 Yr hoffaeth o CullerHug Plus, ymwneud cyfnodol, ymwneud Ysbyt. Ymwneud yw Richard Hughes, rwy'n ymwneud ymwneud ymwneud ymwneud ymwneud, mae'n fawr o'r rhaid o'r rhaid o'r cyfnodol yn Lennx, o'r cymdeithasol, o'r cyfnodol, o'r panel o'r panel o'r gyfrannu cyntafol. Yn ymwneud ymwneud ymwneud ymwneud ymwneud ymwneud ymwyneud, sy'n gweithio o'r cyfnodol yma, ac mae'n meddwl i'r thrin arall, ond mae'r rydych chi'n gweld ffirmraffer, achos mae'r cyfnodol. A gyfnodol ymwneud ymwneud yn ei jyst, un roedd yn y plosio gynnyddol. Mae'r cyfnodol yn ymu'r llibwneud ymwneud ymwneud, eu fus yw'r ardal dyma ym 2011, rwy'n gweithio'r wneud yn llefos rhaid. So far we've sold... ...toddyn we've sold the 3,080 second colour hug, all still hand built in London. So it's been massively more successful than we thought it would be. All these colour hugs are made at home. I used to build the PCBs myself as well. Now that's done by somebody else, but we still do the assembly, the self testing, packaging, sending to the post office, returns, invoicing, tax, all of it ourselves. It's completely like a homegrown business. So you can see it's literally working out of a front room. Now I have an office in the garden, which is slightly more swish. So colour hug inspection, I'm going to do like an introduction to colour science in seven minutes. So this could be fun. So a spectrograph is a device that measures rather than the amount of red, green and blue in an image, it's the device that measures all the frequencies in between in discrete bands. Now that's necessary because we don't use these anymore. These, like a CRT displays we all used to have, used to be fairly similar, so you buy one CRT display, which would usually pretty much match another CRT displays colour output. They'd be basically SRGB modulobrightness, which meant that we didn't have to do an amazing amount of colour management to make images look the same on one display as another display. But then the future caught up with us and we have wide gamut displays, things like the dream colour displays. We have OLED on our phones and tablets and stuff. And all of a sudden you can take a photo now and it looked completely different on one screen to another screen because of all the different, the colour capability of the screen. So you kind of have to do colour management. And if you look at the actual outputs of the devices that we're measuring, gone are the days where we can just assume everything's the SRGB. ColourHug1 assumed that all the devices it was measuring were roughly SRGB, which became a problem as more and more devices deviated from that. So you might see the grey line on the graph is an OLED output of a mobile phone, which if you look at the mobile phone all the colours are very high in contrast, very saturated colours, they pop out at you and it's great for marketing, but it looks really bad for showing realistic photos on. So we need to actually analyse the output of these devices so we can correct the image in a meaningful way. So this is a graph that will destroy any colour scientists in the room. The horseshoe graph on the left basically shows you an SRGB output, which is what you would find on most screens 10 years ago. And the one on the right is a wider gamut display on a modern TFT panel, which basically if you try and squeeze all the colours on the right-hand triangle into the left-hand triangle, the spot colours are going to be different. So red hat, red won't be the correct red, IBM blue will look wrong. And so you have to do actual colour management on devices because of the different gamut ranges. So this is what I really want to build. This is like an industrial spec. I think this is measuring paint for either a weapons or space programme. This is a spectrometer, and it's basically analysing the... It's just firing some light sample and seeing what bounces off. Nothing more complicated. Realistically, this is what I'd like to be building. This is a lab spectrometer, which is the sort of thing you would find in a forensics lab. For instance, maybe analysing blood or something like that. About five, maybe $10,000, so it's an awesome bit of kit, but it's just too expensive for consumer use. And then there's a good old colour monkey. A colour monkey is an awesome device. It's designed for a consumer. It's about £300. It's a really nice piece of kit with some really nice optics inside, but it's completely unfriendly for free software and free hardware. All the hardware inside is completely proprietary. Even the way of driving it, we have to reverse engineer. It's something like this, which is a precision, a piece of equipment, really. Reverse engineering a driver could mean that we're turning on the LED and literally burning the LED out, or powering on the CCD array and causing it permanent damage by trying to reverse engineer it. So it's not an easy thing to use. So I want to really build something that's as good as the colour monkey in terms of performance and resolution for about the same price point, but have completely