Richard Hughes on ColorHug and being an open hardware vendor
Half a year ago the first batch of ColorHug, a open colorimeter by Richard Hughes, started shipping to customers. Today we are talking to Richard about his experience of being an open hardware vendor, and his future plans.
Embracing the color gamut
Richard, recently you passed the 1K mark of shipped units. What's are the most interesting things you've learned from being a hardware vendor so far?
First and foremost, setting up Hughski Limited has been a lot of stress, and a huge amount of fun. There are distinct downsides to being a “hardware vendor” as we've found it is much harder than it looks.
The ColorHug has 29 different parts from 17 different suppliers, all with different lead times and minimum order quantities. Some parts like the IR cut-off filter are custom made and have a 12 week turn around time as they come from a small factory in China. I spend a significant amount of time talking to suppliers and trying to get all the parts in the right place at the right time.
It's also stressful telling people the finished hardware batch will ship “next week” and then finding out a critical component is held up in customs. Luckily everyone has been really understanding as they know I'm not some huge mega-corporation.
As free software developer you've no end of experience of talking to users, but what happens when they are customers too?
Open hardware is even more customer-oriented than working for a closed source сompany, as users want to talk directly to the product designer and usually don't like waiting more than a few days for a really technical answer.
I totally underestimated the amount of time I (and Ania) would be emailing people when I planned out the overheads for ColorHug. I probably spend an hour a day chasing up payments stuck in the PayPal fraud department, and trying to deal with all the technical questions on the mailing list.
Pascal [de Bruijn] has been a super friend in helping with the general color questions on the colorhug-users mailing list, and I owe him a huge debt of gratitude.
Quite a few people email me directly long emails about a specific implementation issue they had when dealing with the ColorHug (e.g. making it work on FreeBSD or using it to calibrate a laser), but I have to gently make it clear that buying a £60 device that's got a couple of pounds profit worked in, doesn't entitle the user to several hours of 1:1 developer time.
That said, if it's a feature is generally useful for more than one person, I'll normally schedule time to work on the issue. Or, I'll ask for patches :)
What's your experience with the post services?
I've found the postage services of the world are not as precise as they suggest. We've had packets sent to Russia going missing and then turning up 10 weeks later at the destination.
We've had about a dozen packets returned as the postal services failed to send them to the correct house, despite them being clearly marked. Out of 1100 packages, just one package has gone missing, and to Royal Mails credit after 6 weeks they sent me a cheque for £46.44 which nearly covers the cost of the parts.
At some point you introduced a feature for remote deactivation of a device that is reported to be missing or stolen and then recently you removed it. Was it really causing that much controversy, or did you stop feeling comfortable about it?
I was never really comfortable with the concept. I took some bad advice from a couple of people who have set up businesses like mine. They suffered heavy losses through mail-fraud, and when I set up Hughski Limited, I wasn't earning any profit on each device, and each “lost” unit was like taking money from my wallet. This was a bitter pill to swallow.
In hindsight I should have known that the open source community isn't your usual group of consumers. After sending over a 1000 units, only 1 has been lost, and there's nothing to suggest that was fraudulent. I think I learnt a valuable lesson and I hope that didn't put too many people off buying the ColorHug when the project started.
What do you think was the most difficult technical challenge you've run into?
I've spent about two weeks of working all the evenings on one particular issue recently. The sensor we use for the ColorHug is a commercial specification light to frequency sensor, which gives us a variable frequency output for the specific set of color pixels selected.
For the 1.1.x firmware versions we worked out the deviceRGB value by counting the number of pulses over a small amount of time. For very dark readings, we just increased the amount of time to the integral maximum time to try and get as much accuracy as possible. Unfortunately, this gave us a very imprecise reading when the number of pulses was less than about 50/integral time. We were just not getting enough data from the sensor to get an accurate low-light result.
What was the solution?
The new experimental measurement mode used in firmware version 1.2.x. It's wonderfully simple, but really hard to implement: we count the number of processor ticks between the individual pulse state transitions.
Basically, we're treating the “digital” frequency sensor like an analogue pulse width sensor. This means we can get accurate readings down to 0.5x the pulse time, and makes the dark readings a lot more precise and accurate.
The new measurement mode isn't yet enabled by default, and after a bit more real-world testing all the existing devices will be changed to default to this measurement mode.
I've uploaded a bit more detail and some graphs to the FAQ.
So, there are no unsolvable issues then? :)
The biggest unsolved technical challenge is still the limited number of CCMX matrices we have, which are used to adapt the primaries in the ColorHug to a specific display technology. We've started work on a “generic” matrix which is much closer to the average monitor than the device calibration primaries, but we're not quite there yet.
Really, we just need more data from people that have expensive photo-spectrometer devices. It's the same problem experienced by any tricolor colorimeter device, so it affects devices like the Huey as well.
On the future
Do you have any design changes planned?
To the ColorHug, no. It's a tried and tested design and we've shipped over a thousand with only one refund so far. We will certainly continue working on the firmware, adding features and fixing bugs.
For the future, I am looking forward to a ColorHug2.
Is it going to be a more professionals targeted device?
Yes, ColorHug2 would be a photospectrometer-style device, and so would be a lot more complex than the original ColorHug.
It'll almost certainly be a lot more expensive too, both because of increased material and labour costs and the fact that I'll sell a lot fewer units, but also because I'll have to buy a lot more expensive calibration equipment to be able to calibrate the calibrator. If you see what I mean.
I'm probably going to start designing the ColorHug2 at Christmas time this year, and hope to get some prototypes ready by Easter. That said, our family gets bigger by one this year, so I'm not promising to stick to those deadlines if my free time gets a bit more sparse.
I'm going to involve the community much earlier at this stage, so I'm sure we'll be discussing the relative merits of different illumination LEDs on the colorhug-users mailing list before I actually spend a few thousand pounds on a prototype.
Do you think there will be a place for a sexier body in ColorHug2 too?
Perhaps. It'll probably still be a black box, but it'll have to be a lot more precise, so probably a custom ABS or 3D printed box.
Maybe get some Blenderheads involved for product design?
I'll be the first to admit I'm not very good at 3D design. I'll do a prototype that is a black square box and then ask the community for better ideas, and then build something that looks a little bit sexy.
After six months in sale and over 1K devices sold worldwide there's still not a single ColorHug on eBay. Do you think it's a good thing or a bad thing?
Well, if it was a really super design, some guy in China would have cloned them and be selling them at half the price I can make them. I know a few people have tried to make the devices, but unless you're building more than 50 or so, it's not really worth it.
In the short term we've got a baby to get ready for, and I've got some stock of ready assembled ColorHug devices that I can sell. Every few weeks we'll still be doing firmware releases to fix niggles and add small enhancements. In the longer term, I'll start shifting my focus to ColorHug2 and start building some prototypes and sending them to key people like Graeme and Pascal.