Mike Harrison at the Superconference: Flying LCD Pixels


Mike Harrison, perhaps better known to us as the titular Mike of YouTube channel mikeselectricstuff, is a hardware hacking genius. He’s the man behind this year’s Superconference badge, and his hacks and teardowns have graced our pages many times. The best thing about Mike is that his day job is designing implausibly cool one-off hardware for large-scale art installations. His customers are largely artists, which means that they just don’t care about the tech as long as it works. So when he gets together with a bunch of like-minded hacker types, he’s got a lot of pent-up technical details that he just has to get out. Our gain.

He’s been doing a number of LCD installations lately. And he’s not using the standard LCD calculator displays that we all know and love, although the tech is exactly the same, but is instead using roughly 4″ square single pixels. His Superconference talk dives deep into the behind-the-scenes cleverness that made possible a work of art that required hundreds of these, suspended by thin wires in mid-air, working together to simulate a flock of birds. You really want to watch this talk.

Did you watch the talk? My favorite mechanical trick that Mike shared was the way he held the LCDs between two PCBs — milling a winding path through them with a 1.6 mm milling to tightly clamp down on the 1 mm thick LCD glass. Getting a black PCB is easy enough, but how do you get black FR4? Well, you could custom order it at great expense, or you could use (wait for it) a black Sharpie marker.

If you’ve never played around with LCDs before, scroll back up and watch the talk, because he also presented tons of details about the electrical requirements of doing so. LCD panels need AC drive, which for ICs that run on a single-sided supply means reversing the drive direction somehow. Mike was able to use a simple microcontroller with a PWM and inverted PWM signal to generate the “AC” and allow for beautiful greyscale on the LCD pixels. Bet your calculator doesn’t do that. (But it could!)

A lot of the details of the design were driven by the fact that these pixels needed to “float” in mid-air. To achieve this effect, they were suspended on two stainless steel wires, and the data and power were sent down the same lines. With up to five meter long wires, Mike was afraid of the data being sent down them radiating out from these unintentional antennas, as well as interference coming in. A 20 kHz Manchester encoding delivered the data. This is a win because it guarantees the maximum possible length of time between positive pulses, which is important because these pulses keep the devices powered. The resulting circuit is so well thought-out that every part is crucial to its success, and is a beauty to behold.

Installation of a large-scale piece like this could be a logistical nightmare. This is not Mike’s first rodeo. He designed in features like power and data OK LEDs that are visible from the vantage point of an installer, making it possible to individually troubleshoot each row of LCDs at a glance.

One of the deepest tips is how Mike maps the device ID of each panel into the controller that needs to know about their physical location. The hard way to do this is program device number one, hang it, program device number two, hang it, and repeat. It’s a bookkeeping nightmare. Instead, the installer presses a “button” on a single device, the controller sends the address down the bus, and the device accepts that address as its own. This means that the addresses can be assigned after everything is set up, which also makes replacement a piece of cake. But the real beauty of this system is the “button” — a tactile switch, magnetic sensor, light sensor — you can use whatever input you’ve got on the node. Mike built an installation that already had accelerometers built in, and tapped each node to lock in the IDs.

Here, none of these were a good match for the minimal hardware on each panel and the high-in-the-air nature of the setup. Mike wanted a system that could be deployed on the end of a pole, but that wouldn’t require electrical contact, which meant something like capacitive sensing. Rather than rely on the tiny capacitive changes of a finger, he mounted a CCFL inverter on a stick, and the LCD panels detected the resulting field. The final touch was that the controller could sense the current each node used, so the LCD pixel could signal that it had received the new ID and the whole procedure could proceed at the speed that someone could wave the wand over the panels. Very slick.

It’s always a pleasure to hear Mike speak on what he’s truly passionate about — clever solutions to hardware problems. I learn something new, and useful, every time.



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *