My main hobby interest is the design and construction of scientific instruments, in particular devices for use in geophysical research. A recent project was the construction of a sensitive seismometer for detecting the microscopic ground motion that results from distant earthquakes. The project was a success and led to a small commercial venture in which I now produce these instruments for amateur seismologists wanting to monitor the shakes and quakes of the tectonic plates that shape our world (ref 1 & photo 2).
The seismometer mechanism is contained inside a clear acrylic cover that protects it from air pressure changes, dust and sudden fluctuations in temperature. The lid is secured with eight knobs, tapped M4 and sealed with an O-ring. While building the first batch of instruments, I soon realised that the seemingly trivial operation of tapping dozens of knobs was taking a disproportionate amount of time and risking a repetitive strain injury. Of course, I could have built or purchased an automatic tapping machine, but neither the time nor the cost required could be justified, given that I was only struggling with making knobs! While repairing our vacuum cleaner I noticed that each threaded hole in the plastic casing had been engineered with shiny brass plugs, tapped to size and incorporated into the moulding. The result is a durable female thread with much greater strength than the parent plastic, and which does not require any manual or machine tapping operation in the factory. This pointed the way towards a solution to my problem.
When I dismantled (and re-assembled!) some other domestic appliances I found that these ‘threaded inserts’ are widely used to create fastenings in items as diverse as mains plugs, TVs, printers and mobile phones (photo 3). More chunky versions are found pressed into modern timber furniture where their purpose is to distribute the stress at important joints. Most inserts are manufactured in brass or steel and designed to be deployed as follows:
Enjoy more Model Engineer reading in the monthly magazine.
Click here to subscribe & save.
- Moulded In, by suspending the insert within the die during the injection process. The domestic electric plug contains an insert incorporated by this process. Such inserts must be closed at one end to prevent flow of molten plastic back into the thread.
- Pressed in Cold, into a suitably sized hole created by or after the moulding or casting process. This is a lower-cost method that does not require modification to the injection mould. Suitable inserts can be open or closed, since blocking flow does not occur.
- Pressed in Hot, into a suitably sized hole cast or drilled into the plastic moulding. This method is termed ‘Heat Staking’ and requires a special tool that heats the insert in order to melt it into the component. Of intermediate cost, this process also permits changes to the insert design without major modification to an existing die. Suitable inserts can again be open or closed at the inward end, but if open, the staking tool needs to block this aperture to prevent flow.
Threaded inserts are not a modern invention but can trace their history back to the days of Bakelite telephones and switchgear in the 1920s, when this revolutionary new plastic proved too weak for tapping holes directly. Since then, inserts have found widespread use in industry but are rarely used in our workshop projects. Trawling the internet for a supplier of small quantities, I came across Anchor Inserts, a British manufacturer with a huge catalogue that extends over a wide range of thread sizes, lengths and materials, with a choice of knurls, ribs and teeth to suit various applications (ref 2). I discussed my problem with Anchor’s technical expert, Austin Wade, who kindly sent a selection of M4 inserts to experiment with (photo 4). The final choice was the 7.9mm long Anchor model BW2015 which could be pressed cold into a 5.7mm diameter hole drilled in the seismometer’s PVC knob using my 1 Tonne arbor press (photo 5). These brass inserts are machined with 4 barbs on the circumference to resist pulling out, combined with a band of axial knurling to resist turning of the insert. Having made the decision, I purchased a pack of 300 and completed a batch of seismometer knobs in record time, changing what was previously a chore into a fun task (photo 6).
Readers of this magazine will be very familiar with the design of Myford and other small bench lathes, equipped with a quick-change toolpost and perhaps even a tailstock turret. In contrast, the specialist lathes used to produce threaded inserts are designed for rapid tool changing and automatic stock feed to permit fast turning, boring, knurling and chamfering to consistent and precise standards. Such specialised machines are unrecognisable to most home engineers and can achieve a rate of production of from 7s to 25s per insert, depending on the part’s size and complexity (photo 7). Making each insert generates similar or slightly more swarf than the weight of the part itself, but all swarf is washed and together with the cutting oil is recovered for recycling. Occasionally Anchor even make special brass inserts with antique threads for the renovation of old Bakelite telephones.
