TI DLP Beagle Bone Cape


Interesting possibilities.
How hard would it be to convert to Raspberry Pi?

2 Likes

Thanks.
Fixed - I think.

Works. 10987654321010

It says they’re sending the video to it over a 24 bit parallel bus with gpios. Does raspberry pi have that many gpios? If so then it seems it wouldn’t be difficult to convert.

Oh goodness yes that would be very different. I skimmed the article and it mentioned gpios but I overlooked any mention of a hardware HDMI peripheral. That unfortunately makes all the difference usually. I thought this was low resolution and moderately low bandwidth.

Ok so the resolution is actually low and the bandwidth is moderate. But relative terms are relatively useless so here’s the math. You end up needing the gpios to toggle at about 10MHz for 640x480, 8 bit color, 3 colors, 30 frames per second, 24 bit data bus (640*480*3*8/24*30 = 9MHz). That might barely be just within the reach of a raspberry pi. Maybe you can pull it off or maybe not. The fact that linux doesn’t support realtime timers may preclude it from being an easy task. If you don’t need full 8 bit color or 30 fps, that could make all the difference. Another thing that will make a difference is what the raspberry pi has in terms of a parallel addressable bus. It’s definitely not small bandwidth though unless you can compromise on the image quality.

Interesting site.
Will add to Committee reference pg.

Hey, not my place to ask that, right? OP asks can a raspberry pi do it and I wanted to know the answer too. :laughing:

SPI + FPGA

I highly recommend Howard Johnson’s High Speed Digital Design: A Handbook of Black Magic. If Electronics has a library, this is a recommended addition to the collection.

I’ve been using the 1K passive probe from that book since the 1990s. As the Zach’s linked article shows, this simple and inexpensive probe gives a cleaner look at clocks and other high speed signals than standard scope probe. The only change I’d recommend is the use of a 950 ohm series resistor instead of a 1K. This yields a 20:1 voltage divider, making it easier to read actual signal levels on the scope.

[Historical footnote] My, how far DMD technology has progressed. I had the opportunity and pleasure to work on the DMD project beginning in 1984 at TI’s (now defunct) CRL (Central Research Labs).

The DMD acronym originally meant Deformable Mirror Device (instead of ‘Digital…’), because the original mirrors were formed by optically reflective and electrically conductive polymer ‘blanked’ that was stretched over the underlying silicon pixel array.

The silicon address circuitry was analog and CCD-based. In essence, it was a ‘backwards’ CCD camera architecture. Serial analog pixel data for a row of the image was shifted into a CCD input shift register. When this register was full, the row of new data was shifted downward into the pixel array, which consisted of columns of CCD registers. Using an architecture dubbed a ‘ramplifier’, the CCD signal at each pixel was amplified to a the level needed to drive the pixel mirror.

Physically, each pixel was concave pit in the silicon with an electrode at the bottom that connected to the output of the ramplifier. The conductive polymer blanket was connected to an appropriate voltage so that there was an electrostatic force of attraction between the pixel electrode and the suspended polymer pixel blanket. This caused the section of the blanket over the concave pit (the pixel) to be deformed and form a dimple. The depth of the dimple was dependent upon the pixel voltage.Thus, it was a Deformable Mirror Device (DMD).

In the next phase of development, the polymer blanket pixel was replaced by a rigid aluminum pixel (with hinges on diagonally opposite corners), with the underlying CCD addressing retained. Even though the new rigid pixel rotated instead of being deformed, the device was still referred to as a ‘DMD’. After the move to the rigid aluminum pixel the DMD was primarily targeted for use in optical processing, with the major application being optical match filter correlators. This emphasis was reflected in the name of the DMD group in TI CRL, the Optical Processing branch.

Most of the customers who funded (Army Missile Command, NASA) us were focused on optical correlators. One of our military demos was referred to as ‘tank on a stick’, where a tank was moved in front of a camera that was feeding an ‘input’ DMD, a second DMD had the Fourier transform of the tank, and a 2nd output camera was placed in the correlation plane. When working properly, moving tank on a stick in front of the input camera caused a bright spot in the corresponding position of the output camera. NASA’s desired application was to assist in docking with the ISS.

One of the tricky aspects of using early DMDs was correctly adjusting the CCD and ramplifier voltages to achieve maximum mirror deflection angle to yield the best optical range. The tricky stuff was with the mirror physics. That is, as the mirror electrode voltage was incrementally increased the pixel deflection angle increased correspondingly - as desired. Equilibrium (the lowest energy state) was reached when the restoring torque of the pixel hinges was equal to the electrostatic force of attraction between the mirror and the pixel electrode. Eventually, the laws of physics and the physical design of the pixel interacted in an undesired manner. That is, at some pixel deflection angle an incremental voltage applied at the electrode caused an incremental increase in the mirror angle. But at this incremental state the lowest energy state is at a greater (than current) deflection angle, because the electrostatic force is proportional to the inverse square of the distance. When this state is reached the pixel continues deflecting downward until it hits something. Unfortunately, this ‘something’ is the attracting pixel electrode. In this case, pixel mirror is a dead short between two voltages and - POOF - the mirror is vaporised and the DMD is toast.

In 1987, the DMD’s inventor, Dr. Larry Hornbeck, recognized that the physics which caused mirrors to be vaporized could be an advantage. Instead of driving the pixels in an analog manner, they would be driven digitally such that the desired mirror operating states were deflected maximally in either the positive or negative direction (angle). The intervening angles would be be transitory, only occurring during pixel flight between the two stable position. Thus, the pixel was bi-stable/digital.

The pixel would be redesigned so that the landing spot for the mirror would be at the same potential as the mirror - no more ‘smoke and no mirrors’ :). The electrode and mirror voltages were set so that after an initial mirror deflection past the energy ‘knee’, the mirrors would land safely. Thus was born the true ‘Digital Micromirror Device’, although that name did not yet exist. Because feature sizes and layer thickness are tightly controlled in semiconductor processing, the corresponding mirror tilt angle was very consistent from pixel-to-pixel and device-to-device - an essential feature for the most efficient optical design.

In 1989 DARPA (Defense Advance Research Projects Agency) awarded TI and partner RCA David Sarnoff Research Center a contract to develop DMD-based high definition display. A final report of this project can be found here, with lots of good images of then-emerging DMD technology. DARPA’s funding was an effort to retain onshore capability in a critical technology, a response to most of the (LCD) display development moving offshore.

At the end 1990, the bulk of the DMD/Optical Processing team was moved from TI’s CRL to a new TI internal organization, Corporate Venture Products. The team size increased dramatically and included a marketing team. It was this marketing team that re-christened ‘DMD’ to be ‘Digital Micromirror Device’. In addition, they created the DLP (Digital Light Processing) brand and promoted in a manner similar to the Dolby brand.

One of the funniest observations regarding DMD technology comes from PC Magazine columnist and industry gadfly John Dvorak. During the development of the DMD HD display at TI, Dvorak had this to say about DMD technology in the August, 1993, edition of PC Magazine.

image

3 Likes

Thanks for a good read!