.jpg Note how the jerk-limited trajectory is slower than the acceleration-limited one. The jounce-limited ("double jerk") is slower still. Nevertheless these are often termed "time optimal" i.e. they are as fast as possible given the constraints. The spectra are fairly quiet from 200-600 Hz. My guess is the mechanical resonances of a machine are all below this, so the significant features of the spectra are a 2 to 3-fold lowering of the peaks around 50 and 200 Hz (?) This is from http://dx.doi.org/10.1007/s00170-011-3842-0 The math for exact-stop planning for jerk- or jounce-limited trajectories is not that hard and appears in many papers. The hard parts are: - blending ("cutting corners" within a tolerance) and/or lookahead (possibly many g-code lines per traj-period?) - how to handle feed-override >100%.(I suspect a feed-override change cannot be applied instantaneously) - non-trivkins (I suspect a jerk/jounce-limited traj-planner needs derivatives of the fwd/reverse kins?) Anders ```

 Re: [Emc-developers] S-curve: my summary From: andy pugh - 2012-05-02 10:09 ```On 2 May 2012 10:56, Anders Wallin wrote: > - non-trivkins (I suspect a jerk/jounce-limited traj-planner needs > derivatives of the fwd/reverse kins?) It is probably possible to determine these simply by running the kins twice with a small offset. In fact I have a feeling that the whole puzzle might be easier in a discrete-time situation (such as the servo thread) than attempts to determine an analytical solution would suggest. My current attempt is a very simplistic planner which basically asks the question "will it be too late to hit my endpoint boundary conditions within the jerk constraint if I don't start now" on every servo thread. I think this can be seen as an attempt to solve the equations using finite-difference methods, but running that in parallel with the motion. I have deliberately not looked at the araisrobo code, because I am doing this for fun :-) -- atp The idea that there is no such thing as objective truth is, quite simply, wrong. ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-02 10:25 ```On Wednesday 02 May 2012 11:56:30 Anders Wallin wrote: > http://www.anderswallin.net/sandbox/trajectory.jpg Are this calculated or measured spectra? Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Viesturs Lācis - 2012-05-02 10:32 ```2012/5/2 Anders Wallin : > One should prefer numbers, experiments, and real data over speculation! > Here's a picture that illustrates these ideas: > http://www.anderswallin.net/sandbox/trajectory.jpg What kind of animal is "jounce" and what does it eat in winter? :) Jerk is rate of change of acceleration. It seems that "jounce" is rate of change of jerk. Is there any practical use of it? Viesturs ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-02 10:51 ```Here is my fft code for octave: http://dl.dropbox.com/u/18713924/fftxyz5_emc.m Record and save in halscope the commanded acceleration and the real acceleration and compare. And we can see what we can achieve with jerc and jounce. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-02 11:45 ```On Wednesday 02 May 2012 12:31:34 Viesturs Lācis wrote: > Is there any practical use of it? http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-02 12:00 ```> http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf http://www.mmrc.iss.ac.cn/~xgao/cnc/vase.html http://www.mmrc.iss.ac.cn/~xgao/cnc/blade.html Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Viesturs Lācis - 2012-05-02 12:24 ```2012/5/2 Joachim Franek : > > http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf Thank You! Not sure that I understand, how exactly introducing "jounce" helps reducing vibrations. Jerk is clear to me - one good way to test it is, when driving a car: when You have to stop, press brake pedal with some force and do not change that force. At the moment, when velocity (car's speed) reaches 0, acceleration instantly jumps from some negative value to zero and everybody in the car will feel it - as if they were released and thrown back into seats. "Jounce" is something new and not yet clear, how it helps... > http://www.mmrc.iss.ac.cn/~xgao/cnc/vase.html That just feels little insane to me - last line number in that file is "116180". I saved that code in a file, I was surprised that its size is only 2,30 Mb. Hmm, maybe I will try running that code someday, to see, how the machine behaves on that. Viesturs ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-02 12:38 ```On Wednesday 02 May 2012 11:56:30 Anders Wallin wrote: > The math for exact-stop planning for jerk- or jounce-limited > trajectories is not that hard and appears in many papers. for 5 axis: http://www.mmrc.iss.ac.cn/~xgao/papernc/02-fan.pdf joachim ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-02 13:12 ```On Wed, 2 May 2012 15:23:25 +0300, Viesturs Lācis wrote: > > Thank You! Not sure that I understand, how exactly introducing > "jounce" helps reducing vibrations. > Jerk is clear to me - one good way to test it is, when driving a car: > when You have to stop, press brake pedal with some force and do not > change that force. At the moment, when velocity (car's speed) reaches > 0, acceleration instantly jumps from some negative value to zero and > everybody in the car will feel it - as if they were released and > thrown back into seats. > "Jounce" is something new and not yet clear, how it helps... If you think out it in terms of orders of differentiation it might help, or not, but I'm game to try... d0 (of motion) = pos d1 = velocity d2 = acceleration d3 = jerk d4 = jounce If you do not have any control over acceleration, so your movement jumps from one speed to the next, then you get lots of vibration in the mechanisms. If you allow some control over the transition you can cut down a lot of the vibration and can increase the accuracy of the machine. You can do this for increasingly higher order and dampen out increasingly higher order harmonics. If you look at the n'th derivative of whatever motion equation you are playing with, then as long as the equation is complex enough to take a n'th order derivative then, you can optimize it for smoothness. The end result is that it willl always tale a little longer to make higher order smooth transitions, but it will reduce more and more vibration in the machine. Does that help? If not then just ignore... EBo -- ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-02 15:13 ```Joachim Franek wrote: > On Wednesday 02 May 2012 11:56:30 Anders Wallin wrote: > >> http://www.anderswallin.net/sandbox/trajectory.jpg >> > > Are this calculated or measured spectra? > The spectrum is so full of noise all the way up and beyond the range of the chart, that it has to be measured on some kind of machine. Jon ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-02 15:22 ```Viesturs Lācis wrote: > Anywy the servo period length can be adjusted as well. I > myself have set servo cycle at 1,5 kHz on D525MW based PC. I have read > that other users have achieved even better numbers. > I have run at 2 and 4 KHz. 2 KHz is easy, with my parallel port-connected hardware, the 4 KHz starts to use up more of the SERVO_PERIOD, but it still works. Jon ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-02 15:52 ```On Wednesday 02 May 2012 17:12:56 Jon Elson wrote: > The spectrum is so full of noise all the way up and beyond the range > of the chart, that it has to be measured on some kind of machine. > > Jon > I found the source of this pictures: http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf And I am not sure what was measured. From this motion profile (0.1 to 0.2 s time to reach max. velocity) I do not expect frequency components in the 800 to 1200 Hz range. Another point is that the spectrum does not look alike sinc. But I may be totally wrong. My experience using a MMA7260 http://www.freescale.com/files/sensors/doc/data_sheet/MMA7260QT.pdf is that you have some spectral components without a move. Switching off the drives makes sometimes a remarkable difference. Conclusion: not all vibrations are movement induced. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-03 01:11 ```Joachim Franek wrote: > And I am not sure what was measured. From this motion > profile (0.1 to 0.2 s time to reach max. velocity) > I do not expect frequency components in the > 800 to 1200 Hz range. > Right, I think this is a bunch of noise from the servo drive, motor, encoder, etc. > Another point is that the spectrum does not look alike sinc. > But I may be totally wrong. > > My experience using a MMA7260 > http://www.freescale.com/files/sensors/doc/data_sheet/MMA7260QT.pdf > is that you have some spectral components > without a move. Switching off the drives > makes sometimes a remarkable difference. > > Conclusion: not all vibrations are movement induced. > Absolutely! Some things like jerk can trigger resonances in the machine and show up in the measured velocity, but there are all sorts of mechanical events in a real machine. There are motor cogging, belt teeth, ball nut balls rolling against the screw and nut, and the linear slide itself. The sum of all this is a very complex spectrum. Jon ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-03 01:26 ```On Wed, 02 May 2012 20:10:20 -0500, Jon Elson wrote: > Absolutely! Some things like jerk can trigger resonances in the > machine and > show up in the measured velocity, but there are all sorts of > mechanical > events in a real machine. There are motor cogging, belt teeth, ball > nut > balls rolling against the screw and nut, and the linear slide itself. > The sum of all this is a very complex spectrum. Hmmm... it would be nice if there was some way to sense the signatures and dump enough meta-info to then do some analysis to ferret out what the root cause might be. To many ideas... not enough time... EBo -- ```

