Hello everyone, I have an audio project for which I would like to use AC / DC converters, 5V and +/- 15V.
I realized the PCB using Hi-Link devices and before going further I realize that my power supplies are not well filtered.
Here are the catches:
AC coupling, 5V in Green and +15V in Blue :

DC coupling 5V in Green, 15V in Blue:

DC coupling 5V in Green, -15V in Blue:

I looked in the literature and found this schematic:

I would like to understand the principle and according to the use of a BJT transistor I will lose 0.6V on the output voltage, is there a FET similar circuit,
or how to get away from the noise of AC / DC converters?
Thank you for suggestions.
Vincent.

 

That's a common mode inductor with a ferrite core. Is your interference common mode or differential mode?
I'd use single inductors with iron powder cores. They tend to be lossier at high frequencies and absorb the interference, turning it into heat, rather than reflecting it around the circuit.
Self-inductance of the capacitors and self-capacitance of the inductors will be a problem at high frequencies. If you can afford the voltage drop, RCRC filters work very well.
This circuit is also pretty good, because it is basically a Sallen-and-Key filter so the output falls off at 12dB/octave. You can use smaller values of C than you could with a bipolar circuit, which tend to be more capacitive than electrolytics.
The FET will drop more like 2V, but do you really need 15V or will 13V do the job? If it's audio then you've only lost 1dB of headroom.
What's the 5V do? If you have a MCU, then a 3.3V linear regulator might remove most of it.
But before you do any of this, the most important thing is the signal routing and grounding on your pcb.

 

Hi and thank you @Ian0, it's in common mode, for the moment I have not yet carried out. I am looking that up.
It may be easier to use inductors with iron power cores. . The schematic with the BJT is (I think) a capacity multiplier.
With a transistor with a gain of 100 that can multiply the value of the capacity by 100 which makes it possible to effectively filter the ripples. The solution indeed with a MOS-Fet will lose more tension.
The +/- 15V is to power the AD75019 circuits and AOPS, the 5V is for a MCU.
Another solution (for more current) is to replace BJT by Darlington, but not necessary in my case.






Ian0

 

hi v.
Check this video for that cap multiplier circuit.
E

 

hi v.
Check this video for that cap multiplier circuit.
E

I just saw it as a 1st order RC filter with an emitter follower on the output as a buffer.
The frequency response still falls at only 6dB/octave, whereas my Sallen and Key version falls at 12dB/octave.
In either case, parasitic capacitances will spoil it!

 

The +/- 15V is to power the AD75019 circuits and AOPS, the 5V is for a MCU.

I wouldn't if I were you - the AD75019 has an absolute maximum supply voltage of +/- 12.7V
15V minus the Vgs of a MOSFET would be about right.
By the way, your C9 is upside down.

And what's a AOPS?

 

hi Ian0,
It was not posted as a fix for his problem, just to explain the circuit operation.

I would like to understand the principle and according to the use of a BJT transistor I will lose 0.6V on the output voltage, is there a FET similar circuit,

E

 

hi Ian0,
It was not posted as a fix for his problem, just to explain the circuit operation.

I would like to understand the principle and according to the use of a BJT transistor I will lose 0.6V on the output voltage, is there a FET similar circuit,

E

Apologies - I understood that but perhaps I could have been clearer with my reply.
I just thought that "capacitance multiplier" seemed a bit of an unnecessarily over-complicated way of looking at it, when it looked like a first order filter with a buffer.
I know one wouldn't use the output of an op-amp as the power supply, but it illustrates my point.
EDIT: I just spotted the op-amp PSU pins are labelled upside down

 

One low tech solution that doesn't suffer from parasitic inductances and capacitances is the good old zener diode.
Series resistor and 12V zener to ground will take out a lot of that interference.

 

I wouldn't if I were you - the AD75019 has an absolute maximum supply voltage of +/- 12.7V
15V minus the Vgs of a MOSFET would be about right.
By the way, your C9 is upside down.

And what's a AOPS?

Thanks Ian0,
between +/-15V I add Voltage regulator to decrease at +/-12V. not in this schematic.
Yes C9 is typo (copy/past) !!!
AOPs is operational Amplifier (in French Ampli-Operationnel ...) Sorry.

 

hi v.
Check this video for that cap multiplier circuit.
E

Thanks @ericgibbs for video link, this remove confusion between 1st order R/C filter and serial-pass transistor use.
Can we consider -3dB attenuation at frequency-cut ? Like normal 1st order filter.

 

Provided that R is small compared to transistor base current and C is large compared to transistor capacitances, then -3dB frequency is the same as a filter.

How much of that switching noise is making it through the 12V regulators?
A different choice of 12V regulator may fix it meaning you don' t need any additional hardware.

 

Provided that R is small compared to transistor base current and C is large compared to transistor capacitances, then -3dB frequency is the same as a filter.

How much of that switching noise is making it through the 12V regulators?
A different choice of 12V regulator may fix it meaning you don' t need any additional hardware.

I have not decided yet to choose the 12V regulator type, a LM7812 get a ripple rejection around 55dB at 120Hz, do you think that is enough?

 

It's the High frequency PSRR that is far more important. LM317 and LM337 are always said to be a huge improvement over 7812/7912, but LM317 is old now - I should think that there are better devices.
A ferrite bead on the input to the regulator, before the capacitor might be enough inductance to deal with the interference you have.
There's no universal answer, it takes experimentation.

 

It's the High frequency PSRR that is far more important. LM317 and LM337 are always said to be a huge improvement over 7812/7912, but LM317 is old now - I should think that there are better devices.
A ferrite bead on the input to the regulator, before the capacitor might be enough inductance to deal with the interference you have.
There's no universal answer, it takes experimentation.

I take what I have in the shelves to make tests.
But you're right, no universal answers !

 

Another remark, I did the scope capture using a large ground on the tip, it would be necessary to do a test with a short ground proximity (paper clip) test to check if they are really due to the AC/DC converter.

 

I take what I have in the shelves to make tests.
But you're right, no universal answers !

What type of ferrite do you think are the best? Axial, SMD ? How to determine value ?

 

What type of ferrite do you think are the best? Axial, SMD ? How to determine value ?

Pretty much guesswork!
And yes, a short ground connection to the right place, is essential to get a reading you can actually believe.
It might be nowhere as bad as you think!

 

Looking at the web, I found a similar way to use the caps multiplier, as you can see, they are placed after the regulator, the dual-self too, there is a good reason for that?