If you’re engineering hardware that has any one
Lots of Big, Costly, Preventable Mistakes
Most of y’all working in startups (and sometimes even larger corporations) are doing it wrong. I know because I am an optical engineer (OE) running an engineering consultancy (SpireStarter.com). Last year, I traveled almost 100% of the time in the US and abroad meeting hardware engineers who needed help with optical engineering. A whopping 85% of the work I did last year was done for free. Why? Because so many of those hardware companies 1) needed to learn what optical engineering was and 2) how it could benefit their specific situations.
Also, most of the engineers I talked to thought an optical engineer was the person who designed their glasses. Breaking News: that is not what we do.
3 Out of 3 Optical Pros Agree
I didn’t want you to just take my word on all this. That’s why I asked engineers at two of the largest optical simulation software providers to chime in. Special thanks for the contributions from Patrick Le Houillier (and Jake Jacobsen,
Video: Optical Simulation Software in Action
Specialties within Optical Engineering
There are a ton of different tracks in optical engineering. Most optical engineers (OEs) will only have experience in a couple. I have experience in an abnormally large number of specialties, but definitely not all. This is why I collaborate with different mixes of specialists depending on the project.
You’ll hear OEs say, “in the end, light is light.” Which means: light follows the same rules of physics no matter the applications. That’s true, but different specialties and even different applications within a specialty can be totally different ballgames in terms of engineering requirements and tools.
The most common two boxes to divide optical engineering specialties into are imaging and non-imaging.
Imaging optics include all optical devices where you’re creating an image. The image could be a picture, a video, or some other representation of real-life that falls on your eyeball. Examples are things like:
- Cameras (including those in machine vision systems)
- Telescopes and Binoculars
- VR Goggles
That scope your proctologist jams up your innards? Yeah, you can thank an optical engineer for that, too.
The term “non-imaging optics” lazily contains all the other types of optical things that aren’t imaging. It includes a LOT of specialties, such as:
- Illumination (aka Lighting)
- Fiber Optics
- Optical Sensors (e.g. IR sensors, light curtains, visible light sensors)
- Pretty much anything else using an LED…
Tools of Optical Engineering
As I mentioned, the tools an OE might use can vary widely depending on application. These include what engineers use to design, simulate, or “debug” in virtual environments, and test in real life.
Optical Simulation Software
Not too long ago, a lot of optical engineering design and analysis was done by hand. Now, we have super fast computers and mind-blowing simulation programs. They can show us just how a design will perform before a tool is ever cut. Work that might have been done with 50 mathematical calculations and pencil drawings on human-sized physical prints is now replaced by millions of calculations on a PC and realistic renderings.
What are Ray Traces?
Most of that simulation work (but not all) is done with ray traces. In a typical simulation, you’ll have a video-game version of all the components of your optical system. So you might have, say, an LED, a lens, some housing geometry, and some sort of detector.
In the simulation, you’ll model a virtual version of each of these. For your LED, maybe you create a source model of the same colors and light distribution and overall amount of light of the diode. Let’s say the lens you want to create is pretty simple and you create that piece right in the simulation software. Maybe the housing is some CAD a mechanical engineer created in SolidWorks. So, you get a .STP file (or whatever format you can use) from the ME and import it into the simulation. Next, you set up a detector in the area where you want light to go so you can see if you succeeded.
Then comes the most difficult part: setting material and texture attributes for all the physical parts. These models should be based on your past experience comparing real life to simulation. (Little changes here can COMPLETELY change your result!)
Last, you might tell that pretend LED you created to turn itself into 1 million light rays based on its model, and you hit “GO”.
When the simulation begins, each of those million rays becomes a calculation, which splits into more rays and calculations each time it hits a surface. Upon completion, you can see with surprising accuracy (if you created your model well) just where all the light rays will end up.
It’s Not Automagic, It’s Still An Art
This explanation might make it seem easy. Let me just say, it’s a lot harder than hitting a button. Each simulation software has its own way of propagating those rays, along with a hundred other differences. It still takes optical knowledge, training, and most importantly, real-world experience to do it right. Even then, it’s really tricky.
