Re: New file uploaded to ap-gto


Ray Gralak
 

Hi Bill,

Someone here suggested trying Lowpass2, and I am running that now. So far DEC looks pretty solid with very, very
little if any corrections.
Of course that's great. However, the algorithms are static and have to be selected. What happens if seeing conditions change?

I know this is changing topics a bit, but it really goes to making autoguiding easier. PHD was supposed to be easy to use, but it's
not so much any more. I've said this before but I think all of the autoguiding techniques are not very sophisticated and sometime
need human intervention to fit changing conditions. One solution could be to use cloud-based machine learning techniques to
automatically adapt the autoguiding parameters based on previously learned behaviors.

Of course this would mean an internet connection would be required. Then again, only lightweight client software would be needed on
a user's computer, making it possible even for low powered devices to use powerful machine learning techniques that wouldn't require
users to figure out many, if any, settings.

-Ray Gralak
Author of APCC (Astro-Physics Command Center): http://www.astro-physics.com/index.htm?products/accessories/software/apcc/apcc
Author of PEMPro V3: https://www.ccdware.com
Author of Astro-Physics V2 ASCOM Driver: https://www.siriusimaging.com/apdriver


-----Original Message-----
From: ap-gto@... [mailto:ap-gto@...]
Sent: Tuesday, May 7, 2019 9:58 PM
To: ap-gto@...
Subject: Re: [ap-gto] New file uploaded to ap-gto



Someone here suggested trying Lowpass2, and I am running that now. So far DEC looks pretty solid with very, very
little if any corrections.

________________________________

From: ap-gto@... <ap-gto@...> on behalf of 'Ray Gralak (Groups)'
groups3@... [ap-gto] <ap-gto@...>
Sent: Tuesday, May 7, 2019 9:54 PM
To: ap-gto@...
Subject: RE: [ap-gto] New file uploaded to ap-gto



Hi Cytan,

Just looking at the source code, the ResistSwitch() algorithm looks like simple proportional control with gain =
aggression value. If that's the case, then it is nearly impossible to screw it up. This means that it has to be
downstream of ResistSwitch() which unfortunately, means that the GTOCP4 firmware when absolute encoders
are
enabled cannot be ruled out as the culprit.
I doubt that very much.

There are many other ways PHD2 could have a bug, for instance incorrectly calculating the calibration values it is
using, incorrectly calculating or rounding up move durations, etc.

-Ray Gralak
Author of APCC (Astro-Physics Command Center): http://www.astro-
physics.com/index.htm?products/accessories/software/apcc/apcc
Author of PEMPro V3: https://www.ccdware.com
Author of Astro-Physics V2 ASCOM Driver: https://www.siriusimaging.com/apdriver

-----Original Message-----
From: ap-gto@... [mailto:ap-gto@...]
Sent: Tuesday, May 7, 2019 5:56 AM
To: ap-gto@...
Subject: Re: [ap-gto] New file uploaded to ap-gto



HI Ray,
Just looking at the source code, the ResistSwitch() algorithm looks like simple proportional control with gain =
aggression value. If that's the case, then it is nearly impossible to screw it up. This means that it has to be
downstream of ResistSwitch() which unfortunately, means that the GTOCP4 firmware when absolute encoders
are
enabled cannot be ruled out as the culprit.

I hope that Andy Glasso who is one of the maintainers of PHD2 is able to shed light on this matter.

cytan

On Tuesday, May 7, 2019, 7:46:57 AM CDT, 'Ray Gralak (Groups)' groups3@... [ap-gto] <ap-
gto@...> wrote:




Hi Cytan,

Let's try something that is not way off left field about PHD2: there may be a bug.
I'm pretty sure there are bugs in both MaximDL and PHD2. Almost every non-trivial program has bugs.

But, even if there is a bug and if PHD2 guided better than MaximDL, my point was that I think PHD2 has stepped
away from its roots. The goal of the initial version was to be simple to operate... and the original version was. I
think
PHD2 has become more complicated than it needs to be, maybe because it's open source and many of the
people
making the changes each had their own ideas on how autoguiding *should* work.

The developers behind MaximDL have been doing this for 25+ years. They have gathered experience using
numerous mount types and operating under a wide range of field conditions. Because the product is commercial
they had resources to buy test equipment like cameras and mounts and hire people to test/develop.

But, as I said previously, I am sure many (most?) are willing to save $500 for software even if it compromises
ease
of use and performance.

-Ray Gralak
Author of APCC (Astro-Physics Command Center): http://www.astro-
physics.com/index.htm?products/accessories/software/apcc/apcc
Author of PEMPro V3: https://www.ccdware.com
Author of Astro-Physics V2 ASCOM Driver: https://www.siriusimaging.com/apdriver

-----Original Message-----
From: ap-gto@... [mailto:ap-gto@...]
Sent: Tuesday, May 7, 2019 4:14 AM
To: 'Ray Gralak (Groups)' groups3@... [ap-gto]
Subject: Re: [ap-gto] New file uploaded to ap-gto



Let's try something that is not way off left field about PHD2: there may be a bug.

