The next method uses one or two discrete transistors to regulate the fan voltage.
In the basic emitter-follower circuit, the voltage at the NPN transistor's emitter is the voltage at its base less about 0.7V, the drop across the base-emitter junction. So a simple potential divider (VR1 & R1) applies 6V-12V to the base of Q1, and about 5.3V-11.3V is supplied to the fan load.
The metal-canned 2N2222A can carry up to 800mA but the loading is ultimately limited by the heat it can dissipate, no more than 500mW. Maximum heating is with the fan running at 6V, the transistor dropping the other 6V, and calculation shows a 2W (167mA) fan is the most powerful it will control.
The E-line ZTX450 and ZTX650 are rated at 1W, so would be good alternatives. They will control fans to around 300mA (3.6W), which covers most 80mm.
The base current is fan current/transistor gain (hfe); with the above transistors, gain is 100 or more so current drawn from the potentiometer is under 3mA; no problem there.
However, using a more powerful transistor to run a more powerful fan is not normally a viable option. The gain will typically be much lower (eg, a 12W 1A BD140 has a gain around 40) so the needed base current very soon gets to a level which will damage the potentiometer track.
One way of increasing the fan power that can be controlled is to use a Darlington power transistor. Now the gain is often 1000 or more, and – with a suitable heatsink – over 50W of heat can be dissipated if necessary.
The disadvantage of the Darlington is that some of the 12V supply is lost over two transistor junctions, not just one. I measured the loss as 1.34V, so my maximum output was only 10.66V (which still beats a 317T circuit).
I used the readily-available TIP122, an NPN power-darlington which will carry up to 5A if fitted with a suitable heat-sink. There are several other TIP alternatives; the TIP110 (2A), TIP120, TIP121 (5A), TIP141, TIP142 (10A), all with gains over 1,000 and all the same B-C-E pin layout (pins down, looking at the label side).
The only other parts needed are a 1k linear potentiometer VR1 (16mm pcb-mounting) and the fixed 0.25W resistor R1.
The 1k R1 shown on the schematic will limit the minimum output to about 4.6V. Alternatively, a 1k2 will just take it down to 5.2V, 1k5 to 5.8V, 1k8 to 6.3V or a 2k2 resistor to about 6.8V, maximum output in all cases, as said, being about 10.6V. (With the +/-20% tolerance on most potentiometers, these minimum voltages are only approximate.)
If, of course, you want to turn the speed right down to zero, R1 can be replaced with a wire link.
There's also room on the board for the 5.04mm (0.2") pitch pcb-mounting connector blocks shown here, which make wiring up easy, or you can just fit wires to a molex power plug and fan fitting.
The stripboard, 9 strips x 12 rows, has just one track break, shown by the red spot between VR1 wiper and Q1 collector. (Remember this is the component-side view.)
Start by making the break – if you don't have the proper track cutter, a few twists by hand with a small (5mm-ish) drill bit does a good job. Then fit the wire links and resistor. Carry on with the taller components – it's easier to mount the transistor on its heatsink, with a thin smear of thermal compound between them, before soldering it to the board. Check the heat-sink doesn't touch any of the wire jumpers and check carefully that the track break is really complete and that there are no solder bridges between tracks. A multimeter or magnifying glass helps.
If controlling a fan or fans over about 2.5W in total a heatsink will be needed, the one shown here is Maplin's RN77J, a 13.5degC/W heat-sink which is plenty powerful but small enough to slide into a drive bay. (More on choosing a heat-sink). Make sure it can't touch any other metal in the case, or sparks will fly (the transistor's metal tab is connected internally to the collector lead, which is connected to the 12V supply).
There's nothing to stop you making up your own Darlington from two ordinary NPN transistors you may already have; Q1 can be any low-power general-purpose, such as a BC107, BC109, 2N3904; Q2 something a bit more meaty like a TIP31 (3A) or TIP41 (6A).
A method that gives the power-handling feature of the Darlington with less voltage loss at full speed is the use of an NPN/PNP pair as shown left.
A low-power NPN transistor Q1 controls a PNP power transistor Q2, and, like the Darlington, the effective gain is the product of the two.
The bulk of the current flows through Q2, and as above, a heatsink is advised.
Following some discussion at SPCR forum on possible weird effects at full speed I did some checks on the original circuit posted here, which was as opposite without R2 in the pot-Q1 base connection. With the potentiometer turned to full and near 12V to the base of Q1, Q2 starts to be shut-down and Q1 draws more current. It's not proved a problem on units I & others have built, but the increased base current could overheat the very last bit of the potentiometer track with some transistor pairings and at highish loads.
Adding a fairly-low-value resistor R2 between the pot wiper and Q1 base drops enough voltage to combat the problem, but it's also advisable to use a moderate-gain power transistor for Q2, such as the 3A TIP32 shown (gain 20), or a BD140 (gain 40), rather than a low-gain one such as the 6A TIP41 (gain 15), and to limit fan load to under 1A (12W). More than ample for the average user.
The method still leaves over 11V to the load with a typical setup. With a single 200mA fan using the parts shown on the above schematic I measured a 0.75V drop at maximum speed, rising to a 1.7V drop when I pushed the envelope a bit with a 1.5A load. Maximum Q1 base current was 0.8mA with the 200mA fan, rising to 3.6mA with the 1.5A load, so sticking to 1A max is recommended.
The circuit can be assembled on a piece of stripboard 16 columns x 9 rows as shown, or a bit smaller if you miss out some of the empty columns (Hint: link j2 will fit in column 8). However, check fits with the Q2 heatsink fitted to ensure it doesn't touch the jumpers; TO-220 tabs are often connected to the collector lead.
Start by making the four track breaks — if you don't have the proper track cutter, a few twists by hand with a small (5mm-ish) drill bit does a good job. Solder up starting with the lowest components (links, then the two resistors) which lets you rest the board upside-down while you solder.
Carefully check the leads on the two transistors before soldering. The Q1 image is the correct way round for a 2N3904 or BC537, but a BC547 and many others will need spinning 180° to suit. TO-18 metal-can transistors like the BC109 fit no problem, as do the flat E-line ones like the ZTX451, but again check the pin-out carefully against a datasheet.
Suitable alternatives for Q1 will be an NPN with a maximum collector current of 100mA or more, and gain 100+, so there are many to pick from apart from the few listed.
For Q2, a PNP with at least 1A rating will allow two or three fans to be connected, so a TIP30 or BD140 are OK; the TIP32 shown has a bit more in reserve and is easy to find.
The only other parts needed are a 1k linear potentiometer VR1 (16mm pcb-mounting), 1k 0.25W resistor R1 (which limits the minimum output to around 5.5V (use 680R to go down to 4V) and the 47R 0.25W resistor R2.
Worst-case heat production (with 12W of fans) is about 3W; a heat-sink is advised for fan wattages over 2.5W, this link gives the sizeing method.
The only downsides I can see to emitter-follower methods of control are
Set against that, they get a bit nearer 12V at full power than the 317T, without going to the expense of low-drop-out regulators.