Optical Software Providers
Sometimes HW managers will ask an OE, “What is the ‘best’ optical software?” That’s like asking a programmer, “What is the best coding language?” The answer will totally depend on the application and, in any case, is loaded with personal bias.
I’m experienced in many of them. Some I’m in love with. Some make my eye twitch thing come back immediately upon viewing. Here I’ll try to keep my biases in check by listing some of the more common names and the applications they are associated with.
Note: these aren’t exclusive uses of each software. With some blood and sweat, many can be hacked to use for different types of optics–I wouldn’t recommend it. But sometimes it can be done–I’m listing how I’ve seen them commonly used. Also, they’re in no particular order.
- Zemax: OpticStudio, LensMechanix* (both are more commonly used for imaging applications and other optics designs not dependent on lighting aesthetics)
- Synopsys: LightTools* (illumination and other non-imaging), CODE V (commonly used for imaging and non-illumination), LucidShape (tailored for meeting the ridiculously extensive design requirements of automotive lighting — I was an automotive lighting engineer, btw)
- Breault Research Organization: (pronounced “Bro”,): ASAP (usually used via scripting environment), APEX* (I’ve seen Breault’s software used for just about every application, especially if pure simulation is what a company wants from SW more than specific design capabilities.)
- Lambda Research Corporation: TracePro (more commonly used for illumination and non-imaging), OSLO (more common for imaging optics or larger optical systems with imaging as a part)
- LTI Optics: Photopia* (most commonly used for luminaire design i.e. lamps for ambient lighting. Also used in other non-imaging applications.)
- ANSYS: SPEOS (most commonly used for illumination.)
- …and a bunch more
* options for either connecting to or add-in available for SOLIDWORKS
Optical Lab Equipment
Optical engineers don’t just work in virtual reality land; we play with physical stuff, too! For things like prototyping and verifying first production, there’s a lot of real-world work going on.
Equipment can range from the size of a hand-held (light) power meter to a multiple-axis robotic measuring machine the size of a room (goniophotometers). Also, optical lab benches where you can align components with high precision are common.
Other times, you need to reverse engineer an assembly or figure out missing specs from an off-the-shelf component you’re trying to use. Getting this data can require specialized machines or even a custom, duct-tapey hacked-together measurement method.
For each type of optical application being developed, there are usually a different combination of tools used. That, of course, adds to the division of knowledge between optics specialties.
A Mechanical Engineering Analogy
I often hear of people trying to mass produce a one-off working optical prototype without doing optical analysis. Oftentimes, they don’t even have all the specs of each component they used. To communicate my feelings on this, I’d like you to picture a mechanical widget design.
This widget has several moving parts. Each part was sanded and whittled down by hand over multiple iterations until they all fit together. It’s made of some random discarded material the widget designer found. Now the designer comes to you and tells you he wants 1,000,000 of these widgets made.
You Say: “Ok, to start, where’s the CAD model to hand off to a factory? What are the tolerances on each part? What materials did you use?”
Widget Designer Says: “I didn’t create a CAD model or do any tolerance analysis on this. I just want to make 1,000,000 of this thing. Here, just take it.”
You Say: “Dude, you really need a CAD model and/or drawings of each part and then figure out the acceptable limits of each dimension. Otherwise, the factories will be making stuff that doesn’t work.”
Widget Designer: “But, this one works. So, I don’t need to do any of that. It’s fine.”
You (me): “WTF?”
This doesn’t make sense in mechanical engineering land, and it makes less sense with optical systems. With optical systems, you can’t see the light rays themselves, so the exact way a system works is often impossible to fully understand without analysis. Plus, if you don’t know the thresholds, dimensions, power, wavelengths, geometry, etc. in your system, you have no idea what you built.
One good check-in question to ask is: “If my vendors disappear tomorrow, do I know enough about my optical system to have other vendors build it?” Often, when asking a hardware startup, the answer is, “ahh…nope.”
