Re: [Emc-developers] 2.7.15 PID Fix

[Jon Elson]

On 03/03/2020 10:13 AM, andy pugh wrote:

We are seeing a number of issues reported to the forum where the fix
to the PID command-deriv input is causing problems with existing
configurations that net that pin to a signal that is not otherwise
connected to a source.

Should we consider a 2.7.16 that reverts this fix, and push the actual
fix to 2.8? (where the config updater script would, maybe, be able to
do something clever with it.)

I've had some reports of odd behavior of PID in very late 
builds, but didn't have enough info

to track it down.

Does anybody have a more specific case of how it is failing?

(Hmm, why would somebody create a net for the command-deriv 
with no source?  Seems kind of

odd.)

Jon


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Rene Hopf via Emc-developers]

> On 3. Mar 2020, at 17:49, Jon Elson  wrote:
> 
> On 03/03/2020 10:13 AM, andy pugh wrote:
>> We are seeing a number of issues reported to the forum where the fix
>> to the PID command-deriv input is causing problems with existing
>> configurations that net that pin to a signal that is not otherwise
>> connected to a source.
>> 
>> Should we consider a 2.7.16 that reverts this fix, and push the actual
>> fix to 2.8? (where the config updater script would, maybe, be able to
>> do something clever with it.)
>> 
> I've had some reports of odd behavior of PID in very late builds, but didn't 
> have enough info
> to track it down.
> 
> Does anybody have a more specific case of how it is failing?
> 
> (Hmm, why would somebody create a net for the command-deriv with no source?  
> Seems kind of
> odd.)

Pncconf and some default configs do that.

> 
> Jon
> 
> 
> ___
> Emc-developers mailing list
> Emc-developers@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-developers


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[andy pugh]

On Tue, 3 Mar 2020 at 16:50, Jon Elson  wrote:

> Does anybody have a more specific case of how it is failing?

https://forum.linuxcnc.org/38-general-linuxcnc-questions/38367-2-7-15?start=0

-- 
atp
"A motorcycle is a bicycle with a pandemonium attachment and is
designed for the especial use of mechanical geniuses, daredevils and
lunatics."
— George Fitch, Atlanta Constitution Newspaper, 1912


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Gene Heskett]

On Tuesday 03 March 2020 11:13:17 andy pugh wrote:

> We are seeing a number of issues reported to the forum where the fix
> to the PID command-deriv input is causing problems with existing
> configurations that net that pin to a signal that is not otherwise
> connected to a source.
>
> Should we consider a 2.7.16 that reverts this fix, and push the actual
> fix to 2.8? (where the config updater script would, maybe, be able to
> do something clever with it.)

Its been what, 2 years since you recommended we disconnect that, which I 
did like this, fixing my problems with something related to following 
errors. From TLM's .hal file:

net x-vel-cmd  <=  joint.0.vel-cmd # => pid.x.command-deriv
and
net z-vel-cmd  <=  joint.1.vel-cmd # => pid.z.command-deriv

I found those nets commented out in the GO704 configs

No pid's are used in the sheldon or 6040.  Should I be now?

Is there now a better way or was that an original stepconfig bug?

Their has been quite a bit of water under the bridge since then.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page 


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Gene Heskett]

On Tuesday 03 March 2020 11:54:09 andy pugh wrote:

> On Tue, 3 Mar 2020 at 16:50, Jon Elson  wrote:
> > Does anybody have a more specific case of how it is failing?
>
> https://forum.linuxcnc.org/38-general-linuxcnc-questions/38367-2-7-15?
>start=0

This thread is at least 2 years newer than the previous disconnect it fix 
I had to do in master.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page 


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Chris Morley]

reverting seems reasonable. In the long run why don't we change it from 'magic' 
to a switch - like a parameter to change the behavior?

Chris


From: andy pugh 
Sent: March 3, 2020 4:13 PM
To: EMC developers 
Subject: [Emc-developers] 2.7.15 PID Fix

We are seeing a number of issues reported to the forum where the fix
to the PID command-deriv input is causing problems with existing
configurations that net that pin to a signal that is not otherwise
connected to a source.

