![]() |
|
|
|
|
|
||||||||||
|
| User list | You are not logged in.
I just finished my new crown using perspex (which i have cut with an laser) and a Manfrotto slide rail. So far anything works great, but I have a question about the horizontal moving speed.
I did a test pano with an 90mm lens , a FOV of 80° to 40°, resulting in 11x4 images. When the head moves horizontal from one capture to the next, it moves the first half way quite fast and then slows down to a very slow speed to finish the second half of the way, which needs several secounds.
I think this is imagined to avoid camera shaking, but it is very time consuming and to my taste it would be enough to slow down the last 10 percent of the way.
Is this a normal behaviour or could it be an issue of my power source (8x 1.2V NiMH AA Batteries)?
Thanks for your help and all the priceless input of this forum!
Best, Helmut
Last edited by SpeedyPan (2009-03-18 11:48:22)
Offline
When moving to positions >6~7° from one point to another, the head accelerates to full speed, and descelerates near the destination; acceleration/desceleration are about the same.
When moving to positions <6~7° from one point to another, it remains at low speed (this is the internal electronic fact).
But I don't understand what you describe... What is the angle of the moves?
PS : the batteries are not in cause (I use teh same config).
Offline
Frédéric
Thanks for your reply. The head moves 7 degrees with this setup. When it moves from one capture to the next, it needs about 1 second for the first 4 degrees, then it slows down and finishes the last 3 degrees in about 5 seconds.
If I get you right, this is a normal behaviour and could not be changed. It would mean, that moving <6 degrees would last longer than moving 7 degrees, right?
Helmut
Offline
Did another test with 14 degrees between the captures. This needs only about 3.5 seconds in total, as the deceleration distance is much less (maybe 0.5 degrees).
Offline
Is it the same for a longer move, like 20°? 5s for the last 3° seems very long...
And, yes, moving 3° can be longer than 8°. I plan to make an external closed loop positionning (see http://trac.gbiloba.org/papywizard/ticket/25), using start/stop commands (which always go to full speed), and controling the position from Papywizard. But it will only be possible with a new design, using threads to control the position.
Offline
At larger movements the decelaration phase is much less, which results in moving 20 degrees is much faster than moving 7 degrees...
Offline
So it seems that you are at the limit. I didn't make precise tests to see what are the critical values...
When I start the external closed loop positioning regulation, I will make a script to measure accelerate and decelerate times, for values from 2° to 10°, to see how to optimize it.
Offline
SpeedyPan wrote:
Is this a normal behaviour or could it be an issue of my power source (8x 1.2V NiMH AA Batteries)?
Yes - that´s normal. I realized it also when i used a 105mm. It starts to move at normal speed and slows down to accurately set the next position. That takes time - but really is accurately.
Alternatively you´ll have to use a more sophisticated gear . . and pay more than 5-7 times the price for a comparable setup like Dr. Clauss "Rodeon" . . ![]()
This one really is precise AND fast . . . and starts at around 3700.-€.
Or the PixOrb at around 6000.-$. This one´s also fast.![]()
best, Klaus
Offline
I'm pretty sure Frédéric will speed it up for us one day ![]()
I already checked out the Clauss system and maybe I will buy one, when I am a famous panorama photographer ![]()
Offline
I can see this as a major drawback.
An ideal shoot would be 3 rows of 15 shots. This is really rough workings but my assumption is that moving from one shot to the next @ 100mm would take about 5 seconds, moving between shots @ 200mm takes around 20 seconds.
3 rows of 15 @ 200mm takes 15 minutes
3 rows of 15 @ 100mm takes 3 min 45 sec
(I know this is rough workings but it's a reasonable guide).
3 min 45 seconds is a long time when trying to capture a landscape but 15 minutes is too long to be useful (most of the time). Basically there is an increase in shoot time of 11 min 15 seconds because the head is working slower with the 200mm lens.
A short term solution could be to move the head forwards and then back by more the 7'. example - to advance by 5' move forwards by 13' and back 8'. I'm not sure if this would be much quicker but I thought I'd put the idea forwards.