Mistakes You Can Avoid with Analysis
A huge benefit to bringing in an optical engineer early on is often a dramatic reduction in number of prototype iterations. For things like light pipes, this can especially be true. Those are the optics that bring light from a circuit board to some other part of your hardware. To keep light leakage to a minimum and to get a nice-looking, even illumination coming out, it’s almost impossible in one take without simulations.
The reason light pipes are attempted to be designed so often without optical analysis may have to do with something Patrick Le Houillier shared with me. A misconception he routinely needs to correct is that people think light in a light pipe flows like water through traditional plumbing, so you just connect A to B. Because, you know, a pipe is a pipe, I guess? (P.S. a light pipe works nothing like a pipe used in plumbing. I very nearly choked to death on my own laughter when Le Houillier told me this.)
The use of the term light piping has caused several engineers and designers not trained in optical engineering to take the term literally and consider light as water into a pipe…Light in a lightguide does not behave like water into a pipe.– Patrick Le Houillier sharing his teaching experience as an optics pro at Synopsys.
Tolerancing and Optimization
Tolerancing and optimization with optical systems when you don’t have an optical engineer is tough.
Similar to mechanical and electronic tolerances, tolerances in optical systems show you all the +/- constraints you need to keep your system within if you want it all to work. So to find the breaking point with every angle, x,y,z-position, current, system temperature, and so on, it makes way more sense to change all these variables in a computer versus in a lab. Then, if you’re able to figure out tolerances, you can build a design that a.) has more wiggle room, and b.) has specs set to the middle of that wiggle room.
If a company isn’t using optical simulations, they might skip these steps altogether. And that’s a recipe for disaster! As a result, companies often build optical systems on a knife’s edge of breaking and don’t know it until it’s too late.
I asked Bob Householder of Zemax if he’d also seen projects where there wasn’t an OE at the start and then optics analysis needed to be brought in later. Here’s what he said:
I have brought into a few projects for imaging or laser work where there was a design, or at least a layout performed by a mechanical engineer or physicist…If an OE was there at the start of the project, the timeline, product cost and performance would have certainly been better.– Bob Householder, Zemax
When Manufacturing Goes to &*#$: Root Cause Analysis
Even a robustly-designed optical system won’t save you from royal screw ups at your vendors. However, if you already built an optical simulation of your system, it’s a million times easier to find the most likely causes of failure. There are a couple ways you can use the simulation to help you.
Backwards Ray Tracing
Because physics and light is awesome, light rays travel the same path forward as they do backward. This means, if you have light going somewhere you don’t want (like blinding a pedestrian), you can figure out where that light comes from in your product. In that case, in your simulation you’d start at where said pedestrian would stand and follow rays back to your device. The spots where those rays land on your widget are possible culprit areas causing the glare.
One of the most spectacular things about optical simulations is you know exactly which variables you’re changing. On the other hand, if you do problem-solving by playing with a physical set-up, you often don’t know what all the variables even are. It’s also difficult to isolate those variables. By that I mean, it’s hard to change just one thing at a time. For example, you might increase the power but that in turn increases your temperature.
So, if your product is giving you a result you don’t like, you can purposely screw up one thing at a time in your simulation. When you mess with a variable, such as: a distance, power going to an LED, a material, a surface finish, etc., and it gives you a shoddy end result close to your real life problem, you know you have a likely suspect.
This is an important ability to have as an optical engineer! Most of the time when there is a system failure, optical design is blamed but has nothing to do with the SNAFU. One time, I was told to diagnose an “optics problem” only to discover the root cause was a failure to create a new part number for an updated version of the driver. So, the old driver was being used by mistake. In case it’s not clear: that has nothing to do with the optics.
Take It From Bob
Here’s Bob Householder’s (from Zemax) 2 cents on this:
Optical simulation software is very useful for root cause analysis as there are many built-in analyses tools and then a robust scripting language to customize analyses.-Your friend and optics pro, Bob Householder, from Zemax
Why Don’t More People Know What Optical Engineering Is?
You may be wondering, “if it’s so important, why do so few hardware nerds know what optical engineering is?”
That’s a great question! There are several reasons. Here are the ones I think contribute the most.
