Posted on

Sneak peek of how streaming light works

OK, so here comes another update and a lo-fi sneak peak on the light streaming capability. Light streaming is basically sampling light from a running video (or the phone camera), amplifying it and blasting it out.



We have been working like dogs for what feels like ages now. If it was easy, they wouldn’t call it hardware. But we are are getting very close to manufacturability. The major issues are sorted and I hope we can announce some definitive dates soon.




Posted on

Up for air

I thought I’d give a little update on the progress we are making. I’ve been heads down in development to the exclusion of almost everything else since the decision was made to change the form factor of the light. We’re now literally seeing the light at the end of the tunnel.

This is going to be a bit geeky, sorry bout that…

Color Temperature

We have a working color temperature range from 1850 to 15000 now.  The low end, 1850, is about the color temperature of candle light and that’s pretty nice. You can adjust the color temperature over the whole range without affecting the brightness of the light. It is technically possible to extend below 1850 but expensive in terms of power.  Expensive because the power is constant over the entire color temperature range and the red emitters max out at low CT. We have to lower the overall light output as we extend the minimum color temperature downwards. 1850 looks like a reasonable compromise.

An alternative would be to decrease the light output as the color temperature decreases. But that would mean that if you change the color temperature, you will have to change your exposure as well. It makes it more messy and increases the chance of error. I don’t like that.

Another nice thing is that you can add filters with no loss of power. You might prefer to stick the light on a Daylight setting and add filters instead of adjusting color temperature directly. If you put a 1/1 CTO* gel on your traditional light you lose 1.1 f-stop. With the light we are developing you lose nothing. Put on any correction filter you like, the light output and your exposure is the same.

How does that work? It works because the microprocessor in the light keeps the output constant regardless of color temperature. If you have the light set to, say, Daylight, then adding a 1/1 CTO filter is exactly the same as changing the color temperature from Daylight (5500 Kelvin) down to ~Tungsten (2900 Kelvin).  Just to be clear: This light uses electronic filters not physical gels. A physical filter works by selectively blocking parts of the light. A filter that cost you 1.1 f-stop, would block a bit more that half the light. Probably 55% or so of the light is lost (1 stop = half the light). What we do instead is simply adjusting the entire spectrum to produce a light of the desired color temperature. Cool, yes?

I’m back next week with more.




* CTO = Color Temperature Orange. A color correction filter used to “warm up” or decrease the color temperature of a light source.


Posted on

Tools for the lean hardware start-up

Early design sketch

If you are into entrepreneurship and start-ups, you probably know that the lean start-up process is all the rage these days. The lean start-up is basically a way to reconcile the different approaches of agile development and user experience design by realigning them to aid business model development. User experience design is a great methodology for making good products, but as with any design process, the problem to be solved has to be known. You can’t solve a problem that you don’t understand. Agile development on the other hand is a method for developing stuff without needing to know what you develop or why.

Early design sketch

Both these methods have their strengths, but they have been at odds with each other. Until now. The new idea is that you do customer development in addition to product development in order to find Product/Market fit. Customer development is basically user experience design on steroids. All this is put into a framework that makes a lot of sense and is very actionable. If you are not familiar with this, it is worth checking it out.

The other side of the lean start-ups is to not spend much money until you have a product that people want and a way to make money from it. A lot of things has changed over the last few years that makes this approach possible. If you develop consumer internet software these days you don’t need to spend money on servers, you buy cheap cloud computing power. You don’t need expensive software licenses, there are low cost or free Open Source alternatives. Etc. The same trend is happening for hardware development but it lags software. Being a lean start-up, I thought I’d tell you a little about the tools and services we are using in developing products.

Development software and hardware

You can do most any kind of software development with free tools if you are willing and able to string together and optimize everything by yourself.  We don’t have the time for that, and we’re using Red Suite from Code Red. It uses a GNU tool chain optimized for the ARM processor we’re using. Code Red is not expensive and is not bad at all! The new debug probe is pretty nice.

The latest trend among chip manufacturers is to sell really inexpensive development boards. Thanks primarily to mobile phones, components are getting smaller and smaller so bread-boarding is often out of the questions. You don’t just pull out your soldering iron to attach six leads to a component that measures 2 by 2 mm.  The best way to work a new part into your design is to buy a development board. I think most manufacturers now charge less than USD 50 for processor development boards. If you don’t need a development board, they will give you samples of most any part for free. Unless they are National Semiconductor who won’t even let you view their promotional material if you don’t register first with a corporate-looking email address. My advice: forget manufacturers who has their head up their ass.

The price is anyway insignificant compared to the investment you have to make to familiarize yourself with a new processor architecture.  We’re using ARM processors from NXP (formerly known as Phillips). The parts are really sweet, but the docs and software drivers are sub-standard and that steals a lot of development time.I hope we can switch to something better like the Texas Instruments Stellaris processors. I we ever find the time! I can say that one of the early mistakes we did was to not take the time to properly investigate the tools (docs and driver code) from the various ARM processor manufacturers.

