Photoart - Kalmar

Making of Vienna 50 Gigapixel

The biggest panorama worldwide ( 21.07.2010 )

About 9 months ago, I was asked by e-mail for testing an automatic panoramic tripod head. At first I was a little surprised because I am doing my panoramas by using a manual head only.
But wondering which progress had been made during the last 3 years, especially concerning software and handling of such devices, contacts were established.

Having done some testing without any problems, in April the time had come. Getting up the Donauturm (Danube Tower, a panoramic tower in Vienna) I had a feeling I never had had before: wouldn’t it be fine to install there all the equipment, to wait that all the rest would be done automatically by the panoramic head and return home after three hours?

Unfortunately, when doing a first evaluation of the single photos, I had to guess that the result would have been defective. There is one thing that cannot be done by an automatic head: thinking.

There were a lot of clipped cars and trains cut into halves to be seen in the pictures, and only one crane had been correctly photographed to be able to be used at the building site without amputation of one of its jib booms. It would have taken months and hard work for glossing over the picture in order to gain some more or less suitable result.
It would of course have been possible to remain all the time of taking pictures in order to release manually whenever there was any scene in motion. But when would that not have been the case? But if you have to remain beside the camera for the whole time only for obtaining good results, such an automatic head will no longer make any sense.

For avoiding the problem of the changing shadows I did not take the pictures by horizontal serpentines as this is done usually, but by vertical lines.
I think that such an automatic head notwithstanding its excellent quality will not be suited for reaching good results. There is, however, one exception: there is no or nearly no living subject to be seen in the picture and no problems concerning the focus will have to be expected (as it was the case of the excellent 27 gigapixel panorama of Paris made by Arnaud Frich).

Having therefore put all my pictures taken into the electronic waste-box and after having taken a deep breath I went up the tower again.

When preparing the project I thought that the greatest obstacle would have been set up there as there is a wonderful visitors’ platform which, however, because of numerous bars, is not suited for taking photographs. Assisted by the personnel of the tower in an outstanding manner, I was fortunately able to bring my camera and myself into position about 50 m above the visitors’ platform.

For obtaining good results of the project many challenges had to be accepted.



There will be always some wind up there.
At first I determined the highest possible shutter speed by taking some pictures. Not before 1/2000 sec and a focal length of 400 mm had been taken (diaphragm 8/650 ASA), the result got  satisfying. Unfortunately, the 1, 4 converter could be used no more because it was too windy.



As shadows will change very rapidly if one begins too early, I waited until 08:45 a.m.
At 10 o’clock 31 degrees had been reached in the sun, 38 at 11, and when I was taking the last picture at 11:43, heat had arisen up to 48 degrees. So I had to take three times as much for going down by some external ladders as I had taken to get up there.


Changing the position
For doing a 360° panorama I had to circulate a diameter of 20 m. Because of numerous antennas, position had to be changed about 8 times.
The point where the position has to be changed has turned out to be decisive.  For instance, the ladder (1) is to be seen in this picture not without any reason. It would not have been any problem to change the position before, but this would have lead to serious consequences, i.e. to parallax displacement because some objects in the near and more distant ones were shot by single photos as it is the case with the multi-storey buildings (2). For this reason position was changed not earlier than these buildings had been shot. The commonly shared opinion that telephoto lenses will not be susceptible to diverging from the nodal point is caused by the fact that, due to the limited focal depth of those objectives, near and distant objects will not appear in the same picture. But if this is the case, divergences will be seen although telephoto lenses were used.
So if you want to violate the law of panorama photography by changing your position while taking pictures, you will have to think this over thoroughly for not getting punished. It was also very important for me to align exactly the camera whenever   position had been changed. To be absolutely sure, I used 3 spirit levels: one was mounted to the central column of the tripod, another one to the head to which afterwards the adapter of the nodal point was attached, and a third one to the camera. Taking pictures was not continued earlier than all the spirit levels had been aligned.


Changing shadows

Earth rotation will of course not stop during work, and therefore also shadows will change. In order to avoid these problems, I did not take the pictures by horizontal serpentines, as commonly used, but by vertical lines. So I was able to avoid nearly all the problems with the changing shadows. It can be seen from the panorama’s interfaces that earth is still rotating (shadows 3 and 4 pointing to different directions).


Some time ago, I ordered a ring with 3.6 ° click stops to be milled, but only for the horizontal rotor, because the vertical one is too small for such fine clickstop grading. But for realising my project click stops were needed for the vertical rotor, too. The simplest solution of the problem was to mount the panoramic head not directly to the tripod, but at first the usual ball head by which the panoramic head could be turned. So the horizontal rotor became the vertical one equipped with click stops.
I made this picture by using a 1.4 converter which later on could not be used because wind was too strong.
By using a panoramic head modified in this manner and with some practice you will be able to work very rapidly. During no more than 3 hours 3600 pictures were taken and all that by manual control. Taking into account the times necessary for changing position and the fact that the AF had to be switched of whenever a new line had begun and switched on when the city was aimed at, one can easily imagine the high working speed offered by such a head.