 Re: [Emc-developers] S-curve: my summary From: Spiderdab <77dab@ti...> - 2012-05-03 03:35 ```Another nice and simple-to-understand link about how to calculate points of a bezier curve: http://cubic.org/docs/bezier.htm maybe it could be useful. ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-03 08:31 ```On Thursday 03 May 2012 03:26:15 EBo wrote: > Hmmm... it would be nice if there was some way to sense the signatures > and dump enough meta-info to then do some analysis to ferret out what > the root cause might be. To many ideas... not enough time... > > EBo -- After getting some experience with the interpretation of spectra you are able to identify the sources of vibrations. To get some impression how it looks like, generate some time series: x(t)=sin(0.01*t) y(t)=sin(0.02*t) z(t)=sin(0.01*t)+sin(0.02*t) and feed this to my octave fft code. You can also use some other fft code. Try to reproduce the results from the fourier web page. If this fits your expectations try to record with halscope some movement data and analyse it. Comfortable would be an integration of fft code in halscope to see instantaneous the result. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Anders Wallin - 2012-05-03 09:39 ```>> The spectrum is so full of noise all the way up and beyond the range >> of the chart, that it has to be measured on some kind of machine. >> Jon > I found the source of this pictures: > http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf > And I am not sure what was measured. From this motion > profile (0.1 to 0.2 s time to reach max. velocity) > I do not expect frequency components in the > 800 to 1200 Hz range. > Another point is that the spectrum does not look alike sinc. > But I may be totally wrong. The authors of that paper are not very clear on what exactly was measured, but I am sure it is not just an fft of the velocity-profile. Rather is a vibration measurement from somewhere on the machine. As such it might have components from the spindle, servo-drive, etc. If someone really wanted to get into this on a hobby level there might be acceleration-sensors (made for e.g. airbag-triggers on cars) which could be used? Simply HAL-scoping commanded/actual position and maybe looking at their spectra might be a good start too. AW ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-03 10:14 ```On Thu, 3 May 2012 12:39:15 +0300, Anders Wallin wrote: >>> The spectrum is so full of noise all the way up and beyond the >>> range >>> of the chart, that it has to be measured on some kind of machine. >>> Jon >> I found the source of this pictures: >> http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf >> And I am not sure what was measured. From this motion >> profile (0.1 to 0.2 s time to reach max. velocity) >> I do not expect frequency components in the >> 800 to 1200 Hz range. >> Another point is that the spectrum does not look alike sinc. >> But I may be totally wrong. > > The authors of that paper are not very clear on what exactly was > measured, but I am sure it is not just an fft of the > velocity-profile. > Rather is a vibration measurement from somewhere on the machine. As > such it might have components from the spindle, servo-drive, etc. > > If someone really wanted to get into this on a hobby level there > might > be acceleration-sensors (made for e.g. airbag-triggers on cars) which > could be used? > Simply HAL-scoping commanded/actual position and maybe looking at > their spectra might be a good start too. Actually a high end smart phone with the new fancy nano gyros + accelerometers might just do the trick. ```

 Re: [Emc-developers] S-curve: my summary From: Lars Segerlund - 2012-05-03 10:25 ```Buy some on sparkfun or similar, with the righte G force range, thera are a lot of them, and hooking one up on SPI or i2c is not complicated. The sensors for airbags are probably not right ... The DIY robotics guys use them a lot as far as I can understand. It will be a bugger to read though ....since you have machine vibration, tool acceleration and the stiffness of the machine ( resonance ? ) ... Check out Variax solution .... laser measurement in realtime of the lenght and deformations in the frame :-D ..... Coooool ...... I was thinking that hooking up a LIM or similar to a reference outside the machine might give better readings ? and simpler signal handling, since it gives absolute position ... It's not uncommon to have a LIM system with amplifiers lying around a machine shop . / regards, Lars Segerlund. 2012/5/3 EBo : > On Thu, 3 May 2012 12:39:15 +0300, Anders Wallin wrote: >>>> The spectrum is so full of noise all the way up and beyond the >>>> range >>>> of the chart, that it has to be measured on some kind of machine. >>>> Jon >>> I found the source of this pictures: >>> http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf >>> And I am not sure what was measured. From this motion >>> profile (0.1 to 0.2 s time to reach max. velocity) >>> I do not expect frequency components in the >>> 800 to 1200 Hz range. >>> Another point is that the spectrum does not look alike sinc. >>> But I may be totally wrong. >> >> The authors of that paper are not very clear on what exactly was >> measured, but I am sure it is not just an fft of the >> velocity-profile. >> Rather is a vibration measurement from somewhere on the machine. As >> such it might have components from the spindle, servo-drive, etc. >> >> If someone really wanted to get into this on a hobby level there >> might >> be acceleration-sensors (made for e.g. airbag-triggers on cars) which >> could be used? >> Simply HAL-scoping commanded/actual position and maybe looking at >> their spectra might be a good start too. > > Actually a high end smart phone with the new fancy nano gyros + > accelerometers might just do the trick. > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: Lars Segerlund - 2012-05-03 10:27 ```I meant LVDT ! ....... sorry ! / Lars.... 2012/5/3 Lars Segerlund : >  Buy some on sparkfun or similar, with the righte G force range, thera > are a lot of them, and hooking one up on SPI or i2c is not > complicated. > >  The sensors for airbags are probably not right ... > >  The DIY robotics guys use them a lot as far as I can understand. > >  It will be a bugger to read though ....since you have machine > vibration, tool acceleration and the stiffness of the machine ( > resonance ? ) ... >  Check out Variax solution .... laser measurement in realtime of the > lenght and deformations in the frame :-D ..... Coooool ...... > >  I was thinking that hooking up a LIM or similar to a reference > outside the machine might give better readings ? and simpler signal > handling, since it gives absolute position ... >  It's not uncommon to have a LIM system with amplifiers lying around a > machine shop . > >  / regards, Lars Segerlund. > > > 2012/5/3 EBo : >> On Thu, 3 May 2012 12:39:15 +0300, Anders Wallin wrote: >>>>> The spectrum is so full of noise all the way up and beyond the >>>>> range >>>>> of the chart, that it has to be measured on some kind of machine. >>>>> Jon >>>> I found the source of this pictures: >>>> http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf >>>> And I am not sure what was measured. From this motion >>>> profile (0.1 to 0.2 s time to reach max. velocity) >>>> I do not expect frequency components in the >>>> 800 to 1200 Hz range. >>>> Another point is that the spectrum does not look alike sinc. >>>> But I may be totally wrong. >>> >>> The authors of that paper are not very clear on what exactly was >>> measured, but I am sure it is not just an fft of the >>> velocity-profile. >>> Rather is a vibration measurement from somewhere on the machine. As >>> such it might have components from the spindle, servo-drive, etc. >>> >>> If someone really wanted to get into this on a hobby level there >>> might >>> be acceleration-sensors (made for e.g. airbag-triggers on cars) which >>> could be used? >>> Simply HAL-scoping commanded/actual position and maybe looking at >>> their spectra might be a good start too. >> >> Actually a high end smart phone with the new fancy nano gyros + >> accelerometers might just do the trick. >> >> ------------------------------------------------------------------------------ >> Live Security Virtual Conference >> Exclusive live event will cover all the ways today's security and >> threat landscape has changed and how IT managers can respond. Discussions >> will include endpoint security, mobile security and the latest in malware >> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ >> _______________________________________________ >> Emc-developers mailing list >> Emc-developers@... >> https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: andy pugh - 2012-05-03 10:27 ```On 3 May 2012 10:39, Anders Wallin wrote: > If someone really wanted to get into this on a hobby level there might > be acceleration-sensors (made for e.g. airbag-triggers on cars) which > could be used? Accelerometers are not especially expensive. http://www.sparkfun.com/products/9332 http://uk.rs-online.com/web/p/accelerometer/7167065/ Well, actually, they can be. The ones I used to glue dozens of to car bodies were £1000 each. -- atp The idea that there is no such thing as objective truth is, quite simply, wrong. ```

 Re: [Emc-developers] S-curve: my summary From: andy pugh - 2012-05-03 10:29 ```On 3 May 2012 11:25, Lars Segerlund wrote: >  It will be a bugger to read though It's a whole career. In fact it used to be all of my job, and is now part of my job. I am an NVH specialist. -- atp The idea that there is no such thing as objective truth is, quite simply, wrong. ```

 Re: [Emc-developers] S-curve: my summary From: Andrew - 2012-05-03 10:38 ```2012/5/3 andy pugh > On 3 May 2012 11:25, Lars Segerlund wrote: > > > It will be a bugger to read though > > It's a whole career. In fact it used to be all of my job, and is now > part of my job. I am an NVH specialist. > > Some guys are trying to get exact 6D position (or exact displacement) of a body (say, machine tool end effector) using accelerometers, gyro and magnitometers. I think the accuracy is very limited here, even 0.1mm is hard to obtain. Acceleration data is noisy, and integrating it twice will give a large error. What do you think about it? Andrew ```

