EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: xaxaxa on August 17, 2018, 05:58:42 am
-
I have been using the AP3419/AP3429 sot23-5/sot23-6 buck converters for generating the 3.3V/1.8V rails in some of my designs and have started noticing significant sawtooth shaped noise on the power rails that are lightly loaded due to the pulse skipping feature. The noise is as low as 8kHz and can't realistically be filtered out. I'm looking for a buck IC with similar specs (1A to 2A output, 5V in, 1-3.3V out) in a similar form factor that DON'T have pulse skipping. Mouser/digikey isn't turning up anything.
Before you say use a linear regulator, the issue is the load on the rail can be unpredictable and wasting some power isn't an option because everything I do is USB powered and I'm very near the power budget.
-
Is this just a niggle, or do you have test data or design parameters showing that it will definitely cause a problem, say with ADC stability or something?
Tim
-
Would an LC filter on the output be sufficient? Maybe more parts, but then you can keep that efficiency with the lighter loads, and you don't have to do much of a redesign.
http://www.analog.com/en/technical-articles/switching-regulator-noise-reduction-with-an-lc-filter.html (http://www.analog.com/en/technical-articles/switching-regulator-noise-reduction-with-an-lc-filter.html)
-
Is this just a niggle, or do you have test data or design parameters showing that it will definitely cause a problem, say with ADC stability or something?
It has already caused problems with a RF synthesizer which showed excessive phase noise; the spectrum shows spurs at +/- 8kHz and its harmonitcs that are only -30dB down.
A LC filter would have to be impractically large to have sufficient attenuation at 8kHz; the inductor would have to be on the order of 2mH.
-
I'm looking for a buck IC with similar specs (1A to 2A output, 5V in, 1-3.3V out) in a similar form factor that DON'T have pulse skipping. Mouser/digikey isn't turning up anything.
TPS562209 maybe? Datasheet says "TPS562209 and TPS563209 always operate in continuous conduction mode, which reduces the output ripple voltage in light load compared to discontinous conduction mode". Low-quantity availability has been bad a few times this year though (as it is right now).
-
What your looking for does sound appropriate for the job, but I don't know my buck converters well enough to recommend anything. With that said, maybe another workaround to consider? What about a fixed load resistor to keep the minimum output from dropping enough to start pulse skipping? I don't know where exactly that would be on this chip, but if you drew a couple mA extra through a fixed resistor and could keep it into a regular duty cycle mode, you may be able to effectively get the behavior you're looking for and while you have excess draw all the time, you don't have the same power consumption as something like an LDO on the output, since it wouldn't be output current dependent.
-
That has the same problem as an LDO: loss of overhead, which is apparently tight. That said, if the primary load can be sensed or controlled, the load resistance can be controlled inversely to it, so that current consumption is constant at the low end. Making the power supply class AB, if you will. ;)
Tim
-
Is it practical to add a load resistor that is just enough to get past the skip mode?
-
I don't know of a switcher with a variable current limit but for 1.5A the sc189 buck regulators are nice, low ripple, and accurate. I've only used fixed versions but you can also use the 1 volt for variable output voltage.
-
Anything with synchronous rectification will be able to run in "forced" continuous conduction mode (CCM) down to zero load. This avoids the minimum-on-time problem that a DCM regulator will have at low loads (which makes pulse-skipping or similar un-avoidable), but you still need to check that they don't go into pulse-skipping automatically anyway for efficiency reasons.
TassiloH found a part that does this, if you need another option then start by looking at "Synchronous" buck converters and then dig into datasheets to check if they will change modes at low load or not.
-
Is this just a niggle, or do you have test data or design parameters showing that it will definitely cause a problem, say with ADC stability or something?
Tim
Pulse skipping is horrible for many analog circuits.
They will introduce a huge amount of noise and other issues.
It sometimes looks a bit that it's only invented to make the efficiency graphs look better.
Not caring about the actual practical performance of the circuit.
-
It has already caused problems with a RF synthesizer which showed excessive phase noise; the spectrum shows spurs at +/- 8kHz and its harmonitcs that are only -30dB down.
A LC filter would have to be impractically large to have sufficient attenuation at 8kHz; the inductor would have to be on the order of 2mH.
Any switching supply - pulse skipping or not, is "no go" to power RF synth. AD appnote talks about LC filters to attenuate LDO (!) noise:
http://www.analog.com/en/analog-dialogue/articles/power-management-design-for-plls.html (http://www.analog.com/en/analog-dialogue/articles/power-management-design-for-plls.html)
For applications like VCO and synthesizers ultra low noise linear regulator shall be used:
http://www.analog.com/en/products/adp150.html (http://www.analog.com/en/products/adp150.html)
-
Maybe the LTC1772B, a no burst mode version of the LTC1772. Constant operating frequency of 550kHz. About 50mA to 2A.
http://www.analog.com/media/en/technical-documentation/data-sheets/1772bfa.pdf (http://www.analog.com/media/en/technical-documentation/data-sheets/1772bfa.pdf)