Author Topic: Designing a Differential and Common Filter Circuit for DC DC Converter Output  (Read 4011 times)

0 Members and 1 Guest are viewing this topic.

Offline MarcusSTopic starter

  • Regular Contributor
  • *
  • Posts: 55
  • Country: au
Hi everyone,

Does anyone have experience designing an output differential and common mode filter using a common mode choke for the output of a dc dc converter power supply.   The dc dc converter I am using is a shielded unit capable of 12V output at 2.5 amps.  It has a switching frequency of 360Khz and will be driving a 12V 300mA load.  I want to put a common mode and differential mode filter to filter out as much noise as possible and have some 1.7mH common mode chokes with 10uH leakage inductance I would like to utilise as a starting point.  The load has an onboard 1000uF electrolytic capacitor that could be used as the dampening capacitor of the filter.

Is anyone familiar with the maths for these kind of filter designs.  I have watched the Biricha videos on you tube but the calculations seem to assume things that I don't have available in the specification of the dc dc converter.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Right, you can't design a filter without all the numbers.

You need:
Converter isolation capacitance
Converter output noise and spectrum, DM and CM
Source impedance (what AC return path is there)
Maximum noise spectrum (i.e., subtract the spectra to get the required attenuation)

You are perfectly welcome to just guess -- indeed, most of us do.  But you need some way to test whether it's done what you think it should do.  You can measure noise in a simple fixture, for example, preferably with a spectrum analyzer but a scope will do in a pinch, with some hand-waving over the waveforms.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline MarcusSTopic starter

  • Regular Contributor
  • *
  • Posts: 55
  • Country: au
Not sure I understand what the converter isolation capacitance is for this dc dc converter. At the very least I don't see anything mentioned in the spec sheet.  The converter is a TDK lambda CCG30-48-12S. 

My calculations tell me with a 40 ohm load (which is assumed from 12V 300mA output power draw from the converter with my intended load) that a 1.7mH common mode choke with 0.56uF capacitors between each leg of the common mode choke and the ground reference to my pcb will give me a cutoff off frequency of 5Khz with a filter Q of 0.7.

If I wanted to make a differential mode filter making use of the leakage inductance of the common mode choke which is 10uH then I would get a cutoff frequency of 900KHz with a filter Q of 0.7 providing the differential mode capacitor is 3300pF.

« Last Edit: November 14, 2019, 03:56:02 pm by MarcusS »
 

Offline MarcusSTopic starter

  • Regular Contributor
  • *
  • Posts: 55
  • Country: au
Also worth mentioning is the 12V dc dc converter is running off the output of a 36V switchmode power supply running off mains.  The 12V dc dc converter is for a secondary rail in the device with the main part of the device running off 36V.  Its an off the shelf item this switchmode power supply but I don't think I necessarily have an ac return path, as the output is not mains referenced from what I can tell.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Could you draw a diagram of the system, including chassis grounding (if applicable), the nature of this 36V supply, the 12V load, and so on?

Is this actually common ground, the 12V is just stepped down 36 and a buck converter would be fine but you're using a module just because..??


I don't think I necessarily have an ac return path, as the output is not mains referenced from what I can tell.

There is ALWAYS an AC return path.  Even if it's battery powered, there is some capacitance from the body of the device to its surroundings.

This is... a bit un-obvious when you haven't seen it before; understanding the common mode equivalent circuit will turn your brain sideways at first. :)

Tim
« Last Edit: November 14, 2019, 06:41:13 pm by T3sl4co1l »
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline MarcusSTopic starter

  • Regular Contributor
  • *
  • Posts: 55
  • Country: au
Power Source is a DTM165PW360C TDK Lambda External Power Supply.  Proposed 12V section shown in the picture.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
It is common ground!  Yeah, don't worry about it.  Tie all the grounds together on a plane and add normal mode filters on either side.  Like 1uH and 2.2uF would be fine.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline MarcusSTopic starter

  • Regular Contributor
  • *
  • Posts: 55
  • Country: au
Are you suggesting to ditch the common mode choke all together. The other option I have is to not tie the dc dc converter output ground to the dc dc converter input ground.  Can you post a sketch of a circuit.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
First let me condense this a bit for presentation purposes; then a little markup:



That one wire (arrow) is a good place to add extra filtering if needed.

You may also put a CMC at the input jack, in case your ground is dirty and a choke is needed.  The power brick should be okay (shouldn't be a source of noise or surge), but it would help with that too.

The important thing is this: if the loads are not common ground (there is no sneak path around the CMC), then the load must be returned to the X'd line.  (Or even if they are, if you want to ensure load current flows through the cable only and not through sneak paths, this must be done.  Not very relevant at 300mA, aside from the CMC, but this can be important for high current loads where the return current causes significant voltage drops across the ground.)

In that case, the CMC would be relevant, and there would be no ground bridge under U2.  Large caps to ground (C5, C7) probably wouldn't be a problem, but probably wouldn't be necessary either.  These would typically be ~nF.  Diff filtering (C10, C11) should be larger, 2.2uF say; the values shown won't really do anything, and C11 (3.3nF) with the ~10uH leakage inductance will lead to a transmission peak at ~880kHz, which may happen to fall near a harmonic (depending on if the DC-DC is variable frequency or relatively stable).