Should we consider a 2.7.16 that reverts this fix, and push the actual
fix to 2.8? (where the config updater script would, maybe, be able to
do something clever with it.)


--
atp
"A motorcycle is a bicycle with a pandemonium attachment and is
designed for the especial use of mechanical geniuses, daredevils and
lunatics."
— George Fitch, Atlanta Constitution Newspaper, 1912


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Jon Elson]

On 03/03/2020 10:54 AM, andy pugh wrote:

On Tue, 3 Mar 2020 at 16:50, Jon Elson  wrote:


Does anybody have a more specific case of how it is failing?

https://forum.linuxcnc.org/38-general-linuxcnc-questions/38367-2-7-15?start=0

So, the pid command-deriv was not used in the past?  Geez, 
that could explain a LOT of issues with

servo tuning before the change.

I use feedback-deriv in my servo systems, would a 
non-functioning command-deriv make that not

work right?

Jon


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Jon Elson]

Looking at the 2.7 docs for the PID component, there is no 
mention of a pid.n.command-deriv
pin.  There is a commandD and a commandDD which seem to be 
outputs available when the

debug mode is turned on.

Jon


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Peter C. Wallace]

On Tue, 3 Mar 2020, Jon Elson wrote:


Date: Tue, 03 Mar 2020 20:54:55 -0600
From: Jon Elson 
Reply-To: EMC developers 
To: EMC developers 
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On 03/03/2020 10:54 AM, andy pugh wrote:

On Tue, 3 Mar 2020 at 16:50, Jon Elson  wrote:


Does anybody have a more specific case of how it is failing?

https://forum.linuxcnc.org/38-general-linuxcnc-questions/38367-2-7-15?start=0

So, the pid command-deriv was not used in the past?  Geez, that could explain 
a LOT of issues with

servo tuning before the change.

I use feedback-deriv in my servo systems, would a non-functioning 
command-deriv make that not

work right?

Jon



No, command-deriv is only used for the FF1 feedforward calculation




___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers



Peter Wallace
Mesa Electronics

(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.



___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Gene Heskett]

On Tuesday 03 March 2020 21:54:55 Jon Elson wrote:

> On 03/03/2020 10:54 AM, andy pugh wrote:
> > On Tue, 3 Mar 2020 at 16:50, Jon Elson  
wrote:
> >> Does anybody have a more specific case of how it is failing?
> >
> > https://forum.linuxcnc.org/38-general-linuxcnc-questions/38367-2-7-1
> >5?start=0
>
> So, the pid command-deriv was not used in the past?  Geez,
> that could explain a LOT of issues with
> servo tuning before the change.
>
> I use feedback-deriv in my servo systems, would a
> non-functioning command-deriv make that not
> work right?
>
> Jon
>
It sure screwed up my stepper systems that used it, following errors 
galore. From my TLM's *.hal file:

net x-vel-cmd  <=  joint.0.vel-cmd # => pid.x.command-deriv
It turns out there is no such signal source as a joint.n.vel-cmd from 
motion.

So one could break the noise injection path as above or put a # in front 
of the net command that constructs it, either way fixes it.

The symptoms were needing to use 1/2" or more FERROR settings. I have one 
axis still set at .5 for FERROR, but it is NOT even present in the 
hal file so its not even used!

Obviously I need to fix that before I do anything critical on TLM. And on 
the GO704 too probably. But that can wait till I have working spindle 
encoders on both the GO704, and the Sheldon. The GO704 should be done 
tomorrow as I've about got it converted from opto indexing to ATS-667 
watching a short piece of nail glued to the spindle for an index pulse. 
Major delay is waiting for jbkwik to set, which takes about a day. The 
optical I had built years ago, the disk came unbolted and sawed the 
opto's in 2. Since all I needed was the index to trigger rigid tapping, 
that is all I'm putting back in.  The A/B stuff is on the rear of the 
motor, and around 100x more accurate.

Based on what I found in my configs just now, I am guessing several we's 
have work to do to optimize that again.

Is there a procedure to do that?

