So the Arduino team has announced an new product, the Arduino Yún.
Looks like a very interesting product. My own LED lighting projects tend to be installed in places where plugging in a USB cable is inconvenient at the least. So one main feature for me will be the programmability of the Yún per WiFi!
While the High Power RGB LED shield only requires three pins from any Arduino or other Microcontroller board (SCL, SDA, GDN) it would be nice to be able to just stack it on top of an Arduino and not have to worry about wiring it up on a breadboard etc.
Looking at some of the photos the Ethernet Jack on the Yún looks very close to the Header and also looks tall enough to interfere with a lot of other shields.
Our led shield is very short and it should fit. Should it not fit a little filing on the interfering edge of the board should do the trick. There are no traces close to that area and its save to file of a mm or so.
I have not posted in a while so a little update is overdue.
After completing the first prototype I received some of the little SMD containers from Adafruit and labeled them. I also prepared a simple little PDF file with about 15 sheets outlining which components were to be installed next. While it still was not as fast as I’d hoped I was able to make 4 more prototypes in about 3 hours. Obviously there is still potential for optimization!
So far, so good. Meanwhile several people have voiced interest in purchasing the shield when it becomes available. A guesstimate is that currently about 75 shields are accounted for. We have requested an updated quote for 100 shields from our chosen manufacturer and we are waiting to receive it to determine the price. Once we have received it we will start a fundraiser on Tindie.com.
I also have not just been sitting around but have worked some on the library. The one thing the library is really missing is a good fading algorithm. For my LED lighting system I developed an adapted 3D Bresenham algorithm.
The original algorithm was developed when most computers did not have specialized hardware to support floating point operations but only integer. While calculating floating point was possible, it was very slow. (Anyone remember Fractint ?) Equivalently most Arduino compatible micro controllers do not support floating point operations using specialized CPU operations thus are very slow when compared to integer operations. The initial Bresenham algorithm was developed to draw a line across a rasterized computer screen and uses only integer math, thus is very fast. For a RGB fading algorithm a 3D version of that algorithm was needed and I found an implementation on the Internet that suited our needs. It had to be modified and taken apart to allow fading several RGB LEDs synchronously but it works very nicely. As I recently found out, depending on what Microcontroller it runs on it can also be blistering fast.
In my own projects I have moved away from the original Arduino’s and have toyed around with a Teensy3. It is a fraction of the size of an Arduino UNO, costs only $20, is US made and employs a Freescale Arm Cortex M4 processor running at up to 98MHz. The new I2C library recently written by user nox771 also better supports the faster I2C hardware on the Teensy3 and supports I2C bus speeds of up to 2.4MHz. Ohh…. I forgot to mention, when using Teensyduino a plugin for the Arduino IDE the Teensy3 is fully Arduino compatible and comes with many libraries that one would have to dig around quite a bit somewhere else to find.
The hardware on our shield supports FM+ mode so up to 1MHz bus speed.
At that bus frequency a complete fade from RGB(0, 0, 0) to RGB(255, 255, 255) using the CIElab lightness corrected library function takes 53 ms!
That is a question I’ve been contemplating about for the last few days. Having produced a first functioning prototype is quite encouraging! The question is now what to do with that success.
While there have been some people that have shown interest in buying one, currently the number possible orders does not exceed 20. That on the other hand would not justify having 50+ boards manufactured as that would include having to lay out about $2000 in advance. We have been in contact with a manufacturer in Germany (where the co-host of this blog is located) , specialized in manufacturing small series. This by the way is not a garage shop but a company that has been doing this for 20+ years and they’d be providing us with a complete turnkey solution. We send them the layout and money and they send us back completed boards. However, that only makes sense if the number of boards is larger than 50 for the first batch and would also really only make sense if there would be several batches of 50+ boards after that.
We could investigate to find a less costly alternative, however, the reduction in cost would have to be rather significant 30%-50% to really make a difference. We have a costed BOM and the price of all the components assuming Newark as the source of supply is about $27 for a qty of 50 boards. That does not leave much room for the labor to actually build the board.
My current thinking is to set up a fund raiser at Tindie and then post on a few forums to determine if there is enough interest to justify having 50+ boards manufactured. Another option would be InMojo. They are also offering assembly and serve as a sales platform for open source hardware.
Meanwhile, as I’ve mentioned above, my early success is really encouraging. Should there not be enough interest I can imagine making these myself depending on what QTY is involved. Making the first prototype was my first venture into SMD and Solder Paste and Reflow Soldering in a Toaster Oven. I started around 6:00PM in the evening getting all the materials ready, building a makeshift stenciling fixture etc. I was done with testing the prototype and had written the blog post by 11:00 PM. I believe it took me a good 120 minutes to populate the board. I had not made an assembly plan and I was doing it on my computer desk going back and forth between the layout and schematic in Eagle to see where the component went and and then I had to use the BOM with the Newark part number and find the bag/reel/tape with the component, and get them out of the packaging to be finally be able to place the thing on the board.
With some better preparation and a little hardware I believe I should be able to get it to well under 30 minutes per board. I have ordered several of these and several of these and with some intelligent, labeling and an assembly plan things should progress much faster.
I have sets for 4 more prototype boards and will see how quickly these can be assembled with better planning within the next week or so.
Once that is completed I intend to send one or more to my partner-in-crime in Germany to do some testing under load. I don’t have enough LEDs to test the thermal side of things and he has managed to burn out one shield already. This should tell us how many LEDs the shield really can sustain! I only use one RGB LED @700mA per shield in my own lighting projects (trippylighting.com) and none of the thermal considerations are of any concern
You can read about it and look at a finished board, however, until you try to pick up these dust-like 0402 capacitors or 0603 resistors with a tweezer and try to place these precisely on a pair of solder pads you really don’t know what you are dealing with.
I started with the smallest components first and about halfway through populating the board I had serious doubts this board would ever work! Nevertheless I decided to continue to at least see if how reflow soldering in the toaster oven would work. While this is certainly not my first electronics project this is my first attempt at a more advanced SMD board and reflowing with a toaster oven was also a first for me.
And again, I’ve read in many places before that the components don’t have to be placed absolutely perfect and the reflow solder process is somewhat forgiving and surface tension will correct some mistakes. There are a number of videos available online that show this, but it’s still thrilling to see it actually happen and have the board come out looking very nicely reflowed.
It is even more thrilling to stack the thing on top of an Arduino connect it to an RGB LED and have it going through a color changing routine immediately. Well…almost. I had placed the diode in the on-board power supply in reverse, but that was quickly fixed with a trusty soldering iron. Anyway here is an image of the first functioning prototype still stacked on top of the Arduino and hot, right out of the oven, so to speak
Before I ordered the 10 prototype boards I had done a good bit of research on to see what people had to say about different vendors. Amongst these was a post that compared the precision of the drill position of two different vendors, namely ITEAD and OSH Park with OSH park being of much better quality.
While the ITEAD quality may be fine for many boards, it won’t work for our boards as can clearly be seen in the image…..ohhh….wait…not in the image I posted a few days ago. But the photo I took with my Canon S90 shows the problem quite clearly. To see it you may have to my other blog @ trippylighting.com (I know, what a shameless plug )
While I am assuming that these boards will work OK for prototyping purposes, I would not want to order production quantities from ITEAD. I believe we’ve mentioned that we have received a quote from a manufacturer and if we decide to hand it over to them, they would provide a turnkey solution and would do everything, including sourcing the components, assembly etc. Having seen a few videos on their web-site I am confident their quality is much higher!
Clearly my product photography skills need some improvements. Snapping pictures with the iPad on the kitchen tablet leaves something to be desired, however, I could not resist posting, now that the prototype PCBs that I had ordered from Itead Studio arrived.
These look much sharper than in the image
The only thing I am still waiting on are the Panasonic Inductors, which were not available from the big three here in the US ( Digikey, Mouser, Newark). I finally found them on the German RS components web site, tracked down a US rep and got a quote for some prototype quantities today. So I hope that it will not take longer than two weeks for them to get here.
