The clock prescaler takes the 60Hz square wave signal from the second power supply comparator and divides it into a 1Hz signal accurate enough to drive the clock. There are two parts – firstly dividing 60Hz by 10 to get 6Hz, then dividing that by 6 to arrive at the 1Hz square wave.
Given I’m in Australia (where 50Hz from the wall is the norm) I’ll need to make some modifications to this part once the clock is functional. For now, I’m going to make the assumption that on 50Hz my clock will run 17% slower than expected. 50Hz / 10 / 5 = 0.83Hz = 17% slower.
How to build it
The prescaler portion of the transistor clock contains 153 components:
41 x ’1N4148′ diodes – These are oddly just referred to as ‘small diodes’ in the manual – no idea why the component name isn’t mentioned.
282 x ’2N3904′ – NPN
2 x ’2N3906′ – PNP
21 x ’221′ – 220 pF
2 x ’104′ – 0.1 mF
31 x 10kΩ (brown-black-orange-gold)
28 x 100kΩ (brown-black-yellow-gold)
Here’s the process I used:
Brought a plastic bending tool. I don’t suggest even attempting to try to build this thing without one of these – it’ll look terrible.
Cut the required number of components from each larger bag and double checked they were correct.
Cut the components from the paper strips. Do not try to just pull the components out – some of them such as diodes are fragile. Do not cut the PNP transistors until you’re ready to insert them – they are easily mistaken for NPN.
Bent the 1N4148 diodes using my bending tool – they are quite small so I cut my own notch in the side of the tool to match the PCB spacing.
Soldered the diodes in place. Diodes are polarised (only work one way) so it’s important they are aligned correctly. One thing which may not be immediately obvious (and isn’t mentioned in the manual) is that the square PCB pads are for the cathodes – the side with the line on it.
Soldered the resistors in place. If you’ve bent them correctly (and consistently) they should stay in place once the board is flipped over – this saves an entire step of having to bend the terminals to make them stay put.
Cut and soldered the PNP transistors – there are just two in the prescaler so make sure to place them correctly.
Soldered the NPN transistors. With the transistors I found it easier to solder the middle terminal first, flipping the board and them aligning them correctly, before soldering the other terminals in place.
Soldered the capacitors in place.
Cut off the spare component legs. Get a decent pair of sharp pliers for this step. You’ll have the excess component legs bloody everywhere after this step.
Align all the components correctly – make it neat!
I then powered up the board and no magic smoke was released. This at least means I don’t have any shorts. If I had an oscilloscope, I’d be able to see a 1Hz (0.83Hz) square wave on the output of the top right of the prescaler. You can see the trace fairly easily where the circuit jumps into the seconds chain. If I measure the voltage on this point however I can see it swing from negative to positive, roughly every second – this is the best I can do until I get one of these.
When you’re at this stage of construction, be careful not to store the board ‘face up’ with the zip-tie underneath. My board developed a noticeable warp from one corner to the other – which I’m hoping will fix itself over time.
The seconds chain is next. I hope to get faster at soldering, and also looking for a way to avoid the metal leads flying everywhere.
The first step to building the transistor clock is sorting the power supply stage. There aren’t too many components in this stage and can be built in around 30 minutes. I have a component bending tool which made installation a lot quicker.
The clock uses a 60Hz signal from the 9V AC 1A power supply, however in Australia we use 50Hz. This is crucially important to the operation of the clock and I will be modifying the prescaler section to divide by 5 instead of 6.
I was happy to discover there were no fires or blown components on applying 9V power to the clock. The capacitor measured 14.3V across its terminals which is good enough to continue onto the next step – the prescaler.
With the October (LAN) long weekend approaching fast, I thought it a good time for a PC rebuild. There are a few things I’ve wanted to do with this rig since I purchased it. My build has the following parts:
Intel Core i7-3820 CPU @ 3.6GHz
Gigabte GA-X79-UD5 Motherboard
32GB Corsair Vengeance DDR3 1600MHz RAM
1200W Corsair AX1200 Power Supply
1GB Gigabyte Radeon 6850 Graphics Card
Intel 120GB 330 SSD
3 x 4TB (12TB) Seagate Barracuda ST4000DM001
8 x 3TB (24TB) Seagate Barracuda ST3000DM001
A few smaller 2TB and 1.5TB drives
The current build works well performance wise with a WEI of 7.3 (if that still means anything). It just needs some love on the construction side – it’s been hastily thrown together with no real thought on cable management or cooling.
24 hour filth
I never turn my machine off. Ever. Even though I’m a clean (non-smoking) person, thousands of hours of constant use takes its toll. You can also see the terrible job I’ve done on the cabling.
Now that all the dust is gone, I can start the wiring.
The quad I have is based on the QAV540G (G == gimbal) so comes with all the gimbal parts as part of the kit. Unfortunately I won’t have control over Yaw with this gimbal as it moves in 2 axes only, but it’s a start.
