The PCBs have arrived back from Chiltern Circuits. There are 8 panels, each containing 28 printed circuit boards.
I’ve made a minor change to the circuit too. The IRFZ44N MOSFET has been replaced by an IRF3205. The new transistor has a lower ‘on resistance’ so a little less power is consumed within the device. The Z44N has an on resistance of 17 milliohms (0.017 ohms), but the 3205 is less than half that at just 8 milliohms (0.008 ohms).
The IRF3205 does have a correspondingly higher gate capacitance (about 3.2 nanofarads), so I’ve reduced the gate resistance from 10k ohms to 4k7. Tests confirm that this has little or no effect on the charge pump circuit – the gate voltage is still over 20 volts relative to ground.
The PWM5 is getting a new PCB. Functionally it’s identical to the stripboard layout and indeed many of the strips have been retained and the pads are mostly laid out in a matrix. Two wire links have been replaced by printed tracks so assembly time will be reduced. A silk screen component layout will be printed on the top surface to help identify which components go where.
In a couple of positions I’ve put additional pads so I can experiment with surface mount devices (SMDs) as an alternative to through hole components. Chiltern Circuits, my local PCB company are making the boards, 224 of them in this first production run. While I’m using them up, I’ll work on the design possibly replacing more of the components with SMDs.
There’s nothing fundamentally wrong with a stripboard PCB – it’s neat, reliable and easily available off-the-shelf. The problem is the amount of time it takes to construct.
First, the 32mm by 36mm piece has to be cut from a larger sheet. The scoring jigs help, but a lot of further cutting is required using a pair of modelling knives. Then the edges have to be sanded down and the track breaks cut using a finger drill bit.
There are 7 resistors, 7 capacitors, 6 diodes, 4 transistors, 2 ICs, an LED and 3 wire links to be soldered in place. The board has no markings to indicate where the components go, so it’s all done from memory.
So it’s time to bite the bullet and design a printed circuit board which can be produced commercially. It need only be a relatively simple, single-sided affair but with the added sophistication of a top silk screen layer indicating where each component goes. This will allow the assembly work to be carried out by a third party.
Once this has been successfully completed, another re-design may be contemplated. That would be to replace many of the parts with surface mount components. These could be machine assembled using a pick and place machine and the remaining through hole components added by hand.
A commercially produced board would of course be more expensive than stripboard, but this has to be offset against the time saved assembling it.
Microchip’s PICkit3 has made programming the 12f683 a great deal quicker and easier. Until recently I was using an old PICkit1 which was slow, had no ZIF socket and required that I run up MPLAB every time I wanted to program a batch of chips.
The PICkit3′s programmer-to-go function is extremely handy. The hex code can be left resident inside the programmer so that no host computer is needed to program a batch of controllers. I just put 5 volts into the mini USB socket, connect up the ICSP connector and push the button.
The 12f683 programs in less than 2 seconds, and by using a ZIF socket, switching chips takes just a few seconds more. Anything which speeds up production is a bonus and the PICkit3 certainly helps shave a few seconds off the assembly time.
Getting a consistent fit between the inner and outer pieces of heatshrink sleeving requires a bizarre production technique involving a kettle of boiling water and a vacuum cleaner crevice tool. Pre-shrinking the sleeving is the only way to ensure the LED lines up with the hole in the outer covering during final assembly.
Firstly the black outer heatshrink sleeves are slid over the crevice tool and dunked into boiling water for a few seconds. Then they’re washed to remove lime scale and dried with a micro fibre towel. The clear inner sleeves follow in a similar manner.
The kettle is nothing special, although the current fashion for wide bases and narrow tops meant doing a bit of searching around. Fortunately cheap kettles aren’t fashionable and a suitable model cost less than a fiver.
The crevice tool was more of a challenge. It had to have a round-ended rectangular cross section and a very slight taper. eBay to the rescue on this one but not before I’d amassed quite a collection of unsuitably shaped vacuum cleaner accessories. The white plastic rectangle on the crevice tool serves as a release mechanism.
Unlike most of my off-grid contemporaries, my storage bank is a rag-bag assortment of old car batteries. Not for me the neat shiny row of brand new, deep-cycle, perfectly matched accumulators. Mine came from old cars, were donated by friends and even found left in a car park. They all work though, more or less.
