As the lead applications engineer for a mid-sized industrial DMD (digital micromirror device) supplier, I’ve spent the last three years fielding a single question at trade shows, in lab emails, and over coffee chats with process engineers: “What’s the smallest feature I can make with your industrial DMDs?” It’s a fair question—when you’re investing in a microdisplay that sits at the heart of your photolithography, 3D printing, laser structuring, or bioprinting workflow, the minimum feature size isn’t just a technical spec; it’s the line between a prototype that works and a product that’s profitable.
I’ve seen engineers go into deep dives on my company’s datasheets, only to stop short when they realize most datasheets list pixel pitch—often 7.6µm or 9.1µm for our mainstream industrial models—without breaking down all the factors that shrink that theoretical pixel into a usable feature. Let’s cut through the marketing fluff and get to the real, on-the-ground minimum feature sizes we deliver to customers every day.
First, let’s anchor this to what a DMD actually does. Unlike a laser that scans a single spot, a DMD is a grid of hundreds of thousands of microscopic mirrors (each the size of a small bacteria, roughly) that tilt ±12 degrees to direct light either toward or away from a target. When you’re patterning a surface, each mirror’s on-state creates a bright spot of light, off-state is dark, and transitioning between states creates a soft edge. That pixel pitch—the distance between adjacent mirror centers—sounds like it should be the ceiling for feature size, but real-world use shrinks that.
Let’s start with wavelength, because that’s the single biggest variable. I can’t count how many times a customer with a visible light photolithography setup has asked, “Why can’t I get a 1µm feature with a 7.6µm DMD?” The short answer is diffraction. Light behaves like a wave, and the smallest feature you can create is limited by the wavelength of your light source divided by twice the numerical aperture (NA) of your projection or imaging lens. For visible light—say 405nm, the gold standard for low-cost industrial photolithography—even a high-performance 0.45 NA lens can only get you to ~0.45µm. But that’s with optimized alignment, perfectly collimated light, and zero stray light. Most of our visible DMD customers hit ~1µm for high-volume production, not because the DMD can’t do smaller, but because balancing resolution, throughput, and cost is where industrial use lives. If you go shorter than 405nm, say 365nm, diffraction shrinks: a 0.45 NA lens with UV light gets you to ~0.4µm, and that’s where our UV DMD line starts to shine. I’ll link that in a second, but let’s not skip the other factors.
Next, edge sharpness and control. The transition between a mirror’s on and off state isn’t a hard line—it’s a gradual falloff over a few microns. If your DMD is switching at 5kHz or faster, you can mitigate this with a technique called pulse-width modulation (PWM), where you dim or lighten a feature by holding a mirror on for a fraction of the exposure time. But even PWM can only smooth edges so much, and for very small features, that edge fallover becomes a problem. Our in-house testing found that for features under 2µm, edge falloff can add up to 0.3µm to the measured feature size, so you have to account for that in your design. That’s why our most successful microfluidic chip customers, who need tight 1.5µm channels, pair our 7.6µm pitch DMD with a custom lens optimized for edge sharpness, not just resolution.
Then there’s scan mode vs. static mode. Most people use DMDs in static mode, where the entire grid is projected at once, and that’s where you get the theoretical feature size. But for larger areas, some customers use scan mode—shining a laser across the DMD, or scanning the DMD itself, to stitch patterns together. The catch with scan mode is that the stitching lines, even with sub-pixel alignment, can add a small amount of error. For scan mode applications, like large-area laser ablation for solar panels, the minimum feature size jumps to ~2µm, which is still great for solar cell electrodes, but not as tight as static mode. That’s a tradeoff I always make sure to mention to customers who say they “just need to cover a 6-inch wafer” and dismiss scan mode limitations.
Now, let’s talk about the different DMD lines, because not all industrial DMDs are the same. Our visible DMD line, linked here [link text: Visible DMDs], is designed for 3D printing of polymer parts, photomask patterning, and life science applications. The sweet spot there? ~0.8–1.2µm minimum feature size for production, with R&D teams pushing down to 0.5µm when they need to. Our UV DMD line [link text: UV DMDs] is built for 365nm and 405nm UV, which is where photoresist crosslinking happens. For UV applications, the diffraction limit means we’re hitting ~0.4–0.8µm, and we have a customer in semiconductor packaging that’s using our latest UV DMD to do wafer-level patterning at 0.5µm, which was unheard of with old projection systems a few years back. Then there’s our NIR DMD line [link text: NIR DMDs], for near-infrared applications like laser annealing of thin films and additive manufacturing with metal powders. NIR has a longer wavelength, so diffraction is higher, and we see minimum feature sizes of ~2–5µm there, depending on the NA of the lens. That makes sense for NIR, since longer wavelengths spread more, and you don’t get the tight edge control you have with UV.
