a brief history of »tx-transform« between 1990 and 1996

by Martin Reinhart

The analogue times

In 1990 I was studying at the University of Applied Arts in Vienna under Bernhard Leitner. Embarrassed at not having come up with a proper semester project, I made an absurd suggestion on a whim. I claimed that if you wanted to drill a hole in the wall, it didn't matter whether the drill or the house rotated. The same would apply, correspondingly, to a photographic image, since it would not matter whether the focal-plane shutter of a camera moved across the film, or the film moved past a slit. I claimed that both had something to do with relativity and luckily it sounded interesting enough to be accepted. I didn't realise at the time what I was saying at all, but having made my claim, I had to find a way to demonstrate it. I started experimenting with slit photography without having any idea that this technique had been in use since the 1940’s. This cluelessness is hard to imagine today, but we still had to wait another 10 years for the beginnings of Internet searches.

At a flea market I bought a used Zenit camera and converted it in such a way that a spool full of film could be pulled past a slot made of cardboard by means of an adjustable motor. Since I had no idea how to calculate the exposure time from the ratio of slit width to film speed and how that in turn would affect the image ratio, I converted my apartment into a test lab. I turned the bathroom into a darkroom and in the living room I set up my brother's model railroad. The converted camera was mounted on one of the wagons and by means of a special control system it was possible to regulate both the speed of the train and the film transport separately. This arrangement allowed me to slowly approach the actual conditions and determine reliable values for the recording. When that was done, I dared to step outside the front door for the first time and take pictures of passing streetcars. These were particularly suitable because they moved in predictable paths and were so large that you couldn't miss them.

Fig.1: One of the first strip photos I took on Vienna’s Mariahilfer Strasse. In this example the camera was static and the film speed was chosen in a way to give an undistorted record of the streetcars passing by.

The results of these recordings amazed me and contained many aspects that I had previously not anticipated in this form. The details were striking: a streetcar driving against the direction of the film was shown mirror-inverted and all passers-by were walking in the same direction – regardless of whether they originally came into the picture from the left or the right. Those that moved too fast in relation to the film speed became thin strokes, those that walked slowly or stopped were stretched out. For me, it was like a glimpse into a new universe that I only slowly began to understand. I was particularly interested in the picture I took of a friend's Dalmatian. The dog was strangely bent and the tail seemed to float in the air. I wondered how this dramatic single image would look as a film sequence? This question engaged me well beyond the already delivered semester project but the most obvious solutions I found were sadly impractical.

Fig.2: My friend Regina’s Dalmatian “Agip” taken with the first, primitive RSA-Camera. The black horizontal lines result from the uneven cut slit mask made from cardboard.

First I came up with a truly analog idea. Since I began to comprehend that taking a picture with my slit-camera was really nothing more than cutting through a flip-book, I developed the model of the “block of information”, which made it possible to visualise how a view of movement along the space axis could be conceptualised. I translated this abstract idea in a quite bold way by having the individual frames of a short filmic sequences processed as single photographic prints. From this stack of about 100 photographs, I then cut away 1mm thin layers at a time with electric impact scissors. After each cut, I took a picture of the cut face with my 16mm Bolex camera – this went on until the block of pictures was shred into thousands of narrow strips. The two films that I made in such a way were technically poor, but they were the first moving RSA images I had ever produced. They at least proved that my approach was correct and led to visually interesting results.

Fig.3: A test for the first RSA films which were made by slicing a stack of photographs. The model is Isabella Parkinson.

Since cutting up photo stacks had turned out to be time-consuming and not particularly practical, I figured next that all I had to do was to set up hundreds of slit cameras side by side, and each camera would capture a slightly different image of reality that way. This idea came to me when I looked at Eadweard Muybridge’s famous 1879 Palo Alto set-up where he, for the first time, captured the movement of a racehorse in a chronophotographic sequence. But since it was impossible for me to built even a second slit-camera, I came up with a new idea: why not let a scene repeat over and over again and move the camera? So I started to capture movement from a video monitor and called the result of this technique RSA sequences, which was short for the untranslatable German term “Raumschnittabfolgen” (Spatial Section Sequences).

At the time, in the summer of 1992, I applied for art funding at the Federal Chancellery of Austria (BKA) and described the process in my proposal as follows:

(1) Information block

Without going into the narrative structures of the cinema film (flashback, parallel plot, etc.), filmic movement, and by this we also mean the film's own movement, knows only one direction – from the first to the last frame of a strip. This layering of information along a temporal vector can be shown most vividly in the form of the flip-book. In this children's toy (as in film-historical methods: Mutoscope etc.), the illusion of movement is created by a rapid succession of individual layers of time. However, I do not want to go into the physiological basis of this optical fusion here, but rather to emphasise the physicality of the flip book. The flip book, like the film reel, contains the totality of all spatial aspects of movement and a manifest temporal order - I therefore symbolically call it an information block.

Since the 1940s, a special photographic technique has been used to determine the exact time of the run-in (especially in horse races and other races): photo finish photography. In this technique, a continuously moving film strip is moved past a slit in a special camera. The spatial plane covered by this slit is identical to the finishing line. The object moving past this line literally records itself. An essential parameter in these shots is the relationship between the angular velocity of the object and the film speed: if the film runs too fast, the object is compressed, if it runs too slow it is stretched , if the object remains in the plane of the shot, it runs in a horizontal stripe pattern. When evaluating the target shot photos, we then simply (in reality, of course, it is a little more complicated) look for the point left on the film by the first, and thus fastest, object (e.g. the snout of a horse); the effective moment of crossing the target line can be determined with the (milli)metre measure – the time can be measured in length!