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page 


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Gene Heskett]

On Tuesday 03 March 2020 22:22:25 Peter C. Wallace wrote:

> On Tue, 3 Mar 2020, Jon Elson wrote:
> > Date: Tue, 03 Mar 2020 20:54:55 -0600
> > From: Jon Elson 
> > Reply-To: EMC developers 
> > To: EMC developers 
> > Subject: Re: [Emc-developers] 2.7.15 PID Fix
> >
> > On 03/03/2020 10:54 AM, andy pugh wrote:
> >> On Tue, 3 Mar 2020 at 16:50, Jon Elson  
wrote:
> >>> Does anybody have a more specific case of how it is failing?
> >>
> >> https://forum.linuxcnc.org/38-general-linuxcnc-questions/38367-2-7-
> >>15?start=0
> >
> > So, the pid command-deriv was not used in the past?  Geez, that
> > could explain a LOT of issues with
> > servo tuning before the change.
> >
> > I use feedback-deriv in my servo systems, would a non-functioning
> > command-deriv make that not
> > work right?
> >
> > Jon

> No, command-deriv is only used for the FF1 feedforward calculation

Humm, Peter, none of my stuff uses that pid.n.command-deriv input now, 
but I have used FF1 in my stepper configs, so how about some examples of 
correct net useage statements to set that up right, once and for all?

Thanks Peter.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page <http://geneslinuxbox.net:6309/gene>


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Peter C. Wallace]

On Tue, 3 Mar 2020, Gene Heskett wrote:


Date: Tue, 3 Mar 2020 23:21:54 -0500
From: Gene Heskett 
Reply-To: EMC developers 
To: emc-developers@lists.sourceforge.net
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On Tuesday 03 March 2020 22:22:25 Peter C. Wallace wrote:


On Tue, 3 Mar 2020, Jon Elson wrote:

Date: Tue, 03 Mar 2020 20:54:55 -0600
From: Jon Elson 
Reply-To: EMC developers 
To: EMC developers 
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On 03/03/2020 10:54 AM, andy pugh wrote:

On Tue, 3 Mar 2020 at 16:50, Jon Elson 

wrote:

Does anybody have a more specific case of how it is failing?


https://forum.linuxcnc.org/38-general-linuxcnc-questions/38367-2-7-
15?start=0


So, the pid command-deriv was not used in the past?  Geez, that
could explain a LOT of issues with
servo tuning before the change.

I use feedback-deriv in my servo systems, would a non-functioning
command-deriv make that not
work right?

Jon



No, command-deriv is only used for the FF1 feedforward calculation


Humm, Peter, none of my stuff uses that pid.n.command-deriv input now,
but I have used FF1 in my stepper configs, so how about some examples of
correct net useage statements to set that up right, once and for all?



If its working for you I would just leave the PIDs command derivative 
pin unconnected so its calculated internally.




Thanks Peter.

Cheers, Gene Heskett
--
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
- Louis D. Brandeis
Genes Web page <http://geneslinuxbox.net:6309/gene>


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers



Peter Wallace
Mesa Electronics

(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.



___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Jon Elson]

On 03/03/2020 09:22 PM, Peter C. Wallace wrote:


No, command-deriv is only used for the FF1 feedforward 
calculation
But, I DO use FF1 in practically all my configs, and it does 
seem to work.  So, the problem is linking a signal to the 
command-deriv pin, but then NOT feeding it the correct 
derivative.  I can't see any way to fix this in PID.
The new 2.7.15 PID is not wrong, it just exposes an error in 
other code (the configs) that hadn't been right for

some time.
I guess without linking to command-deriv then PID calculates 
this internally by difference between this command and
the most recent command.  If you link a signal to 
command-deriv but do not provide the correct value, PID

would misbehave for sure.

Not clear what to do in this case, but fixing the broken 
configs and the autoconfig program that creates these
seems important.  Documenting the problem is critical, and 
explaining what the user needs to do to

make his system run right.

As for the docs, the generic 2.7 docs do NOT show any 
reference to pid.x.command-deriv

Am I looking in the wrong place, or has it been removed?

Jon


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Peter C. Wallace]

On Wed, 4 Mar 2020, Jon Elson wrote:


Date: Wed, 04 Mar 2020 10:44:26 -0600
From: Jon Elson 
Reply-To: EMC developers 
To: EMC developers 
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On 03/03/2020 09:22 PM, Peter C. Wallace wrote:


No, command-deriv is only used for the FF1 feedforward calculation

But, I DO use FF1 in practically all my configs, and it does seem to work.