Cold pressing of inserts is only appropriate for compliant thermoplastics, such as PVC or ABS and not for brittle materials such as acrylic, or when the hole to be threaded is near the edge of a part, risking a crack or more serious breakage due to the forces involved. In such cases the process of heat staking that was mentioned earlier is the preferred option. It is important to recognise that this method can only be used for the class of materials termed ‘thermoplastics’ (which can be remitted), as opposed to ‘thermoset’ plastics which are formed from ingredients that react and fuse together during the high temperature moulding process (and which cannot be remelted). It is this type of thermal stability which makes thermosets ideal for moulding kettles, plugs, hairdryers and other hot appliances. As you can imagine, heat staked inserts are more intimately bonded to the host part and therefore can sustain greater loads in a finished assembly. I have a number of projects planned in my workshop where this will be a requirement, and so I decided to design my own heat staking tool. As the project developed it diversified considerably, eventually resulting in two more devices which have proved valuable additions to my workshop. As ever, I have tried to keep the cost to a minimum and to use recycled parts wherever possible.
The Heat Staking Tool
The main purpose of this tool is to hold and then heat the threaded insert to a sufficient temperature that it enters the pilot hole with little force, while plastic flows into the flutes and knurls to complete the bond. Hence, key requirements are the means of raising and monitoring the temperature, and guiding the insert into the hole. I decided to limit my ambitions by using my existing pillar drill to drive the insert, leaving only the problem of heating the insert and measuring the temperature. In this context the figure of interest is the ‘Glass Transition Temperature’ or Tg for the thermoplastic, which is the point at which the material softens to a rubbery state, rather than fully melts to a liquid. Values of Tg for the common engineering plastics which we generally encounter in our workshops are listed in table 1.
Figures for a comprehensive list of plastics are given in ref 3. Before embarking on the construction of your own heat staking tool it is important to note that some plastics give off hazardous fumes when heated and so provision should be made for operation in a well-ventilated space. The risks associated with heating certain plastics are detailed in ref 4.
Glancing at the table above, it can be seen that most thermoplastics we are likely to handle have transition temperatures below about 130°C, with only acetal and polycarbonate requiring a much higher value. However, since these materials are actually chosen for their toughness, they are unlikely to really need a threaded insert fastening, and I therefore decided to exclude these higher transition temperatures in the tool’s specification. So what heat source should we use? A gas burner offers power but lack of fine control, leaving an electrical heater as the ideal candidate. Rather than dismantling my wife’s hair dryer in the search for suitable wire, I decided that the more harmonious approach was to purchase a set of aluminium-bodied power resistors with a high temperature rating and mount these on a solid block of brass to form a compact heat source. Such resistors are flanged for chassis-mounting, are relatively inexpensive, and can be found rated to a temperature of 200 C, which suits this project perfectly. For a supply of V volts, connected to a resistor of value R ohms, the power dissipation is V2/R watts and the current through the component is simply V/R amps. An array of identical resistors each of R ohms, can be wired in combinations of serial and parallel such that the total resistance is still R, while increasing the power dissipation (fig 1). Of course, we still need a power supply able to drive current through the resistor array, and one which must be controllable to maintain a set temperature. A suitable component that can be obtained for zero or very low cost is the ATX power supply found in desktop PCs. Despite their electronic sophistication, even a new 600 Watt ATX unit can be bought for under £20 and they can even be salvaged free from your local amenity site or computer repair shop. I extracted an Enlight ATX unit from a broken Windows 95 computer, reckoning that the 150 watt output would be ample to raise the heat stake to a sufficient temperature (photo 8).
Although different models vary in rated power, all modern ATX supplies conform to an agreed standard with regards to colour coding of the wires and the voltages that they carry. This is to ensure that components such as the motherboard, DVD drive and graphics cards will interface correctly with voltages that match their individual specifications. The typical ATX unit has a bunch of cables that are duplicated and branch to several connectors, and photo 9 shows one of these with the wires labelled accordingly (identical wires run to other connectors).