It is a matter apart not to loose concentration during the whole time with temperatures exceeding by far 40°.


Technical specifications


I am always asked which camera, optical lenses, programs and other hardware I am using. As it can be seen from the above photo, the panorama has been taken by a 5dMk and a 400mm/5, 6. But camera and optical lenses will influence quality by no more than 5% because the most important influence is being exerted by the weather.
Or put in other way: when the weather is fine, much better pictures will be made by nearly every combination of cheap camera and cheap objectives than by the most expensive devices. Very good results will also be obtained from using a 1, 4 converter if the shutter speed will be increased to a high extent. As to the picture above: 100% crop, 1/2000 sec, diaphragm 8 (i.e. aperture completely opened), dead calm. It has been correctly focussed.
If the shutter speed is too low, many gigapixel pictures made by a 1.4 converter will get blurred. However, it is a different matter if a 2.0 converter is used, but I should recommend keeping your fingers off. It will be much more reasonable to buy a camera with a cropsensor. As to the camera, you should choose a more expensive one, because low-price cameras have been designed neither for shooting at once several thousand pictures nor for doing it several times.


Raw or Jpg

I am using Raw only, which has turned out to be advantageous. When taking your picture in this moment you will not have to think about white balance.

Tripod/panoramic head

A solid tripod is a must. That’s all.
As to the panoramic head, many different solution are available, going from cheap ones of 150€ to those very expensive mechanized VR solutions. I am using a Nodalninja 5, which, however, is not meant as a recommendation. Each stable head will be suited for such a project.
As mentioned above, those expensive mechanized solutions will make an advantage only if one will accept all the limitations originating from the motor.
There is another disadvantage: When such projects are being dealt with, combining the motor head with the auto focus will be somewhat difficult. There is a simple reason: In such great panoramas you will always find some spots where the AF cannot find any point for focussing (for instance, when realizing my project I had to manage some problems caused by the waters). There is one exception: you will make sure in advance that there will no such problems at all.
As manual focussing was done for each line of the Paris panorama, considerable improvement has been gained compared to the traditional   method of not changing the AF during the whole period of photographing.
But this method will not come up to the possibility of focussing perfectly each single picture.
I should like to advise you to spend money rather for expensive optics than for expensive automatic solutions. But if you nevertheless want automatic photographing without getting only an accumulation of pixel, you will have to think over thoroughly where the camera should be stationed, or be prepared for long nights with the photoshop. 



For assembling so many photos into one panorama only, so-called stitching software will be needed. The following programs have prevailed in recent years: PTGui, Autopano and Hugin. It would not be serious to recommend one of them, because equal results can be obtained from all these programs if handled correctly. You should try them at least for 3 months before buying. Test versions of suitable programs are always being available.
There is further information you should have: Realising a landscape panorama (city panoramas included) will easy for stitching software, because dimension is not important. Therefore it does not matter if a city panorama has to be computed with 10 or 10.000 single photos. Unfortunately this does not apply for the photographer, too. It will be always him and not the software to be the principal cause for not obtaining any result at all or a very bad one. No program will, therefore, be able to bring a perfect result if such quantities of dates have to be dealt with.
The typical sources of error are:

  • Control points attached to moving elements (clouds, cars...)
  • No control points at all in overlapping pictures (most frequent stitching error)
  • Elements that have been focussed in one picture but not in the other overlapping one

If there are phantom pictures the wrong position if your automatic head has been chosen. In such a case no program will be helpful

Another tip concerning great panoramas: before stitching you should take the time for finding out any eventual source of error. A forecast is being offered by each program before stitching what kind of result is to be expected. If the prevision is an unfavourable one try to find the reasons in order to prevent your computer from useless suffering.



At first good news: Such a panorama can be done by nearly every computer.
Now the bad news: the older the model the more patience will be needed. Up from 10 gigapixels you should use a 64 bit system with at least 8 GB Ram and 4 core CPU. And you will need much of space on your hard disk for computing a panorama of such a dimension (at least 2 TB). All the programs will shift their information as temporary files to the hard disk.


PC or Mac, Intel or AMD

All this is of no importance.


Assembling the single pictures

Each photographer will have his own workflow.
When realising such great projects, I never will download the whole panorama at the same time, only for the simple reason that misplaced control pints will creep in. To find them out, I will usually download no more than500 pictures which will be optimized before the next download. So it is much easier to find out problems and to solve them. But everyone will use his own method. After having set all control points I divide the whole into 6 tiles, 60° each of them. It would be useless to have computed the whole panorama at the same time because there is no program available for opening such a large file. And if so, would you really want to know what time will be needed to open on your computer a file of 150 GB? Each tile is of about 25 GB – a considerable amount of data.


How to present such panorama in the internet

This can be done by using the Krpano viewer and its tools.  All the small tiles used for the internet can easily be computed from the great tiles and assembled to a great panorama. Apart from all the many possibilities offered by this small viewer, it is the customer service which will go beyond any expectation as high it might be.


Julian Kalmar