T
three_jeeps
Guest
On Sunday, October 6, 2019 at 11:56:12 PM UTC-4, George Herold wrote:
Hmmm I would not say that derivative control is 'bad'. It depends on plant
characteristics. From a purely mathematical perspective slow systems
(those with a fair amount of inertia (relatively) can benefit from a small
amount of derivative action.
PID controllers are, in general, used a lot when plant dynamics are not
known, e.g. a good math model. They can be adjusted in the field to more
easily address real world issues: e.g, inertia, windage, friction,weight.
There have been lots of papers written about approaches to tune PID loops,
automated and adaptive PID tuners, etc.
In reality, noise coupled into a system can create instability largely due
to too much derivative action (roughly speaking remember the derivative of
a step function is a Dirac delta function and in practical terms is
infinity. It all depends on how the digital implementation of the
derivative function).
A good control engineer has a grasp of the mathematical foundations (all based on math and physics) as well
as the coding/implementation aspects as well as the practicalities of the
real world implementations. Complications arise when there are nested
loops and tuning is affected due to interactions of the loops. After all, it IS a 'system' thing....
I believe the example you described is known as the 'inverted pendulum'.
(Made by Quanser) It is a common control lab hardware used in many
universities both for doing classical control theory as well as state space
approaches to control. There are some interesting you tube vids around it.
J
On Sunday, October 6, 2019 at 10:54:10 PM UTC-4, pcdh...@gmail.com wrote:
On Friday, October 4, 2019 at 9:19:04 PM UTC-4, George Herold wrote:
On Friday, October 4, 2019 at 3:24:59 PM UTC-4, John Larkin wrote:
On Fri, 4 Oct 2019 12:19:48 -0700 (PDT), jjhudak4@gmail.com wrote:
On Wednesday, October 2, 2019 at 2:02:14 PM UTC-4, bitrex wrote:
On 10/2/19 1:59 PM, bitrex wrote:
On 10/2/19 1:49 PM, John Larkin wrote:
On Tue, 1 Oct 2019 17:55:09 -0700 (PDT), George Herold
gherold@teachspin.com> wrote:
John (Larkin), You offered to send me a 'standard' NDA. I wanted to
take you up on that. Maybe on dropbox? Or my email (now) is
ggherold@gmail.com
Thanks
George H.
Here is our starting-point NDA.
https://www.dropbox.com/s/b924j4e9jbbven2/NDA%20Draft%20-2%20..docx?dl=0
This has been mangled by multiple lawyers, and other companies usually
sign this, sometimes with minor revs. It's pretty much the standard
Silicon Valley NDA.
We have had some legal doings with a giant company who signed this
basic form when they needed stuff badly. They later discovered that we
take it seriously.
One trick companies will do is to sign the NDA, let you do a lot of
work for them, show them how it's done, and then do a big prior-art
search to justify stealing the designs. I could name names. About all
you can do then is walk away and concentrate on working with people
who have ethics.
It happens and unless the job is heading towards mid five figures
there's not much to do but write it off. I rarely have any work that
pays that much so far.
Clue: if their engineers look eager to do it themselves, they probably
will.
In the 21st century there are online avenues both for contract
work/employment and for just contracting to do tutoring/general
education. If someone seems super-eager to do it themselves then there
is no reason not to just acknowledge that and re-direct to the proper
department, they often accept.
You can charge hourly for just talking, which while perhaps not quite as
well-paid or as emotionally satisfying as designing something, is pretty
easy money. Sometimes they decide they'd rather you do something once
they fully realize it's beyond them so it turns out into being an
elongated pitch session, but you get paid for it.
PID controllers are apparently a hot area of interest and there are
software guys who will pay just to talk with someone who seems like
they've successfully implemented one.
Really??? I'll throw my hat in the ring. Successfully eh? I bet they got tripped up by the issues associated with discretization of an analog model. Yes, numbers in digital computers can and do roll over causing amusing results when integral windup all of a sudden becomes negative. Then there is always the sampling time jitter caused by numerous task scheduling and badly done scheduling analysis. Oh, we'll just throw a few digital filters at it...that just made it worse....lol yes, Nyquist criteria must be paid attention to....
There are a few fun corner cases: bumpless transfer, slew windup,
various bits saturating, nonlinearities, things like that.
The "d" part of PID is usually bad news.
Hmm where's Tim Wescott. He sorta showed me that there are times where
the D is paramount... Like the thermometers you stick in your mouth and they
give you the final temp based on the slope.
George H.
The D is mostly useful when you have a slow second-order plant, such as a current-controlled motor. Angular acceleration is roughly proportional to current, so a bit of D can help a lot.
Thanks, As usual I shake my head yes, but I think I'd have to build
something for it to have real meaning. About a year ago my son was
asking about control loops for his quad copter.. I pleaded ignorance of
motors. He has joined the rocketry club at college and is helping them
build drones... so maybe he can teach me about it. :^)
Tim W. made that video with a motor and blade on a balanced stick...
I should go watch that again.
George H.
In more complicated situations such as thermal control, the phase shift increases without bound with frequency, so adding D makes things worse rather than better.
Cheers
Phil Hobbs
Hmmm I would not say that derivative control is 'bad'. It depends on plant
characteristics. From a purely mathematical perspective slow systems
(those with a fair amount of inertia (relatively) can benefit from a small
amount of derivative action.
PID controllers are, in general, used a lot when plant dynamics are not
known, e.g. a good math model. They can be adjusted in the field to more
easily address real world issues: e.g, inertia, windage, friction,weight.
There have been lots of papers written about approaches to tune PID loops,
automated and adaptive PID tuners, etc.
In reality, noise coupled into a system can create instability largely due
to too much derivative action (roughly speaking remember the derivative of
a step function is a Dirac delta function and in practical terms is
infinity. It all depends on how the digital implementation of the
derivative function).
A good control engineer has a grasp of the mathematical foundations (all based on math and physics) as well
as the coding/implementation aspects as well as the practicalities of the
real world implementations. Complications arise when there are nested
loops and tuning is affected due to interactions of the loops. After all, it IS a 'system' thing....
I believe the example you described is known as the 'inverted pendulum'.
(Made by Quanser) It is a common control lab hardware used in many
universities both for doing classical control theory as well as state space
approaches to control. There are some interesting you tube vids around it.
J