Amongst the forest of wires, those of particular interest to us are the +12V (yellow), Ground (black), +5V Standby (purple) and Power ON (green). When you press the start button on the front of your computer a micropower watchdog circuit on the motherboard acts to connect the green Standby wire to Ground, activating the +12V (and 3.3V, etc) circuits and starting up the PC. The only exception is the +5V Standby voltage which is always ON, since it is needed to constantly power the watchdog unit. Hence, a fortunate aspect of the ATX design is that it enables the control of a large current supply by simply triggering the Power ON line. Furthermore, this trigger circuit can be supplied from the permanent 5V Standby line which easily delivers sufficient current.
In my design the array of heater resistors is connected between the +12V and ground wires, with the power being varied from about 10% to 95% by a Pulse Width Modulation (PWM) circuit that cycles the ATX on and off at about 3Hz (fig 2). Higher frequencies would be possible but the advantage of this low rate is that the ATX fan can be heard pulsing, providing an audible confirmation of the PWM cycle.
<SPLIT>
The next challenge was to decide on a method for measuring the temperature of the heat stake, using a sensor that could operate from about 20°C to 130°C, or slightly higher to provide an allowance for thermal gradients from the sensor to the tip of the heat stake. This temperature range is somewhat problematic, since it occupies the region between low-cost easy to use thermistors (to about 80°C) and more expensive, harder to interface thermocouples (to 1000°C or more). Platinum resistance sensors of the type used to monitor solar panels are a good option but are expensive both to buy and to build the circuits that display the temperature. Similar limitations apply to the newer technologies of infrared thermometers, while cheap liquid crystal film sensors are not available with high temperature ratings.
Eventually I discovered that the thermal characteristics of a silicon diode could provide the basis for a simple and inexpensive sensor that spanned the temperature range required. When passing a current, these components display a near-linear drop in voltage of 2mV/°C which can be amplified and scaled to give a temperature reading. Common diodes such as the 1N4148 or 1N5817 are less than 3mm in diameter, cost pennies, and can withstand temperatures up to 160°C when packaged in glass or ceramic. My circuit (fig 2) is based on a 1N5817 Schotty diode which happened to be handy, but other silicon diodes should also be suitable, provided you check the temperature rating on the relevant data sheet. My circuit uses an AD623 instrumentation amplifier to multiply the voltage difference between a trimmer potentiometer VR1 and the diode, the output then passing through a 100 microamp moving coil meter in series with a second trimmer, VR2. I removed the panel from the meter and replaced the scale with one marked 0 to 160°C designed and printed from FastCAD. Both the PWM and temperature circuits were assembled on Veroboard (photo 10), and then cleaned in turpentine to remove solder flux which can slowly absorb moisture and create faults. As mentioned earlier, power to both circuits is provided from the ATX +5V Standby supply.
Calibration of the sensor circuit was carried out by attaching the silicon diode to a laboratory mercury thermometer with shrink sleeve, then heating the pair with a hot air gun in steps to increasing temperatures (photo 11). VR1 and VR2 were adjusted after each cycle until readings on the moving coil meter matched those on the glass thermometer, after which the circuit was calibrated and ready to use. If at any time the diode happens to fail then it will be a simple matter of swapping it for another 1N5817 whose electrical characteristics should, theoretically, be identical.
I mounted both circuit boards and the meter inside an old Verobox and attached this to the front of the ATX unit by self-adhesive spacers so as not to impede the cooling airflow. As mentioned earlier, only five lines are needed from the power supply to provide 5V, 12V, Ground and the On/Off line that connects to the PWM control board. For good measure I decided to use pairs of wires for the 12V (yellow) and Ground (black) to ensure adequate current carrying capacity and these were taken with the other two (purple and green) to a terminal block inside the Verobox from where connections were made to the two circuit boards (photo 12). I also added wiring from the 5V supply to a green LED on the front panel to show when ‘all subsystems are in power-on status’, although I stopped short of adding a label with this text! The PWM control potentiometer was fitted to the front with a Bakelite bezel salvaged from a vintage radio. A standard Europlug socket was mounted on the side of the Verobox to supply current to the heat stake, while a pair of colour-coded, 4mm sockets were also added to connect to the temperature sensor. Alternatively, a single 4-way connector with adequate current rating could be used in place of this combination which simply made use of components scrounged from my electrical parts bin. Finally, all the unwanted wires from the ATX supply were pulled back inside the unit, cut short and their ends made safe with layers of shrink sleeve. The final assembly is shown in photo 13.
