1
\$\begingroup\$

I have a 12.3V power rail, which is sourced from a 2A 12V AC wall adapter.

The 12.3V rail supplies a 3A 4V SMPS, which sometimes sees large load fluctuations from almost 0A through to 2A instantaneously, and back again.

During the load fluctuations, the 12.3V rail can see a voltage drop between 100mV - 600mV depending on load, along with switching noise from the SMPS.

Example of 100mV voltage drop for 0.5ms on 'unfiltered' 12V rail with switching noise.

Example of voltage drop on unfiltered 12V rail with switching noise

Originally, for a separate 5V supply I was just running a linear regulator straight off the 12.3V rail, for around 40mA @ 5V. Although some amount of high frequency switching noise was getting through the regulator and onto the 5V rail.

I decided to add a low pass LC filter, between the 12.3V rail and the linear regulator. That did an excellent job, and almost completely removed the switching noise, and slowed the dV/dt during the load transients sufficiently that the linear regulator could still provide a stable 5V. The transients were still detectable, although were reduced almost into the noise floor.

Example of implementation and testing of 'filtered' 12V rail, showing voltage drop, and effects of under-damped low pass filter causing the extra oscillation (ringing).

Example of implementation of low pass filter

Despite this the 'filtered' ~12V rail still would drop 100mV - 600mV exactly like the 'unfiltered' 12.3V rail.

The problem is that I need to draw power from the 'filtered' ~12V rail to power a 12V sensor. The sensor has non-polarity variable current output and will be at the end of an 75m+ cable.

If I connected the 12V sensor directly to the 'filtered' ~12V power rail, then the 100mV - 600mV voltage drop occurs, surely would it not propagate in a wave down the length of the cable? and might (should?) affect the stability of the current output of the sensor, and any derived readings.

I'm struggling to determine the best approach to achieve a stable 'filtered' ~12V rail, from the 'unfiltered' 12.3V rail.

I've simulated a number of low pass topologies, and then build up the circuits, although I've been struggling to get the simulations to match the real world filters.

Approaches I've considered include 1st and 2nd order low pass RLC filters and extra stages, using high speed switching diodes to isolate the 'unfiltered' rail while it's at a lower voltage than the 'filtered' rail (during the voltage drop). I've also looked at emitter follower capacitance multiplier circuits and even looked at just simple RC low pass filters.

My (self imposed) constraints are to stick with ceramic capacitors (lifetime & reliability), although Tantalum caps can also be considered. I've used miniature SMD components (minimum 0603) to limit loop area and emissions.

The 'filtered' rail needs to run at a minimum of 12V, and I suppose the extra 0.3V could be dropped to achieve superior regulation if needed.

For my approximate model to simulate the 'unfiltered' 12.3V DC rail is to use a square wave with 300mV peak to peak at 12V DC offset, running at 200Hz with 90% duty cycle. Although, the load that causes the voltage drop occurs at a regular frequency, in reality it occurs after random intervals rather than continuously.

Example simulation of low pass filter for 'filtered' 12V rail, problems include large over-voltage at startup, and unstable oscillations during changing load. Despite that, during steady state ripple is less than 10mV.

Example of simulation of low pass filter

My question is are there any low pass topologies that I may not be aware of, or something that can be done to improve the existing filter?

Kind regards,

Edit: Potentially better solution. I've come up with the following circuit, which is a two stage RLC filter, that borders slightly on the under-damped side, although at maximum load will still provide atleast 12V, while also sticking to the restriction of only using Tantalum or Ceramic caps

enter image description here

\$\endgroup\$
  • \$\begingroup\$ how about using 10 ohms resistor and 1,000uF capacitor? \$\endgroup\$ – analogsystemsrf Mar 31 at 3:19
  • \$\begingroup\$ A voltage regulator acts as a variable negative resistance (impedance) at te input, which makes using a passive LC filter problematic (since they are fixed in value), if you mean to match the impedances. You can still have it, as you saw, but with inherent drawbacks which, ultimately falls on you to choose and accept the lesser evil. \$\endgroup\$ – a concerned citizen Mar 31 at 8:21
  • \$\begingroup\$ @analogsystemsrf Thanks, I've tried to do this without electrolytics, and the largest Tants in reasonable form factor are only 100uF - 220uF. \$\endgroup\$ – rcsandell Mar 31 at 10:40
  • \$\begingroup\$ @aconcernedcitizen I know what you are saying, as the linear regulator is a constant voltage source, the current it draws will change with the load, and because of the fixed value resistor in the RLC loop the 'filtered' 12V rail will always have voltage drop when the current draw increases without active regulation. I definitely don't mind figuring out some kind of active regulation solution, I've already tried a L7812 12V lin reg, although the dropout voltage was too high for 12.3V, and hence then why I tried the emitter follower capacitance multiplier, although that isn't a regulator either. \$\endgroup\$ – rcsandell Mar 31 at 10:47
  • \$\begingroup\$ I've use the following RLC low pass simulator to workout a critically damped filter, which completely removed the startup oscillations [link] (sim.okawa-denshi.jp/en/RLCtool.php). @analogsystemsrf is completely right, in that slapping on 1000uF cap definitely does the job and removes the ripple down to less than 10mV. Just wondering if there are any active or alternative filter methods that could be used instead? \$\endgroup\$ – rcsandell Mar 31 at 11:05
1
\$\begingroup\$

I considered that, although I only have access to the 12.3V coming in from the 220V AC transformer. I was considering a buck-boost converter, although that seemed overkill, and would also probably be expensive, add more noise on the 'unfiltered' rail, and might cause other unforeseen problems

15V AC-DC SMPS is commonly cheap nowadays, adding buck-boost would complicate even more. Depending on your current demands, for example take a look at Aliexpress here, the price can be as low as below 3 USD for 15V 2A supply. While it might be fun for you learning about the filter, I suggest that you shouldn't over-complicate yourself, I don't know if you have cost or time limits in your project.


The 12.3V rail supplies a 3A 4V SMPS

Does the 4V (or 5V?) supply that you have can handle 15V input voltage? If it does, then you can use the 15V SMPS, put a filter between before the second stage supply and then parallel it with LM7812.

I don't know that in the end you'll be using SMPS or linear regulator for the +5V rail.

It's all is depends on your needs.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
  • \$\begingroup\$ Brilliant, thanks. I'll use the 15V supply and run the 4V SMPS rail directly off it, then will also have the low pass filter in parallel to the 4V SMPS, with the L7812 and the L7805 after the filter. \$\endgroup\$ – rcsandell Apr 1 at 5:24
  • \$\begingroup\$ @rcsandell alright, so you have three separate rails. We're relying on the Filter and PSRR of the Linear Regulator. Try to put the SMPS also after the filter if the +4V rail will interfere with your +12V rail. Report back if you have any problem. \$\endgroup\$ – Unknown123 Apr 1 at 20:16
  • \$\begingroup\$ This paper from TI might help you designing the power filter. \$\endgroup\$ – Unknown123 Apr 1 at 20:32
  • \$\begingroup\$ Brilliant, thank you. \$\endgroup\$ – rcsandell Apr 9 at 14:37
  • 1
    \$\begingroup\$ Waiting for the 15V power supplies to arrive, although tested everything with lower voltages and seems to work just fine. \$\endgroup\$ – rcsandell Apr 10 at 16:21

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.