On Saturday, May 21, 2016 at 10:04:20 AM UTC+12, Tim Wescott wrote:> On Fri, 20 May 2016 16:29:39 -0500, Tim Wescott wrote: > > > On Fri, 20 May 2016 10:56:33 -0700, gyansorova wrote: > > > >> On Friday, May 20, 2016 at 5:57:43 PM UTC+12, Tim Wescott wrote: > >>> On Thu, 19 May 2016 15:40:51 -0700, gyansorova wrote: > >>> > >>> > On Friday, May 20, 2016 at 7:49:55 AM UTC+12, Tim Wescott wrote: > >>> >> On Thu, 19 May 2016 10:54:03 -0700, gyansorova wrote: > >>> >> > >>> >> > On Thursday, May 19, 2016 at 4:58:38 PM UTC+12, Tim Wescott > >>> >> > wrote: > >>> >> >> On Wed, 18 May 2016 14:44:23 -0700, gyansorova wrote: > >>> >> >> > >>> >> >> > On Thursday, May 19, 2016 at 9:35:10 AM UTC+12, Phil Hobbs > >>> >> >> > wrote: > >>> >> >> >> On 05/18/2016 05:12 PM, Tim Wescott wrote: > >>> >> >> >> > On Wed, 18 May 2016 10:05:16 +1000, Clifford Heath wrote: > >>> >> >> >> > > >>> >> >> >> >> On 18/05/16 01:54, Tim Wescott wrote: > >>> >> >> >> >>> On Tue, 17 May 2016 08:15:42 -0500, John S wrote: > >>> >> >> >> >>>> For your next demo, use an electromagnet to lift a metal > >>> >> >> >> >>>> ball and hold it suspended. Sense the height with a > >>> >> >> >> >>>> light sensor. Use PID to achieve stability. > >>> >> >> >> >>> I've done that. You need a honkin' big electromagnet to > >>> >> >> >> >>> make it work with a plain steel load. > >>> >> >> >> >> > >>> >> >> >> >> Could you use a smaller electromagnet below, partly > >>> >> >> >> >> canceling the pull of a rare-earth magnet from above? > >>> >> >> >> > > >>> >> >> >> > Yes. IMHO, a thingie that's hanging suspended from a point > >>> >> >> >> > looks more impressive than a thingie that's hanging > >>> >> >> >> > suspended between two points. > >>> >> >> >> > > >>> >> >> >> > A thingie that's floating _above_ a point is more > >>> >> >> >> > impressive yet, but if you do that then all of a sudden you > >>> >> >> >> > have to control it in three dimensions instead of one (or > >>> >> >> >> > perhaps six if you can't make it inherently stable in > >>> >> >> >> > rotation). You can float an aluminum pan above an > >>> >> >> >> > electromagnet and have it be stable, but you can't do the > >>> >> >> >> > same thing with plain old magnets. > >>> >> >> >> > > >>> >> >> >> > > >>> >> >> >> You can, actually, if the object is sufficiently diamagnetic, > >>> >> >> >> such as pyrolytic graphite. I have a little demo on the > >>> >> >> >> shelf over my lab bench that levitates a small sheet of > >>> >> >> >> graphite over four NdFeB magnets arranged in a quadrupole. I > >>> >> >> >> posted a video a few years back. > >>> >> >> >> > >>> >> >> >> When physicists visit, I give them a spiel about room > >>> >> >> >> temperature superconductors and the Meissner effect...the > >>> >> >> >> size of the double-take goes linearly with how much physics > >>> >> >> >> they know. ;) > >>> >> >> >> > >>> >> >> >> Cheers > >>> >> >> >> > >>> >> >> >> Phil Hobbs > >>> >> >> >> > >>> >> >> >> -- > >>> >> >> >> Dr Philip C D Hobbs Principal Consultant ElectroOptical > >>> >> >> >> Innovations LLC Optics, Electro-optics, Photonics, Analog > >>> >> >> >> Electronics > >>> >> >> >> > >>> >> >> >> 160 North State Road #203 Briarcliff Manor NY 10510 > >>> >> >> >> > >>> >> >> >> hobbs at electrooptical dot net http://electrooptical.net > >>> >> >> > > >>> >> >> > PID is quick and nasty but good to get something working. > >>> >> >> > Proper classical way is by lag-lead controllers and > >>> >> >> > integrators all cascaded and by using a Bode plot, phase > >>> >> >> > margin etc. > >>> >> >> > >>> >> >> Terminology, terminology. > >>> >> >> > >>> >> >> Give me a transfer function of your "lag-lead controllers and > >>> >> >> integrators all cascaded". > >>> >> >> > >>> >> >> It's of the form > >>> >> >> > >>> >> >> a2 * s^2 + a1 * s + a0 > >>> >> >> H(s) = ---------------------- > >>> >> >> s ( s + b ) > >>> >> >> > >>> >> >> right? > >>> >> >> > >>> >> >> So -- that's the transfer function for a PID, with a > >>> >> >> band-limited derivative term. Plain ol' PID. Nuthin' special > >>> >> >> about it except for the fancified language you want to use to > >>> >> >> describe it. > >>> >> >> > >>> >> >> Same thing. Different words. Words are nothing. Reality is > >>> >> >> everything. PID = "lag-lead controllers and integrators all > >>> >> >> cascaded", > >>> >> >> only "PID" is shorter. > >>> >> >> > >>> >> >> Now, a PID that's _tuned_ using the seat-of-the-pants method -- > >>> >> >> that can be improved on with the classical Bode plot method > >>> >> >> using gain & phase margins. But it's still a PID -- or a > >>> >> >> lead-lag controller