Here are a couple of options for generating proper 48V phantom power from a 9V supply. Before we get stuck into the details I should point out that the quiescent current for the first option is about 27mA and for the second about 52mA, so don’t be thinking you can run this from a PP3/1604/6LR61 battery – it will become exhausted in no time. Rather, these are circuits you might consider running from a substantial rechargeable battery pack or a 9V wall wart or stompbox supply.

If you are not familiar with phantom powering, you should read my previous article on the subject here.

The first circuit is a capacitive charge-pump design based on a TC1044S charge pump. I chose the Microchip product over the industry standard 7660 device because it is cheap, it works with a supply voltage up to 12V, and you can double the switching frequency of the original to 20kHz. Here’s the schematic:-


The TC1044S – along with D3, D4, C7 and C8 – forms a conventional charge-pump voltage doubler as figure 8 on page 6 of the Microchip data sheet.  In this case we have added a further six stages of charge pumpage to give a no-load voltage of about 63V at the cathode of D16.  However, as soon as you load up the output of the charge pump, the voltage starts to droop – so we need no-load voltage above 48V and linear regulator to iron out the droop.

The regulator is an LM317AHV (Fairchild) or LM317HV (National Semiconductor), which is the high-voltage version of the ‘317.  Also, the LM317 is a floating design so – given that the LM317 drop-out voltage is about 1.5V – as the input voltage drops below 49.5V the output tracks at approximately 1.5V less than the input voltage. Consequently, if you do run the circuit on batteries then as the input voltage drops below 9V, so the output voltage falls off in step. D3~D16 can be any small signal Schottky diode (the lower the Vf, the better). D1 and D2 can be almost any general rectifier at all, e.g. 1n914, 1n4148, 1n4001~7.  The BAV21 is a beefy 1N4148.

There are some drawbacks to this design.  Firstly, it has significant switching losses.  Whilst the quiescent current for a 1-stage charge pump is about zero, the quiescent current for the above design (bearing in mind that the LM317 has a minimum load current of 3.5mA to maintain output regulation) is 27mA @ 9.2V.  Also when you put a significant load on the output, the conversion efficiency for the load current is about 65%, which is very poor.  Secondly, in order to provide 10mA at 48V (as per IEC61938 spec.), you need an input voltage of at least 9.5V.  Thirdly, it will only do one ‘lot’ of phantom power, and fourthly, it switches at 20kHz, which may be a problem for you if you’re doing high-quality digital audio.

So, cue design number two.  This is straight off the Texas Instruments WEBENCH online design package.  I specified a switcher with 8.5V to 14.5V input, 48V output at 20mA.  Here’s the schematic: –


It is worth noting that Ti have a number of switching devices which are less overkill than the above circuit, but – without exception – all of the devices are only available in tiny surface-mount packages and as such, not Stompville style. So this design has the good old LM2585 at its heart.  The switching frequency is 100kHz, which is good for digital audio sample rates up to 48kHz.  If you’re sampling at 96kHz, good luck to you.  There’s not much to say about the circuit but it is worth noting that you have a wide choice of suitable inductors.  Basically you will likely get away with any inductor that has inductance of 0.56 ~ 1.5 mH providing the inductor has a current rating in excess of 0.5A and a d.c. resistance below about 2 Ohms.  Note that an iron-powder toroidal-cored inductor will have better EMC performance than a ferrite-cored bobbin inductor and is therefore preferable.  I used a Bourns 2124-H-RC from Farnell (1mH, toroidal, 0.4 Ohm, 1.3A), because it was the best-value option they had in stock on the day.  I have a Panasonic ELC12D561E (560 uH ferrite bobbin inductor) to try as an absolute-minimum-spec. part, but I haven’t got round to trying it yet. Also, D1 can be any Schottky diode with a Vrrm of 100V (minimum) and a current rating of 1 A (minimum).

As you would expect, this design works flawlessly, maintaining a steady output of 48V/10mA from a supply between about 3.6 ~ 24V d.c. (You can go higher, but 24V was the highest voltage I had available when I wrote this article).

Here are photo’s of the two designs.  The PCB’s are designed to fit a Hammond 27134PSLA or 1590B stomp box in a similar manner to my re-Stomp re-amping project: –


Note that the LM2585 design has a surface-mount LM2585S on the rear.  This is just about the smallest surface-mount item I am prepared to solder by hand!

Finally, here’s a table comparing the specification and relative performance of the two designs:-



  1.  The TC1044S design has a startup time of several seconds if all the capacitors are discharged.
  2. Input voltage range for IEC61938 compliance.  The TC1044S circuit will run at voltages down to 1.5V. For a 9V d.c. input you can get about 7mA at 48V output.  The LM2585 design will probably accept an input voltage in excess of 24V.
  3. The TC1044S circuit will produce more than 10mA, but the input voltage will need to be increased accordingly to maintain output voltage.  The TC1044S IC has a maximum input voltage of 12V, so running the circuit off an unregulated 12V supply (even a 12V charger supply that outputs 13.8V) is not an option.
  4. The LM2585 design is based around a load current of 20mA. If you want to pull significant current from the LM2585 design, you may need to change the inductor, the smoothing capacitors and the compensation network.  Best to go back to the Ti WEBENCH software and see what you get.
  5. This is calculated by taking the quiescent (no load current) as a baseline and then calculating the efficiency of the conversion of an additional 10mA of output current at 48V.

Two final caveats:

Firstly, the LM317HV data sheet specifies a standard value for R1 of 240 Ohms.  This low value is specified because the LM317 requires a minimum output current of 3.5mA for proper regulation. In this design, R1 is increased to 1k2 Ohm and PR1 and R2 increased accordingly.  This gives a quiescent current through the R1-PR2-R2 network of approximately 1mA at 48V.  However, the LED current is 2.1 mA @Vf 1.7V. On the sample I had, 3.1 mA (i.e. 1 mA + 2.1 mA) quiescent load current was enough to get the regulator working properly.  So, if you build the TC1044S design, don’t omit the LED and if you find that the LM317 is not regulating, reduce the value of R3 to increase the LED current.

Secondly, pay careful attention to the voltage ratings of the electrolytic capacitors as shown on the schematics.  If you over-voltage an electrolytic it will overheat and its useful life will be severely shortened. It may even explode!  svfavicon.png


Leave a Reply

Set your Twitter account name in your settings to use the TwitterBar Section.