It will work because if the command-deriv pin is not connected the derivative 
will be calculated internally


So, the problem is linking a signal to the command-deriv pin, but then NOT 
feeding it the correct derivative.  I can't see any way to fix this in PID.
The new 2.7.15 PID is not wrong, it just exposes an error in other code (the 
configs) that hadn't been right for

some time.


I guess without linking to command-deriv then PID calculates this internally 
by difference between this command and
the most recent command.  If you link a signal to command-deriv but do not 
provide the correct value, PID

would misbehave for sure.


Yes, you will get a 0 FF1 PID term



Not clear what to do in this case, but fixing the broken configs and the 
autoconfig program that creates these
seems important.  Documenting the problem is critical, and explaining what 
the user needs to do to

make his system run right.




As for the docs, the generic 2.7 docs do NOT show any reference to 
pid.x.command-deriv

Am I looking in the wrong place, or has it been removed?


It was added early in 2.8 and it works the same as feedback-deriv
(value calculated internally if unconnected)




Jon


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers



Peter Wallace
Mesa Electronics

(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.



___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[andy pugh]

On Wed, 4 Mar 2020 at 16:55, Peter C. Wallace  wrote:

> > As for the docs, the generic 2.7 docs do NOT show any reference to
> > pid.x.command-deriv
> > Am I looking in the wrong place, or has it been removed?
>
> It was added early in 2.8 and it works the same as feedback-deriv
> (value calculated internally if unconnected)

I don't think that is right, as the problem we are seeing is with
2.7.15 systems.

-- 
atp
"A motorcycle is a bicycle with a pandemonium attachment and is
designed for the especial use of mechanical geniuses, daredevils and
lunatics."
— George Fitch, Atlanta Constitution Newspaper, 1912


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Peter C. Wallace]

On Wed, 4 Mar 2020, andy pugh wrote:


Date: Wed, 4 Mar 2020 17:20:02 +
From: andy pugh 
Reply-To: EMC developers 
To: EMC developers 
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On Wed, 4 Mar 2020 at 16:55, Peter C. Wallace  wrote:



> As for the docs, the generic 2.7 docs do NOT show any reference to
> pid.x.command-deriv
> Am I looking in the wrong place, or has it been removed?

It was added early in 2.8 and it works the same as feedback-deriv
(value calculated internally if unconnected)


I don't think that is right, as the problem we are seeing is with
2.7.15 systems.


It was first added to 2.8 and them merged into 2.7 before the
2.7.15 release AFAIK

Peter Wallace
Mesa Electronics

(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.



___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[andy pugh]

On Wed, 4 Mar 2020 at 17:26, Peter C. Wallace  wrote:

> It was first added to 2.8 and them merged into 2.7 before the
> 2.7.15 release AFAIK

I think that they were added in 2010:
https://github.com/LinuxCNC/linuxcnc/commit/44586b831e1ef5e826003190ab317cda76b6e8e3#diff-912d11dfa52738be509f63e4ff1922a4
(And that commit also includes documentation of the pins)

-- 
atp
"A motorcycle is a bicycle with a pandemonium attachment and is
designed for the especial use of mechanical geniuses, daredevils and
lunatics."
— George Fitch, Atlanta Constitution Newspaper, 1912


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Peter C. Wallace]

On Wed, 4 Mar 2020, andy pugh wrote:


Date: Wed, 4 Mar 2020 17:52:09 +
From: andy pugh 
Reply-To: EMC developers 
To: EMC developers 
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On Wed, 4 Mar 2020 at 17:26, Peter C. Wallace  wrote:



It was first added to 2.8 and them merged into 2.7 before the

2> 2.7.15 release AFAIK

I think that they were added in 2010:
https://github.com/LinuxCNC/linuxcnc/commit/44586b831e1ef5e826003190ab317cda76b6e8e3#diff-912d11dfa52738be509f63e4ff1922a4
(And that commit also includes documentation of the pins)


I meant the patch that made the command derivative behavior match the feedback 
derivative behavior (and the manual), that was quite recent.