Whilst not particularly difficult, it can take some time to get the parts of the gimbal aligned correctly. Make sure the motors are mounted corrected and all parts are oriented as per the manual. I must have put it together at least three times incorrectly before getting all the parts to line up. I haven’t worried too much about the wiring for the motors just yet.
Attaching the gimbal to the frame
This was relatively straightforward. Below you can see my temporary motor connections to the controller (soldered). I’ll be adding some plugs and socket shortly. It doesn’t matter which order the wires at attached, but I just made sure the middle wire on the motor matched the middle solder join on the controller.
After applying some power the gimbal just needed a few tweaks in SimpleBGC to get it to move correctly. I’ve adjusted the power for the motors to ’100′ (out of 255) which seems to give them enough torque for now without them getting too hot. The IMU provides real-time feedback of its position back to the controller, then moves the motors in the opposite direction to keep them level. You can see the first test run below.
Ironically, YouTube has detected that my ‘video may be shaky.’ Still needs some work I think.
The gimbal build portion of the quad has been severely delayed due to technical problems with the hardware. I purchased two AlexMos Brushless Gimbal Controller with IMUs from GetFPV but they both worked terribly. Flimsy wiring, along with persistent IO errors meant they both went back for replacements. Six weeks later the replacements arrived and they worked much better. Below you can see the controller in action (green is good) along with the tiny Inertial Measurement Unit on the top left. The IMU will be mounted to the actual gimbal and provide inertia tracking and stabilisation.
I really can’t recommend the 8-bit AlexMOS boards based on my experiences – whilst they seem well-built I’ve a huge number of problems getting them to work. Go the 32-bit version instead with the added benefit of a third yaw axis.
I mounted the controller and knocked together a custom power cable to connect it to the dirty frame. I made some modifications to the plug and socket to make them sit flush on the frame. I probably shouldn’t have used a plug as it’s one more thing that can break apart in the air, but it’s not essential to safe flight so I think it’s OK.
With the controller board mounted, powered and at not throwing ten thousand IO errors per second, next up will be construction of the gimbal itself and mounting of the IMU.
On my quadcopter, I’ll be using the OpenPilot CC3D Flight Controller from GetFPV. I also considered the OpenPilot Revo, however I’m not 100% keen on waiting forever for the GPS module to be released. I’d like to upgrade to something from DJI a little later but for now the CC3D fits the bill.
The biggest challenge with mounting the CC3D onto the QAV540G clean frame is finding a way to deal with all the extra wiring. Not wanting to simple bunch them all up, I did my best to make it as neat as possible, which can be seen below. The ESCs are connected as follows:
Front left – channel one (‘front’ as in the front of the quadcopter)
Front right – channel two
Back right – channel three
Back left – channel four
It’s not hugely important which ESC is connected to each channel but it makes calibration easier as channels do not need to be re-assigned manually within the CC3D calibration software.
I’ve also disconnected the positive and negative wires from channels two to four as the controller only needs power from one channel (channel one in this case).
The bundle of wires with the toroids (the green things) needs more work to make it neater and possibly reduce the length a little bit, but for now it’s good enough for me. I should probably get a protective housing for the controller as well as some stage.
With the four frame legs attached to the dirty frame base plate, it’s time to install the ESCs (Electronic Speed Controllers). The ESCs control the speed and direction of the motors and provide a solid interface for the flight controller.
Adding length to the ESC motor cables
The default cables that came with my ESCs wasn’t long enough to be useful. I wanted the cable to make its way at least through the aluminium arms, and also have the ability to modify the plug/socket configuration without pulling the dirty frame apart. My first idea was to cut open the heat-shrink and unsolder the wires directly. I would have had to put a lot of heat into the solder joints to remove them and though maybe this wasn’t a good idea. The last thing I wanted was to fry an ESC with excess heat. I decided against this method and re-added some fresh heat-shrink material. I needed a new plan.
The right way
With the ESCs safely shrouded in heat-shrink material, I brought some similar gauge wire to add to the length. In my view there is really only one way to extend wires, that is:
Use an identical gauge and color of wire
Strip 10mm off of each end of wire – do not ‘twist’ the ends just yet. If you don’t have a proper wire stripping tool, stop now and go buy one. It’ll set you back around $20-30 for a decent one.
Push the wires ‘into’ each other so their individual strands intertwine and overlap entirely.
Twist the wires together in opposite to directions to make sure they won’t come apart for the next step.
Solder them together, making sure you apply enough heat for the solder to really penetrate into the wire.
Apply heat-shrink. Double check there are no stray wire strands piercing through which could short on adjacent wires.
Attaching the ESCs to the frame
With the motor wires on the ESC extended, we can now look at soldering the power wires onto the dirty frame PCB. I test fitted the ESCs into the exact position I wanted them then shortened the red and black power wires just enough. Be sure to make sure the the + and – pads correspond to the correct wires as it’s easy to get them the wrong way around. Add solder to both the pad and the wire before applying enough heat to solder them together.
After repeating this four times I attached the dirty frame top plate and another round of Loctite. I’m done with the ESCs. Next up, I’ll be installing the motors.