How to recude noise in push-pull type converter/inverter?

[Mithun_K_Das]

For high efficient DC/DC converter push-pull type converter is the most suitable for low power segment. But in this type of converter, noise produces due to high frequency switching even the transformer is winded the best. This noise create harmful effects in micro-controller circuits such as ADC. Now how can this noise can be removed? I tried with adding a snubber (R+C) between two drains. But I found this is not enough. Or I'm using the wrong value. So, I'm expecting suggestions from experts.

Thanks in advance.

 

[KlausST]

Hi,

i made good experience with an appropriate layout.
For sure the errors cause by theDC/DC switching depend on many parameters.

A cut in the power plan around the DC/DC circuit can help to keep the noise away from the rest of GND.
Short traces for the switching power lines, good, HF capacitors with low ESR, multiple vias from the capacitors to GND...are some hints for "wired" coupling.

Shielded inductors, or a shileding of the complete DC/DC circuit is good for inductive coupling effects.

Especially with ADCs you can synchronize the DC/DC frequency to the ADC sampling frequency. With that most of the HF noise is canceled, but giving a DC offset in the ADC reading
Sometimes it is possible to switch off the DC/DC in times (maybe 1 ms or less) when the ADC is running it´s sampling/conversion...

To give you a magnitude: for a handheld measurement device (low power) it is possible to achieve +/- 1 LSB of noise with a 16 bit ADC, 3VRef, 50kHz bandwidth.. without any shielding
If you give us more information about: switching voltage, current, frequency, ADC resolution, bandwidth..and a picture of your layout (with some description) we can better assist you..


Good luck
Klaus

 

[mtwieg]

RC snubbers should help quite a bit, if they're designed properly. Another often overlooked method is using an RF screen between transformer windings. While it will increase leakage inductance a bit, it can cut down on EMI tremendously. Also if your power FETs are heatsinked to the chassis, then you can add RF screens there too.

 

[Mithun_K_Das]

Here is the test result. This is found while the converter is on. The signal is taken at the ADC pin of the micro-controller.

 

[KlausST]

Hi,

Most probably this noise is influenced in the scope's lines. So the ADC will see a way more cleaner signal.
Do some ADC readings and analyse the digital values.

Good luck
Klaus

 

[mtwieg]

Agreed, depending on how your probes are connected, they may show you much more interference than what the circuit actually sees in their absence. Using a short pigtail ground connection can help mitigate this.

 

[mrinalmani]

Shielded inductors, or a shileding of the complete DC/DC circuit is good for inductive coupling effects

Please could you explain more about shielding the entire DC/DC circuit. How is it shielded? By aluminium foils? Steel sheets? Does it include shielding the H-Bridge also?
Instead, will it not be reasonable to shield the MCU itself?

 

[Mithun_K_Das]

No no bro. The prob is connected ok. And when the converter is off, it shows a smooth voltage. But when the converter is on the voltage comes like this.

Also, the ADC is displaying the voltage when the converter is off But, it goes crazy when the converter is on. Once it shows 9V,then it shows 40V(say). This values comes randomly. That means the ADC is affected by this noise.

 

[KlausST]

Hi,

Do you use a proper ground plane?
Clean, decoupled reference voltage?
RC filtering of HF signal directely at the ADC input?

Expect less than some mV at the ADC input.

My opinion: it is either a schematic error (wrong values) or most probaly a layout problem.

Show us your layout, with a little description..

Klaus

 

[FvM]

The problem has two sides:
1. The power stage and it's "noise" (switching transient) generation power
2. The measurement circuit and it's interference susceptibility

We can expect that both sides are contributing to the bad result.

 

[aashitech]

Two simple to do suggestions >>
(1) Try Putting decoupling capacitors at micro controller power pins as close as possible ... try parallel combination of 10nF and 100nF.
(2) Put one RC filter at ADC pin input... R value ca be anything in between 100-680 Ohms and C == 100nF.

in context of point no. 2 above ..if you are measuring battery voltages then obviously you will be using some voltage divider .. in that case try putting a electrolytic capacitor of say 1~4.7 uF from resistor divider junction to ground...hopefully that will average out/ remove the noise you are having.

 

[Mithun_K_Das]

I tried with adding RC filter circuit (R=100R & C=0.1uF) results not so good. Then I tried with adding a electrolyte capacitor, 0.22uF,0.47uF,1uF,10uF,47uF,470uF. All results bad. Also I tried with taking a average value from 5 data. That was little more good but not enough. voltage fluctuation in the display still happening although lower than before.

 

[mrinalmani]

Temporarily powering the MCU from a different source, if possible, will confirm whether the noise is electromagnetic or simply electrical.

 

[Mithun_K_Das]

I found a fault in my design. I put the micro-controller near the SMPS transformer by mistake. Other design of the same product is not making a disturbance like this one. The nose has almost eliminated. In the new PCB design, I put the MCU far from the DC/DC converter section.