--
atp
"A motorcycle is a bicycle with a pandemonium attachment and is
designed for the especial use of mechanical geniuses, daredevils and
lunatics."
?? George Fitch, Atlanta Constitution Newspaper, 1912


Peter Wallace
Mesa Electronics
___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[andy pugh]

On Wed, 4 Mar 2020 at 18:07, Peter C. Wallace  wrote:

I meant the patch that made the command derivative behavior match the
> feedback
> derivative behavior (and the manual), that was quite recent.
>

Ah, sorry. I thought you were explaining the statement that the pins are
not in the docs.
( Which appears not to actually be true:
http://linuxcnc.org/docs/2.7/html/man/man9/pid.9.html )


-- 
atp
"A motorcycle is a bicycle with a pandemonium attachment and is designed
for the especial use of mechanical geniuses, daredevils and lunatics."
— George Fitch, Atlanta Constitution Newspaper, 1912

___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Jon Elson]

On 03/04/2020 02:14 PM, andy pugh wrote:

On Wed, 4 Mar 2020 at 18:07, Peter C. Wallace  wrote:

I meant the patch that made the command derivative behavior match the

feedback
derivative behavior (and the manual), that was quite recent.


Ah, sorry. I thought you were explaining the statement that the pins are
not in the docs.
( Which appears not to actually be true:
http://linuxcnc.org/docs/2.7/html/man/man9/pid.9.html )




It took me a while to find where you got this link from :

http://linuxcnc.org/docs/2.7/html/man/man9/pid.9.html

But, in the same html tree, there is :
http://linuxcnc.org/docs/2.7/html/hal/rtcomps.html#_pid
Which has fewer pins.  These should be harmonized.


Jon


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Gene Heskett]

On Wednesday 04 March 2020 12:52:09 andy pugh wrote:

> On Wed, 4 Mar 2020 at 17:26, Peter C. Wallace  wrote:
> > It was first added to 2.8 and them merged into 2.7 before the
> > 2.7.15 release AFAIK
>
> I think that they were added in 2010:
> https://github.com/LinuxCNC/linuxcnc/commit/44586b831e1ef5e826003190ab
>317cda76b6e8e3#diff-912d11dfa52738be509f63e4ff1922a4 (And that commit
> also includes documentation of the pins)

Sort of. But it does provide some hints as to how to utilize it to 
improve control, If you have encoder feedback on axis axis then you 
could feed the feedback.deriv pin with the velocity of that axis's 
encoder. But from reading that, I fail to deduct a source for cmd.deriv. 
I am envisioning a differential, for instance between current command 
and previous command or maybe a sum2 with one input set for a gain of -1 
and an rc lag from the command input to the other input giving a 
differential output according to the charge across the cap. Since we 
don't have a cap in hal components, we'ed have to use a low pass.  But I 
quickly run out of imagination as to the rc time constant to synthesize.

just discussing the howto's would I think be educational for those of us 
who never got the transcendental stuff in school 70+ years ago.

Help?

Thanks

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page 


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[andy pugh]

On Thu, 5 Mar 2020 at 03:14, Gene Heskett  wrote:

> encoder. But from reading that, I fail to deduct a source for cmd.deriv.

It's a feed-forward input, and the source would be the velocity output
from the motion controller for that joint.
(ie, the motion system outputs both a current position _and_ current
velocity for each joint actuator)

-- 
atp
"A motorcycle is a bicycle with a pandemonium attachment and is
designed for the especial use of mechanical geniuses, daredevils and
lunatics."
— George Fitch, Atlanta Constitution Newspaper, 1912


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Peter C. Wallace]

On Wed, 4 Mar 2020, Gene Heskett wrote:


Date: Wed, 4 Mar 2020 22:12:47 -0500
From: Gene Heskett 
Reply-To: EMC developers 
To: emc-developers@lists.sourceforge.net
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On Wednesday 04 March 2020 12:52:09 andy pugh wrote:


On Wed, 4 Mar 2020 at 17:26, Peter C. Wallace  wrote:

It was first added to 2.8 and them merged into 2.7 before the
2.7.15 release AFAIK