 Re: [Emc-developers] S-curve: my summary From: Lars Segerlund - 2012-05-03 10:40 ```Me thinks LVDT ...... / Lars Segerlund. 2012/5/3 Andrew : > 2012/5/3 andy pugh > >> On 3 May 2012 11:25, Lars Segerlund wrote: >> >> >  It will be a bugger to read though >> >> It's a whole career. In fact it used to be all of my job, and is now >> part of my job. I am an NVH specialist. >> >> Some guys are trying to get exact 6D position (or exact displacement) of a > body (say, machine tool end effector) using accelerometers, gyro and > magnitometers. I think the accuracy is very limited here, even 0.1mm is > hard to obtain. > Acceleration data is noisy, and integrating it twice will give a large > error. > What do you think about it? > > Andrew > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: andy pugh - 2012-05-03 10:46 ```On 3 May 2012 11:37, Andrew wrote: > Acceleration data is noisy, and integrating it twice will give a large > error. > What do you think about it? I think you are correct. You will get limited accuracy, and also steady-state drift unless there is some secondary absolute feedback. -- atp The idea that there is no such thing as objective truth is, quite simply, wrong. ```

 Re: [Emc-developers] S-curve: my summary From: Andrew - 2012-05-03 11:12 ```2012/5/3 Lars Segerlund > Me thinks LVDT ...... > > Or linear encoders 2012/5/3 andy pugh > I think you are correct. You will get limited accuracy, and also > steady-state drift unless there is some secondary absolute feedback. I think that adding one or two struts with linear encoders should improve the accuracy much. Andrew ```

 Re: [Emc-developers] S-curve: my summary From: dave - 2012-05-03 14:29 ```On Thu, 3 May 2012 12:27:21 +0200 Lars Segerlund wrote: > I meant LVDT ! ....... sorry ! > > / Lars.... > > 2012/5/3 Lars Segerlund : > >  Buy some on sparkfun or similar, with the righte G force range, > > thera are a lot of them, and hooking one up on SPI or i2c is not > > complicated. > > > >  The sensors for airbags are probably not right ... > > > >  The DIY robotics guys use them a lot as far as I can understand. > > > >  It will be a bugger to read though ....since you have machine > > vibration, tool acceleration and the stiffness of the machine ( > > resonance ? ) ... > >  Check out Variax solution .... laser measurement in realtime of the > > lenght and deformations in the frame :-D ..... Coooool ...... > > > >  I was thinking that hooking up a LIM or similar to a reference > > outside the machine might give better readings ? and simpler signal > > handling, since it gives absolute position ... > >  It's not uncommon to have a LIM system with amplifiers lying > > around a machine shop . > > > >  / regards, Lars Segerlund. > > > > > > 2012/5/3 EBo : > >> On Thu, 3 May 2012 12:39:15 +0300, Anders Wallin wrote: > >>>>> The spectrum is so full of noise all the way up and beyond the > >>>>> range > >>>>> of the chart, that it has to be measured on some kind of > >>>>> machine. Jon > >>>> I found the source of this pictures: > >>>> http://www.mmrc.iss.ac.cn/~xgao/papernc/jounce20111122.pdf > >>>> And I am not sure what was measured. From this motion > >>>> profile (0.1 to 0.2 s time to reach max. velocity) > >>>> I do not expect frequency components in the > >>>> 800 to 1200 Hz range. > >>>> Another point is that the spectrum does not look alike sinc. > >>>> But I may be totally wrong. > >>> > >>> The authors of that paper are not very clear on what exactly was > >>> measured, but I am sure it is not just an fft of the > >>> velocity-profile. > >>> Rather is a vibration measurement from somewhere on the machine. > >>> As such it might have components from the spindle, servo-drive, > >>> etc. > >>> > >>> If someone really wanted to get into this on a hobby level there > >>> might > >>> be acceleration-sensors (made for e.g. airbag-triggers on cars) > >>> which could be used? > >>> Simply HAL-scoping commanded/actual position and maybe looking at > >>> their spectra might be a good start too. > >> > >> Actually a high end smart phone with the new fancy nano gyros + > >> accelerometers might just do the trick. I think the first thing I'd try would be a mirror on the machine and a laser at an angle. The trick will be a sensor with enough sensitivity to get good signals. First step might be audio output, the ear is pretty good at frequencies a machine is likely to vibrate out. Just thinking out of the box. Dave > >> > >> ------------------------------------------------------------------------------ > >> Live Security Virtual Conference > >> Exclusive live event will cover all the ways today's security and > >> threat landscape has changed and how IT managers can respond. > >> Discussions will include endpoint security, mobile security and > >> the latest in malware threats. > >> http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > >> _______________________________________________ Emc-developers > >> mailing list Emc-developers@... > >> https://lists.sourceforge.net/lists/listinfo/emc-developers > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. > Discussions will include endpoint security, mobile security and the > latest in malware threats. > http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ Emc-developers > mailing list Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-04 03:12 ```EBo wrote: > Hmmm... it would be nice if there was some way to sense the signatures > and dump enough meta-info to then do some analysis to ferret out what > the root cause might be. To many ideas... not enough time... > By running the machine at different speeds, you can sort out some resonances from periodic vibrations from motors and screws. A dynamic signal analyzer can excite the machine with pure sine waves of varying frequency and plot the amplitude/phase response and learn a lot more about what is going on. Jon ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-04 03:17 ```Anders Wallin wrote: > If someone really wanted to get into this on a hobby level there might > be acceleration-sensors (made for e.g. airbag-triggers on cars) which > could be used? > Simply HAL-scoping commanded/actual position and maybe looking at > their spectra might be a good start too. > Years ago there were vibration sensors from the oil exploration industry on the surplus market. You might be able to find these, still, at flea markets. Basically a magnet on springs inside a coil. A speaker with a bit of mass added will work fine, too. You need a modest amplifier to pick up the signal from the voice coil. head-arm motors out of a scrapped disk drive could also be used. Jon ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-04 03:22 ```On Thu, 03 May 2012 22:11:46 -0500, Jon Elson wrote: > EBo wrote: >> Hmmm... it would be nice if there was some way to sense the >> signatures >> and dump enough meta-info to then do some analysis to ferret out >> what >> the root cause might be. To many ideas... not enough time... >> > By running the machine at different speeds, you can sort out some > resonances > from periodic vibrations from motors and screws. A dynamic signal > analyzer > can excite the machine with pure sine waves of varying frequency and > plot > the amplitude/phase response and learn a lot more about what is going > on. That is half of what I was envisioning when I wrote the above. Another part after a characterization of a basic setup would be to compare that with data from running a real part -- and flagging which parts seem to cause the most feedback (vibration ,initera, etc.). Just brainstorming here, but could it be used as a tool to help tune a machine? EBo -- ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-04 03:35 ```On Thu, 03 May 2012 22:16:53 -0500, Jon Elson wrote: > Anders Wallin wrote: >> If someone really wanted to get into this on a hobby level there >> might >> be acceleration-sensors (made for e.g. airbag-triggers on cars) >> which >> could be used? >> Simply HAL-scoping commanded/actual position and maybe looking at >> their spectra might be a good start too. >> > Years ago there were vibration sensors from the oil exploration > industry > on the surplus > market. You might be able to find these, still, at flea markets. > Basically a magnet on > springs inside a coil. A speaker with a bit of mass added will work > fine, too. > You need a modest amplifier to pick up the signal from the voice > coil. > head-arm > motors out of a scrapped disk drive could also be used. this would be fun to experiment with... EBo -- ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-04 06:34 ```On Friday 04 May 2012 05:21:57 EBo wrote: > That is half of what I was envisioning when I wrote the above. Another > part after a characterization of a basic setup would be to compare that > with data from running a real part -- and flagging which parts seem to > cause the most feedback (vibration ,initera, etc.). Just brainstorming > here, but could it be used as a tool to help tune a machine? > > EBo -- > Yes. Knowing the resonances of the mechanics you can choose rotation speed of spindle times number of cutting edges not to hit this. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-04 06:48 ```On Fri, 4 May 2012 08:33:51 +0200, Joachim Franek wrote: > On Friday 04 May 2012 05:21:57 EBo wrote: >> That is half of what I was envisioning when I wrote the above. >> Another >> part after a characterization of a basic setup would be to compare >> that >> with data from running a real part -- and flagging which parts seem >> to >> cause the most feedback (vibration ,initera, etc.). Just >> brainstorming >> here, but could it be used as a tool to help tune a machine? > > Yes. Knowing the resonances of the mechanics you > can choose rotation speed of spindle times number of > cutting edges not to hit this. The tools to do this would be a cool addition... ```

 Re: [Emc-developers] S-curve: my summary From: Kent A. Reed - 2012-05-04 14:17 ```On 5/4/2012 2:48 AM, EBo wrote: > On Fri, 4 May 2012 08:33:51 +0200, Joachim Franek wrote: >> On Friday 04 May 2012 05:21:57 EBo wrote: >>> That is half of what I was envisioning when I wrote the above. >>> Another >>> part after a characterization of a basic setup would be to compare >>> that >>> with data from running a real part -- and flagging which parts seem >>> to >>> cause the most feedback (vibration ,initera, etc.). Just >>> brainstorming >>> here, but could it be used as a tool to help tune a machine? >> Yes. Knowing the resonances of the mechanics you >> can choose rotation speed of spindle times number of >> cutting edges not to hit this. > The tools to do this would be a cool addition... > > Vibration analysis and the sensors to drive it have been in the toolkit for advanced manufacturing for several decades. When I was still at NIST (claiming no credit, I was in a different line of work than the manufacturing engineering guys), major sensor-technology themes included detection of tool breakage and other faults, predictive maintenance, as well as active compensation for toolwear, differential thermal expansion, machine characteristics, etc. Google is your friend. And, I agree, the resulting tools are very cool. They contribute directly to quality and productivity. Regards, Kent ```