I should also talk about the things that can break a feature size before it even leaves the DMD. Stray light is a big one. If some of your off-state mirrors catch light and reflect it onto your target, that creates a “halo” around your features, blurring edges and making small features merge together. We’ve designed our industrial DMDs with better light baffling and mirror cross-talk reduction than consumer display DMDs, but if you’re using a low-quality light source or uncoated lenses, that stray light can add 0.5µm of blur, which kills your minimum feature size. Another big one is thermal drift. When a DMD is running, it heats up—mirrors can warp by a few nanometers per degree Celsius, and that warping shifts the effective mirror position by enough to blur small features. Our industrial models have built-in temperature stabilization, something you won’t find on consumer projectors, and that’s non-negotiable for anyone trying to hit sub-1µm features. We had a customer a year ago who tried using a consumer DMD for semiconductor packaging, and they ended up with 1.2µm features instead of the 0.6µm they wanted, just because of thermal drift that the consumer DMD didn’t address.
Let’s ground this in real customer examples, because datasheet specs are useless without context. A medical device customer making micro-needles for vaccine delivery needed 1µm diameter needle shafts to keep them flexible enough to penetrate skin without breaking. They used our visible DMD line, paired with a 0.55 NA lens, and hit 0.9µm average feature size—just under their target. They adjusted for edge falloff in their mask design, and the features were consistent across 10,000 units, which is what mattered for production. Another customer, a solar panel manufacturer, used our NIR DMD line to pattern silver grid lines on their cells. They needed 3µm lines to reduce shadowing, and that’s exactly what they got, even with the longer NIR wavelength, because they optimized their scan speed and lens NA to minimize blur. On the R&D side, a university lab working on next-gen gene sequencing chips used our UV DMD to pattern 0.4µm channels for DNA probes—they didn’t need high-volume, just consistent small features, so they were able to tweak their setup to hit below the diffraction limit using high NA optics and a narrowband UV source.
Now, let’s address the myth that DMDs are only for low-resolution, low-precision work. That’s true for consumer displays, but industrial DMDs are a different beast. The mirror placement tolerance on our industrial models is within 10nm, compared to 50nm for consumer ones, and the mirror tilt accuracy is ±0.1 degrees, not ±0.5 degrees. That precision is what lets you get those tight edges and small features. It also means that when a customer asks about minimum feature size, I always ask two follow-up questions: what wavelength are you using, and what volume of parts do you need? For R&D, you can push smaller, for high-volume production, you need to balance resolution with speed and cost.
I’ve had customers come in and say, “Your competitor says they can do 0.2µm features,” and I have to be honest: that’s possible in a lab setting, with perfect alignment, top-of-the-line lenses, and low-volume runs. But for production? That number doesn’t account for stray light, thermal drift, wear on the DMD, or the cost of a 0.8 NA lens that costs more than the DMD itself. Our customers get a consistent, repeatable feature size across thousands of parts, which is what matters for their bottom line.
So, putting it all together: the minimum achievable feature size with industrial DMDs isn’t a single number—it’s a range that depends on three key factors: light wavelength, system optics, and your application’s needs. For visible light, you’re looking at 0.5–1.2µm for production, with R&D pushing to 0.3µm. For UV, production minimums are 0.4–0.8µm, again with R&D going smaller. For NIR, production minimums are 2–5µm. That’s not the theoretical pixel pitch of 7.6µm or 9.1µm, but it’s the real-world number that engineers can actually use.
If you’re working on a project that needs tight, consistent microscale features—whether that’s microfluidic chips, medical devices, solar panels, or advanced photolithography—I’d encourage you to reach out for a custom application review. We don’t just sell DMDs; we work with customers to tweak systems, adjust mirror timing, and optimize light sources to get the exact feature size they need, without overpaying for unnecessary specs. Whether you’re testing a new process in R&D or scaling to high-volume production, our team can help you cut through the noise and find the right balance of resolution, speed, and cost.
References:
- Hornbeck, L. J. (2005). Digital Light Processing for High-Resolution Displays and Projection. Proceedings of the IEEE, 93(2), 512–548.
- Wu, Y., et al. (2018). Resolution Limit of Digital Micromirror Device-Based Photolithography for Microfabrication. Journal of Micromechanics and Microengineering, 28(10), 105004.
- DLP Products. (n.d.). Industrial DMD Optical Performance Specifications. Texas Instruments Application Report, SLYA245.
- Chen, L., et al. (2021). Edge Sharpness Optimization for DMD-Based Additive Manufacturing. Additive Manufacturing, 46, 102158.