I think that they were added in 2010:
https://github.com/LinuxCNC/linuxcnc/commit/44586b831e1ef5e826003190ab
317cda76b6e8e3#diff-912d11dfa52738be509f63e4ff1922a4 (And that commit
also includes documentation of the pins)


Sort of. But it does provide some hints as to how to utilize it to
improve control, If you have encoder feedback on axis axis then you
could feed the feedback.deriv pin with the velocity of that axis's
encoder. But from reading that, I fail to deduct a source for cmd.deriv.
I am envisioning a differential, for instance between current command
and previous command or maybe a sum2 with one input set for a gain of -1
and an rc lag from the command input to the other input giving a
differential output according to the charge across the cap. Since we
don't have a cap in hal components, we'ed have to use a low pass.  But I
quickly run out of imagination as to the rc time constant to synthesize.

just discussing the howto's would I think be educational for those of us
who never got the transcendental stuff in school 70+ years ago.

Help?

Thanks



Normally you would just connect the joint velocity pin to the PID command 
derivative





Cheers, Gene Heskett
--
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
- Louis D. Brandeis
Genes Web page <http://geneslinuxbox.net:6309/gene>


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers



Peter Wallace
Mesa Electronics

(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.



___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Gene Heskett]

On Thursday 05 March 2020 07:16:20 andy pugh wrote:

> On Thu, 5 Mar 2020 at 03:14, Gene Heskett  wrote:
> > encoder. But from reading that, I fail to deduct a source for
> > cmd.deriv.
>
> It's a feed-forward input, and the source would be the velocity output
> from the motion controller for that joint.
> (ie, the motion system outputs both a current position _and_ current
> velocity for each joint actuator)

Thanks Andy. Since I might have to dig into that, I may as well "do it 
right".  But this is the signal I was instructed to disconnect a couple 
years ago to get rid of my instant following errors, so it sounds as if 
my default setting for FF1=1 is wrong. Whats the proper starter value 
for all the FF#'s then?

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page 


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Gene Heskett]

On Thursday 05 March 2020 09:25:01 Peter C. Wallace wrote:

> On Wed, 4 Mar 2020, Gene Heskett wrote:
> > Date: Wed, 4 Mar 2020 22:12:47 -0500
> > From: Gene Heskett 
> > Reply-To: EMC developers 
> > To: emc-developers@lists.sourceforge.net
> > Subject: Re: [Emc-developers] 2.7.15 PID Fix
> >
> > On Wednesday 04 March 2020 12:52:09 andy pugh wrote:
> >> On Wed, 4 Mar 2020 at 17:26, Peter C. Wallace  
wrote:
> >>> It was first added to 2.8 and them merged into 2.7 before the
> >>> 2.7.15 release AFAIK
> >>
> >> I think that they were added in 2010:
> >> https://github.com/LinuxCNC/linuxcnc/commit/44586b831e1ef5e82600319
> >>0ab 317cda76b6e8e3#diff-912d11dfa52738be509f63e4ff1922a4 (And that
> >> commit also includes documentation of the pins)
> >
> > Sort of. But it does provide some hints as to how to utilize it to
> > improve control, If you have encoder feedback on axis axis then you
> > could feed the feedback.deriv pin with the velocity of that axis's
> > encoder. But from reading that, I fail to deduct a source for
> > cmd.deriv. I am envisioning a differential, for instance between
> > current command and previous command or maybe a sum2 with one input
> > set for a gain of -1 and an rc lag from the command input to the
> > other input giving a differential output according to the charge
> > across the cap. Since we don't have a cap in hal components, we'ed
> > have to use a low pass.  But I quickly run out of imagination as to
> > the rc time constant to synthesize.
> >
> > just discussing the howto's would I think be educational for those
> > of us who never got the transcendental stuff in school 70+ years
> > ago.
> >
> > Help?
> >
> > Thanks
>
> Normally you would just connect the joint velocity pin to the PID
> command derivative
>
Ok, but thats the signal I disconnected around 2 years back to get rid of 
an instant following error for a movement of 1/10 the MIN_ERROR. So  
hooking it back back up now does what to my FF# settings?

> > Cheers, Gene Heskett
> > --
> > "There are four boxes to be used in defense of liberty:
> > soap, ballot, jury, and ammo. Please use in that order."
> > -Ed Howdershelt (Author)
> > If we desire respect for the law, we must first make the law
> > respectable. - Louis D. Brandeis
> > Genes Web page <http://geneslinuxbox.net:6309/gene>
> >
> >
> > ___
> > Emc-developers mailing list
> > Emc-developers@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/emc-developers
>
> Peter Wallace
> Mesa Electronics
>
> (\__/)
> (='.'=) This is Bunny. Copy and paste bunny into your
> (")_(") signature to help him gain world domination.
>
>
>
> ___
> Emc-developers mailing list
> Emc-developers@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/emc-developers


Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page <http://geneslinuxbox.net:6309/gene>


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Peter C. Wallace]

On Thu, 5 Mar 2020, Gene Heskett wrote:


Date: Thu, 5 Mar 2020 10:42:19 -0500
From: Gene Heskett 
Reply-To: EMC developers 
To: emc-developers@lists.sourceforge.net
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On Thursday 05 March 2020 09:25:01 Peter C. Wallace wrote:


On Wed, 4 Mar 2020, Gene Heskett wrote:

Date: Wed, 4 Mar 2020 22:12:47 -0500
From: Gene Heskett 
Reply-To: EMC developers 
To: emc-developers@lists.sourceforge.net
Subject: Re: [Emc-developers] 2.7.15 PID Fix

On Wednesday 04 March 2020 12:52:09 andy pugh wrote:

On Wed, 4 Mar 2020 at 17:26, Peter C. Wallace 

wrote:

It was first added to 2.8 and them merged into 2.7 before the
2.7.15 release AFAIK


I think that they were added in 2010:
https://github.com/LinuxCNC/linuxcnc/commit/44586b831e1ef5e82600319
0ab 317cda76b6e8e3#diff-912d11dfa52738be509f63e4ff1922a4 (And that
commit also includes documentation of the pins)


Sort of. But it does provide some hints as to how to utilize it to
improve control, If you have encoder feedback on axis axis then you
could feed the feedback.deriv pin with the velocity of that axis's
encoder. But from reading that, I fail to deduct a source for
cmd.deriv. I am envisioning a differential, for instance between
current command and previous command or maybe a sum2 with one input
set for a gain of -1 and an rc lag from the command input to the
other input giving a differential output according to the charge
across the cap. Since we don't have a cap in hal components, we'ed
have to use a low pass.  But I quickly run out of imagination as to
the rc time constant to synthesize.

just discussing the howto's would I think be educational for those
of us who never got the transcendental stuff in school 70+ years
ago.

Help?

Thanks


Normally you would just connect the joint velocity pin to the PID
command derivative


Ok, but thats the signal I disconnected around 2 years back to get rid of
an instant following error for a movement of 1/10 the MIN_ERROR. So
hooking it back back up now does what to my FF# settings?



It requres somewhat different tuning because of differening delays
in the TPs idea of velocity and velocity calculated in PID via DP/DT
Changing anything now is likely a waste of time





Cheers, Gene Heskett
--
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law
respectable. - Louis D. Brandeis
Genes Web page <http://geneslinuxbox.net:6309/gene>


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers


Peter Wallace
Mesa Electronics

(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.



___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers



Cheers, Gene Heskett
--
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
- Louis D. Brandeis
Genes Web page <http://geneslinuxbox.net:6309/gene>


___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers



Peter Wallace
Mesa Electronics

(\__/)
(='.'=) This is Bunny. Copy and paste bunny into your
(")_(") signature to help him gain world domination.