free software and completely free hardware. So after doing this talk a few times, people have said, but you know you can make a spectral device using a bit of broken DVD and a webcam, right? And yeah, you can. You can actually use a broken piece of DVD and a high resolution webcam for finding out the absorption spectra of the sun or finding out emission spectra of a candle or something. And it's a really good demo in a classroom, but it's completely unsuitable for creating an ICC profile. There's just not enough resolution. It's extremely stable. And it's not precise enough. It's non-linear in almost every respect. So it's very, very difficult to get any meaningful results from this. So it's kind of discounted. The other issue with trying to build a spectra is people say, oh yeah, but you can go to a company like Hamamatsu where they will provide you a two centimetre by one centimetre spectra device. Sure, but it only has a resolution of say 20 or 30 nanometres, which means when you try and measure a really spiky spectral output, like this is the red phosphor on a CRT display, but CCFL backlight would be something similar as well. If you're binning this into, say, 20 nanometre bands, you completely miss some of the spectral information which makes the ICC profile you produce almost worthless. So actually anything you need, the bare minimum for a spectra device is probably something like five nanometres for any kind of sensible results, which means discounting a lot of the commercially available inexpensive spectra devices that exist. Now the other problem with designing a spectra is you need to illuminate it. Now ColourHug 1 and ColourHug 2 were simple colour emitters. They measured the amount of green, blue and red emitted from a source. But for something like paint or dye, paper or wood, you need to actually shine an illuminant onto the sample itself to be able to read a reading back, which is awesome because then all of a sudden you can have one device which can profile your display and can also profile your printer and paper combination as well. Then you start talking about illuminance. Now illuminance is really just a posh word for light source. Now here you can see there's the helium vapour light which is making everything yellow and very sort of orangy. So it kind of shows that the illuminant can make the colour look very different to what it actually is. With modern LEDs, although the LEDs spectrally are still very spiky, actually this will have a much higher CRI index, how close it is to daylight compared to spiky illuminant like fluorescent. But it does make a difference if you're choosing paint colour and you know that it's going to be in a theatre like this with fluorescent lighting, you don't want to be choosing blue sea covers that look purple under the lighting that you're going to be using it in. So actually caring about the illuminant is important for measuring this stuff. So I don't have the ability to make a spectrograph myself. So I can make the electronics, I can do the software, I can do the packaging, I can do all of this, but I haven't got decades of experience in optical assemblies which would let me build the spectrograph itself. So I've been asking other companies, a middle quote, this is a quote for just 10 units for £18,000 just for the actual optics itself. Now raising the order quantity does make this drop significantly, but you can see this isn't a small amount of money and this is just a hobby. So this is the reason I've been making a small amount of money on each colour hug 2 sold so I could fund the development for colour hug plus. And actually even if I try and match the £300 RP price of a colour monkey with a colour hug plus, I only make just under £15 on each one which is a tiny, tiny profit on a £300 device. This is after all the taxes and VAT and everything else is taken out. So there's actually only a tiny amount of profit so you have to sell an awful lot of colour hug plus is to make out your initial investment back. So it's really, what I'm trying to express is this is basically a hobby. This isn't my professional business. There's no corporate backing for this stuff. It really is me standing there trying to produce something that's as good as a commercial product with basically no money. So I launched colour hug 2. Colour hug 2 was a successor to colour hug 1. We went from using a $2 sensor to a $15 sensor. And we're also trying other stuff in the colour hug 2 that we'll use for colour hug plus. Like this chip down there is a temperature chip, precision temperature chip so I can do temperature compensation of the sensor. Not so essential for colour hug 2, massively essential for colour hug plus. That's a technical try. The top chip is a high-speed DNA-able SRAM because for instance to fill a colour hug 2 it will do display, attack and decay and refresh rate calculations faster than USB will take the data so we buffer them. For colour hug plus it becomes essential for even sampling. We have to use the buffer as well. So it's a