 Re: [Emc-developers] S-curve: my summary From: andy pugh - 2012-05-04 14:24 ```On 3 May 2012 11:27, andy pugh wrote: > Accelerometers are not especially expensive. Thinking about it, in the context of a machine tool a strain gauge applied to the frame would tell you all you need to know, and they are very cheap. Other uses would include measurement of cutting forces and frame deflection. http://uk.rs-online.com/web/p/strain-gauges/0632180/ -- atp The idea that there is no such thing as objective truth is, quite simply, wrong. ```

 Re: [Emc-developers] S-curve: my summary From: Kent A. Reed - 2012-05-04 14:30 ```On 5/4/2012 10:24 AM, andy pugh wrote: > On 3 May 2012 11:27, andy pugh wrote: > >> Accelerometers are not especially expensive. > Thinking about it, in the context of a machine tool a strain gauge > applied to the frame would tell you all you need to know, and they are > very cheap. Other uses would include measurement of cutting forces and > frame deflection. > > http://uk.rs-online.com/web/p/strain-gauges/0632180/ > It's getting a bit long in the tooth, but this 1989 survey of sensor options may still be useful http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA220405 Regards, Kent ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-04 15:11 ```On Fri, 04 May 2012 10:17:42 -0400, Kent A. Reed wrote: > On 5/4/2012 2:48 AM, EBo wrote: >> On Fri, 4 May 2012 08:33:51 +0200, Joachim Franek wrote: >>> On Friday 04 May 2012 05:21:57 EBo wrote: >>>> That is half of what I was envisioning when I wrote the above. >>>> Another >>>> part after a characterization of a basic setup would be to compare >>>> that >>>> with data from running a real part -- and flagging which parts >>>> seem >>>> to >>>> cause the most feedback (vibration ,initera, etc.). Just >>>> brainstorming >>>> here, but could it be used as a tool to help tune a machine? >>> Yes. Knowing the resonances of the mechanics you >>> can choose rotation speed of spindle times number of >>> cutting edges not to hit this. >> The tools to do this would be a cool addition... >> >> > Vibration analysis and the sensors to drive it have been in the > toolkit > for advanced manufacturing for several decades. > > When I was still at NIST (claiming no credit, I was in a different > line > of work than the manufacturing engineering guys), major > sensor-technology themes included detection of tool breakage and > other > faults, predictive maintenance, as well as active compensation for > toolwear, differential thermal expansion, machine characteristics, > etc. > Google is your friend. > > And, I agree, the resulting tools are very cool. They contribute > directly to quality and productivity. Yes, those projects are uber cool, and I am familiar with some, but not all, of the tools you mention. My real point was "it would be REALLY cool if EMC/LinuxCNC has these analytical tools in its arsenal." I would LOVE to be wrong, but I do not remember hearing anything about these in the forums, documentation, etc. BTW, thanks for the pointers ```

 Re: [Emc-developers] S-curve: my summary From: Kent A. Reed - 2012-05-04 16:07 ```On 5/4/2012 11:11 AM, EBo wrote: > Yes, those projects are uber cool, and I am familiar with some, but not > all, of the tools you mention. My real point was "it would be REALLY > cool if EMC/LinuxCNC has these analytical tools in its arsenal." I > would LOVE to be wrong, but I do not remember hearing anything about > these in the forums, documentation, etc. I'd guess this is due to lack of awareness that such tools could be incorporated. The EMC/LinuxCNC users are such a diverse bunch. At one end of the spectrum we have researchers using the available codebase to implement and test all manner of ideas (including some being discussed in this mail list). At the other we have punters like me with pedestrian table-top mills who just are just proud to be able to get it to make make chips fly on home projects. In between fall all manner of hobby/commercial/industrial users with all manner of machinery. I think demand would be greater if some principal benefits could be enunciated for the users group, out of which goals for new work items might be shaped (and dare I say it, prioritized). > BTW, thanks for the pointers You're welcome. Finding references about all I can contribute. Y'all are working faster than I can think:-) Regards, Kent ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-05 02:19 ```EBo wrote: > Just brainstorming > here, but could it be used as a tool to help tune a machine? > Yes, definitely. Some of the lower frequency stuff could be extracted with Halscope dumps. Jon ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-05 02:27 ```EBo wrote: > Yes, those projects are uber cool, and I am familiar with some, but not > all, of the tools you mention. My real point was "it would be REALLY > cool if EMC/LinuxCNC has these analytical tools in its arsenal." I > would LOVE to be wrong, but I do not remember hearing anything about > these in the forums, documentation, etc. > > I have a dynamic signal analyzer, but have not had stellar results using it. But, it seems that one could build such a thing in HAL. You'd need a swept sine wave generator to stimulate the system, and record the response of the system, both amplitude and phase to the stimulus. Just add the sine wave excitation to a constant position and feed this to the PID, and record position. Then, feed that to a Fourier analysis. Jon ```

 Re: [Emc-developers] S-curve: my summary From: Karl Cunningham - 2012-05-05 06:02 ```On 05/04/2012 07:26 PM, Jon Elson wrote: > EBo wrote: >> Yes, those projects are uber cool, and I am familiar with some, but not >> all, of the tools you mention. My real point was "it would be REALLY >> cool if EMC/LinuxCNC has these analytical tools in its arsenal." I >> would LOVE to be wrong, but I do not remember hearing anything about >> these in the forums, documentation, etc. >> >> > I have a dynamic signal analyzer, but have not had stellar results using it. > But, it seems that one could build such a thing in HAL. You'd need a swept > sine wave generator to stimulate the system, and record the response of > the system, both amplitude and phase to the stimulus. Just add the sine > wave excitation to a constant position and feed this to the PID, and > record position. Then, feed that to a Fourier analysis. Drawing on another one of my hobbies... Here is a description of some techniques for this sort of thing, including a link to a writeup from HP (from the 1960s) on using their wave analyzer. http://bnordgren.org/seismo/lpmeas3.pdf This has to do with seismometer design, but I think the principles are very similar. I have a few geophones (mentioned earlier in this thread) which might be usable for collecting vibration data. I don't have a good high-sample-rate ADC, but perhaps some scope traces would suffice. If someone has any specific tests they would like done with this on a 3-axis knee mill, I could give it a go. Karl ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-05 06:22 ```On Fri, 04 May 2012 22:44:32 -0700, Karl Cunningham wrote: > On 05/04/2012 07:26 PM, Jon Elson wrote: >> EBo wrote: >>> Yes, those projects are uber cool, and I am familiar with some, but >>> not >>> all, of the tools you mention. My real point was "it would be >>> REALLY >>> cool if EMC/LinuxCNC has these analytical tools in its arsenal." I >>> would LOVE to be wrong, but I do not remember hearing anything >>> about >>> these in the forums, documentation, etc. >>> >>> >> I have a dynamic signal analyzer, but have not had stellar results >> using it. >> But, it seems that one could build such a thing in HAL. You'd need >> a swept >> sine wave generator to stimulate the system, and record the response >> of >> the system, both amplitude and phase to the stimulus. Just add the >> sine >> wave excitation to a constant position and feed this to the PID, and >> record position. Then, feed that to a Fourier analysis. > > Drawing on another one of my hobbies... Here is a description of some > techniques for this sort of thing, including a link to a writeup from > HP > (from the 1960s) on using their wave analyzer. > > http://bnordgren.org/seismo/lpmeas3.pdf > > This has to do with seismometer design, but I think the principles > are > very similar. > > I have a few geophones (mentioned earlier in this thread) which might > be > usable for collecting vibration data. I don't have a good > high-sample-rate ADC, but perhaps some scope traces would suffice. First off. Thank you for the reference!!! This is a great information and a better idea ;-) Added with a little of simple statistics on top of fft/wavelet/etc. and I think a lot of mileage can be gained if I do not miss my bet... > If someone has any specific tests they would like done with this on a > 3-axis knee mill, I could give it a go. Where are you located? I'm currently working as a contractor at NASA's Goddard Space Flight Center, and currently have access to an old MaxNC-10 and some newer electronics (like a Gecko-520+motors AND Li's FPGA NURBS beta hardware _ motors + drives, etc.) So if we are within a couple hundred miles of each other then maybe we can trow some tests together... EBo -- ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-06 00:20 ```Karl Cunningham wrote: > On 05/04/2012 07:26 PM, Jon Elson wrote: > >> You'd need a swept >> sine wave generator to stimulate the system, and record the response of >> the system, both amplitude and phase to the stimulus. Just add the sine >> wave excitation to a constant position and feed this to the PID, and >> record position. Then, feed that to a Fourier analysis. > I have a few geophones (mentioned earlier in this thread) which might be > usable for collecting vibration data. I don't have a good > high-sample-rate ADC, but perhaps some scope traces would suffice. > > If someone has any specific tests they would like done with this on a > 3-axis knee mill, I could give it a go. > Actually, as I hinted above, I think this could be done with a script and features built into LinuxCNC. There is a waveform generator as part of HAL, and a module that will record traces from the RT component of Halscope into a numerical file. So, the only thing needed is a little piece of code to extract amplitude and phase info from the halscope trace and create a spreadsheet data file from the same measurement over a range of frequencies. So, if you have a machine with encoder feedback to the PC, you should be able to do this. Assuming a 1 KHz servo cycle, you should be able to detect resonances up to almost 500 Hz. Jon ```

