5
\$\begingroup\$

There have been questions regarding connection of grounds of an isolated DC-DC converter. However I could not find anything that discusses whether the two grounds (when connected) should only be connected at one point.

I have connected the input and output returns of the isolated DC-DC converter using a jumper at a single point, and consequently created two split grounds on the bottom layer (as shown in the image of the layout). Is this advisable? Just to be clear, there are no other analog circuits on this board (ADC, DAC, etc), only the typical components that accompany a DC-DC converter: fuse, TVS diode, filtration caps, and LC filter.

Thank you!

enter image description here

enter image description here

\$\endgroup\$
  • 4
    \$\begingroup\$ I don't think there is any problem with just connecting the planes unless you are concerned about ground currents leaking across the barrier in undesirable places. But the real question is why would you spend the money on an isolated converter when you're just un-isolating it? \$\endgroup\$ – Daniel Jan 11 '16 at 20:46
  • \$\begingroup\$ @Daniel I actually have a pretty good reason (at least as far as I know). My power supply is a Li-Ion battery that supplies 12V-16.8V while one of my boards requires a regulated 12V input. I could not find a non-isolated DC-DC converter that would output regulated 12V when its input is also 12V. On the other hand, it is fairly easy to find isolated DC-DC converters that can output 12V with an input range from 9V-18V or 9V-36V. \$\endgroup\$ – Curious Cat Jan 12 '16 at 13:24
  • 2
    \$\begingroup\$ Ah. If you can find a module, the kind of converter you need is Buck-Boost or SEPIC converter. And be careful about the output of your isolated supply. It may not have as stable of an output as you expect under a range of loads! (ie may require additional regulation) \$\endgroup\$ – Daniel Jan 12 '16 at 17:14
  • \$\begingroup\$ Moved my answers over to an answer, and added some warnings about Buck-Boost converters. \$\endgroup\$ – Daniel Jan 12 '16 at 20:59
  • \$\begingroup\$ @Daniel I removed my previous comment as the Buck-Boost topology would not work for my circuit (as the ground is shared elsewhere). \$\endgroup\$ – Curious Cat Jan 13 '16 at 17:11
3
\$\begingroup\$

It's fine to connect like that, it just means you will have a non-isolated converter.

A single ground-plane should be fine too since this is just a power-converter board. The usual reasons for a split ground plane (or more generally, a star topology of grounds) are where you have different signal-types in different circuit blocks and you want to ensure that ground-currents in one block do not induce a voltage in the ground of another block.

You may find that it's cheaper to buy a non-isolated converter though than to make one from an isolated converter.

\$\endgroup\$
  • 1
    \$\begingroup\$ Yes, you are absolutely right: non-isolated converters are cheaper than the isolated ones, but it is very difficult to find non-isolated converters that will output regulated 12V with input voltage ranging from 12V-16.8V. \$\endgroup\$ – Curious Cat Jan 12 '16 at 13:27
  • 1
    \$\begingroup\$ If you're willing to actually build a switching converter instead of plugging in a potted module (I'm usually not!), then a quick google indicates that an LT3957 or LT3759 can operate a SEPIC converter, which will handle that. I'm sure there are others. \$\endgroup\$ – William Brodie-Tyrrell Jan 12 '16 at 23:38
3
\$\begingroup\$

I don't think there is any problem with just connecting the planes unless you are concerned about ground currents leaking across the barrier in undesirable places.

If you want a non-isolated converter than can convert up or down, the kind you need is a Buck-Boost or SEPIC converter. Note that the Buck-Boost will actually generate a negative voltage with respect to ground!

If that is not a problem for your circuit (it doesn't need to interact with anything else) you can just exchange the -V and GND terminals and get a positive supply. But take care because your new ground is no longer "ground"!

If you decide to stay with the isolated supply, be careful about the output regulation. It may not have as stable of an output as you expect under a range of loads! (ie may require additional regulation)

\$\endgroup\$
  • 2
    \$\begingroup\$ I think a SEPIC has a positive output voltage and that the buck-boost and Cuk are inverted. A Cuk however does not have discontinuous output current, so it's a little easier to filter and achieve EMC compliance. I've no idea on availability of pre-packaged control modules for those topologies though! \$\endgroup\$ – William Brodie-Tyrrell Jan 12 '16 at 23:27
  • \$\begingroup\$ The Buck-Boost converter would not work as the ground is shared elsewhere in the circuit. @WilliamBrodie-Tyrrell There are other complications if I was to go with a SEPIC converter. As I need a 10W output, all the SEPIC chips I found would overheat at my max load (according to my quick thermal calculation scribbles). Also, I could not find a single SEPIC that is capable of 100% duty cycle (case where Vin=Vout=12V); maximum duty cycle numbers for higher end SEPIC converters are 90%. I have decided to stick with the isolated converter as the prototype I have built is operational. \$\endgroup\$ – Curious Cat Jan 13 '16 at 17:29
  • \$\begingroup\$ Sorry to hear that. Sometimes you have to work around practical concerns in annoying ways... \$\endgroup\$ – Daniel Jan 13 '16 at 21:01
  • 1
    \$\begingroup\$ LT3759 allows you to use whatever external FET switch you want, for as much current as you want. A SEPIC can never run at 100% duty-cycle because it has a series capacitor. For a voltage ratio of 1:1, the SEPIC duty cycle will actually be very close to 50%. See page 20 of this datasheet: cds.linear.com/docs/en/datasheet/3759fc.pdf still, that's much more work than plugging in a potted isolated module! \$\endgroup\$ – William Brodie-Tyrrell Jan 14 '16 at 0:00
  • \$\begingroup\$ @WilliamBrodie-Tyrrell That is an excellent chip. However I am debating using that chip as I don't have the means to solder the exposed pad; I could place large vias on the pad in order to solder it from the bottom but I am a bit skeptical since I haven't done it before. After some more searching I found LTC1871 that is quite similar in operation to LT3759 and should be able to handle the output power levels that I need without overheating. As you said, it is definitely more work to setup (but that's the bulk of the fun)! \$\endgroup\$ – Curious Cat Jan 14 '16 at 17:00

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.