whole load of technical things I've introduced early for colour hug 2 that we're using occasionally just so for colour hug plus we know it works. I've been prototyping colour hug plus for I guess a couple of years now testing different ADCs, DAX loads of different stuff and I've found a design that I can produce which is good enough. There's no way I can make something that's going to be a lab-grade spectro but I can do something that's good enough for an ICC profile. This is an original spaced prototype. The spaced prototype this was originally using a crossed CT design using a large optics element. This is just a space thing from a workup where everything would go. We didn't go with this. We actually went with a much more expensive sensor with an optical holographic linear array on it which is much more expensive but much smaller which meant I could reuse a lot of the tools I built for colour hug 2 for colour hug plus. So this is the actual colour hug plus PCB. This is one of the first prototypes that came out. It looks very, very similar to a colour hug 2 with a few more holes, few more chips. The digital side on one side is almost exactly the same. So I know the digital side works completely. The analog side is the other side which is completely different to a colour hug 2 and it's actually rammed just the same as the top of the digital side is. I would have had colour hugs pluses to show as prototypes but I made a small mistake with the analog design involving a difference of opinion with voltages in me and a vendor. So I have to re-spin the analog side before I can actually get some prototypes out so it's very, really close. The physical design is very, very similar to colour hug 2 in that there's a foam pad at the top, slightly thicker, ABS lids, there's a glass, a circular piece of glass with a reflective coating on. There's a 3D printed element. There's a sensor itself. There's the PCB with two illuminants. The incandescent illuminant which provides a wide band illuminant like the sun, like the CRI100 and UV source as well which allows us to do things like... For instance, paper has a coating on it that fluoresces under UV light which makes the paper look more blue than it would do otherwise which makes your eyes get tricked into thinking it's brighter white which looks awesome when you look at the paper but then you print stuff and the print looks wrong because you haven't taken into account the fluorescence of the paper. So we have to use the extra illuminant which we can switch in, which the colour monkey can't do so we can work out how much of that is fluorescing and then a simple case and that kind of fits together really well. It was tricky. The 3D part actually shields the sensor from stray light internally and the internals are completely matte black as well. So the finished article it looks almost identical to a colour hug 2 in size slightly deeper. This allowed me to reuse all the existing tools I built for punching the USB hole, the status LED and it means that the NRE is kept at absolute minimum. This is my prototype that's sitting at home. The glass with the reflective coating, the 3D element, the optical assembly which on its own I think is about 140 pounds when you buy 100 of them at a time so it's kind of expensive and the digital side finished, some screws and the lid. Ignore the quality of the cutting just hacking up for a demo. The finished one will have a beautiful interved curved surface and the optical assembly, the 3D printed bit will have like a nice matte flat finish rather than the moment but it kind of shows that it does all fit in the box and it kind of works. So all of this stuff is completely open hardware and open source. The firmware itself is GPLv2+, it's written for the XC8 compiler. I'd love to use SDCC, the free compiler but it mis-compiles the code when it's better I'll switch to that. The model itself is written in open SCAD which is again open in a CERN licence and the hardware is written in GDA tools like PCB and GSCAM which is again all CERN or GPLv2+. So it's completely open source. Now this is my current plan so color 1 we've phased that completely is one left in stock in case anyone desperately needs one. Colorhug 2 we've phased in the end of 2015 and we're still selling. Colorhug ALS is an ambient light sensor that I designed mainly for the kernel hackers so they could have a device for testing the kernel stuff and the user space bits that sit on top but we're still selling it as a crude ambient light sensor mainly monochrome. And Colorhug Plus which I wanted to bring in like I wanted to actually bring it up to speed last year but I kind of prioritised putting a roof on the house and we've got a three-year-old little girl so I kind of prioritised being a dad then Red Hat then Colorhug which I think is probably the same thing to do. So Colorhug Plus will definitely happen but probably not quite this year maybe end of this year. That's my talk. Thank you for listening and I've got five minutes left for questions. Thank you.