The heart of the heat stake is an array of four 6.8 Ohm, aluminium-clad power resistors, each rated at 25W and connected in parallel to give an effective resistance of 1.7 Ohms. When the PWM-controlled supply from the ATX unit is at its maximum of 95% (a peak of 11.4V), the power dissipation in the array will be 11.42/1.7 = 76.4 Watts, which I guessed would be sufficient to raise the temperature to at least 130°C. It certainly burnt my fingers during a bench test! In fact the calculated peak power will be an over-estimate due to dissipation in the wires carrying current to the array, and for this reason I used a hefty 4-core cable to minimise such losses.
Fig 3 shows the completed heat stake as a solid model created in Alibre Design (ref 5), both as an opaque assembly and one in which some of the parts are rendered transparent to show internal detail. I found the process of working with a solid modelling program very beneficial, since it provides assurance that all parts will fit correctly, it can reveal internal drillings, and it provides a realistic image that ensures that a design simply ‘looks right’. The downside is the steep learning curve involved with such complex software, although in my experience I found it possible to create useful solid models of several projects within a day or two of starting out.
The heat stake comprises four main components (fig 3 & photo 14). At its heart is a solid brass block of 1 1/8” square section stock, 50mm long, onto which are fixed the four power resistors. A 3mm diameter hole is bored diametrically through this block to provide a cavity to house the silicon diode sensor (photo 15). The lower end of the block is tapered 45 degrees and threaded M6 into which a choice of staking tips can be fitted depending upon the application. Photo 16 shows two tips which I have made: one to fit the Anchor M4 insert mentioned previously, and another for heat-staking plastic pillars into a domed profile. The heated brass block is insulated from the upper part of the assembly by a block of Thermalite foam cement, which Linton Wedlock described in MEW198 as an effective insulator inside his casting furnace. This block is designed to be 50mm square in section and 60mm long, but was initially cut a few millimetres oversize using a wood saw under a steady flow of water from a garden hose. After thorough drying, the block was glued to the aluminium top plate, which had been CNC profiled from 5mm thick plate (photo 17) and roughened on the underside with a coarse file. The adhesive I used was Geocell Quickgrip, a gap-filling multipurpose adhesive available from builders’ merchants which can tolerate temperatures up to 80°C. Tests on a Thermalite sample demonstrated that the bond to aluminium was much stronger than the foam cement itself.
Once the Quickgrip had cured, the top spigot was screwed onto the assembly, enabling the Thermalite to be turned back to the design length of 60mm, taking care to protect the lathe bed from cement dust (photo 18). Next, four 3mm holes for the power and sensor wires were drilled through the Thermalite using those in the top plate as a guide (photo 19). The Thermalite insulator block was then finished to 50mm size and square by rubbing on a smooth brick pavior immersed in a bowl of water, which I found gives a nice finish to this rather friable material (photo 20). The same process was also used to produce the 45° chamfer at the end, judging this angle by eye. Photo 21 shows the finished Thermalite and brass blocks, with M4 stud inserts and matching holes, ready for jointing. This was accomplished by mixing a pure cement-water slurry which was pushed into the Thermalite’s holes using a matchstick, then pressing it onto the studded brass block to form a tight bond. Before this can be done it is important to thoroughly moisten the Thermalite, otherwise strong capillary suction from the foam cement will instantly dry the slurry, making it impossible to bring the parts together and weakening any bond that results. Further slurry was applied with an old credit card to give a smooth rendered top coat. After a week of delayed drying under damp tissue, all cement surfaces were buffed with fine sandpaper and then painted with black stove paint to give a pleasing finish.
In this design it is important to ensure maximum heat transfer to the brass block from the aluminium bodied resistors. This was achieved by lapping the block and resistors with wet and dry paper on my DIY surface table (MEW159), and applying a film of heat-conducting paste before fixing tight with M3 screws. I used a non-curing silicone compound from Rapid Electronics which can apparently increase heat transfer by up to 50% (ref 6).
