There are four basic ways of getting controlled power to the fan motor using semi-conductors – adjustable linear voltage regulators, transistor emitter-follower systems, pulse width modulation (PWM) control and switched-mode power supplies (SMPS).
The Other Voltages page showed how a fixed voltage regulator like the LM7805 could be rigged to give a variable output, and this is just what Zalman do in their little Fanmate II controller. This regulator sets the minimum output at 5V, maximum is about 1.5V below the ingoing supply level. Maximum load will depend on the size of heatsink fitted up to the chip's 1A limit. The Fanmate only has a small sink, limiting loads to 0.5A (6W) so you may want to build a more powerful version and/or make one more suited to panel mounting.
The circuit is straight from the Texas Instruments datasheet and I've included the recommended small capacitors – these should be fitted near to the regulator input and output pins to prevent it from oscillating under some circumstances. The Zalman circuit doesn't bother with them, but it only saves a few pence.
Output voltage is given by the formula
where Iq is typically about 4mA (0.004A) and VR1' is that fraction of VR1 in use, so with R1 = 1k and VR1 = 1k, the maximum theoretical output of over 12V allows for some adverse tolerance on the resistors and regulator. As said, you'll be lucky to get over 10.5V from a 12V supply.
I've used a piece of stripboard 9 strips wide, 15 holes long which nicely suited the screw-terminal blocks and heatsink used, a Maplins RN77J rated at 13.5°C/W. Bolt the regulator to the sink with a thin smear of thermal compound in between before assembly. With a different sink do a dry-run first to check everything fits.
The heatsink will be electrically connected to the regulator's 'common' pin so must not touch the links or other case metal. No track breaks are needed.
Capacitors with 5mm pitch leads are a perfect fit; the 330nF may be hard to find (Rapid Electronics have them, 10-3266) but you can use a 220nF or 470nF instead if necessary (which Maplin have in multilayer polyester film, DT99H or DS80B).
The terminal blocks I get from Rapid; Maplin sell similar 2-way 5.08mm pitch but theirs are much more expensive. Or it may suit you to solder wires direct to the board and take them to a molex supply and fan connector.
Check the finished board for any errors or solder bridges between tracks, fire up and test.
|Rg1||LM7805 5V regulator|
|C1||330nF (0.33uF) ceramic or film.|
|C2||100nF (0.1uF) ceramic or film|
|R1||1k 0.25W 5% or better|
|VR1||1k lin, 16mm PCB-mounting potentiometer|
|Misc||TO220 heat-sink & paste, connectors, control knob to suit.|
The more commonly used (but slightly more expensive) adjustable regulators such as the LM317 or L200 are essentially similar, the LM317 being at heart a 1.25V regulator, the L200 a 2.77V device.
Their advantages over the 7805 include higher current handling (up to 1.5 amps with a 317T, 2A with the L200), a much lower quiescent current Iq, and a lower minimum voltage, though that's of little interest in a fan controller.
Most published regulator circuits give an output down to the minimum possible – with fan control there's little point in going below about 6V, as (a) the fan may not start running at lower voltages, and (b) you're wasting a lot of potentiometer track that could be used for better sensitivity. Adding a fixed resistor in series with the pot achieves this.
The LM317 has a Reference Voltage of 1.25V between the output and adjustment pins, and with the circuit shown, in similar vein to the previous example,
As said, the Iq value is very much lower (typically 50uA = .00005A) and can usually be ignored in calculating the output.
With the values shown in the circuit above, the calculated output will be from 6.5V up to 11.7V, ideal for fans. The resistor combinations for other output voltages can be calculated with this 317 Calculator; for example, making R1 = 200R and R2 = 680R will give a 5.5V–11.8V range with the 1k pot.
As pointed out earlier, some volts are lost across the regulator transistor junctions, so from a 12V input the output is limited to around 10v in any case (10.25V on mine, with a 0.2A fan). I've designed for 11.7V because potentiometers have a pretty broad tolerance range (±20% is common), so there's a bit of leeway if the actual pot resistance is on the low side of nominal.
|C1||100n (0.1uF) 16V ceramic or mylar film.|
|C2||1uF 16V tantalum bead|
|R1||240R 0.25W 5% or better|
|R2||1k 0.25W 5% or better|
|VR1||1k lin, 16mm pcb mounting potentiometer|
|Misc||TO220 heat-sink & paste, control knob|
The circuit was built on a piece of veroboard, 7 strips wide, 15 holes long. Two track cuts as shown by the red spots.
A 0.25" (6mm) drill bit can be used to make these, a few turns by hand will remove enough copper to break the track.
I used a slim conductive polymer potentiometer for VR1 but a 16mm PCB-mounting carbon type will fit; the link (and resistor, if necessary) will go under the pot barrel.
The 317T requires a heat-sink – a 25°C/W type will run fans rated at up to about 10W (at 12V) as at the worst case it will be dropping 6V at 0.42A, i.e. 2.5W.
I like the Maplin RN77J type shown on the photo at the bottom of the page, they're rated at 13.5°C/W but very neat (only 20mm tall), or their clip-on KU50E is cheaper and at 25°C/W is suitable for a lot of fannage. (More on choosing a heat-sink)
Fit the heatsink and dry-assemble the other components to check there's no interference, then cut the tracks and start soldering, lowest components (jumper links, then R2, etc). If there is any danger of the heatsink touching the j1 link it can be moved to columns 2, 4 or 14.