with an integrator, all cascaded, if that's > >>> >> >> the only way you can retain your sanity. But no matter what > >>> >> >> words you use, the resistors and caps do not change: the > >>> >> >> description does not change the circuit, > >>> >> >> or the code, or the pneumatic cylinders, or whatever you use to > >>> >> >> implement the controller. > >>> >> >> > >>> >> >> -- > >>> >> >> > >>> >> >> Tim Wescott Wescott Design Services http://www.wescottdesign.com > >>> >> >> > >>> >> >> I'm looking for work -- see my website! > >>> >> > > >>> >> > Wrong! It is more than ordinary PID. Of course you can emulate a > >>> >> > PID that way as you point out, but that is just the starting > >>> >> > point. Much os this stuff is not generally not understood by the > >>> >> > great unwashed and does not appear in the text books. Here goes. > >>> >> > > >>> >> > What is an ideal Bode plot? That is the first question to ask > >>> >> > > >>> >> > Second question is what limits the bandwidth of say an > >>> >> > electro-mechanical system? > >>> >> > > >>> >> > An idea Bode plot would be a vertical straight line 9except the > >>> >> > close look system woudl be unstable) > >>> >> > > >>> >> > Limiting factor on bandwidth is structural resonance(s). > >>> >> > > >>> >> > > >>> >> > Hence you need multiple integrators at low frequency to make your > >>> >> > Bode plot as steep as possible. Likewise you can use multiple > >>> >> > phase-advances too but with caution since you disrupt the > >>> >> > structural resonance. (phase advance has high frequency gain of > >>> >> > course). > >>> >> > > >>> >> > So a PID is rather limited. The order of a controller could be > >>> >> > much higher depending on just how many integrators you can > >>> >> > squeeze in at low frequency and how many phase advances. If you > >>> >> > can't fit an integrator in then you need a phase-lag which has > >>> >> > less phase problems. > >>> >> > > >>> >> > Anyway, this is the art of the control engineer which has been > >>> >> > around since the 70s or so but is kind of lost with todays PID > >>> >> > thinking. Nothing wrong with PID of course if you are happy with > >>> >> > the result and like second best, but you can do so much better. > >>> >> > > >>> >> > There is not enough time or space to explain all of it. I only > >>> >> > know it myself because an old hard disk designer explained it to > >>> >> > me thirty years ago. I and you know the theory already, it's just > >>> >> > the art of it all that is pretty smart. > >>> >> > It's like lego, putting in extra terms here and there and > >>> >> > watching your phase budget - push up bandwidth like a violin > >>> >> > string. None of it is in textbooks because such people don't > >>> >> > write text books, they like keeping it trade secret! > >>> >> > >>> >> Semantics again. With the exception of a double integrator (which > >>> >> isn't always a good idea) you're describing my "PID with > >>> >> decoration" > >>> >> and saying "that's not PID". > >>> >> > >>> >> And just as a note: multiple lead-lags only work to the extent that > >>> >> the high-frequency behavior of the plant is consistent and > >>> >> well-mannered. You can easily design monstrously frustrating lab > >>> >> queen if you don't pay attention to this. Systems that only work > >>> >> at room temperature, or when they're freshly maintained, or when > >>> >> the operator doesn't bang on them too hard are not systems that > >>> >> make the people who sign checks happy. > >>> >> > >>> >> -- > >>> >> > >>> >> Tim Wescott Wescott Design Services http://www.wescottdesign.com > >>> >> > >>> >> I'm looking for work -- see my website! > >>> > > >>> > Double integrators are always a good idea. The have an enormous gain > >>> > at low frequencies and hence reject disturbances. > >>> > >>> Whilst, and at the same time, causing enhanced disturbance response > >>> somewhere else. See Bode's Sensitivity Integral for details. > >>> > >>> > What you are talking about is process control of relatively slow > >>> > processes - which is fair enough. > >>> > >>> I'm so happy for you that you're psychic. But maybe you need to brush > >>> up on the mind-reading. Read on... > >>> > >>> > I am talking high speed elec mech drives like a hard disk head. > >>> > Speed is of the essence and you must maximize the bandwidth. > >>> > >>> I'm talking about high precision electromechanical drives that > >>> actually have to work over the full military temperature range, in > >>> high vibration environments, while mounted onto a variety of > >>> platforms. > >>> > >>> -- > >>> Tim Wescott Control systems, embedded