Fig.4:The “Information Block” as an illustration in my 1992 funding application. The 3D-Model was rendered by my friend Norbert Bayer, who was the only person I knew at that time that had access to a SGI Indigo workstation.

(2) Spatial Section Sequence

This artificial word is made up of three parts, which I would like to explain further. In contrast to the time cuts of flip-books (or the individual frames in cinema films), I understand a spatial cut as a cut through the information block along a spatial plane. Space, as stated earlier, does not refer to our three-dimensional surrounding world, but to the physicality of the information block. The word sequence refers to the way in which these sectional images are strung together, namely starting from the left and going to the right side of the recording space. However, since I do not have a large number of simultaneously launched target cameras, but only one, I use a trick: I repeat the movement and position the camera at a different location for each run. Afterwards, I link the resulting images according to their original order.

Fig.5: Early motion tests filmed from a screen, May and December 1992

Fortunately the art funding was granted to full extend and so I was able to intensify my work on a new RSA recording system over the next two years. The first step included the building of a reliable slit camera that could hold more film. I had the opportunity to take a closer look at the only photo finish camera installed at the Vienna horse racing track and, with the help of friends, to collect all kinds of used camera parts. Together with Herwig Petzny, the head of the metal workshop in our department at the university, we built a fully functional camera within a year, which, even from today's perspective, combined some interesting technical solutions. The bigger problem for me, however, was the construction of the electronic motor control. It's hard to imagine today what a challenge that was in the early 1990s. Together with my friend Toni Poeltl, we etched and assembled our own circuit boards and built our own encoders, power supplies and driver circuits to control stepper motors. But the effort was worth it: after about 18 months building time I had a very precise camera with which the film transport speed could be set in 100 steps and which had an optical viewfinder, interchangeable and very accurate slit masks and a magazine that could take 60m of 35mm film.

Fig.6: Building the electronics for the RSA camera together with Anton Poeltl, 1993

But the camera was only one component of the full set-up. It was mounted on a linear guide driven by a stepper motor in order to film one vertical line from a Trinitron monitor. For the first film, which to this day is the only one made this way, I made a short video sequence that should be played over and over again in a loop. This turned out to be impossible with conventional video players, like the U-Matic machines we had at the university. In these days looping a video was only possible with the then new video disc recorder, but luckily there was a company in Vienna that had such a machine – a device which cost as much as a small car. I had them record my clip and then played it back onto tape several hundred times. This ten second clip was made in such a way that it showed a tiny white square at its end. A light sensor was stuck to the screen at this exact spot and thus gave the signal to advance the camera one step along it’s linear track.

Fig.6: On the left side the complete RSA set-up: in the foreground the slit-camera mounted on a linear guide – in the background the PC that controlled the stepper motor, a Philips Video Disc player and a Sony Trinitron monitor.

In this way the whole width of the monitor was scanned line-by-line and exposed on the moving 35mm filmstrip. Although this method of recording worked as planned, it still was very time-consuming and costly. Also, at that time, it was not possible for me to recopy the images that were taken with the RSA camera accurately. Trying to photograph them from the light table with my 16mm camera meant such a loss of quality that in the end the effort didn't really pay off. With the completion of this one and only RSA test film, my attempts with analogue solutions ended and a new era began.

Fig.7: Input and transformed output frames from the RSA test film, around 1994.

The early computer days

Around 1996, the first video capture solutions for Windows PCs came onto the market and at the same time hard disks became bigger. So it suddenly was conceivable to make transformations on the computer. Since the graphic cards at that time were by far not powerful enough to display video, the signal was processed via a separate card and output directly to a PAL monitor. The same component also captured the incoming PAL video and directly saved it to SCSI hard discs. The COMO system I used produced files that could only be decoded together with the original hardware and this also is the reason why I can no longer access any of the projects from that time. Still I was finally able to export all the single frames that my video recording was composed of – also something that was unthinkable before. My friend Toni Poeltl, who had already helped me build the camera, wrote a DOS programme that could rearrange the columns of these PAL image sequences in such a way that they were transformed according to my ideas. Toni called his programme "tx-transform" and the process is known by that name to this day.

Fig.7: Scheme of the column rearrangement between input and output images as it is used in the »tx-transform« process.

Not only the name remained unchanged, but also the basic re-ordering logic that we came up with is still the very same today. As you can see from the scheme above the process is surprisingly simple: the first transformed image is composed from all first vertical lines of all input images, the second image from all the second lines, and so on. What becomes clear here is that the the information actually stays the same and is only re-arranged – nothing is added and nothing gets lost. It also means that when you transform an already transformed sequence again, the original sequence is restored in a lossless way. At the time we worked with PAL images which have a resolution of 720 x 576 pixel. Since I wanted the two axes of the “Information Block” equally long, I used sequences with 720 input images, which at 25p is about 28 seconds. This way the input and the transformed output files would have both the same aspect ratio and the same duration.

Fig.9: One still from the very first tx-transform sequence I have rendered on a computer. Processing the clip took more than 30 hours. The model is Isabella Purtauf.

In the beginning, rendering these sequences was quite time consuming since exporting the single frames took some hours and the transformation process almost 2 days. Also the 500MB SCSI discs I needed were outrageously expensive and I could only afford a single one. So I needed to burn stacks of CDs to store the results. Apart form this technical difficulties it took some time until I understood how I could influence the results so that they did not seem totally random. I understood that clean and interesting images could only be produced if the shooting was done under very controlled conditions and so I started to build a whole set of electric turntables and tripod heads. By end of 1996 I had produced a number of quite impressive results and considered the project to be completed. It was not until I met Virgil Widrich in May 1997 that a new era began. But that is another story...