I looked at the source code:

https://github.com/OpenPHDGuiding/phd2/blob/master/guide_algorithm_resistswitch.cpp


The member function that does the dirty work is


double GuideAlgorithmResistSwitch::result(double input)


Unfortunately, I don't see anything obviously wrong with the ResistSwitch algorithm right now. It looks like the
ResistSwitch algorithm just takes the input (which I assume is the error) and multiplies it by the aggression
setting
and returns it back to the caller.

cytan


On Monday, May 6, 2019, 10:48:12 PM CDT, 'Ray Gralak (Groups)' groups3@... [ap-gto] <ap-
gto@...> wrote:




What's interesting about this discussion is that "PHD" is supposed to be simple to use, but I think the PHD2
developers lost track
of that by adding too many features and knobs.

Now, MaximDL is the easy tool to use. And, apparently MaximDL works better.

-Ray Gralak
Author of APCC (Astro-Physics Command Center): http://www.astro-
physics.com/index.htm?products/accessories/software/apcc/apcc
Author of PEMPro V3: https://www.ccdware.com
Author of Astro-Physics V2 ASCOM Driver: https://www.siriusimaging.com/apdriver

-----Original Message-----
From: ap-gto@... [mailto:ap-gto@...]
Sent: Monday, May 6, 2019 8:34 PM
To: ap-gto@...; andy.galasso@...
Subject: Re: [ap-gto] New file uploaded to ap-gto



Adding Andy Glasso to the thread here. There has got to be a way to set PHD2 to work similarly to your
experience
with MaximDL.

________________________________

From: ap-gto@... <ap-gto@...> on behalf of chris1011@... [ap-gto] <ap-
gto@...>
Sent: Monday, May 6, 2019 7:00 PM
To: ap-gto@...
Subject: Re: [ap-gto] New file uploaded to ap-gto



Max move doesn't come into play for small corrections on the order of 1 arc sec.


Rolando




-----Original Message-----
From: Bill Long bill@... [ap-gto] <ap-gto@...>
To: ap-gto@... <ap-gto@...>
Sent: Mon, May 6, 2019 8:41 pm
Subject: Re: [ap-gto] New file uploaded to ap-gto




Hey Roland,

Did you try lowering the MxDEC value in PHD? That will allow you to change the maximum size of a
correction,
without changing the aggression. I use 2000 and have used 1500 in the past with good success. The default is
2500
and at 28% aggressiveness as you mentioned, it would only apply up to a 700 move. Maybe put the
aggression
back to 100% and set MxDEC to 1500?

________________________________

From: ap-gto@... <ap-gto@...> on behalf of chris1011@... [ap-gto] <ap-
gto@...>
Sent: Monday, May 6, 2019 6:27 PM
To: ap-gto@...
Subject: Re: [ap-gto] New file uploaded to ap-gto


Hi Cytan,


There is no way that the encoders have anything to do with causing Dec aggressiveness to behave this way.
The
encoders work perfectly with MaximDL at the proper settings of 100% aggressiveness.


And yes, you are getting the same instability as I was getting.



Rolando




-----Original Message-----
From: cytan299@... [ap-gto] <ap-gto@...>
To: ap-gto <ap-gto@...>
Sent: Mon, May 6, 2019 7:59 pm
Subject: Re: [ap-gto] New file uploaded to ap-gto




Hi Rolando,
I'm still experimenting with my AP1100AE with PHD2 settings and saw this on Saturday night. I'm glad that
you
may be seeing what I saw and I can confirm what you saw. Please see:

https://groups.yahoo.com/neo/groups/ap-
gto/files/AP1100AE%20and%20PHD2%20DEC%20settings%20%28weird%20behavior%29/
<https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgroups.yahoo.com%2Fneo%2Fgroups%2F
ap-
gto%2Ffiles%2FAP1100AE%2520and%2520PHD2%2520DEC%2520settings%2520%2528weird%2520behavior%
2529%2F&data=02%7C01%7C%7Cf4ceac89efaf44e58d0d08d6d2901818%7C84df9e7fe9f640afb435aaaaaaaaaaa
a%7C1%7C0%7C636927913648814470&sdata=wxgXrSHNJhXmMmp2cw3E77gpgZE34mu8DF%2BVoqf5%2BH
Y%3D&reserved=0>


When I had DEC aggression set to 100%, the DEC behaviour is clearly very odd. Then I turned it off to see
whether
the seeing was that bad. Clearly not! I then set DEC aggression to 50% and it looked a lot better.

Perhaps its the absolute encoders and PHD2 DEC guiding not playing well together?

cytan











Join main@ap-gto.groups.io to automatically receive all group messages.