Cheers
Will
Last edited by london-will (2009-03-18 16:55:43)
Offline
london-will wrote:
I can see this as a major drawback.
An ideal shoot would be 3 rows of 15 shots. This is really rough workings but my assumption is that moving from one shot to the next @ 100mm would take about 5 seconds, moving between shots @ 200mm takes around 20 seconds.
3 rows of 15 @ 200mm takes 15 minutes
3 rows of 15 @ 100mm takes 3 min 45 sec
(I know this is rough workings but it's a reasonable guide).
3 min 45 seconds is a long time when trying to capture a landscape but 15 minutes is too long to be useful (most of the time). Basically there is an increase in shoot time of 11 min 15 seconds because the head is working slower with the 200mm lens. Any advice on reducing this time would be invaluable!
Cheers
Will
Will,
Welcome to the forum.
Off the top of my head, what if you were to shot column by column rather than row by row?
Or, and this is a bit more work.
If you consistently shot 3 rows of 15 shots with the same lens for your desired pano FOV then you could make a custom presets for the 200mm whereby you shoot every other position in one direction the reverse to take the alternates, for each row.
http://trac.gbiloba.org/papywizard/wiki … Presetmode
..............
BTW for those who have not yet seen Will's really splendid panos:
http://www.willpearson.co.uk/
Last edited by mediavets (2009-03-18 17:01:38)
Offline
Do you mean putting the camera in landscape position rather then portrait? I hadn't thought of that....! Nice one, I'll give it a try.
Another idea I had -
A short term solution could be to move the head forwards and then back by more the 7'. example - to advance by 5' move forwards by 13' and back 8'. I'm not sure if this would be much quicker but I thought I'd put the idea forwards.
Offline
london-will wrote:
Do you mean putting the camera in landscape position rather then portrait? I hadn't thought of that....! Nice one, I'll give it a try.
No, that's not what I meant but it's a good idea. Worth a try.
I meant to chnage the order in which the matrix is shot so that it shoost column by column rather than row by row (with the camera in portait orientation) my (off the top of my head) thinking was that the pitch variation between rows might be greater than the 'magic' 6 degrees.
Another idea I had -
A short term solution could be to move the head forwards and then back by more the 7'. example - to advance by 5' move forwards by 13' and back 8'. I'm not sure if this would be much quicker but I thought I'd put the idea forwards.
My idea about using a custom preset, seeks to achieve the same goal:
"Or, and this is a bit more work.
If you consistently shot 3 rows of 15 shots with the same lens for your desired pano FOV then you could make a custom presets for the 200mm whereby you shoot every other position in one direction the reverse to take the alternates, for each row.
http://trac.gbiloba.org/papywizard/wiki … Presetmode"
This is seeking to achieve the same goal - ie. to make the yaw variation between shots greater than 6 degrees.
A custom prest is simply a list - in XML format - of shooting positions, they can be in any order. This would conventionally be used for defining shooting positions when shooting sphericals with a FE lens, but can be used for any other purpose.
Last edited by mediavets (2009-03-18 17:50:09)
Offline
I like this idea and want to give it a try.
Does anyone know how to get a new preset into the Nokia 770?
Offline
SpeedyPan wrote:
I like this idea and want to give it a try.
Does anyone know how to get a new preset into the Nokia 770?
If your Nokia has a removable card then plug your Nokia in to a PC via USB- the removable card appears as another disk drive to the PC.
Copy the custom preset file to the removable card.
When in Papywizard load the Preset file via File/Load preset...
Preset files loaded this way remain available until Papwizard is closed. Such preset files can have a filename in the form <anyname>.xml.
Installing custom presets permanently is a bit more difficult and they must all be in one XML file with a special name which must be placed in a specific location in the filesystem - on a 770 you would need to install GPE-Filemanager to get access to the filesystem.
Try the other way first - it's easy.
In fact I like to use my presets this way - keeping a 'library' of presets on one of the two removable cards on my N800 (yes, N800 supports two removable cards) and loading them on a per session basis as required.
Last edited by mediavets (2009-03-18 17:49:20)
Offline
Both axis run at the same speed; the 6° limitation is the same on yaw and pitch.
Another solution is to take one picture every n, so each move is >6°. Then, we go back to take the others... Like parsing all white squares of a checkboard, before parsing the black ones. But we can do more complex things, depending of the needs.
What user parameters do you think we should add for such feature? The real question is: what are the drawback of such feature we could avoid by a user choice?
I'll try to implement something like that quickly...
Offline
I'm finding this really tricky.... In a practical world, shooting in a linear way (a, b, c, d, e,) is better to capture a scene, if the scene changes retouching easier.
examples would include -
- Shooting sky where clouds move.
- Shooting landscapes where clouds cast shadows or the sun comes in and out.
- Shooting at ground level where cars / busses etc move in and out of shots.
- Shooting in the blue hour / sunsets when the light changes constantly.
From experience shooting a,b,c,d, can handle these situations better then shooting one picture every n (white squares of a checkboard). I know that shooting every n would result in a faster shoot time but I'm not sure it would result in a better render. I suppose its a payoff between shoot time and getting the best results.
Offline
fma38 wrote:
Both axis run at the same speed; the 6° limitation is the same on yaw and pitch.
Another solution is to take one picture every n, so each move is >6°. Then, we go back to take the others... Like parsing all white squares of a checkboard, before parsing the black ones. But we can do more complex things, depending of the needs.
That's what I was trying to explain/suggest but using a custom preset to achieve this sort of thing.
I think one would choose to shoot by row top down and might offer options to skip by 2, 3, or 4 shots, then reverse (and reverse again as required) on each row fill in before proceeding to the next row?
You might have to add additional shots to make it work - so they divide appropriately -but the trade off in terms of overall time may be worth it?
Or maybe not, for the scenarios that Will described so well.
............
Of course the best (but very costly) solution is to use two cameras on a special variant of the RodeonVR head as planned for the 20 Gigapixel pano of Paris project:
http://www.autopano.net/blog/2008/12/15 … igapixels/
See a short video of this RodeonVR 'beast' on this page (scroll down):
http://www.autopano.net/blog/2008/12/15 … igapixels/
Even so I imagine that there must be limits as to how long it will take to shoot effectively at each position, because factors over and above speed of movement of the head will play a part, such as time allowed for the whole rig to settle after movement, mirror lockup and all that, especially when shooting very hi-res very large matrices/mosaics with long focal length lenses?
I seem to recall it took something like 1 1/2 hours for Gerard to shoot the 13 Gigapixel New York pano, and how long was it for shooting the Yosemite pano?
Last edited by mediavets (2009-03-18 20:32:16)
Offline
I did a simple test with 10 images from 0 degree to 45 degrees.
Test No. 1 with 0, 5, 10, 15, 20. 25, 30, 35, 40, 45 degrees needed 103 seconds.
Test No. 2 with 0, 25, 5, 30, 10, 35, 15, 40, 20, 45 degrees needed 65 seconds, which means it is about 37 percent faster. In other words... a shot normaly lasting 20 minutes could be done in 12.5 minutes.
I'm not sure, if this "speed-up-trick" really would end up in a worse stitch because the total time needed is so much less.
Both tests were done with 0.5 seconds shooting time and 0.3 seconds stabilization time.
So for me Will has found a really good workaround to speed up panos with long lenses. Thanks for that, Will!
Last edited by SpeedyPan (2009-03-18 21:45:16)
Offline
Thanks for taking the time to do this experiment.
The results are very interesting.
Offline
Did one more test with 17 shots over 80 degrees.
Normal order: 180 seconds
Optimized order: 98 seconds
So it is 45 percent faster...