But with everything common ground, there's no common mode, just make sure the ground is solid under it, and bring out the connections to the same side so that C8 and C10 are somewhat nearby (and again, C10 and C11 should be larger values, and the CMC can be replaced with a single 1uH inductor on the 12V line).

Cheers,
Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline MarcusSTopic starter

  • Regular Contributor
  • *
  • Posts: 55
  • Country: au
I think I may have confused things a bit by not showing that the minidsp power input doesn't use the common ground, it is intended to connect across C11.  The link between the input and output Vin - and Vo- of the 12V dc dc converter doesn't need to be there either but I am curious if there is any harm in doing so.

According to the calculations I don't understand how the differential filter has a transmission peak at 880kHz.  That is indeed the frequency Fc but I expected for the differential mode filter of 10uH and 3300pF that would give me a Q of 0.7 so that shouldn't be a peak.

What may also complicate things is that the rca audio outputs of the minidsp hd (which would be referenced via the common mode choke to the power supply ground in this schematic) would in operation essentially also reference to the 36V ground of the audio amplifier which is in a different chassis and runs off a completely different power supply.

Basically meaning V- of the miniDSP is connected to the RCA output grounds on the minidsp board.  Those RCA outputs and therefore grounds connect to the audio amplifier RCA inputs and therefore grounds and those RCA input grounds in the audio amplifier are connected directly to the gnd of the separate 36V power supply powering the audio amplifier.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Oh, so it's audio, and there is a common ground elsewhere?  So there would probably be some interest in minimizing ground loops to preserve audio quality?

The 300mA load is pretty small, I would route the 36V to the power amp and connect the converter there (with filtering as needed).  This keeps the loop smaller than having a separate power supply section, and keeps the loop current low.

According to the calculations I don't understand how the differential filter has a transmission peak at 880kHz.  That is indeed the frequency Fc but I expected for the differential mode filter of 10uH and 3300pF that would give me a Q of 0.7 so that shouldn't be a peak.

Taking the circuit at face value, there's no load (or, well, the 4.7k I guess), so the Q can be quite high.  I don't know what kind of load a miniDSP board is like so I might assume the worst.

If it's a typical load, it probably has a few uF onboard, in which case having a C11 doesn't really matter.  You would still have to figure out the Q factor though.

You generally want a filter impedance much less than the load resistance, because the filter impedance (with a factor of Q higher or lower, depending on which equivalent resonance the filter happens to manifest) is how much AC voltage (supply ripple) appears at the filter for a given change in load current.  For a better than 10% regulation under 300mA load step change, you need less than 4 ohms, and so on.

Electrolytics of modest values (a few uF) have ESR comparable to this (an ohm, or fractional ohms), making them appealing to ensure low Q.  A smaller ceramic can be placed in parallel to maintain good HF performance.  (The high-Q (ceramic or film) capacitance should be at most a third of the lossy bulk capacitance.  So, a 10uF electrolytic with <= 3.3uF ceramic in parallel, should be pretty well behaved.)

Tim
« Last Edit: November 16, 2019, 05:19:16 am by T3sl4co1l »
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline MarcusSTopic starter

  • Regular Contributor
  • *
  • Posts: 55
  • Country: au
The 300mA load I have quoted is what the minidsp draws and is also what it says on the datasheet.  I have measured it and it is pretty constant and doesn't change much.  The first thing on the minidsp board is a 1000uF electrolytic so I haven't added any additional electrolytics on the load side of the dc dc converter.  The primary reason being is that the 12V dc dc converter has a 1200uF maximum capacitive load rating.

So in other words the common mode and differential mode filters are being designed for a 300mA load at 12V.  The 4700 ohm resistor is for some additional damping but I may get rid of it completely.   It was basically there to stop the filter ever having no load at all and is not necessary with the minidsp connected as it will be on the final pcb.  With my calculations, the Q of the filters with a 40 ohm load (e.g. 12V @ 300mA) is about 0.7 for both the differential and common mode filters.  Cutoff frequency is about 900Khz in differential mode and 5Khz in common mode.  Do you agree with my maths on that one.

Probably also worth mentioning is the filter design doesn't use electrolytics.  The 0.56uF capacitors are film caps and the capacitors that have pF values (C11 and C10) are all ceramics.   I have just used the wrong symbols on my schematic and I probably should have used a non polarised symbol.

C10 is meant to be the "pi" capacitor.  C11 + leakage inductance of the common mode choke is the differential mode filter and C6, C7 and the common mode choke that is 1.7mH in common mode is the common mode filter.

The other thing that is important is the amplifier is in a completely different chassis running off a completely different power supply and will remain that way.
In real world operation due to the fact the audio signal from this dsp chassis connects to the audio input of the amplifier chassis, there will be a connection between the DSP V- and the amplifiers power supply ground.

Hopefully that clears things up a bit.
« Last Edit: November 16, 2019, 07:29:15 am by MarcusS »
 

Offline MarcusSTopic starter

  • Regular Contributor
  • *
  • Posts: 55
  • Country: au
Also C8 across the input of the 12v dc dc converter on the 36V side is also a ceramic.  The 2x 330uF capacitors on the input side of the dc dc converter are nippon chemicon low esr electrolytics.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf