Of the almost endless things we know to be true but can’t see, one that impresses me greatly is the refreshing and cycling of electronic devices. It’s not difficult for devices to exceed our “flicker fusion” thresholds — around 8kHz for touch, between 16 and 30 Hz for vision, and around 20khz for audition (OK, that’s just above human auditory capabilities, not accurately any kind of fusion, but stick with me, damnit).

One of the premises of much ubicomp literature is that the interfaces between and operations of machines will be invisible, and in some sense then, incomprehensible now — taking a form we cannot imagine or manifesting in an extrasensory fashion by virtue of speed (gigahertz), medium(radio), complexity, and so on. What’s great is that like so much previous, we can devise instruments and interfaces to bring it into the realm of the sensory. I have a simple demonstration — you can see fluorescent tubes cycling in the image below. An ordinary camera brought it into the realm of human detection.

Now, I have an inkling that the proper design of ubiquitous systems will include affordances to bring their operations and interactions into the realm of human perception — most crucially in cases involving encounters with novel systems as well as transactions involving critical information. But that is the subject of a future post.

Now that I’ve outlined some reasons for wanting a grayscale (color-wheel-free) projector, here’s how to make one. It’s simple, and the procedure is generalizable onto pretty much any DLP/DMD device.

First, turn off and unplug your projector. In this case, the projector is a Dell 5100MP.

Remove these screws. Don’t miss the studs on the DVI connector, they hold the top on. Take it off.

This is the internal layout of the projector. We’re really only interested in the color wheel right now. It is held in by two screws. Unscrew them. Remove it by gently prying at the base, and grasping the little metal clip with a needlenose pliers. Wear gloves so that you do not contaminate the optics with finger oils.

Now that the color wheel is free, inspect it. Note the dichroic coating and the clear section. The clear section is used to increase the overall luminance of the image. When you adjust “brightness” or go into “brightness mode” on your projector, you are adjusting how much light the mirror array reflects through this clear section.

The color wheel is a very simple device, built much like a hard drive platter. It is a series of optical filters mounted on a brushless DC motor (BLDC). Under the clear section, there is a piece of black tape (blanking patch) which subtends the same angle as the clear slice. An infrared detector watches this tape to signal a complete cycle. In a future article, I will examine the exact relationship between this tape strip/detector and the actual refresh rate of the projector. In combination with a NI-DAQ, it might be a good way to use commodity DLP systems for time-critical research. It might also be possible to inject a timing signal in place of the optical detector to control the refresh rate manually.

Now that your color wheel is out of the optical path, you can simply put it out of the way. I secured the color wheel assembly to the lamp box with double-sided tape and covered it with a lexan blister pack. Fire up your projector to make sure you haven’t ruined everything.

Ahh, that’s what I wanted to see. A grayscale image unperturbed by color wheel tomfoolery.

DLP projectors are interesting things. They are based on MEMS technology — in essence, a million or more tiny controllable mirrors, one for each pixel. But the mirrors are not colored. To get the final output color, the image is actually refreshed three times per frame, and a colored filter is spun over each sub-frame. It’s above the flicker fusion threshold, so normally you can’t see it with your eyes. But a camera with a sufficiently high shutterspeed can:

Note the filtered light being reflected back into the lampbox.

There are many situations in which this kind of projection is undesirable, particularly temporal experiments where one would like to present a stimulus at some exact moment. The next two images illustrate the inter-frame refreshing of pixels. Remember, each pixel is a little mirror being turned on (toward the lens) and off (away from the lens) in a time sequence. To make a pixel darker, it is off longer. To make a pixel brighter, it is on longer.
It’s somewhat easier to see than to describe. Imagine a fixed number of events in a sequence. “1″ represents on. “0″ represents off.

Darker:
00000100

Brighter:
11110101

Mid-Gray:
10101010

This is PWM — pulse width modulation. Because of the way PWM works, it is not possible to guarantee with suitable precision when a pixel might be “on” or “off”. Note the seemingly random patterns in the next two images, which were captured at the edges of the red/green and red/blue refresh intervals. The large blocky artifacts on the desktop are due to JPG compression, but the small artifacts (particularly in the toolbox) are from PWM.

It is actually possible to see the effects of color wheel operation simply by making a fast saccade across a projected image. If the projector is a DLP(DMD) device, you will be able to perceive rainbows at bright edges.

Now that I’ve explained some of the problems of DLP devices, the next post will explain a simple solution that will solve some of them.

Make sure your interface degrades gracefully.

The Marantz PMD620 has a remarkable interface design feature — the “level” LED indicators at the top of the unit. With two LEDs and three states, they convey all the necessary information about recording levels, leaving the OLED screen free for other purposes and obviating the need for bouncing level meters or excessive attention to the display. Note that the same effect could not have been achieved with a single multi-color LED, as in bright, outdoor conditions, it would be difficult to discriminate the color.

Some pictures and explanation:

In the first state, the LEDs are unlit. This indicates that the signal level is less than ideal.

In the second state, the first LED indicator is lit. This indicates that the signal is at or above an acceptable level. (Thankfully, this level can be pre-set by the user).

In the third state, the red LED lights up to indicate that the signal is clipping.

I’m going to be taking this blog in new directions, starting today.

For Wolfgang’s birthday, I virtually blew up his house.

Bigger and Bigger.

edit: this post will return later.

Shaun Fort and I played a show at the 2003 Fargo Winter Carnivale, held in the Fargo Theatre. I recently had the pleasure of revisiting the show with Shaun (though we’ve maintained occasional contact, we’ve also more or less gone our separate ways). Of all the things I think stand out about that show (risky live electronic music performance, hand-built hardware, and a standing ovation to our public war protest) the one that I’m now the most fond of was our integration of state surveillance footage.

In the days preceding the show, we collected video from ND and MN interstate traffic camera systems. The end result was a pan-state time lapse view leading right up to the show, and the closest we could come to realtime monitoring using the state surveillance apparatus. This approach — infusing data local to the performance venue — has become central to my approach, from last summer’s “attack” on the Plains Art Museum to the scale-accurate photogrammetry-derived 3D model of downtown Fargo I produced for my recent show at Dempseys. Note: that’s a test render…

Andy Filer took some beautiful photos of the Carnivale, this one in particular inspired this post.