 Re: [Emc-developers] S-curve From: Yishin Li - 2012-05-06 11:54 ```On Tue, May 1, 2012 at 3:34 AM, dave wrote: > On Mon, 30 Apr 2012 13:41:12 -0400 > EBo wrote: > > Sorry all of your text went away when I hit reply. > I just posted this in a reply to Andy. > > caris.mech.ubc.ca/Publications/sonja.pdf > > But alway remember: Good, Fast, Cheap ... pick any two. ;-) > > I tried Sonja's approach for final deceleration and approaching target. However, I can not get the same accel/vel curves while approaching target. Has anyone tried Sonja's method? -Yishin ```

 Re: [Emc-developers] S-curve: my summary From: Karl Cunningham - 2012-05-06 14:33 ```On 05/04/2012 11:22 PM, EBo wrote: > On Fri, 04 May 2012 22:44:32 -0700, Karl Cunningham wrote: >> On 05/04/2012 07:26 PM, Jon Elson wrote: >>> I have a dynamic signal analyzer, but have not had stellar results >>> using it. >>> But, it seems that one could build such a thing in HAL. You'd need >>> a swept >>> sine wave generator to stimulate the system, and record the response >>> of >>> the system, both amplitude and phase to the stimulus. Just add the >>> sine >>> wave excitation to a constant position and feed this to the PID, and >>> record position. Then, feed that to a Fourier analysis. >> >> Drawing on another one of my hobbies... Here is a description of some >> techniques for this sort of thing, including a link to a writeup from >> HP >> (from the 1960s) on using their wave analyzer. >> >> http://bnordgren.org/seismo/lpmeas3.pdf >> >> This has to do with seismometer design, but I think the principles >> are >> very similar. >> >> I have a few geophones (mentioned earlier in this thread) which might >> be >> usable for collecting vibration data. I don't have a good >> high-sample-rate ADC, but perhaps some scope traces would suffice. > > First off. Thank you for the reference!!! This is a great information > and a better idea ;-) Added with a little of simple statistics on top > of fft/wavelet/etc. and I think a lot of mileage can be gained if I do > not miss my bet... > >> If someone has any specific tests they would like done with this on a >> 3-axis knee mill, I could give it a go. > > Where are you located? I'm currently working as a contractor at NASA's > Goddard Space Flight Center, and currently have access to an old > MaxNC-10 and some newer electronics (like a Gecko-520+motors AND Li's > FPGA NURBS beta hardware _ motors + drives, etc.) So if we are within a > couple hundred miles of each other then maybe we can trow some tests > together... I'm a bit further away -- San Diego. If this can be made to work, it's possible these techniques could also be used to help tune PID coefficients. Or at least point out where there is marginal loop stability. Karl ```

 Re: [Emc-developers] S-curve From: dave - 2012-05-06 14:38 ```On Sun, 6 May 2012 19:53:58 +0800 Yishin Li wrote: > On Tue, May 1, 2012 at 3:34 AM, dave wrote: > > > On Mon, 30 Apr 2012 13:41:12 -0400 > > EBo wrote: > > > > Sorry all of your text went away when I hit reply. > > I just posted this in a reply to Andy. > > > > caris.mech.ubc.ca/Publications/sonja.pdf > > > > But alway remember: Good, Fast, Cheap ... pick any two. ;-) > > > > > I tried Sonja's approach for final deceleration and approaching > target. However, I can not get the same accel/vel curves while > approaching target. > > Has anyone tried Sonja's method? > > -Yishin Hi, Any idea what is limiting the accel, etc? Of course, the real question is how smooth is it and then can you speed it up? Thanks for trying it. Dave > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. > Discussions will include endpoint security, mobile security and the > latest in malware threats. > http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ Emc-developers > mailing list Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve From: Yishin Li - 2012-05-06 15:40 ```On Sun, May 6, 2012 at 10:38 PM, dave wrote: > On Sun, 6 May 2012 19:53:58 +0800 > Yishin Li wrote: > > > On Tue, May 1, 2012 at 3:34 AM, dave wrote: > > > > > On Mon, 30 Apr 2012 13:41:12 -0400 > > > EBo wrote: > > > > > > Sorry all of your text went away when I hit reply. > > > I just posted this in a reply to Andy. > > > > > > caris.mech.ubc.ca/Publications/sonja.pdf > > > > > > But alway remember: Good, Fast, Cheap ... pick any two. ;-) > > > > > > > > I tried Sonja's approach for final deceleration and approaching > > target. However, I can not get the same accel/vel curves while > > approaching target. > > > > Has anyone tried Sonja's method? > > > > -Yishin > > Hi, > Any idea what is limiting the accel, etc? Of course, the real question > is how smooth is it and then can you speed it up? Thanks for trying it. > > Instead of follow the whole Sonja's algorithm sets, I just tried to limit the final accel with sin template. (3.2' (modified 3.2)): a(t) = 0.5 * acc * sin(PI * t /dtmax - PI/2) - 0.5 * acc (3.4) dtmax = 0.5 * PI * acc / jmax I think I made mistakes for integrating 3.2' for vel(t) and pos(t). I would like to learn how to use octave to integrate 3.2'. I'm newbie to octave. Any guide would be appreciate. -Yishin ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-05-06 15:47 ```On Sunday 06 May 2012 17:40:25 Yishin Li wrote: > On Sun, May 6, 2012 at 10:38 PM, dave wrote: > > > On Sun, 6 May 2012 19:53:58 +0800 > > Yishin Li wrote: > > > > > On Tue, May 1, 2012 at 3:34 AM, dave wrote: > > > > > > > On Mon, 30 Apr 2012 13:41:12 -0400 > > > > EBo wrote: > > > > > > > > Sorry all of your text went away when I hit reply. > > > > I just posted this in a reply to Andy. > > > > > > > > caris.mech.ubc.ca/Publications/sonja.pdf > > > > > > > > But alway remember: Good, Fast, Cheap ... pick any two. ;-) > > > > > > > > > > > I tried Sonja's approach for final deceleration and approaching > > > target. However, I can not get the same accel/vel curves while > > > approaching target. > > > > > > Has anyone tried Sonja's method? > > > > > > -Yishin > > > > Hi, > > Any idea what is limiting the accel, etc? Of course, the real question > > is how smooth is it and then can you speed it up? Thanks for trying it. > > > > > Instead of follow the whole Sonja's algorithm sets, I just tried to limit > the final accel with sin template. > (3.2' (modified 3.2)): a(t) = 0.5 * acc * sin(PI * t /dtmax - PI/2) - 0.5 * > acc > (3.4) dtmax = 0.5 * PI * acc / jmax > I think I made mistakes for integrating 3.2' for vel(t) and