strange deep zoom glitch (help!)

[claude]

Should be a regular-looking minibrot with rings like the bottom right corner all around, but instead it's all spangly and glitchy, not to mention the wrong shape.

Anyone have any ideas why this might occur and how to fix it?  Or even seen anything like it in another location?

This is rendered with series approximation (2048 coeffecients of extended-exponent double) and distance estimation, it's dark because the filaments are very dense.

I've half a mind to fake the ending of the 2x zoom sequence with an appropriately scaled and rotated view of the top-level continent, as only the last 7 frames (including this one) are affected...

[quaz0r]

2048 coefficients  

when you guys try to debug locations like this, do you test manually limiting how far the series approximation goes, to see when it does in fact render properly?  maybe you can perform some kind of analyses and figure out a more proper criterion for when to bail on the series approximation   

[Chillheimer]

...and if you zoom deep enough, just before reaching infinity.. you will find..
Pacman!

(sorry, I have nothing helpful to contribute..)

[quaz0r]

Code:

                  ______
           _____/      \\_____
          |   _     ___   _   ||
          |  | \     |   | \  ||
          |  |  |    |   |  | ||
          |  |_/     |   |_/  ||
          |  | \     |   |    ||
          |  |  \    |   |    ||
          |  |   \. _|_. | .  ||
          |                   ||
          |     Here lies     ||
          |    Chillheimer    ||
          |                   ||
          |   He contributed  ||
          |        The        ||
          | PacMan Conjecture ||
          |    to humanity    ||
          |                   ||
          |                   ||
  *       |  *   **    * **   |**      **
   \)).....//.,(//,,..,,\||(,,.,\\,.((//

[Adam Majewski]

what are results from other programs ?
Can you post parameters ?

[claude]

what are results from other programs ?
Can you post parameters ?

I haven't tried any other programs, but I believe it is too deep for MandelMachine, which leaves Kalles Fraktaler as the only(?) published program that can do it, and that doesn't do distance estimation (as of now) so it's not useful to me.at this time.

Location is Kalles' e10000 location: http://www.chillheimer.de/kallesfraktaler/e10000.kfr with radius between 1e-10000 (glitch first appears) and 1e-10002 (image in my first post).

[PieMan597]

You could maybe try putting the initial reference in the minibrot?

[claude]

It should be already, but maybe there's a bug (with insufficient precision perhaps?) - thanks for the idea of something to look into..

[quaz0r]

picking better reference points is of course a good thing, but shouldnt we be the most concerned that incorrect results can be produced?  i think this should be the focus   

[PieMan597]

I'll try loading the coordinates into Mighty Mandel, and see what I get.

[claude]

I think it's fixed (still some frames to render but looking promising, I can see a babybrot a few pixels, wide enough to tell it isn't pacman).  thanks PieMan597 for pointing me to the problem.  quaz0r, don't worry - this was a case of user error aka garbage in garbage out... (I had the manual precision signifier set too low, so the primary reference wasn't as good as it should have been).. my code doesn't yet check for every stupidity...

[quaz0r]

good news