And that’s the problem, some work better than others. Testing them though was a problem. Attaching a 55 watt headlight bulb shows that the battery will hold a charge and deliver a few amps of current. But to check the capacity, the bulb needs to be left there, certainly for hours, possibly for days and constantly monitored. Impractical, to say the least.
So step in the Sealey BT102 Battery Tester. This is a convenient little device which gives a very quick and apparently accurate indication of the condition of the battery.
Start by connecting the battery clips to the battery terminals and the 4 digit LED display initially shows the battery voltage accurate to one hundredth of a volt. Press enter and you can input the battery type and Cold Cranking Amps (CCA) in either SAE, DIN, EN, IEC or CA. The test runs and instantly shows the actual measured CCA and a good/bad assessment on red, yellow and green LEDs. By comparing the measured CCA to the value printed on the battery label you get an idea whether the battery should be kept or discarded (recycled).
Nothing is said in the instructions about how the tester actually works, but I get the sense that the device has been well designed and I’m inclined to trust it. An LCD rather than LED display would have better as the display is difficult to read in bright daylight, but this is a minor gripe really.
So I’ve used the BT102 to weed out a couple of poorly performing batteries and satisfy myself that the rest are still usable.
The tester has no internal battery as it uses the battery under test to power the electronics.
There’s something rather satisfying about turning harvested sunlight back into its native form, and an LED torch is one of the best ways to do this.
I’ve standardised mostly on AA rechargeables to store energy outside of my lead-acid battery bank, so my choice of torch reflects this.
The LED Lenser M5 torch uses a single AA cell, stepping the voltage up internally to power the LED. This works well, although as the cell becomes discharged, the microcontroller functions can become erratic.
The M5 is exquisitely small. It’s about as small as it could be, given that it has to house an AA battery, a microcontroller switch and a LED with focusable lens. It’s substantially smaller than LED Lenser’s other 5 series (single AA) torches; the P5, T5 and the bulky L5.
The variable focus lens is a joy to use. It slides forwards a few millimeters to focus the beam into a bright square spot. The square shape is a reasonably well focussed image of the LED itself, in fact you can almost make out the connecting wires. Slide it back for a large clean circular flood beam of light, ideal for finding your way round the house at night.
The microcontroller functions are sensibly simple. There are 3 modes; bright, dim and strobe, the latter useful for dazzling hostile assailants (particularly those with photosensitive epilepsy). Dim is similar to strobe except that the rate of switching is faster – too fast for the eye to see. When you wave the torch from side to side in dim mode you can see the switching process quite easily. In bright mode, the LED pumps out 108 lumens, apparently.
So how bright is 108 lumens? It’s not something that can be easily described… Certainly do throw away that nasty old torch which uses a 3 volt bulb and a couple of C cells, but don’t gouge out you car’s headlights and replace them with LED Lenser M5s.
The battery lasts for an hour or two in bright mode and around 4 to 6 hours in dim.
This is probably my favourite LED torch. It’s bright enough for most tasks, easy to operate and requires just the one AA cell. It’s not the brightest LED torch available, but a sensible compromise between brightness and running time has been chosen, I think.
When the four lengths of red, yellow and black wire come off their reels, they have a considerable curl in them. I needed to find a way to lay them straight for a few days before production – it makes soldering them to the PCB a lot easier.
The solution is a set of panels of Corex, a fluted plastic board used to make estate agent ‘for sale’ signs. The wires lie straight, the curl is removed.
Click the image for a closer look.
So far, around 300 controllers have sold to locations around the globe. I’m keeping a list of countries that I’ve posted to. Here it is.
- United Kingdom
- Ireland
- Sweden
- Finland
- Denmark
- Latvia
- Netherlands
- Belgium
- Bulgaria
- France
- Switzerland
- Spain
- Portugal
- Italy
- Malta
- Malaysia
- Singapore
- Philippines
- Indonesia
- Australia
- Canada
- United States
and now…
- Greece
- Tahiti
- Austria
- Turkey
- Germany
- New Zealand
The (not so) Elusive £1 per Watt
Well now it seems we’re not far away from that goal. On eBay recently I spotted my favourite size solar panel, the 80 watt at just £89.99 plus carriage. Perhaps it’s time to grow my system.