Check the polarity of the tantalum electrolytic C2 before soldering – there's usually a plus sign (+) and a line by the positive lead.
One warning – the tab of the 317T is connected to the regulator's output pin, so don't allow the heat-sink to touch any grounded metal case parts, circuit-board links, etc, or a short-circuit will occur.
Photo shows an earlier design to use up some rather long heat-sinks – I had to bend the regulator legs to fit a drive bay, the Mark#2 is simpler to build and neater.
A major drawback of the 317T and similar standard regulators is that 1.7–2 volts are lost from the 12V supply, so the fan can only run up to about 85% of its full speed and flow rating. Although this site is aimed at keeping noise down, there may be occasions when you want to turn the wick right up.
You can get much nearer the 12V input with the newer low-dropout regulators, and with the controller shown below my voltage loss running a 200mA (2.4W) fan was only 130mV (12V in, 11.87V out, 99% full speed!).
They cost a fair bit more than the humble 317T – Rapid Electronics price for a MIC29152 is £3.10 compared to 38p for a 317T – but when you factor in the other parts, especially with a fancy control knob, it's not such a big difference overall, and well worth it IMHO.
I've used Micrel's MIC29152BT regulator, which is rated at 1.5A and comes in a 5-pin TO-220 case. If you need more current, the MIC29302BT is identical but rated to 3A, or the similar MIC29502BT will carry 5A. You may need a bigger heat-sink with the bigger boys. See the heat-sink page for the details.
Unlike the 317, the Reference Voltage is across R2, not R1, and the calculated voltage output is
With the resistors shown, the calculated output range works out at 6.9V to 12.5V, giving a bit of headroom if component tolerances run the wrong way. (The ±20% tolerance band on pots can play havoc with calculations – if your "10k" pot is below about 9k you won't get the full potential. The ratio of (R1+VR1):R2 must be over 8.7:1 – increase R1 to 12k or reduce R2 to 2k)
The two capacitors ensure stability, particularly useful with long supply leads. Some alternative resistors are given below to suit different minimum voltages.
|C1||100n (0.1uF) 16V ceramic|
|C2||10uF 16V aluminium electrolytic|
|R1||10k 0.25W 5% or better (see below)|
|R2||2k2 0.25W 5% or better (see below)|
|VR1||10k lin pot, 16mm pcb mounting (see below)|
|Misc||TO220 heat-sink (20–25degC/W) & paste, control knob|
The stripboard is 7 rows x 14 columns, two track breaks as shown in red. Bend the regulator leads carefully. I found the best way was to support the outer legs as near the case as possible with a thumb-nail and bend outwards slightly, then support again on the other side (I use a dart tip, also very useful for curling resistor or diode leads for vertical mounting) and bend in to fit. The 2 & 4 leads need less bend, again support with a dart or fine screwdriver. Test for fit in the stripboard and fasten on the heatsink.
|Output Range with various resistors|
|R1||VR1||R2||Min output||Max output|
Dry-run assemble the components to ensure nothing fouls on the potentiometer or heat-sink, then make the track breaks and solder up, lowest components (links) first, finishing with the regulator. Check carefully, especially for any solder bridges; a multimeter with buzzer continuity check is handy for this, or a magnifying glass.
You can use a range of other resistors in the R1–VR1–R2 chain to give different minimum outputs, or to suit available potentiometers; some examples are given in the table right, or you can work your own values out using the above formula. It's no problem using 100k or 470k pots, just multiply each resistor by 10. For the minimum output, 1.24V, use a wire link instead of R1.
Of course, you won't get over 12V output from a 12v supply, but the headroom allows for some component tolerance.
This regulator can also be used with high-value NTC thermistors in a thermal controller, using the thermistor in place of R2 and a suitable pre-set potentiometer for VR1. For a 100k @25°C thermistor (Maplin CR05F), R1+VR1 set to 470k would give outputs of 5.8V at 20°C, 7V at 25°C, 9V at 31°C and full-speed at 37°C, so R1=430k, VR1=100k would be on the button for fine tuning.
If you want to check different thermistors, there's an Excel spreadsheet download for the L200 regulator that can be adapted, with the complex formula shown here incorporated.
There's also a MIC2941ABT low drop-out regulator, rated at 1.25A and a fair bit cheaper (£1.97), but the pin-out is totally different, as shown right, so it won't work in the above layout. It also has no big brothers as alternatives, though 1.25A is adequate for most fan-control purposes.
Otherwise the circuit is virtually the same, the only difference being that the 2941 "Shutdown" pin is connected to ground but the 29152 "Enable" pin is connected to Vin for normal "On" operation.
The resistor values used above will give the same results, within a few millivolts.
The 220nF (0.22uF) capacitor C1 is recommended if supply leads are over 10" long (which covers every PSU I've seen) and C2 can be any value over 10uF, aluminium or tantalum.
Other needed parts are as the above list, and again the alternatives will work as detailed.
Rapid Electronics and RS Components stock all the Micrel LDO regulators mentioned, Maplin don't do any.
(to be continued)