7
\$\begingroup\$

Working on a power supply design, some of these integrated buck/boost all-in-one dc/dc switching supplies have selectable switching frequency and some can synchronize to an external clock.

What benefit can be found from synchronizing my switching supply to my system clock (or fraction of)?. For references there is only one clocked ic in my device (a DDS chip) and the rest are either asynchronous logic or just basic analog components.

I'm specifically looking at the LTC3115(datasheet) as my master regulator, with daisy chained 79/78XX style linear regulators for the rails.

\$\endgroup\$
9
\$\begingroup\$

If your switching power supply frequency is close to your system clock but not exact, you could get mixing (from nonlinearities), perhaps causing interference in your signal band in analog circuitry. Usually it's the difference ("beat") frequency, not the sum, that can come to haunt you.

Mixing is a nonlinear operation (multiplying) and the sum and difference signals derive directly from the trig identity:

enter image description here

If they're synchronized, usually such problems will be minimized (perhaps a DC offset at worst).

\$\endgroup\$
  • 1
    \$\begingroup\$ +1 See for example this theremin.us/203/203.html schematic of a theremin, in which frequencies ~60-100KHz are heterodyned to audible frequencies. It's quite likely there would be noise from the switch-mode converter somewhere in this range were it not synced to one of the oscillators. \$\endgroup\$ – Chris Johnson Jan 28 '14 at 0:12
  • \$\begingroup\$ Ok, I can see this, one question though. Is supply heterodyning a concern when the two frequencies are close multiples? As in 500KHz vs 5 MHz. My intuition says that the the answer is no, harmonic coupling will exist but the amplitude of the heterodyne will be really small and it will average out to just a DC supply bleed through. \$\endgroup\$ – crasic Jan 28 '14 at 3:36
  • \$\begingroup\$ Fourier analysis might be more reliable than intuition. If you break (say) two square waves down into harmonic content you can predict the amplitude of each component of the mixed signal from the above equation. \$\endgroup\$ – Spehro Pefhany Jan 28 '14 at 3:53
4
\$\begingroup\$

For completeness I would like to note that switching regulators that exhibit a SYNC terminal usually have that capability for another reason. While synchronizing to the MCU clock can be advantageous as described above, this is very rarely needed and is thus seldom done.

The real reason regulators can be synchronized is for the case when you have multiple regulators in your design. Then, by feeding all of them the same clock frequency but with different phase, you can make sure they are not switching at the same time, which reduces input voltage and current ripple, eases input filtering requirements, and reduces EMI phenomena.

Another common reason is simply to make sure that all your regulators are switching at the same frequency, even if in-phase. This in turn allows you to focus your filtering efforts at a much narrower frequency band than if everything did its own thing at a who-knows-what frequency.

\$\endgroup\$
  • \$\begingroup\$ +1 for reduces EMI phenomena \$\endgroup\$ – mike65535 Jan 31 at 17:09
  • \$\begingroup\$ A good example of this is 8 or 16 phase CPU power supplies for computer motherboards. \$\endgroup\$ – MadHatter Jan 31 at 18:34

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.