pos(t). > > I would like to learn how to use octave to integrate 3.2'. > I'm newbie to octave. Any guide would be appreciate. > > -Yishin Octave is primary for numerical math. Try maxima, a CAS (computer algebra system). http://maxima.sourceforge.net/ Joachim > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers > ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-06 16:54 ```On Sun, 06 May 2012 07:33:39 -0700, Karl Cunningham wrote: > On 05/04/2012 11:22 PM, EBo wrote: >> On Fri, 04 May 2012 22:44:32 -0700, Karl Cunningham wrote: >>> On 05/04/2012 07:26 PM, Jon Elson wrote: >>>> I have a dynamic signal analyzer, but have not had stellar results >>>> using it. >>>> But, it seems that one could build such a thing in HAL. You'd >>>> need >>>> a swept >>>> sine wave generator to stimulate the system, and record the >>>> response >>>> of >>>> the system, both amplitude and phase to the stimulus. Just add >>>> the >>>> sine >>>> wave excitation to a constant position and feed this to the PID, >>>> and >>>> record position. Then, feed that to a Fourier analysis. >>> >>> Drawing on another one of my hobbies... Here is a description of >>> some >>> techniques for this sort of thing, including a link to a writeup >>> from >>> HP >>> (from the 1960s) on using their wave analyzer. >>> >>> http://bnordgren.org/seismo/lpmeas3.pdf >>> >>> This has to do with seismometer design, but I think the principles >>> are >>> very similar. >>> >>> I have a few geophones (mentioned earlier in this thread) which >>> might >>> be >>> usable for collecting vibration data. I don't have a good >>> high-sample-rate ADC, but perhaps some scope traces would suffice. >> >> First off. Thank you for the reference!!! This is a great >> information >> and a better idea ;-) Added with a little of simple statistics on >> top >> of fft/wavelet/etc. and I think a lot of mileage can be gained if I >> do >> not miss my bet... >> >>> If someone has any specific tests they would like done with this on >>> a >>> 3-axis knee mill, I could give it a go. >> >> Where are you located? I'm currently working as a contractor at >> NASA's >> Goddard Space Flight Center, and currently have access to an old >> MaxNC-10 and some newer electronics (like a Gecko-520+motors AND >> Li's >> FPGA NURBS beta hardware _ motors + drives, etc.) So if we are >> within a >> couple hundred miles of each other then maybe we can trow some tests >> together... > > I'm a bit further away -- San Diego. > > If this can be made to work, it's possible these techniques could > also > be used to help tune PID coefficients. Or at least point out where > there > is marginal loop stability. Sitting down together at a machine might be a bit logistically challenged... We could meet half way in between when I travel back to central Lousiana, but still logistically challenged. It is also to bad that I no longer have access to the 14bit 300Ms/s ADCs I got to play with a little in the past. That would probably tell us a little bit ;-) But seriously, I think it would be nice if we could come up with some various instruments and examples as an add on toolbox. EBo -- ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-05-06 17:06 ```On Sunday 06 May 2012 17:47:28 Joachim Franek wrote: > > On Sunday 06 May 2012 17:40:25 Yishin Li wrote: > > On Sun, May 6, 2012 at 10:38 PM, dave wrote: > > > > > On Sun, 6 May 2012 19:53:58 +0800 > > > Yishin Li wrote: > > > > > > > On Tue, May 1, 2012 at 3:34 AM, dave wrote: > > > > > > > > > On Mon, 30 Apr 2012 13:41:12 -0400 > > > > > EBo wrote: > > > > > > > > > > Sorry all of your text went away when I hit reply. > > > > > I just posted this in a reply to Andy. > > > > > > > > > > caris.mech.ubc.ca/Publications/sonja.pdf > > > > > > > > > > But alway remember: Good, Fast, Cheap ... pick any two. ;-) > > > > > > > > > > > > > > I tried Sonja's approach for final deceleration and approaching > > > > target. However, I can not get the same accel/vel curves while > > > > approaching target. > > > > > > > > Has anyone tried Sonja's method? > > > > > > > > -Yishin > > > > > > Hi, > > > Any idea what is limiting the accel, etc? Of course, the real question > > > is how smooth is it and then can you speed it up? Thanks for trying it. > > > > > > > > Instead of follow the whole Sonja's algorithm sets, I just tried to limit > > the final accel with sin template. > > (3.2' (modified 3.2)): a(t) = 0.5 * acc * sin(PI * t /dtmax - PI/2) - 0.5 * > > acc > > (3.4) dtmax = 0.5 * PI * acc / jmax > > I think I made mistakes for integrating 3.2' for vel(t) and pos(t). > > > > I would like to learn how to use octave to integrate 3.2'. > > I'm newbie to octave. Any guide would be appreciate. > > > > -Yishin > > Octave is primary for numerical math. > Try maxima, a CAS (computer algebra system). > http://maxima.sourceforge.net/ > > Joachim > jjf@...:~\$ maxima Maxima 5.22.1 http://maxima.sourceforge.net using Lisp GNU Common Lisp (GCL) GCL 2.6.7 (a.k.a. GCL) Distributed under the GNU Public License. See the file COPYING. Dedicated to the memory of William Schelter. The function bug_report() provides bug reporting information. (%i1) (amax / 2) * sin((%pi * t / dtmax) -(%pi / 2) ) + amax /2 ; %pi t amax cos(-----) amax dtmax (%o1) ---- - --------------- 2 2 (%i2) integrate(%o1,t) ; %pi t amax dtmax sin(-----) amax t dtmax (%o2) ------ - --------------------- 2 2 %pi (%i3) integrate(%o2,t); 2 %pi t amax dtmax cos(-----) 2 dtmax amax t (%o3) ---------------------- + ------- 2 4 2 %pi (%i4) (%i4) fortran(%o2); amax*t/2.0E+0-amax*dtmax*sin(%pi*t/dtmax)/%pi/2.0E+0 (%o4) done (%i5) fortran(%o3); amax*dtmax**2*cos(%pi*t/dtmax)/%pi**2/2.0E+0+amax*t**2/4.0E+0 (%o5) done Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-06 17:18 ```Karl Cunningham wrote: > > If this can be made to work, it's possible these techniques could also > be used to help tune PID coefficients. Or at least point out where there > is marginal loop stability. > Absolutely! A Bode plot of gain and phase vs. frequency tells everything about loop performance and stability, as well as pointing out additional resonances in the system. Automatically generating tuning parameters from this data is a bit above my mathematical abilities, but getting a Bode plot would be a great guide to manually tweaking the PID settings. I think it would be fairly easy to write a decent tuning guide with step by step procedures for obtaining much improved tuning. Jon ```

