I'd like to build a project where I'm essentially controlling analog signal generation in the audio range from a microcontroller. In other words, an analog synth with digital control. I'm at the stage where I've prototyped and tested most of the core circuits on a protoboard, using simply two batteries (+ some caps) for a -9V/+9V split power supply. Now that I'm about to move to layout a PCB, and adding the microcontroller, I'd like to also have a more robust PSU. The requirements are:

  1. +12V, -12V outputs for the analog side (I can basically live anything between 9 to 15 volts, bipolar, actually)
  2. +3.3V for the microcontroller
  3. output current a few hundred mAs. I don't know the exact power requirements yet, but since I'm basically just doing signal processing, I don't expect there will be very much power needed.
  4. Input from a DC wall wart. Wall wart because I'm not a professional, so I don't want to deal with mains voltage, DC simply because they're easier to find than AC wall warts
  5. low noise in the audio range

I'm currently thinking of using something along the lines such as http://www.linear.com/product/LT3471, http://www.ti.com/product/tl497a or similar (suggestions are welcome), more or less just following the datasheet schematics to get the various voltages. I'm asking for general comments, is this a viable route to go, and some specific things:

  • will the switched-mode supplies produce problems with noise? Should I use a linear regulator after the inverters to reduce that noise, or will filter caps be enough?
  • the LT3471 would be nice since it has two outputs, but as I'm going to build this with a soldering iron, I suppose the leadless package would produce problems/be impossible to solder? Is there a similar part in an easier package? (I do have experience soldering SMD, and will need to do that anyway for the uC I've chosen to use, so SMD is okay)
  • what else should I know?
  • Is there an easier way that I'm just missing completely? :) I know about virtual grounds, but as far as I understand, splitting a +24V supply would also not be without its problems, when there's a reasonably complicated circuit to connect behind it. Also, 24V wall warts aren't necessarily easy to find, I'd probably need to cannibalize some laptop supply, etc.

Consider a pre-assembled multiple output DC-DC converter module. I'm guessing you'd be looking for one that outputs +/-15V from 9-18V in? You can probably even find one that has a third output with 3.3V. They're usually not overly cheap, but they do simplify the design process, they at least make claims to meet a particular noise spec, and you can choose one that you're able to hand solder easily - either through-hole or leaded SMT. They're readily available from the big vendors (Digikey claim over 200,000 DC-DC converter modules on their books).

For example, the LT3467 for the +/-15V part, and an LTC1174-3.3 for 3.3V can be used (or LTC1164HV-3.3 for a little bit more flexibility in input voltage), according to the schematics suggested in their datasheets.

  • \$\begingroup\$ That is precisely what I've been looking for. It seems that it's difficult to find what I need though, typically what I find is either in a difficult package, or only outputs maybe 50mA, or something other thing is irreparably wrong. For example the LT3471 would be perfect, except for the package. I suppose I should just look harder? You wouldn't happen to have any specific suggestions? \$\endgroup\$ – Timo Oct 9 '14 at 1:11
  • \$\begingroup\$ Looked up the Digikey catalogue: browsing to DC-DC converters, and filtering for 2 outputs and inverting gives only eight hits, all LT products in a DFN package... so if the solution exists, I don't know where to look for it. \$\endgroup\$ – Timo Oct 9 '14 at 1:19
  • \$\begingroup\$ @Timo digikey.com.au/product-search/… and look at the "Power Supplies - Board Mount" section. \$\endgroup\$ – markt Oct 9 '14 at 2:05
  • \$\begingroup\$ Also, have a look at cds.linear.com/docs/en/datasheet/3467afe.pdf page 12 in particular. \$\endgroup\$ – markt Oct 9 '14 at 2:17
  • \$\begingroup\$ Ok, so I just didn't know where to look for! I was only looking under the IC's section, "PMIC - Voltage Regulators - DC DC Switching Regulators". Somehow just overlooked the power supplies section :) Specifically, the LT3467 seems quite promising. It should be simple enough to add a small DC/DC converter parallel to it to then generate the 3.3V for the uC. I'm gonna look into this a bit more carefully in the evening, and will probably accept your answer. Maybe edit to add the suggestion for LT3467 there, as an example of a specific solution? \$\endgroup\$ – Timo Oct 9 '14 at 12:27