software and circuit design I'm > >>> looking for work! See my website if you're interested > >>> http://www.wescottdesign.com > >> > >> Well you are not doing your best with a PID, that is for sure. I assume > >> speed is not an issue in terms of transient response and it's a don't > >> care much about disturbance rejection. PID is ok for those that know > >> little and want a fast and simple solution - I haven't a problem about > >> that. You don't want to be using it for fast electro-mech drives > >> though. > >> Sure it will work, not the best solution by a long way though. Then > >> when you get to system with more than one input and output then you > >> have to throw away the PID and the lag-lead methods too. > > > > You have no clue what you are talking about. Try actually reading what > > I've been saying. > > Let me rephrase that. You are obviously a good controls engineer, but > you have trouble reading what other people write. Perhaps you've never > worked outside of the industry you're in, and it has caused ossification > of some of your brain cells. Try actually reading what I've been saying. > > Note the part that hasn't been changed. Reading comprehension often > improves if you actually read what you're "reading". > > -- > > Tim Wescott > Wescott Design Services > http://www.wescottdesign.com > > I'm looking for work -- see my website!I don't like arguing with other engineers because they are usually as smart if not smarter than yourself! I have a great respect for your work, let's just leave it at that.
Stability and Insanity
Started by ●May 17, 2016
Reply by ●May 22, 20162016-05-22
Reply by ●May 24, 20162016-05-24
On Tuesday, May 17, 2016 at 8:05:57 AM UTC-4, Evgeny Filatov wrote:> On 17.05.2016 12:14, herrmannsfeldt@gmail.com wrote: > > On Monday, May 16, 2016 at 11:41:55 PM UTC-7, Tim Wescott wrote: > >> I've just spent over two weeks getting ready to do my next video. > > > > (snip) > >> I am, at this point, not only wondering if it was worth it, but > >> questioning my sanity in carrying on even when the going went beyond > >> tough to just plain crazy. > > > > (snip) > >> * Built an oscillator that uses the pendulum as its resonator (this is > >> where stability comes in -- is an oscillator stable? How is it stable? > >> What if it's showing chaotic behavior?). This was astonishingly > >> frustrating, and didn't finally work until I carefully modeled the > >> pendulum as a resonator AND took the coil inductance into account in the > >> circuit. This part too about a week. > > > > There is an old saying about the best way to learn something is to (try to) teach it. > > > > I suspect you now know it better than ever. > > > > Sounds like fun. > > > > Reminds me about Huygens, who learned about coupled oscillators working > > with many clocks on the same wall. Also, it is supposed to be that women living > > in the same house will phase lock their periods. > > > > Or circadian rhythm. The fun part is that circadian clocks actually > operate within each cell of an organism, and can be modeled as a > negative feedback loop. > > Quoting "Molecular Biology of the Cell" by Alberts (6th ed.), pp. 876-878: > > "Having a circadian clock enables an organism to anticipate the regular > daily changes in its environment and take appropriate action in advance. > Of course, the internal clock cannot be perfectly accurate, and so it > must be capable of being reset by external cues such as the light of > day. Thus, circadian clocks keep running even when the environmental > cues (changes in light and dark) are removed, but the period of this > free-running rhythm is generally a little less or more than 24 hours. > External signals indicating the time of day cause small adjustments in > the running of the clock, so as to keep the organism in synchrony with > its environment. Following more drastic shifts, circadian cycles become > gradually reset (entrained) by the new cycle of light and dark, as > anyone who has experienced jet lag can attest." > > ... > > "In Drosophila and many other animals, including humans, the heart of > the circadian clock is a delayed negative feedback loop based on > transcription regulators: accumulation of certain gene products switches > off the transcription of their own genes, but with a delay, so that the > cell oscillates between a state in which the products are present and > transcription is switched off, and one in which the products are absent > and transcription is switched on." > > GeneSortof related: http://www.google.com/patents/US6497718 Just google and the strangest things may be found - lol! Clay
