Three days are left till the reward for mining Bitcoin blocks goes down in half, to 12.5 coins per block, from 25. It is the second time that this happens, as in November of 2012 the reward went down to 25 from 50. Next (and final) halfing will take place in another four years. The mechanism is integral to the core Bitcoin design and has been implemented in the node code since day one. To the people in the field this is so important, that they track it by the hour.
Mining blocks is seriously expensive, and the Bitcoin crowd thinks that it’s a good thing. Cause that prevents the ‘bad guys’ from doctoring several blocks in a row and stealing a bunch of coins via “double spending”. [Will cover the details at some future date, as there is no big urgency with this.]
The largest cost in Bitcoin mining is electric power, both to run the data centers full of specialized computers, and to cool them. Most mining is done in China, using subsidized power. But the ‘carbon footprint’ of the Bitcoin industry is nothing short of atrocious. Some compute farms are in Iceland, using geothermal power, while saving on the AC bill.
Bitcoin miners get paid in two ways – a reward’ of so many coins per block and a fee for recording transactions in the block. So far the reward has been by far the predominant part of miners’ income. Bitcoin price has been rather volatile recently, but at $600 the reward of 25 coins is worth $15,000. By comparison, the miner recording fees, in the best case at current rates, could add up to around 0.5 BTC ($300) for the maximum size 1 Mb block.
The fee is usually quoted in Satoshi / byte, where a Satoshi unit is 1/100,000,000 BTC. Today, a fee that gets your transaction recorded without delay is around 50 Satoshi. For the shortest 250 byte Bitcoin transaction, it works out to around 7.5 cents. Not all transactions are this short – the ones with many ‘inputs’ and ‘outputs’ can be several kilobytes long.
The Bitcoin marketplace has been running recently at around 200,000 transactions per day. By design, a Bitcoin block is mined about every ten minutes, for 144 per day. So, less than 1400 transactions are recorded per block, on average, not 4000. Partly this is due to larger transactions, but also because many miners do not fill their blocks fully, out of fear of duplicating a transaction that another miner already put in a block of their own, and having the entire block ‘orphaned’ and not getting paid at all. Smaller miners may fill as little as 1/4 of a 1 MB block.
All this is important because if the miners have to compensate for the loss of 12.5 BTC per block by increasing transaction fees, they would have to raise them twenty-fold. Is the Bitcoin community prepared for a $2 transaction fee?
 Learn more about Bitcoin and the details of mining and rewards.
Wireless charging is cool. Because anything wireless is cool. Next best thing to not having to do it at all.
So, people have tried to do wireless power for over a century. Think Tesla – the man, not the car. He wanted to send electricity across distances without any wires. At the moment we’d settle for charging the phone by putting it on a special pad, and avoiding three seconds of fiddling with plugging in the cable.
The Qi standard, of which we’ll talk further later, is built into the latest Google Nexus 6. And, there is a whole bunch of cheapo pads offered on eBay, one branded TechMatte. Having invested $11, I had a pleasure of taking it apart.
The top pops open without any tools, thanks! Here comes the Coil. Tesla would be happy, initially at least. The square piece of plastic that the coil is resting on is stuck onto the back of the case with two-sided tape, which one pry off to see the interesting part of the PCB.
The main ASIC is labeled GPMQ8005A, and it comes from a Shenzhen company Generalplus Technology. A QI Compliant Wireless Power Transmitter (no surprise there). I could not find an official quote on it, but on Alibaba there was an offer for $2.80. There are two TI (ON Semi) LM324 Quad Opamps, at $0.10 these days. Two Michrochip TC4427 MOSFETS, at $0.90, Alpha & Omega Semi AO4606 30V complementary MOSFET, say $0.20. Or maybe I got them all wrong. Will microscope them again at work.
On the back of the unit is says: Input 5V – 1.5A, Output 5V – 1A. In the context of a wireless charger, the Output labeling is wrong in so many ways.
Oh, the dang thing does not work. Have to take the phone out of the case. Finding the one spot on top of the pad where the green LED goes on is a game that grows old quickly. And, if you find it, by using an App that monitors charging you will discover that it does not charge worth a damn. Cannot even keep-up with the power consumption of the screen being on.
