Dual Supply (Bipolar) from an SMPS DC Adapter (Wall Wart)

Thread Starter

reynolds087

Joined Jul 22, 2020
35
Does anyone know how I can create a dual supply from an SMPS single supply DC adapter? This is for an audio circuit. It needs to be able to deliver at least 1 amp, ideally more than that. It also needs to filter out all of the high frequency switching noise from the SMPS stage. I've been trying to use a push pull configuration with BJTs and a "virtual ground" but still seeing a lot of noise on my scope. I put a couple of RC low pass circuits in series to the base of the transistors, but the noise is getting through.
 

Ian0

Joined Aug 7, 2020
4,821
I’ll make two assumptions if it for audio:
1) It doesn’t have to be perfectly symmetrical, within half a volt or so will do.
2) There will always be a load on it, because of supply currents of op-amps etc.
If so, you can use a Čuk converter open-loop. Just drive the FET with a 50:50 mark space squarewave and the negative supply will track the positive within one diode drop.
Čuk is by far the quietest switcher, because both input and output currents are continuous.
48BE4568-93C8-43C8-A9D4-108FB98B96A0.jpeg
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
Thanks, I will study this topology. I appreciate it. Do you have any advice for removing the noise from the previous stage? The source DC voltage that I am using is very noisy. It shows like 2.5VPP on my scope. I don't know how to filter it while maintaing low output impedance.
 

Ian0

Joined Aug 7, 2020
4,821
Once you have positive and negative supplies, you can use a common mode choke. Follow it by capacitors to earth. Why do you need to maintain a low impedance?
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
Once you have positive and negative supplies, you can use a common mode choke. Follow it by capacitors to earth. Why do you need to maintain a low impedance?
Maybe I am using the wrong terminology. When I said low output impedance, I meant a lot of current available to power the load, without a lot of wasted heat and power loss in the filtering stage.

I am curious how do you select a common mode choke?
 

sparky 1

Joined Nov 3, 2018
701
It is difficult to find because the marketing terminology for search has so many flavors leading to the wrong product,
I can try to improve on the search terms but this is real adventure to find:

DC-DC Boost Converter Positive and Negative Dual Output 3V-6V To 5V-32V 12V/24V Non-isolated Power

DC-DC Boost Buck Converter with Positive and Negative Voltage Regulator Output Voltage ±12VDC,
a dual power filter board, A common mode and EMI noise suppressor for dual polarity supply.
 
Last edited:

Thread Starter

reynolds087

Joined Jul 22, 2020
35
I tried putting a common mode choke in series to the DC supply, but the noise is still out of control. I've attached an image. The frequency is constantly changing, so just ignore that part.

I even referenced the negative side to earth, but it didn't help.
 

Attachments

Thread Starter

reynolds087

Joined Jul 22, 2020
35
I kept the 1.5mH common mode choke, and added five x 1000uF capacitors and 4 x 47uH inductors in a multi-stage pi filter, and now I'm getting the attached results across a 15K resistor. It's better, but I'm kind of stumped why the noise isn't attenuated more. Is this normal?
 

Attachments

Ian0

Joined Aug 7, 2020
4,821
1. Make sure that you are not picking up anything with the loop of wire formed by the scope earth clip. Use the little coil spring that came in the packet with your probe for the earth connection.
2. Speed up the timebase and fiddle with the trigger level until you can clearly see the shape and frequency of the waveform you are trying to eliminate.
3. Get rid of the electrolytic, and use 10nF to 1uF multilayer ceramics.
electolytics have far too much resistance and inductance at the frequency you are trying to eliminate.
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
1. Make sure that you are not picking up anything with the loop of wire formed by the scope earth clip. Use the little coil spring that came in the packet with your probe for the earth connection.
2. Speed up the timebase and fiddle with the trigger level until you can clearly see the shape and frequency of the waveform you are trying to eliminate.
3. Get rid of the electrolytic, and use 10nF to 1uF multilayer ceramics.
electolytics have far too much resistance and inductance at the frequency you are trying to eliminate.
So what values would you recommend for filtering this noise? I think I have a couple of 100uF ceramic capacitors, but can't breadboard them because they are large smd form factor. I have the regulars axial ones up to 1uF I believe. I also have a couple of big toroids that I can wind up to about 25mH.

I'm also curious if I drive the base of a transistor with the filtered output, can I buffer the current like a capacitance multiplier?
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
You can get leaded MLCs up to and beyond 1uF, but first you need to know what you are trying to remove.
I'm not really sure how to find the frequency of the noise. It seems totally random. Also I'm looking for more generalized principles for removing high frequency noise rather than specifically targeting this one power supply, because I'd like to have an adaptable circuit to use with any off the shelf switching power supply.
 

Ian0

Joined Aug 7, 2020
4,821
Unfortunately, there is no such thing as a universal interference filter.
Your supply will have a repetitive waveform between 65kHz and 150kHz. Set your timebase to 2.5us, AC couple, and start looking.
 

Thread Starter

reynolds087

Joined Jul 22, 2020
35
Unfortunately, there is no such thing as a universal interference filter.
Your supply will have a repetitive waveform between 65kHz and 150kHz. Set your timebase to 2.5us, AC couple, and start looking.
I am curious why a low pass filter wouldn't simply remove everything above the cutoff frequency. Ideally I just want as close to flat DC as I can get. There isn't a low pass filter technique to accomplish that? It sounds like you are recommending a band pass to basically notch out the problem frequency Is that correct?
 

Ian0

Joined Aug 7, 2020
4,821
Because there is no such thing as a perfect capacitor or inductor.
A capacitor has a strip of metal foil wound up into a spiral. A wound conductor is an inductor.
An inductor has layers of windings on top of each other. Conductors close together separated by insulation are capacitors.
The frequency at which a capacitor or inductor stops being whatever it is opposed to be and starts being the opposite is its self-resonant frequency.
The larger the capacitor or inductor, the lower the self-resonant frequency, so the lower you place your cutoff frequency, the lower the frequency at which the filter stops working.
At high frequencies, your series-L parallel-C low pass filter becomes a series-C parallel-L high-pass filter and lets all the interference through.

If you wind toroidal inductors with a single layer of wire, you get the highest possible SRF.
Iron powder cores are excellent, because they are not very good inductors at high frequency - they become lossy. A really good hf inductor reflects signals is does not pass, and they go off and annoy another part of the circuit. A lossy iron powder core absorbs the interference and gets warm, but the interference is dissipated instead of reflected.
Also, don’t underestimate the humble ferrite bead. It is surprisingly good at removing very high frequency interference.
Also don‘t forget that any interference on audio >20kHz can’t be heard. Have you ever seen the mess that is the output of a class-D amplifier?
 
Top