Meanwhile my Panasonic InR toaster oven nicely follows a ROHS temperature profile. PID controlled with the Arduino Reflow Oven Shield
The Mylar stencil arrived weeks ago from Pololu and I hope to be able to post about the first fully assembled prototypes within the next 4 weeks.
Reducing the PCB in size certainly has some benefits , however, the reduction of surface area can also easily reduce thermal performance, if care is not taken. The LED driver chip LT3496 has a rather large (compared to the size of the chip ) thermal pad on the underside that is reflow soldered onto the top ground plane of the PCB. Also thermal vias are placed directly under the pad and surrounding area to help conduct the heat to the bottom ground plane. Both planes serve as heat spreaders. So the question that interested me was how much ground plane is necessary and how many vias are needed, what size should they have and how densely (or not) should they be spaced ? Read the rest of this entry
So what progress have we made ?
While PeterJ has been able to find a very promising manufacturer specialized in small series production I have been busy with adding a few smaller improvements to the shield design.
- I reduced the board size by almost 20 %. The main reason for this was that PCB pricing is usually calculated in $/area and this board having 2oz copper is more expensive already compared to the regular 1oz that are available from the usual (here in the US) hobby/low volume PCB houses such as OSH Park and BatchPCB. One can easily see the impact of the board size and compare prices using the usual online calculation tools. I based my pricing on the CustomPCB.com site online tools. They offer 2oz copper thickness, however this is only interesting for a larger number of boards. If you only need a hand full you’ll have to look somewhere else. We are loking at an initial batch of 50 so that was not a problem.
- The Murata Inductors are not available anymore and had to be replaced. The Panasonic inductors that were suggested by the manufacturer that PeterJ contacted are in fact better than the Murata. Not only are they shielded but have a roughly 30% reduced RDC which means in essence that they transform a lot less energy into heat. The shielding helps reducing EMI and the shorter signal paths of the smaller board help reducing EMI as well.
- The barrel jack has been replaced with a screw terminal for greater flexibility.
- I added 2 solder jumpers on the top side of the board to be able to disable the pull-up resistors. In my own project I am using 5 shields on about 4m of CAT5 Ethernet cable. The 1.8k pull-up resistors, while still in spec, are already on the lower side and when using several shields one can quickly run into I2C problems. I had to remove the pull-up resistors on 4 of my 5 shields and jumpers would have definitely made that easier. Particularly if one later on decides to use the shield standalone. Hand soldering another set of 0402 package resistors back onto the board is not my cup of tea and in this case solder jumpers just make for a better design.
- I added another set of two solder jumpers to the bottom side of the board to provide compatibility with some of the more recent additions to the family of Arduino Boards. The old shield was compatible to the Duemilanove and UNO R2. The new shield, will also accommodate the UNO R3, which has the newer headers that offer more contacts. The jumpers also allow to “switch” over to the newer SDA/SCL contacts that the Arduino team now has conveniently placed exactly on the opposite side of the newer boards, e.g. the Leonardo and Due.
- With the exception of the thermal vias under the LT3496 all other vias will be tented.
The only thing that will not really be Arduino compatible is the mounting holes. These do not follow the usual Arduino pattern anymore, but I doubt that is a problem. If a user wants to stack the shield onto an Arduino, the holes are not of much use anyway and if the shield will be used standalone, compatibility with the Arduino hole locations is a rather useless feature.
Whether the first batch of 50 boards that we have in mind will be of the newer design has yet to be decided. It may not happen. I am rather reluctant to submit an untested design to manufacturing even when the changes are not really that significant. More than 20 years of engineering experience have taught me that lesson I’ll need to see how quickly I can assemble a prototype and test it. Below is a screenshot of the improved version.
Open hardware does not only mean that it’s free for anyone to tinker with it, modify it and use it, it also means that everybody is free to put it into production. Usually, you’d do that once you have made significant modifications to it, like, maybe, built an arduino like uC into a larger design.
It also means that you can take a project that has been abandoned for some reason and keep it alive. That’s what I’m trying to do here. While Neuroelec is away for reasons noone seems to know, there’s still demand for his excellent high power LED shield and thus I’m trying to not only scratch my own itch (I need a few more) but also, maybe, provide them to the others who are looking for a simple solution to drive 700mA LEDs en masse.
Thankfully, Neuroelec has made all required documentation and software available, either through his blog or through google code, this way, I think, I should have everything needed to start another small production run. My problem is: despite the fact that I’ve tinkered with electronics for many years, I’ve never actually had anything produced in even a small batch. I’ve sent out e-mails to request prices, but what I usually got was board prices, not full production prices. I’ll have to keep looking, but I’m hopeful that I can provide a number of fully populated boards no later than January.
There are many different power LEDs from manufactures. Even the same LEDs mounted on different boards. I have some high power LEDs that I bought for the LED shield. Here is a brief review on these for the people who consider to buy one of those.
350mA White LED
The LED is one of most common 350mA white LED and is copycat of famous Luexeon I. I bought it from Sparkfun but it is likely sourced from China. It come with MCPCB. Measured Vf=3.18 at 350mA
700mA Cree XR-E Q2
Cree is a big name in the high power LED industry. They are simply cutting-edge in term of LED chip technology along with Philips. They are making one of the most efficient LED in the market. XR-E is also quite efficient 700mA line. “Q2” is the bin codes which indicate basically brightness LED among the same XR-E. There are quite a bit of price difference depending on bin code. Better bin code means brighter and better efficiency. (P2 bin – 67lm, Q2 – 87lm, Q5 – 107lm, all at 350mA, minimum flux). You can buy XR-E P4, Q2, Q5, R2 bins from dealextreme. Measured Vf=3.62 at 700mA
700mA Cree XR-E Color
These are red, green, blue LEDs that are same series with white XR-E. Bought them from dealextreme as well. they are efficient and bright but don’t have any bin code. Color LED have slightly different wavelength depending on bin code. Since they are three separate LED, they are not good for RGB color mixing. Lens from Sparkfun can be used with these three LEDs for color mixing though. If you know any place I can buy Cree RGB mounted on single MCPCB, please let me know. Measured Vf at 700mA are red = 2.16V, green = 3.65V, blue =3.2V
3W (350mA x 3) RGB LED
This is basically 3 LED dies (Red, Green, Blue) combined in a chip. Each color is rated 350mA. I bought it from sprakfun. Almost same kind are widely available from many places. Since 3 different LED dies are in a chip, it has very good color mixing. Quility of build is just fine. Solder points are a bit too close. Be careful not to make a short. Measured Vf at 350mA are red=2.36, green = 3.24, blue = 3.18
10W (350mA x 3 x 3) RGB LED
When you want to have a powerful RGB LED that has a good color mixing, this is what your looking for. This LED is exactly 3x 3W RGB LED above. This LED has 3x 350mA in series for each color. This is not exactly 10W more less 8.8W, but still very bright. Quality of build is very good. It is cheaper than similar 700mA Lumiled comination. Measured Vf at 350mA are red=6.39V, Green=9.38V, Blue=9.22V
700mA Philips Lumiled Color
Philips is another big name in LED market along with Cree, Osram, SSC and so on. They make very efficient 700mA. Unlike Cree color led, you can easily find RGB LEDs that are mounted on single MCPCB or FR-4 PCB. Believe or not red big PCB (from Sparkfun) and small white star MCPCB (from ledsupply) have same kinds of LEDs. They have all Three RGB 700mA Lumileds. A difference in LEDs is one from Sparkfun has royal blue rather than ragular blue. Royal blue is the way how Philips call deeper blue (shorter wave length). MCPCB has smaller thermal resistance than FR-4 PCB. Small thermal resistance means cooler LED (about 10 Celsius degree difference). Due to the larger size of PCB (Sparkfun), it has poorer RGB color mixing.