Various flavors of Lithium batteries are very popular now, Lithium-ion (Li-ion), and Lithium polymer (LiPo, or Li-poly). They do hold more charge (specific energy) than Nickel Metal Hydride (NiMH) ones, some almost twice as much, for the same weight. They can be recharged more times, and maybe even charge faster. You will find Lithium batteries in every cell phone and quadcopter.
But NiMH AA batteries are each sporting as much as 2500mAh at 1.2V (nominal). It does take 3-4 of them to get to the LiPo cell voltage of 3.6V. But the NiMH batteries hold that 1.2V level through just about the whole discharge cycle, which the Lithium ones do not. Most importantly, Lithium batteries are now viewed as a serious fire hazard, because they can get mighty hot, if not charged correctly. So hot, that some incidents were reported, of laptops catching fire. Now each Lithium battery cell and each device using them must have warning labels. There are rules about having them on planes or in postal parcels, as if they were some kind of a high explosive bio-weapon, which they are not. But, NiMH are considered safe, by default.
NiMH batteries are not as environmentally suspect, as the NiCd were, because of serious toxicity of Cadmium. And modern battery designs have no ‘history’. Remember, how once in a while we had to discharge NiCd batteries all the way in order to restore them, to make sure they did not ‘remember’ a smaller amount of charge? No more, with NiMH. One does have to charge NiMH with some care, so as not to overcharge them, as they can heat up a bunch, and also not to trickle charge them forever.
I found in my garage a 15-year-old Kodak NiHM charger, and also bought a Lenmar 8 Port Battery Charger over a year ago. Brands had shifted and diminished, but let’s find out what changed technologically in fifteen years. You will see that there are systematic differences, but also surprising similarities. There are four screws holding the Kodak unit together, tiny hex screws, and I was too lazy to look for a set of hex wrenches. A couple of small flat screwdrivers worked.
Lenmar went even further. There are six safety screws on the back, of a slightly unusual variety, with two slots on the sides. Not something common, like the star with a pin sticking out in the middle, which I do have bits for. Well, two different sizes of scissors did the job in the end.
The Kodak plugs straight into the wall outlet, but Lenmar has a walwart that brings out 12VDC. The beauty of such a setup is that you can have a walwart and a plug specific to every country and set of regulations, and the main circuit is the same, and only has low voltage. So everyone does this now.
But why would Lenmar use the funky screws, as inside there is no high voltage whatsoever. And these screws must have cost at least a dime a pound more than the common variety. Would those extremely pragmatic folks in Shenzhen ever do this on their own, or was this the one design contribution by their US counterparts?
The mechanicals of battery contacts and the utilitarian color and layout of the PCB top feel very similar between the two units. There is a bit of a difference in how actual electronics are accomplished. Besides the whole 110V AC issue that the Kodak has to contend with, they used a load of discrete components and a couple of ICs, TI LM324 quad Op Amps, which are a dime apiece these days.
Lenmar sports two Michrochip TC4427 MOSFETS, $0.90 each, though they must be getting them a whole lot cheaper. Also, one can see an Alpha & Omega Semi AO4606 30V complementary MOSFET, maybe $0.20.
So, now you know. The more things change, the more they stay the same.
Outdoor solar lighting is a wonderful idea. No wires, can stick them anywhere. And ‘solar’ is still cool, and this seems like a perfect application – charge during the day, light-up when it’s dark.
We grew tired of $5 outdoor solar lights breaking after two weeks, and $15 ones staying on for a couple of hours after dark, which in the winter means about 9pm.
So, let’s find out what the higher end has to offer. Malibu Solar Spotlight is not cheap, at $50 at Home Depot. It promises 10 hours of light, at 54 Lumens. That’s about one twentieth of the light that a 100W incandescent (old style) bulb would put out. But LED light shines in one direction and looks really bright. Remember that time you shone a new LED flashlight into your eye, despite all the warnings from the manufacturer, your mom, and the priest.
The device does come with a separate solar panel unit, connected to the light fixture by a black wire of about 3′. The angle of the panel adjusts with a screw, so you can place and aim the panel where sun does shine. The panel is relatively sizable at about 20″sq. and labeled on the back at 2W.
Alibaba lowball prices for bare 2W panels are around $3. Or they quote $0.70/W, making the 2W panels about $1.50. Now, Amazon has ones at that rating for $10, and packaged as a product, in a case, 2W panels are $30. May be able to say more after reading up on monocrystalline, polycrystalline, and amorphous silicone types used in solar panels, the number of cells, etc.
The light comes from an array of nine LEDs, making them 6 Lumens each. These are old school, rounded top ones. Nowadays, a single 3V LED emitter, the one that’s square and has a light yellow color, will put out way more that 60 Lumens, and cost under $0.50 on Alibaba.
The sturdy cast metal outer housing feels good in the hand. The plastic standing pole and spike do not. The electronics subsystem has an ON switch on the back of the unit. A maze of wires towards the front does not inspire confidence in this aspect of the design from Asia.
It also houses two Fullriver 900mAh 18500 LiFePO4 3.2V batteries, in parallel. Fullriver is a serious industrial battery brand from China. The LiFePO4 chemistry is known for fast charge and constant voltage, and used for rechargeable tools and even electric cars. Equivalents of these batteries could cost $5-10 each retail. Off brand lowball prices on Alibaba are around $1.
As to the 10 hours at 54 Lumens claim? It would take some 3Wh to charge each battery, at 100% charger efficiency. After allowing for some inefficiency, we’d need some equivalent of 2W over 4 hours from the panel. Should be doable on a bright sunny day. Now, even imagining that we can get the full 900mAh out of each battery over 10 hours, it would give us some 0.6W to work with. So, could this setup, with these commonplace LEDs produce 90 Lumens per W? Not very likely…
Oh, the first unit we bought at Home Depot went right back – DOA. But the second one’s been working for a year now, and is still impressively bright well after midnight. As to ten hours – not really, more like 4-5.
NETGEAR has been selling a line of cute little Internet camera systems called VueZone for a couple of years now. Actually, the company known for WiFi access points and routers acquired a San Diego firm Avaak to get the product.
The way the setup works is that several tiny battery-operated cameras talk wirelessly to the central base station plugged into the home Internet gateway, like a cable modem. VueZone uses same 2.4GHz ISM (Industrial, Scientific and Medical) band as does WiFi, Bluetooth, and some cordless phones. A proprietary signaling scheme is likely used, to optimize power consumption of sending video frames to the base station.
The electronics come in a tight package, with a camera assembly connected via a flex cable to the PCB, as does the small motion detection circuit. The central processor is Texas Instruments MSP430F5524, a low power 25MHz 16-bit processor with 64K Flash and 6K SRAM, $5.50.
Wireless connectivity is provided via Nordic Semiconductor nRF24L01, a low power 2.4GHz RF transceiver, $2.20. There is also Cypress CY62138CV30LL 2Mbit SRAM, $3, and Xilinx XC2C32A, $1.60. Component selection of the camera assembly and the motion detection circuit warrant a second look.
The mechanical design is neat, though it looks like a puzzle when one tries to put the thing back together. In the end the pieces did fit, and were all accounted for. The sliding power switch and the battery door lock are somewhat of a kludge, though they do mostly work. The magnetic mounting semi-sphere is brilliant and gimmicky. Pretty cool, in the end.
One design choice was to use two 3V “Lithium” batteries, in parallel, whether because of discharge characteristics or physical dimensions, Shorter but fatter than AA, these batteries are known as 123 and built with the LiMnO2 chemistry. Rated at 1500mAh, they are used in cameras and medical equipment, but despite the word lithium in the title, they are not rechargeable. And 123 batteries are not cheap, with the Energizer EL123 at least $1.50 in bulk. A comparable quality alkaline AA Energizer E91 is roughly $0.50 each in bulk.
