# Filtering analog power supply

I'm working on a project which runs off a 50V DC power supply. There is a high power H-bridge being switched at 250Khz powered straight of the 50V. There's also the analog control electronics running of 12V. The 12V has to be created from the main 50V supply. Like this:

Obviously there's noise being created in the 50V by the switching of the H-bridge which I'd like to prevent making it's way to the 12V supply. Which is the preferred method for this?

Things I've considered:

• A low pass RC filter before the linear regulator.
• A low pass LC filter before the linear regulator. Hard to damp properly as the analog section's input impedance is obviously not fixed.
• Use a linear regulator with high PSRR (some are designed to have more then others). Aug 15, 2017 at 13:28
• 50 V - 12 V is 38 V so plenty of headroom to first regulate down to 24 V and then down to 12 V. That by itself already improves the line regulation. How much current does the 12 V supply need to deliver ? If it is not so much an RC filter is a good option. An LC filter can also be an option but watch out for the LC resonance frequency. Also the inductor might need to be large and/or expensive. Aug 15, 2017 at 13:30
• We need to know how much current the +12V will use. Most likely not much, or you would use a switching regulator? Also dropping 50V to 12V will require a high-voltage regulator, your choices will be limited. Also, how much noise do you require on the +12V? Separate supplies for microcontroller and analog? etc. Aug 15, 2017 at 13:32
• @peufeu: Current demands are 75mA max. I'm using a LM317 configured for 12V output, it allows for a 40V I/O voltage differential. There's no micro-controller, mainly op-amps. Aug 15, 2017 at 13:43
• Ouch. Absolute maximum rating for LM317 is 40v I/O differential (as you say). Ask yourself: at turn on, what voltage does LM317 see at its input? Chances are, it is greater than 40V, while its output is zero, or near zero. Aug 15, 2017 at 14:58

OK, let's go.

Most regulators have crummy HF PSRR, so we begin by adding a ferrite bead (I checked the ratings), Murata provides all spice models so they're already loaded into my simulator which is convenient. A cap to ground (1µF 100V MLCC) and a series resistor to prevent ringing. This gives a nice HF PSRR boost (see graph).

Next we have two choices. LM317 is out, as ratings will be exceeded at power-up when output voltage is zero (caps discharged) and input voltage is 50V. Thus we have to use LM317HV, or another HV regulator, or hack one together from some spare parts.

Red is LM317, black is roll-your-own. Top is PSRR (ferrite included), bottom is output impedance. Three traces for each (0mA, 20mA, 75mA output current).

DIY custom job wins on PSRR due to BD139 having low capacitance, also it has more feedback, but requires big output cap. Also it could be unstable if you screw it up. But let's be honest, LM317HV will be sufficient.

Note the ESR of output caps matter, I've specced cheapo Panasonic general purpose ones, plus a 100nF ceramic.

I'd use a switcher to get 15V from 50V, then LC filter (ferrite/resistor/cap) then LDO or LM317, that would do the job without a heat sink. If output impedance is not critical, a switcher followed by LC filter would be enough. Opamps work fine with a few ohms impedance in the supply, they usually have huge LF PSRR, it's the HF noise they have trouble with.

• Are you using micro-cap simulator dude? Aug 15, 2017 at 16:41
• Yes I love the GUI in that simulator, it's very user-friendly. Aug 15, 2017 at 16:50
• I agree and I would recommend it to anyone. Aug 15, 2017 at 16:51