Electronics and test equipment

For creating schematics and printed circuit boards, you need PCB layout software. Cadsoft Eagle takes the price for the worst reverse-polish-notation user interface ever. A few examples: “Cut” does not cut anything, it copies. “Group” does not group anything, it selects. The entire user interface is backwards and based on verb-noun task order instead of noun-verb. (Noun-verb means that you first select something and then do something with it. This is the way all other software in the galaxy works. E.g. you click something and move it around. Verb-noun means you do it the other way around. In Eagle you first invoke the move command, then you group (select, remember) something, then you invoke the ‘move group’ command, then you move it.) The trouble is that Cadsoft Eagle is sturdy, relatively inexpensive and there are no obvious competitors in this mid-range. But I swear it’s going in the trash the day I can afford something better. If you want one of the big PCB layout packages you are looking at USD 5000 to 10.000. There is some free PCB layout software out there but I haven’t found anything good. With PCB software you invest a lot in your component library. You may be looking at weeks of work redrawing components if you switch. Eagle is probably popular because it has a free version for non-commercial work.

You’ve probably noticed that the price of consumer electronics has dropped incredibly. Big screen TVs that cost USD 5000 ten years ago now cost USD 500. The same is true for some professional equipment, but for a slightly different reason. Precision test equipment like oscilloscopes used to require years of work and divine inspiration by analog design geniuses. But silicon manufacturers have steadily improved the range and precision of integrated circuits to a point where any engineer can basically build a decent scope with a handful of IC’s. Scope manufacturers are going to disagree violently. But the fact is that a USD 2000 scope does the same job as a USD 20.000 scope from a few years ago. That also means that the main differentiator is no longer hardware, it is software. It’s no longer a matter of can the scope do something, but how well can it do it? This general trend towards better user experience is certainly key to lean hardware start-ups.

On the other hand, scientific test equipment like spectrometers, calibrated light sources and all that good stuff is just incredibly expensive. But there is not really a way around it. We are not just building products, we are building amazing products. The quality of light is paramount for a photographic light source and we need to be able to measure it’s composition in detail. So just find a way to pay for this sort of gear and try to look happy.

Assembly and manufacturing

Atoms are the new bits, but the atoms still need manufacturing, assembly and shipping. There is very little volume production done in Europe any longer. It is probably the same for the US. All local manufacturing companies now have a Chinese partner. Prototypes and low volume is still done locally, but any volume production is done in a ‘low cost country’. Its simple math. The production cost is just not comparable at all. Local manufacturers has turned into project managers for factories in lower cost areas. Interestingly the recent recession has expedited this trend and turned into an advantage for hardware start-ups. The recession forced factories to make their production lines more flexible in order to handle shorter production runs. This makes it possible for hardware start-ups to get low volume production done at reasonable prices. It is no longer the case that you can beat your competitors by simply being the biggest and leverage the advantage of huge scale . Well, let me rephrase that a little: If you are in a commodity business, where the only difference between your competitor and yourself is price, the biggest will win. The classic examples are electricity or gas. It’s not like there is a markup on premium electricity. We’re not in a commodity business and we’re probably less price sensitive. On the other hand, the market is much more global now. Great stuff spread faster, bad stuff dies faster.

3D tools and services for prototyping

Shapeways is a 3D printing service that can 3D print a great variety of materials and colors and we use them for enclosure prototypes. Compared to the traditional 3D printing and mode-building services you find in most major areas, web-based services like Shapeways and i.materialize cost half or less, and can deliver a much wider range of materials, but turn-around times are weeks instead of days. They are mainly set up to handle personal fabrication. Our experience is that it takes 3 weeks before you get your stuff back. That is a little too long and makes them less useful if you are in a hurry to get a product together.

Ponoko (RazorLAB in the UK) is a great laser cutting service. We’ve been using them for creating light guides and acrylic enclosure parts. It is great for prototyping even if acrylic does not have the optical qualities we want in the final product. Ponoko just added 3D printing to their offering yesterday. Personal fab is on a roll and you can take advantage of that.

We’re using 3D software to design the enclosures. The thing to understand about 3D modeling software is that there are two types of modeling. One see the world as made up by a gazillion solid objects put together in all sorts of ways. The other see the world as a big surface made out of a gazillion mesh points. There are no going back and forth between them. If you want to make a 3D printed part, you need a mesh file. If you want to make a CNC machined part, you need a file with solid geometry. I’m painting the situation in black and white here, but that’s the basic premise.

For mesh modeling there is no beating SketchUp. Developed by @Last and acquired by Google a few years ago. It’s simply great. Easy to use, well thought out and very capable in the latest version. And being free, you can’t beat the price. For serious work, you eventually want features in the paid Pro version, but the price is not bad at all. SketchUp has also become the software of choice for the whole ‘maker’ movement.

Solid modelers can be crazy expensive. I haven’t settled on a solid modeling package yet, but I’m currently trying out Alibre Design and Rhinoceros both are in the low end of the price range. They are decent but nothing to get excited about. FreeCAD is Open Source and free but I just can’t get the hang of it. Some times you just get the feeling that life is too short to try and work out how to use some introvert software that took years to develop and yet the developer(s) did not spend more than 15 minutes on the ‘getting started’ file. Developing incredibly capable software that is just too frustrating to use. What a waste of talent!

You may ask why not simply 3D print everything and forget the solid modeling mess? Well, currently 3D printers don’t make things that work; they make things that look like things that work. At some stage you need to make the real thing. You want to test physical properties like fit, strength, heat dissipation, how it feels in the hand.  CNC machining can help you do that.

That was some of the tools and services we use. I hope this might have been useful to some of ya. Hit the comments if you feel like it.