Optical Engineers = Unicorns
I often hear engineering leaders even within large corporations complaining they have extreme difficulty finding optical engineers. There aren’t a lot of us. Therefore, even if you’ve worked in a company that both needed an OE and knew what an OE was, you may have never worked alongside one.
OE’s are so rare, sometimes even they don’t know they are one. When there’s a dire need for an OE in a corporation, I’ve seen management force an EE to learn how to use optical engineering software. On more than one occasion, I’ve met an “electrical engineer” at an optics software user’s meeting. It would often come out that the EE exclusively did OE work. Then, I would have to break it to them: “dude, uhhh… you’re an optical engineer.”
No, I’m not…I…holy crap, you’re right. I…I’m an optical engineer.”Electrical Engineers Around the Country Developing Optical Systems
Take It From Patrick
Patrick Le Houillier from Synopsys totally backs me up with this unprompted observation:
…we know that a lot of industries will hire mechanical and/or electrical engineers to do this kind of work either because they cannot find any, think they can design products without them, or don’t even know this type of engineer exists.– Patrick Le Houillier from Synopsys, Optics Expert Who Has Seen Some Things
It’s Sort of a New Thing
Teaching optical science as a field of study only began in 1929 in the US. (That’s only 90 YEARS, man!) Kodak and Bausch & Lomb dumped a bunch of cash into the University of Rochester, and said, “we would like to purchase some optical engineers”. Those 2 companies were making advanced tech revolving around optics at that time, and they could afford to help create an entire Institute of Optics.
Most other companies at that time didn’t need to develop high-tech optics stuff. As time went on and more tech began to involve optics, the need developed, but cost was still prohibitive. Optical engineers were and are rare. And then new optics development tools were developed, but the cost of those were prohibitive for a long time. For these reasons, a lot of optics dev was outsourced to optics manufacturers for a long time.
Today, the need for optical engineering is ubiquitous. Although OE tools are still expensive, they are not as comparatively costly as they once were. The main reason left is: paradox. Lack of awareness of the need for optical engineers early in a company’s history begets lack of awareness later on.
How to Get You Some Optical Science (or an Optical Engineer)
- Hire a full-time Optical Engineer. If you a.) have the budget for one and b.) can actually find one (we’re a rare breed, remember!), this is your best bet. If you can lure an OE who also has experience in the optics specialties you need help with, then congratulations, you’ve caught yourself a unicorn! A lot of times, one or both of these conditions prove too difficult for even big-name companies to meet. In that case, you can try option 2 or 3.
- Hire a part-time or freelance Optical Engineer. Pro: you have an OE! Cons: still can be hard to find, and if you go through a headhunter, you may get a lower level of service. If you go through a placement agency, you should also know enough about optical engineering to know exactly what tasks to request of your OE. Otherwise, if you directly find an experienced freelance engineer, you may be able to snag one who can hold your hand and take the lead a bit more. Other big con: when the contract is up, you could lose access to valuable project knowledge.
- Find an optical engineering service provider. Pros: You have an OE! …Maybe more than 1 if a technical roadblock pops up. You may also get access to knowledge in multiple optics specialties. When the contracts are over, there’s a higher likelihood of access project knowledge if you need it in the future. Other pros: things like training, software and hardware will usually all be covered in the rates you pay, so there are fewer peripherals to worry about. Plus, they typically provide a whole lot of hand-holding, sometimes even education, if needed. (And considering how many HW nerds think OE is for making eyeglasses, this benny is key.) Cons: still not as awesome as having your own FT engineer. Also, rates and payment structures among optical engineering firms vary widely and some are difficult for smaller start ups to work with.
Still have lingering questions about optical engineering? If they’re super general, ask in the comments below, and I’ll respond. If you need some technical optics jargon explained in layman’s terms, send me a message on my company site HERE. I’ll get it added to the layman’s glossary if there are enough requests. If you have specific engineering or education needs involving optics, also shoot me a message through my company’s site. Either we can give you a free preliminary consultation, or I’m happy to send you to the best alternative resource for your needs.
Just please, please, start using optical science in your product development involving optics. It’s there to make your life easier! It’s science! It’s engineering! You love these things. Love an OE.