Here are the preset data:
<shoot>
<pict yaw="0." pitch="0." />
<pict yaw="5." pitch="0." />
<pict yaw="10." pitch="0." />
<pict yaw="15." pitch="0." />
<pict yaw="20." pitch="0." />
<pict yaw="25." pitch="0." />
<pict yaw="30." pitch="0." />
<pict yaw="35." pitch="0." />
<pict yaw="40." pitch="0." />
<pict yaw="45." pitch="0." />
<pict yaw="50." pitch="0." />
<pict yaw="55." pitch="0." />
<pict yaw="60." pitch="0." />
<pict yaw="65." pitch="0." />
<pict yaw="70." pitch="0." />
<pict yaw="75." pitch="0." />
<pict yaw="80." pitch="0." />
</shoot>
<shoot>
<pict yaw="0." pitch="0." />
<pict yaw="20." pitch="0." />
<pict yaw="5." pitch="0." />
<pict yaw="25." pitch="0." />
<pict yaw="10." pitch="0." />
<pict yaw="30." pitch="0." />
<pict yaw="15." pitch="0." />
<pict yaw="35." pitch="0." />
<pict yaw="45." pitch="0." />
<pict yaw="55." pitch="0." />
<pict yaw="40." pitch="0." />
<pict yaw="50." pitch="0." />
<pict yaw="60." pitch="0." />
<pict yaw="75." pitch="0." />
<pict yaw="65." pitch="0." />
<pict yaw="80." pitch="0." />
<pict yaw="70." pitch="0." />
</shoot>
Offline
I did some Merlin "speed" tests.
1 row of 10 shots 30% overlap/time value 0,1 sec/stab.dely 0 sec.
50mm FF 197 degr FOV: time= 1:17
80mm FF 124,5 degr FOV: time= 1:03
100mm FF 99,9 degr FOV: time= 0:58
110mm FF 90,9 degr FOV: time= 1:00
120mm FF 83,4 degr FOV: time= 1:17
130mm FF 77,0 degr FOV: time= 1:24
140mm FF 71,5 degr FOV: time= 1:34
150mm FF 66,8 degr FOV: time= 1:44
160mm FF 62,6 degr FOV: time= 1:53
170mm FF 58,9 degr FOV: time= 1:55
250mm FF 40,1 degr FOV: time= 1:26
400mm FF 25,1 degr FOV: time= 1:01
and 1 row, 30% overlap/time value 0,1 sec/stab.dely 0 sec.
FOV 90
50mm Time= 0:32
80mm Time= 0:47
100mm Time= 0.52
120mm Time= 1:24
150mm Time= 2.36
200mm Time= 3.10
400mm Time= 3:51
It seems to me that the Merlin/PW is at best (speed) with a ±100mm@FF for mosaics
BTW, All started from "home" position.
Frederic, can we have that in 1.6 automatic or button, so after shooting go to "home" ???
Last edited by bigwade (2009-03-19 01:14:16)
Offline
Thanks for all these tests!
SpeedyPan, could you try this scheme:
<shoot>
<pict yaw="0." pitch="0." />
<pict yaw="10." pitch="0." />
<pict yaw="20." pitch="0." />
<pict yaw="30." pitch="0." />
<pict yaw="40." pitch="0." />
<pict yaw="50." pitch="0." />
<pict yaw="60." pitch="0." />
<pict yaw="70." pitch="0." />
<pict yaw="80." pitch="0." />
<pict yaw="75." pitch="0." />
<pict yaw="65." pitch="0." />
<pict yaw="55." pitch="0." />
<pict yaw="45." pitch="0." />
<pict yaw="35." pitch="0." />
<pict yaw="25." pitch="0." />
<pict yaw="15." pitch="0." />
<pict yaw="5." pitch="0." />
</shoot>Is it faster?
Offline
Good Morning Frédéric
Yes, it is a little bit faster: 93 seconds.
It could be some seconds faster if there were not a slow 5 degree step between 80 and 75 degrees.
By the way... I was able to reproduce bigwades test results on my system.
Best, Helmut
Last edited by SpeedyPan (2009-03-19 08:55:39)
Offline
Powered by PunBB
© Copyright 2002–2005 Rickard Andersson
|
CHOOSING KOLOR Why choose Kolor? Which solution to choose? Download a trial Where can I buy? Education |
SOFTWARE Autopano Pro Autopano Giga Panotour Panotour Pro XnView |
ACCESSORIES Training DVD Panobook PROJECTS Paris 26 Gigapixels Yosemite 17 Gigapixels |
COMMUNITY Forums Blog |
COMPANY About Kolor Corporate blog Resellers Contact |
PRESS Press center Press review TOOLS My account |