Two hexagonal hoops of solid copper wire (stripped mains cable) were formed by hand, then soldered to the resistor array to join them electrically in parallel (photo 22). The 1N5817 diode leads were trimmed to length and insulated with silicone tube (glass beads could instead be used) and inserted into the brass heater block. Wires from the 4-core cable were then passed through the Thermalite block and connected to the heaters and sensor, the cable being secured to the top plate tang using epoxy and heat shrink tubing. The completed heat stake is shown mounted in my pillar drill in photo 1 and on the bench in photo 23.
Tests showed that with the ATX at maximum power (10 on the Bakelite dial), the heat stake reached 150°C after about 10 minutes, which easily exceeds my target of about 130 C. The next experiment was to evaluate the unit’s performance in actually heat staking a set of inserts and for this I drilled a pattern of 5.6mm diameter holes in some scrap industrial PVC and fitted the M4 insert tip to the unit, using conductive paste to ensure optimum heat transfer. 90% power was then applied, raising the temperature to an indicated 140°C and the tip of the stake lowered into the first brass insert. After a few seconds, the surrounding PVC became soft and the insert could be emplaced with only light pressure on the quill, proving that the device was working as anticipated (photo 24). This was a very satisfying result, especially since the total build cost of all the equipment was under £20 and made use of a surplus ATX power supply and other scrap components. Since finishing the main project, I soon discovered that this unit has other applications, which include spot welding incompatible plastics, micro-injection moulding and the forming (staking) of plastic rivets. For example, Photo 25 shows a small PVC pillar that has been fused to an aluminium bellcrank using a doming tool mounted in the heat stake. I have used the same tool to melt and mould a small acrylic lens for a 1:12 scale dolls’ house telescope and to create a ‘domed rivet effect’ on a plastic model kit.
The controlled ATX unit can be used to power other useful workshop tools, and in the second part of this article I will describe two other applications which have proved very handy in my workshop.
<SPLIT>
In the previous instalment I described the design and construction of a controlled power supply based on a ATX unit from defunct PC, and which formed the heart of an inexpensive heat staking machine. In my workshop I use this device to mould-in threaded metal inserts, saving the time and fatigue associated with tedious hand tapping. However, I soon found myself thinking of other applications which could make use of this power source and this second article describes two more devices which I hope will be of interest to MEW readers.
I have described how my ATX unit provides a variable power supply with a maximum output of 150W, although this could easily be increased to 600W or more by swapping it for a more hefty unit for little extra cost. I have gone on to use the existing unit to power a simple hot-wire cutter for shaping foam wings for a model glider, the wire in this case being steel trace line from the local angling shop. I imagine this design of controlled supply could also be used as a variable speed DC motor driver.
Workshop Hot Plate
Many of my projects involve encapsulating electronic circuits and bonding various parts with epoxy resins. This winter has already been so cold that my toes have become superconducting, and these resins are taking forever to cure. The answer has been to build a simple hot plate of the type found in chemistry laboratories where they are used to accelerate chemical reactions, although I’m now free to confess that in our school we used one to melt our teacher’s Mars bar!
I am not providing any detailed drawings of the design, since the my unit was cobbled together from a steel box that was once a burglar alarm, power resistors (from Rapid Electronics) and a slab of 3.3mm thick aluminium plate cut from the control panel of a 1960s electron microscope. By the way, if you are fortunate enough to find an electron microscope at your local tip or in your neighbour’s dustbin, then do take it home because it will provide a treasure trove of precision parts!
After repainting the alarm box a subdued shade of grey, the next step was to cut the aluminium plate to a size of 330 x 230mm, allowing a margin to accommodate some U-profile silicone rubber edging that would act as an insulator. This material was unearthed from the scrap box marked ‘Rubber Department’, although I am sure that the door seal from your wife’s fridge freezer would do just as well. The next step was to estimate the power needed to warm the plate to a peak temperature of, say, 60°C which would be sufficient to cure resins, speed chemical reactions, but stop short of setting fire to my workbench. The required thermal calculations could doubtless have been performed using advanced finite-element software on a supercomputer, taking into account convection currents and radiative loss from the aluminium plate and box. However, my BBC micro was not up to that task and so the job was programmed into my on-board analogue computer (i.e. brain) which crunched the numbers and output a guesstimate of 50 to 100W.