 Re: [Emc-developers] S-curve: my summary From: Jon Elson - 2012-05-06 17:27 ```EBo wrote: > > It is also to bad that I no longer have access to the 14bit 300Ms/s > ADCs I got to play with a little in the past. That would probably tell > us a little bit ;-) > A digital scope with some kind of computer connectivity (network, USB thumb drive, USB device) should do fine. Since the LinuxCNC servo rate is only one KHz, a trace with a 10 KHz sampling rate should be quite adequate. But, really, recording encoder position with Halscope is quite good. Jon ```

 Re: [Emc-developers] S-curve: my summary From: dave - 2012-05-06 17:33 ```On Sun, 06 May 2012 12:54:28 -0400 EBo wrote: > On Sun, 06 May 2012 07:33:39 -0700, Karl Cunningham wrote: > > On 05/04/2012 11:22 PM, EBo wrote: > >> On Fri, 04 May 2012 22:44:32 -0700, Karl Cunningham wrote: > >>> On 05/04/2012 07:26 PM, Jon Elson wrote: > >>>> I have a dynamic signal analyzer, but have not had stellar > >>>> results using it. > >>>> But, it seems that one could build such a thing in HAL. You'd > >>>> need > >>>> a swept > >>>> sine wave generator to stimulate the system, and record the > >>>> response > >>>> of > >>>> the system, both amplitude and phase to the stimulus. Just add > >>>> the > >>>> sine > >>>> wave excitation to a constant position and feed this to the PID, > >>>> and > >>>> record position. Then, feed that to a Fourier analysis. > >>> > >>> Drawing on another one of my hobbies... Here is a description of > >>> some > >>> techniques for this sort of thing, including a link to a writeup > >>> from > >>> HP > >>> (from the 1960s) on using their wave analyzer. > >>> > >>> http://bnordgren.org/seismo/lpmeas3.pdf > >>> > >>> This has to do with seismometer design, but I think the principles > >>> are > >>> very similar. > >>> > >>> I have a few geophones (mentioned earlier in this thread) which > >>> might > >>> be > >>> usable for collecting vibration data. I don't have a good > >>> high-sample-rate ADC, but perhaps some scope traces would suffice. > >> > >> First off. Thank you for the reference!!! This is a great > >> information > >> and a better idea ;-) Added with a little of simple statistics on > >> top > >> of fft/wavelet/etc. and I think a lot of mileage can be gained if > >> I do > >> not miss my bet... > >> > >>> If someone has any specific tests they would like done with this > >>> on a > >>> 3-axis knee mill, I could give it a go. > >> > >> Where are you located? I'm currently working as a contractor at > >> NASA's > >> Goddard Space Flight Center, and currently have access to an old > >> MaxNC-10 and some newer electronics (like a Gecko-520+motors AND > >> Li's > >> FPGA NURBS beta hardware _ motors + drives, etc.) So if we are > >> within a > >> couple hundred miles of each other then maybe we can trow some > >> tests together... > > > > I'm a bit further away -- San Diego. > > > > If this can be made to work, it's possible these techniques could > > also > > be used to help tune PID coefficients. Or at least point out where > > there > > is marginal loop stability. > > Sitting down together at a machine might be a bit logistically > challenged... We could meet half way in between when I travel back to > central Lousiana, but still logistically challenged. > > It is also to bad that I no longer have access to the 14bit 300Ms/s > ADCs I got to play with a little in the past. That would probably > tell us a little bit ;-) > > But seriously, I think it would be nice if we could come up with some > various instruments and examples as an add on toolbox. > > EBo -- > Meet at Wichita? Stuart is a great host. Maybe I'm being too simplistic but hal should do this. ... generate a sine wave feed to the amp. Monitor response with halscope. Then plot. How you adjust parameters from the bode I'm not certain but this is really a pretty bright group and someone can provide guidance. :-) Dave > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. > Discussions will include endpoint security, mobile security and the > latest in malware threats. > http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ Emc-developers > mailing list Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: Joachim Franek - 2012-05-06 18:09 ```Joachim Franek PIBF (Physik Ingenieurbüro Franek) Ernst-Reuter-Str. 8 63486 Bruchköbel 06181-907353 USt.-IdNr: DE187574683 On Sunday 06 May 2012 19:33:30 dave wrote: > On Sun, 06 May 2012 12:54:28 -0400 > EBo wrote: > > > On Sun, 06 May 2012 07:33:39 -0700, Karl Cunningham wrote: > > > On 05/04/2012 11:22 PM, EBo wrote: > > >> On Fri, 04 May 2012 22:44:32 -0700, Karl Cunningham wrote: > > >>> On 05/04/2012 07:26 PM, Jon Elson wrote: > > >>>> I have a dynamic signal analyzer, but have not had stellar > > >>>> results using it. > > >>>> But, it seems that one could build such a thing in HAL. You'd > > >>>> need > > >>>> a swept > > >>>> sine wave generator to stimulate the system, and record the > > >>>> response > > >>>> of > > >>>> the system, both amplitude and phase to the stimulus. Just add > > >>>> the > > >>>> sine > > >>>> wave excitation to a constant position and feed this to the PID, > > >>>> and > > >>>> record position. Then, feed that to a Fourier analysis. > > >>> > > >>> Drawing on another one of my hobbies... Here is a description of > > >>> some > > >>> techniques for this sort of thing, including a link to a writeup > > >>> from > > >>> HP > > >>> (from the 1960s) on using their wave analyzer. > > >>> > > >>> http://bnordgren.org/seismo/lpmeas3.pdf > > >>> > > >>> This has to do with seismometer design, but I think the principles > > >>> are > > >>> very similar. > > >>> > > >>> I have a few geophones (mentioned earlier in this thread) which > > >>> might > > >>> be > > >>> usable for collecting vibration data. I don't have a good > > >>> high-sample-rate ADC, but perhaps some scope traces would suffice. > > >> > > >> First off. Thank you for the reference!!! This is a great > > >> information > > >> and a better idea ;-) Added with a little of simple statistics on > > >> top > > >> of fft/wavelet/etc. and I think a lot of mileage can be gained if > > >> I do > > >> not miss my bet... > > >> > > >>> If someone has any specific tests they would like done with this > > >>> on a > > >>> 3-axis knee mill, I could give it a go. > > >> > > >> Where are you located? I'm currently working as a contractor at > > >> NASA's > > >> Goddard Space Flight Center, and currently have access to an old > > >> MaxNC-10 and some newer electronics (like a Gecko-520+motors AND > > >> Li's > > >> FPGA NURBS beta hardware _ motors + drives, etc.) So if we are > > >> within a > > >> couple hundred miles of each other then maybe we can trow some > > >> tests together... > > > > > > I'm a bit further away -- San Diego. > > > > > > If this can be made to work, it's possible these techniques could > > > also > > > be used to help tune PID coefficients. Or at least point out where > > > there > > > is marginal loop stability. > > > > Sitting down together at a machine might be a bit logistically > > challenged... We could meet half way in between when I travel back to > > central Lousiana, but still logistically challenged. > > > > It is also to bad that I no longer have access to the 14bit 300Ms/s > > ADCs I got to play with a little in the past. That would probably > > tell us a little bit ;-) > > > > But seriously, I think it would be nice if we could come up with some > > various instruments and examples as an add on toolbox. > > > > EBo -- > > > Meet at Wichita? Stuart is a great host. > > Maybe I'm being too simplistic but hal > should do this. ... generate a sine wave feed to the amp. Monitor > response with halscope. Then plot. How you adjust parameters from the > bode I'm not certain but this is really a pretty bright group and > someone can provide guidance. :-) > > Dave For one frequency from siggen you get a response from the drive: also a sine function. Recording both with halscope you have to measure ampitude and phase: use levenberg-marquardt form octave for this. https://staff.ti.bfh.ch/sha1/Octave/index/f/leasqr.html http://en.wikipedia.org/wiki/Levenberg%E2%80%93Marquardt_algorithm https://engineering.purdue.edu/kak/courses-i-teach/ECE661/HW5_LM_handout.pdf (Look for other libs etc the wiki page: implementations.) Result from lm-algo is amplitude and phase shift. Repeat this step for other frequencies and plot aplitude and phaseshift as function of frequency. Joachim ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-06 20:47 ```On Sun, 6 May 2012 10:33:30 -0700, dave wrote: > On Sun, 06 May 2012 12:54:28 -0400 > EBo wrote: > >> Sitting down together at a machine might be a bit logistically >> challenged... We could meet half way in between when I travel back >> to >> central Lousiana, but still logistically challenged. >> >> It is also to bad that I no longer have access to the 14bit 300Ms/s >> ADCs I got to play with a little in the past. That would probably >> tell us a little bit ;-) >> >> But seriously, I think it would be nice if we could come up with >> some >> various instruments and examples as an add on toolbox. > > Meet at Wichita? Stuart is a great host. > > Maybe I'm being too simplistic but hal > should do this. ... generate a sine wave feed to the amp. Monitor > response with halscope. Then plot. How you adjust parameters from the > bode I'm not certain but this is really a pretty bright group and > someone can provide guidance. :-) Might could work, but I would have to check with work, etc. How long do you think a hack session would take? This would take some planning, and I would have to see if I could any machines, or just help in the hackathon. Would Stuart be willing to host? EBo -- ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-06 20:59 ```On Sun, 6 May 2012 20:08:59 +0200, Joachim Franek wrote: > Joachim Franek > > PIBF (Physik Ingenieurbüro Franek) > Ernst-Reuter-Str. 8 > 63486 Bruchköbel > 06181-907353 > USt.-IdNr: DE187574683 Meeting half way would likely get a very wet and/or COLD! Reykjavík anyone 8-) > For one frequency from siggen you get a response from > the drive: also a sine function. Recording both with halscope > you have to measure ampitude and phase: > use levenberg-marquardt form octave for this. > https://staff.ti.bfh.ch/sha1/Octave/index/f/leasqr.html > > http://en.wikipedia.org/wiki/Levenberg%E2%80%93Marquardt_algorithm > > https://engineering.purdue.edu/kak/courses-i-teach/ECE661/HW5_LM_handout.pdf > (Look for other libs etc the wiki page: implementations.) > > Result from lm-algo is amplitude and phase shift. > > Repeat this step for other frequencies and > plot aplitude and phaseshift as function of frequency. These look like very good suggestions and references. Now if I could make enough time around work to play with this it would be fun, but work beckons -- back to the radar processing... ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-06 21:13 ```On Sun, 06 May 2012 12:26:42 -0500, Jon Elson wrote: > EBo wrote: >> >> It is also to bad that I no longer have access to the 14bit 300Ms/s >> ADCs I got to play with a little in the past. That would probably >> tell >> us a little bit ;-) >> > A digital scope with some kind of computer connectivity (network, USB > thumb drive, > USB device) should do fine. Since the LinuxCNC servo rate is only > one > KHz, a > trace with a 10 KHz sampling rate should be quite adequate. But, > really, > recording encoder position with Halscope is quite good. That is assuming that you have a machine with servo motors with an encoders, and my machines are all steppers with various types of drives. Also, sometime back in the tread we were talking about attaching various sensors to a machine instead of singling on hal. That being said, a recordable digital oscope should be fine, and for on the cheap we might also be able to get away with the DAC's on an Arduino (10 bits, and I thought I read somewhere they can be sampled as high as 77KHz by tweaking the clock downscaler, although I am not sure if the Arduino is up to the task -- would be cool if it was though. ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-06 21:14 ```On Sun, 06 May 2012 12:17:36 -0500, Jon Elson wrote: > Karl Cunningham wrote: >> >> If this can be made to work, it's possible these techniques could >> also >> be used to help tune PID coefficients. Or at least point out where >> there >> is marginal loop stability. >> > Absolutely! A Bode plot of gain and phase vs. frequency tells > everything > about loop performance and stability, as well as pointing out > additional > resonances in the system. Automatically generating tuning parameters > from this data is a bit above my mathematical abilities, but getting > a > Bode plot would be a great guide to manually tweaking the PID > settings. I think it would be fairly easy to write a decent tuning > guide > with step by step procedures for obtaining much improved tuning. that would be cool. ```