___
Emc-developers mailing list
Emc-developers@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/emc-developers

Re: [Emc-developers] 2.7.15 PID Fix

[Gene Heskett]

On Thursday 05 March 2020 10:54:49 Peter C. Wallace wrote:

> On Thu, 5 Mar 2020, Gene Heskett wrote:
> > Date: Thu, 5 Mar 2020 10:42:19 -0500
> > From: Gene Heskett 
> > Reply-To: EMC developers 
> > To: emc-developers@lists.sourceforge.net
> > Subject: Re: [Emc-developers] 2.7.15 PID Fix
> >
> > On Thursday 05 March 2020 09:25:01 Peter C. Wallace wrote:
> >> On Wed, 4 Mar 2020, Gene Heskett wrote:
> >>> Date: Wed, 4 Mar 2020 22:12:47 -0500
> >>> From: Gene Heskett 
> >>> Reply-To: EMC developers 
> >>> To: emc-developers@lists.sourceforge.net
> >>> Subject: Re: [Emc-developers] 2.7.15 PID Fix
> >>>
> >>> On Wednesday 04 March 2020 12:52:09 andy pugh wrote:
> >>>> On Wed, 4 Mar 2020 at 17:26, Peter C. Wallace 
> >
> > wrote:
> >>>>> It was first added to 2.8 and them merged into 2.7 before the
> >>>>> 2.7.15 release AFAIK
> >>>>
> >>>> I think that they were added in 2010:
> >>>> https://github.com/LinuxCNC/linuxcnc/commit/44586b831e1ef5e826003
> >>>>19 0ab 317cda76b6e8e3#diff-912d11dfa52738be509f63e4ff1922a4 (And
> >>>> that commit also includes documentation of the pins)
> >>>
> >>> Sort of. But it does provide some hints as to how to utilize it to
> >>> improve control, If you have encoder feedback on axis axis then
> >>> you could feed the feedback.deriv pin with the velocity of that
> >>> axis's encoder. But from reading that, I fail to deduct a source
> >>> for cmd.deriv. I am envisioning a differential, for instance
> >>> between current command and previous command or maybe a sum2 with
> >>> one input set for a gain of -1 and an rc lag from the command
> >>> input to the other input giving a differential output according to
> >>> the charge across the cap. Since we don't have a cap in hal
> >>> components, we'ed have to use a low pass.  But I quickly run out
> >>> of imagination as to the rc time constant to synthesize.
> >>>
> >>> just discussing the howto's would I think be educational for those
> >>> of us who never got the transcendental stuff in school 70+ years
> >>> ago.
> >>>
> >>> Help?
> >>>
> >>> Thanks
> >>
> >> Normally you would just connect the joint velocity pin to the PID
> >> command derivative
> >
> > Ok, but thats the signal I disconnected around 2 years back to get
> > rid of an instant following error for a movement of 1/10 the
> > MIN_ERROR. So hooking it back back up now does what to my FF#
> > settings?
>
> It requres somewhat different tuning because of differening delays
> in the TPs idea of velocity and velocity calculated in PID via DP/DT
> Changing anything now is likely a waste of time

I get that impression from some too slow to see it moves. The pid vels 
look sane, but the pid_n.error gradually builds up to slightly 
over .0001002xxxmm and stays there bouncing around like its moving from 
stiction, with an error polarity opposite to the velocity. Backlash comp 
too high by .0001 maybe.

In any event I suspect thats very close to the theoretical resolution 
given its steppers at either 8 or 16 microsteps.

If it won't move with similar errors at G0 speeds when I get to where I 
can see it move, I'll check the docs for enlightenment.

Thanks Peter.

> >>> Cheers, Gene Heskett
> >>> --
> >>> "There are four boxes to be used in defense of liberty:
> >>> soap, ballot, jury, and ammo. Please use in that order."
> >>> -Ed Howdershelt (Author)
> >>> If we desire respect for the law, we must first make the law
> >>> respectable. - Louis D. Brandeis
> >>> Genes Web page <http://geneslinuxbox.net:6309/gene>
> >>>
> >>>
> >>> ___
> >>> Emc-developers mailing list
> >>> Emc-developers@lists.sourceforge.net
> >>> https://lists.sourceforge.net/lists/listinfo/emc-developers
> >>
> >> Peter Wallace
> >> Mesa Electronics
> >>
> >> (\__/)
> >> (='.'=) This is Bunny. Copy and paste bunny into your
> >> (")_(") signature to help him gain world domination.
> >>
> >>
> >>
> >> ___
> >> Emc-developers mailing list
> >> Emc-developers@lists.sourceforge.net
> >> https://li