The 25W resistors used in the heat stake design are available in a range of low-resistance values and can be connected in parallel or series arrays to produce the required power output, as shown in fig 1 of the previous article. A number of configurations are clearly possible, the key requirement being that the heat output should be uniformly distributed over the top plate, with no hot spots. Hence, use of a single heater resistor was ruled out, and instead I opted to use an array of six, 10 Ohm resistors connected in parallel, giving an effective resistance of 1.7 Ohms which is the same as the parallel array of four 6.8 Ohm resistors used in the heat stake. Consequently, the peak power output will be identical, namely 76.4 Watts, a figure that matched my requirements.
Once again, the base of each resistor was lapped flat with wet and dry paper then fixed in a grid to the underside of the aluminium plate using 6BA brass countersunk screws and nuts (photo 26). A thin film of conductive silicone paste was applied to ensure maximum heat transfer between each resistor and the plate. These resistors were then wired together in parallel and a 1N5817 silicon diode temperature sensor fastened in contact with the underside of the plate by means of a simple metal fixture. The resistor array and sensor were wired up to a 4-core cable (2 conductors for the heater; 2 for the sensor), this cable being terminated once again with a pair of 4mm plugs for the sensor and a Europlug for the heater current. The finished hot plate is shown in photo 27.
Tests showed that when maximum power was delivered from the ATX supply, the hot plate settled to a temperature close to 70°C, which comfortably exceeds my design brief. To provide a check on the surface temperature I obtained two self-adhesive two liquid crystal thermometer strips which were attached to aluminium backing plates (photo 28). These handy strips span the ranges 30 to 60°C and 60 to 90°C and change colour according to temperature (ref 7).
Apart from speeding the cure of resins, the hot plate has proved very useful in several other roles, such as drying out circuit boards after cleaning, and maintaining the temperature of metal blacking and electroplating solutions (photo 29). It was quick and inexpensive to make, and I am sure that other readers will find even more applications for such a device.
ThermoCoaster with Dual-Sensor Technology!
This device was conceived as the solution to the biggest problem faced by us model engineers toiling through the winter months – our tea gets cold! You all know the problem: the Domestic Engineer delivers a steaming brew to the Model Engineer who is in the middle of setting up a job. Once it is completed, however, he is dismayed to find that his mug of life-giving elixir is cold, although her biscuits are still edible! Biscuits with cold tea – what could be worse? Faced with this dilemma, my solution was to use the ATX power supply to heat a flat metal coaster fitted with a diode temperature sensor. But hey, why not go one better and include a second sensor that monitors the actual drink temperature? As the mug is placed on the coaster it will trigger a microswitch that swaps to a sensor that makes contact with the mug’s side. That way I could be absolutely certain that my brew would be at the perfect temperature before raising it to my lips and imbibing. Thus was invented MEW’s world famous ThermoCoaster with Dual-Sensor TechnologyTM!
In keeping with my eco-philosophy, the coaster was mainly built using recycled materials, the total cost of £2 coming from the diode sensor and a single 6.8 Ohm resistor. This will produce a peak output of 19 Watts with the ATX control set to maximum, a figure I guessed sufficient to maintain my tea at a drinkable 60°C. The aluminium coaster plate was cut from the same microscope control panel, and the microswitch was salvaged from an old laser printer. A nice piece of pale oak was removed from a scrapped kitchen cabinet and CNC machined to a curvy form that comprises the main body of the coaster. Photo 30 shows the underside of the coaster plate with the diode sensor, resistor and black microswitch fitted: a small plunger projects through a hole in the plate to operate the switch (and swap sensors) whenever the mug touches down.
The second diode is potted inside a plastic and aluminium module mounted between a pair of spiral cable reliefs supplied by Bulgin to fit their Buccaneer connectors. Wires from the diode to the switch pass down through these spirals which have sufficient springiness to press the sensing module into contact with the drinks mug (photo 31). The complete unit is finished underneath with a plastic cover turned from PVC on the lathe, and is shown being deployed on my workbench in photo 32. The ThermoCoaster works wonderfully, and can heat my tea to scalding temperature, with a red LED flashing to show the input power level. Surely, this is a prime candidate for the most loony project of the year? But that’s no reason for you not to make one!