I've had some pretty good success with dual 15V wall-warts in series. That makes 30V total with a center tap. They're already isolated from the mains for safety reasons, so you're not going to short them by doing that. Then you can regulate one down to 3.3V or use a third wall-wart for your uC. If you're particularly concerned about regulation, you can use slightly higher wall-warts with local regulators.

Don't forget local filtering on the PCB's, but it sounds like you know that.

  • \$\begingroup\$ Thanks, that could be a choice to get quickly up and moving with prototyping, but if I ever get to making this something practical, having 2 or 3 wall warts for a single device could be a bit of a pain :) \$\endgroup\$ – Timo Oct 8 '14 at 20:46
  • \$\begingroup\$ For one-off projects, I'm usually okay with multiple wall-warts. But you're right, in somewhat large quantities they can be a pain. If you're okay with separating the power and functional PCB's though, you can use the wall-warts for testing and use the same functional PCB with a purpose-designed power supply later. \$\endgroup\$ – AaronD Oct 8 '14 at 20:53
  • \$\begingroup\$ I'd be more than a little concerned about your wall wart approach if they're earthed. I certainly wouldn't try it in Australia, for example. \$\endgroup\$ – markt Oct 8 '14 at 21:32
  • 1
    \$\begingroup\$ If they're earthed, then you're right, it won't work. Generally though, if they're 2-prong, they're isolated, and if they're 3-prong, they're earthed. I'd be scared of a 2-prong that wasn't isolated, even if it was the sole power source. \$\endgroup\$ – AaronD Oct 8 '14 at 21:44
  • 1
    \$\begingroup\$ Wall wart isolation from mains shouldn't be taken for granted in any country. A whole lot of the world's wall warts come from china, and isolation is expensive. \$\endgroup\$ – Chintalagiri Shashank Oct 9 '14 at 11:08

Although you mentioned needing hundreds of mA, I am going to assume you don't need that much on the negative side.

Generate your +12v using the boost converter you linked to, or similar.

Then use the ICL7660 from intersil to generate -12v. The 7660 is a voltage inverter that takes as input a positive voltage from 1.5v to 12v, and produces as output a negative voltage that mirrors its input. It can output up to 45 mA at -12v.

It is available in a DIP package from Digi-Key for $1.83 (and also various SMD packages).

I have been using this chip (to generate -5v from a +5v supply) in various products without any problems.

  • \$\begingroup\$ This seems like a very relevant suggestion also. I'll admit I haven't really measured the current on the separate rails on the protoboard, nor have I simulated it, so you could be right that the need on the negative rail isn't that much. OTH, most of the analog side is pretty symmetric, in which case one would expect the current draw to be close to symmetric also? Although I'm probably accepting markt's answer above, since the LT3467 seems to match the original requirements better, I'd upvote your answer, but I don't have enough reputation, sorry. \$\endgroup\$ – Timo Oct 9 '14 at 12:31
  • \$\begingroup\$ @Timo I made that statement about the current requirement for the negative rail being much less than the positive from my own past experience, without any information about your circuit; if you have an easy way of measuring the negative rail current I would do so. \$\endgroup\$ – tcrosley Oct 9 '14 at 14:45
  • \$\begingroup\$ Unfortunately I don't, since I've had the various sections built up on a protoboard one or two at a time, so the PCB will be the first time they're all running at once. In hindsight, I should of course have measured the current used by each section while they were up :/ I think I'm just going to overdesign this in terms of the current, and if I end up building more than just one, I can then downgrade the PSU to something simpler/cheaper, once I know the actual power requirements. \$\endgroup\$ – Timo Oct 9 '14 at 14:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.