 Re: [Emc-developers] S-curve: my summary From: dave - 2012-05-07 01:02 ```On Sun, 06 May 2012 16:47:20 -0400 EBo wrote: > On Sun, 6 May 2012 10:33:30 -0700, dave wrote: > > On Sun, 06 May 2012 12:54:28 -0400 > > EBo wrote: > > > >> Sitting down together at a machine might be a bit logistically > >> challenged... We could meet half way in between when I travel back > >> to > >> central Lousiana, but still logistically challenged. > >> > >> It is also to bad that I no longer have access to the 14bit 300Ms/s > >> ADCs I got to play with a little in the past. That would probably > >> tell us a little bit ;-) > >> > >> But seriously, I think it would be nice if we could come up with > >> some > >> various instruments and examples as an add on toolbox. > > > > Meet at Wichita? Stuart is a great host. > > > > Maybe I'm being too simplistic but hal > > should do this. ... generate a sine wave feed to the amp. Monitor > > response with halscope. Then plot. How you adjust parameters from > > the bode I'm not certain but this is really a pretty bright group > > and someone can provide guidance. :-) > > Might could work, but I would have to check with work, etc. How long > do you think a hack session would take? This would take some > planning, and I would have to see if I could any machines, or just > help in the hackathon. > > Would Stuart be willing to host? > > EBo -- > Might could! Oh! That is right out of TX/NM. :-) D > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. > Discussions will include endpoint security, mobile security and the > latest in malware threats. > http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ Emc-developers > mailing list Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```

 Re: [Emc-developers] S-curve: my summary From: EBo - 2012-05-07 01:16 ```On Sun, 6 May 2012 18:02:44 -0700, dave wrote: > On Sun, 06 May 2012 16:47:20 -0400 > EBo wrote: > >> On Sun, 6 May 2012 10:33:30 -0700, dave wrote: >> > On Sun, 06 May 2012 12:54:28 -0400 >> > EBo wrote: >> > >> >> Sitting down together at a machine might be a bit logistically >> >> challenged... We could meet half way in between when I travel >> back >> >> to >> >> central Lousiana, but still logistically challenged. >> >> >> >> It is also to bad that I no longer have access to the 14bit >> 300Ms/s >> >> ADCs I got to play with a little in the past. That would >> probably >> >> tell us a little bit ;-) >> >> >> >> But seriously, I think it would be nice if we could come up with >> >> some >> >> various instruments and examples as an add on toolbox. >> > >> > Meet at Wichita? Stuart is a great host. >> > >> > Maybe I'm being too simplistic but hal >> > should do this. ... generate a sine wave feed to the amp. Monitor >> > response with halscope. Then plot. How you adjust parameters from >> > the bode I'm not certain but this is really a pretty bright group >> > and someone can provide guidance. :-) >> >> Might could work, but I would have to check with work, etc. How >> long >> do you think a hack session would take? This would take some >> planning, and I would have to see if I could any machines, or just >> help in the hackathon. >> >> Would Stuart be willing to host? >> >> EBo -- >> > > Might could! Oh! That is right out of TX/NM. :-) if NM is an option, How about Albuquerque? ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-05-07 19:23 ```> jjf@...:~\$ maxima > > Maxima 5.22.1 http://maxima.sourceforge.net > using Lisp GNU Common Lisp (GCL) GCL 2.6.7 (a.k.a. GCL) > Distributed under the GNU Public License. See the file COPYING. > Dedicated to the memory of William Schelter. > The function bug_report() provides bug reporting information. > > (%i1) (amax / 2) * sin((%pi * t / dtmax) -(%pi / 2) ) + amax /2 ; > %pi t > amax cos(-----) > amax dtmax > (%o1) ---- - --------------- > 2 2 > (%i2) integrate(%o1,t) > ; > %pi t > amax dtmax sin(-----) > amax t dtmax > (%o2) ------ - --------------------- > 2 2 %pi > (%i3) integrate(%o2,t); > 2 %pi t > amax dtmax cos(-----) 2 > dtmax amax t > (%o3) ---------------------- + ------- > 2 4 > 2 %pi > (%i4) > > (%i4) fortran(%o2); > amax*t/2.0E+0-amax*dtmax*sin(%pi*t/dtmax)/%pi/2.0E+0 > (%o4) done > (%i5) fortran(%o3); > amax*dtmax**2*cos(%pi*t/dtmax)/%pi**2/2.0E+0+amax*t**2/4.0E+0 > (%o5) done > > small correction: jjf@...:~\$ maxima Maxima 5.22.1 http://maxima.sourceforge.net using Lisp GNU Common Lisp (GCL) GCL 2.6.7 (a.k.a. GCL) Distributed under the GNU Public License. See the file COPYING. Dedicated to the memory of William Schelter. The function bug_report() provides bug reporting information. (%i1) (amax / 2) * sin((%pi * t / dtmax) -(%pi / 2) ) + amax /2 ; %pi t amax cos(-----) amax dtmax (%o1) ---- - --------------- 2 2 (%i2) integrate(%o1,t); %pi t amax dtmax sin(-----) amax t dtmax (%o2) ------ - --------------------- 2 2 %pi (%i3) %o2 + c0; %pi t amax dtmax sin(-----) dtmax amax t (%o3) - --------------------- + ------ + c0 2 %pi 2 (%i4) integrate(%o3,t); 2 %pi t amax dtmax cos(-----) 2 dtmax amax t (%o4) ---------------------- + ------- + c0 t 2 4 2 %pi (%i5) diff(%o4,t); %pi t amax dtmax sin(-----) dtmax amax t (%o5) - --------------------- + ------ + c0 2 %pi 2 (%i6) diff(%o5,t); %pi t amax cos(-----) amax dtmax (%o6) ---- - --------------- 2 2 Joachim ```

 Re: [Emc-developers] S-curve From: Joachim Franek - 2012-05-07 19:41 ```On Tuesday 01 May 2012 10:32:12 Joachim Franek wrote: > On Monday 30 April 2012 22:42:45 andy pugh wrote: > > For my simple exact-stop and constant-speed requirements, I am > > planning on an approach that says, on every servo cycle: > > If I was to decrement acceleration by the max jerk from the current > > velocity and acceleration, how long would it take for v = 0? > > v(t) = v0 + at -j.t^2/2 => t = -v0 + sqrt(a^2 + 2.v0.j) / 2.v0 > > This is equation 5.12 and 5.14?. Maxima gives different result to 5.14: (%i1) vh + as * t -jm * t * t /2; 2 jm t (%o1) vh - ----- + as t 2 (%i2) solve(%o1,t); 2 2 sqrt(2 jm vh + as ) - as sqrt(2 jm vh + as ) + as (%o2) [t = - ------------------------, t = ------------------------] jm jm Joachim ```

 Re: [Emc-developers] S-curve: my summary From: Karl Cunningham - 2012-05-07 23:43 ```On 05/05/2012 05:19 PM, Jon Elson wrote: > Karl Cunningham wrote: >> On 05/04/2012 07:26 PM, Jon Elson wrote: >> >>> You'd need a swept >>> sine wave generator to stimulate the system, and record the response of >>> the system, both amplitude and phase to the stimulus. Just add the sine >>> wave excitation to a constant position and feed this to the PID, and >>> record position. Then, feed that to a Fourier analysis. >> I have a few geophones (mentioned earlier in this thread) which might be >> usable for collecting vibration data. I don't have a good >> high-sample-rate ADC, but perhaps some scope traces would suffice. >> >> If someone has any specific tests they would like done with this on a >> 3-axis knee mill, I could give it a go. >> > Actually, as I hinted above, I think this could be done with a script > and features > built into LinuxCNC. There is a waveform generator as part of HAL, and > a module > that will record traces from the RT component of Halscope into a numerical > file. So, the only thing needed is a little piece of code to extract > amplitude and > phase info from the halscope trace and create a spreadsheet data file > from the same > measurement over a range of frequencies. So, if you have a machine with > encoder > feedback to the PC, you should be able to do this. Assuming a 1 KHz servo > cycle, you should be able to detect resonances up to almost 500 Hz. I agree that should work with what's in Linuxcnc. I'm game to try them on the machine I have access to. It is a 3-axis machine with servos and encoders (~40K counts/inch) on the lead screws. It does have some backlash (0.003" in X and 0.009" in Y), and I'm curious how that affects table motion. This is partly why I was thinking of using a geophone to measure high-frequency components. It might be interesting to also see how backlash affects table motion in Linuxcnc. Karl ```

 Re: [Emc-developers] S-curve: my summary From: Kenneth Lerman - 2012-05-08 14:01 ```The following link describes a development board for a network analyzer chip, with a sweep signal generate, ADC, and USB interface. I got one for \$66 at Digikey. The only negative for us is that you might want to swap out the crystal to get a lower sweep frequency. http://www.analog.com/static/imported-files/user_guides/UG-364.pdf Regards, Ken On 5/6/2012 1:26 PM, Jon Elson wrote: > EBo wrote: >> It is also to bad that I no longer have access to the 14bit 300Ms/s >> ADCs I got to play with a little in the past. That would probably tell >> us a little bit ;-) >> > A digital scope with some kind of computer connectivity (network, USB > thumb drive, > USB device) should do fine. Since the LinuxCNC servo rate is only one > KHz, a > trace with a 10 KHz sampling rate should be quite adequate. But, really, > recording encoder position with Halscope is quite good. > > Jon > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Emc-developers mailing list > Emc-developers@... > https://lists.sourceforge.net/lists/listinfo/emc-developers ```