The base station has several lights and a button used to put the unit in the state to ‘pair’ with the camera. Not surprisingly, the base station runs off a wallwart and connects to the router Ethernet. I would like to see what processor was picked for the job of encapsulating the video stream, before sending it to the server. But, I could not pop it open and punted in the end. If anyone cares, let me know and I’ll take a saw to it.
Netgear recently revamped their VueZone offering, eliminating the four camera setup and raising prices on other configurations. This is perhaps in preparation for switching over to a brand new and incompatible system, an HD camera set named Arlo and priced at $350 for a two camera configuration. Fortunately, there is a brisk trade in VueZone combos and individual pieces on eBay. I picked-up the two camera setup for $75, in brand new condition.
Looking through a camera at what’s happening outside was first done by Nazis in 1942, to keep an eye on the immense V-2 rockets. Now, you too can look outside, and using your smartphone. Something that Nazis could not do.
And your camera will be inside the doorbell, so you can see people at the door, particularly their index finger. Video doorbells, the WiFi ones, have been attempted for several years now. DoorBot raised $1 million from VCs over two years ago, and is finally shipping, having been re-branded as Ring. And there’s been crowdfunded Chui, i-Bell, Smartbell, etc. They look different but somehow they all cost $199 and work with an iPhone.
A big reason for placing an outdoor surveillance camera in the space of a regular doorbell is that the two wires coming out of the wall carry 16-24V AC, from a transformer rated at 10-30VA. Just think of the doorbell buttons that glow yellow. While WiFi provides wireless data connectivity, getting power outside does take an effort. Bringing power from inside requires drilling the wall. An outdoor outlet with a wallwart and a long cable would at best be unsightly and potentially hazardous. Batteries would run out eventually or get destroyed in particularly hot or cold weather, and the unit would have to be removed to install new batteries.
One camera doorbell is actually shipping and was well received – SkyBell. On Indiegogo they raised a hefty $600 thousand two years ago. So, let’s look at what’s inside.
The plastic cover comes off after removing one tiny Allen screw. Not surprisingly, the PCB is round and has a camera setup hanging off it. The main processor is ST Microelectronics STM32F407 in LQFP 100 pin package. This little embedded monster sport a 32-bit ARM Cortex-M4, running up to 168MHz, with 1MB of Flash. Some of the peripheral options include USB and Ethernet MAC, in addition to the usual UART, SPI, I2C.
The STM32F407VGT6 part is around $10.* They could have used a lower end part, with an ARM M3 processor, at 120MHz and no FPU, like STM32F217VG. But in the same package and memory size, the price quote is about the same between the two devices. As to power consumption, SkyBell is running wired. But if power consumption were a concern, these parts are not particularly low power. Full on, at 168MHz the ‘407 takes 93mA and ‘217 at 120MHz takes 61mA. A low power ARM processor Stop mode, that retains SRAM but has peripherals off, can get below 0.5 mA nominal.
Despite having all kinds of power from the doorbell transformer, SkyBell includes a 300mAh Lithium Ion battery (about $1 on Alibaba in quantity). If the idea behind the backup battery was for the device to stay up through a short power outage, the WiFi AP and the router would likely be down during an outage anyway and SkyBell has no local storage.
Included is a Linear Technology LT3990 power regulator, over $3. WiFi is implemented via a leadless QFN package with no clear manufacturer markings. A U.Fl cable connects it to Taoglas FXP74 external antenna, $6.
Video is implemented via OmniVision OV7740 1/5″ CMOS VGA image sensor, which is $5-10 with a lens on Alibaba. Plus Conexant CX93510-11Z VGA JPEG encoder with 128K frame buffer, $6. A flex cable connecting the two says 650nm. The lens is marked ZXZ001.
The PCB is unevenly smeared with some goop, which would offer no protection against the elements. The back plate attaches to the (stucco) wall with two screws. This $199 device is conveniently located within easy reach, to be pried off with a screwdriver and ripped out in two seconds flat. The plastic case with loosely connected lens assembly and the button will not withstand a good punch. The rain can get in from several places in the back, as there is no provision for even a simple moisture seal. I suppose, a good amount of silicon caulking inside and around the rim of the unit could offer some protection.
* Prices are planning estimates, according to Octopart, quantity 10K.