Small desk supply

While the original purpose for the supply was a lowish cost replacement for smaller analogue mixing consoles it can also be used for many other uses where +/- 17 volts (other voltages can be set) and +48 volts and 12V is required.

The supply uses a voltage multiplier to generate the 48 volt supply so a readily available 15-0-15 Volt transformer may be used.

While the specifications ultimately do depend on the Transformer used and heat-sinking, the power supply is capable of delivering:

+12V at 1 Amp (lamp supply)

+16.88V at 1.5 Amps

-16.88V at 1.5 Amps

+48.66V at 0.5 Amps

Circuit description

The circuit uses standard run of the mill voltage regulators, with a TL783 for the 48 volt supply, and a LM317 and LM337 for the main positive and negative supply rails.

IC1 and IC2 are the negative and positive voltage regulators the output voltage is determined by R11 and R12 for the negative rail and R21 and R22 for the positive rail using the formula Vout = 1.25 x (1+ Rx2/Rx1)

I went for a fixed rather than adjustable design, for a couple of reasons.

Firstly minor differences between positive and negative rail do not really matter for op-amp based circuitry as it uses the 0v rail as a reference and well designed circuitry in general will assume DC offsets between stages anyway and will compensate for that. Secondly adding a trim-pot does increase setup complexity and one noisy trim-pot can cause major problems or damage to equipment connected to the supply (and yes I have seen damage caused by a dodgy pot in a power supply).

The only unusual circuitry is the voltage multiplier consisting of D1, D2, C5, C6 and C1+2.

If we presume a 30V centre tapped supply, on the first half cycle (I.e. the upper AC input is positive and the lower Negative) C5 charges via D1, up to 42.3V (30V*1.414) on the second half cycle (lower AC input going positive) the negative terminal of C5 goes to 21.21V – now D1 is off but D2 conducts, discharging C5 into C6 so that the positive terminal of C6 charges up to somewhere approaching 63.5 Volts while the multiplier does have a lot of ripple this more than enough to provide our 48 volt rail.

While the TL783 is rated at 750mA I found that in this circuit, ripple and hum reach unacceptable levels above 650mA or so. However 500mA should be more than enough to provide phantom power for 24 channels (while a shorted channel will draw 17mA from the phantom power supply, most devices/microphones draw much less than 10mA, typically 2 or 3mA) If you do require more current from the 48V supply a 10uF decoupling capacitor between the junction of R31 and R32 and Ground will improve ripple rejection of the TL783.

The LM317 and LM337 showed that they could happily provide a rock steady supply, up until they shut down due to over current at over 1.5 Amps.

Squirting noise into your earth.

One thing you need to take care of is ensuring that any earth used for signal references do not have current flowing through them as this current flow will cause a voltage drop across the earth, resulting in noise. This is where the concept of star Earthing comes from, this is where each part of the circuit has a separate track back to a common earth point which is not shared with any other circuitry.

The worst noise current source is the supply decoupling capacitors, ironically intended to reduce noise in the power supply. This is because linear voltage regulators (such as the ones used in this project) are high gain amplifiers which essentially work by comparing the output voltage with a reference voltage. When the output drops below the reference the regulator turns the output on, if the output exceeds the reference the regulator turns the output off.

This means that the output of a voltage regulator can potentially be very noisy and this is what capacitors C11, C21, C31 and C41 are for, to smooth out this noise.

While this noise voltage may only be in the order of millivolts the current through any decoupling capacitors can be in the order of amps which can result in significant noise on any earth tracks particularly is we are dealing with microphone level signals.

The good thing is that as we are talking millivolts of noise on the supply, almost any resistor in series will reduce this noise current to negligible levels which is why most mixing desks decoupling circuits use resistor as well as capacitors, as shown in the diagram on the left.

The circuit fits on a single PCB, with the voltage regulators and indicator LEDs across one edge so that LEDs and Heatsink can be attached to the outside of any case or bracket and connections are on the opposite side of the PCB.
The PCB fits on to a bracket made from a 125mm piece of 50mm by 80mm extruded angle aluminium at least 2.5mm thick, as per the diagram on the following page. over 200mA or so the regulator will need additional heat sinking .

Either a dual 15V transformer or a centre tap 30V transformer can be used, however these are wired differently.

The different options are shown here, note your transformer may not use the same colour code as shown here, so you will need to confirm windings before connecting.

Note that the Centre tap diagram shows the transformer windings labelled 0-15-30 and this seems to be the way most transformers are labelled (according to a quick, very unscientific survey I conducted of Transformers I found lying around), but they may also be marked 15-0-15 where 0 is the centre tap and 15 is the end of the winding.

The Frame ground terminals are provided for convenience, they are otherwise isolated from the circuitry on the PCB. If you use a toroidal transformer I strongly recommend that the Power supply ground is connected to frame ground at some point (such as via the terminals provided on the PCB) as the construction of a toroidal transformers means that their insulation is inadequate for use in double insulated power supplies.

Choosing a Transformer

Your supply Transformer needs to be at least 15V for a 17 Volt supply, the reason for this is that filter capacitors charge up to peak voltage rather than the RMS voltage and the peak voltage in this case is 15 x 1.414 = 21.21 Volts, however then you need to allow for a volt or two of ripple and a couple of volts for the regulator.

Keeping the unregulated DC supply as low as practical, means your voltage regulators will run cooler and require smaller (i.e. cheaper) heat-sinks as power in the output stage is current times the voltage drop across the regulator, so at 1.5 Amps using a 15V transformer, a 17 volt regulated output will result in (21.21-17) x 1.5 = 6 Watts ,whereas a 17 volt transformer the regulator would dissipate (17 x 1.414 – 17) x 1.5 = 10.56 Watts.

The trade off here is, that this assumes your AC supply is reasonably stable. Since the LM317/337 require up to 2.5V across them in order to function reliably at 1.5 Amps (lower current requires less forward voltage) then for 17 Volts output we need to allow for a power supply of at least 19.5 Volts, if we then add another volt to allow for supply ripple we want 20.5 volts on the unregulated side.

This means, that a 15V transformer will work fine, until the AC supply drops by as little as 5%. In Australia the AC supply is specified as 230V +10% / -6% so going by worst case conditions a 15V transformer will be cutting things a bit fine, whereas a 16V transformer would allow for a comfortable 10% sag in power.