Linear power supply spikes & SMPS power suply spikes

Thread Starter

Doros

Joined Dec 17, 2013
137
Hello everybody,

I am working for weeks now to make a decent power supply for my workbench, with your help, and finally my linear one gives the output you can see. With some spikes, for which I can not identify the frequency.

How can I get rid of them? Propably using a low pass filter, but I am not sure what kind of capacitor to use. Ceramic, electrolytic etc.

I have also an SMPS, which with a voltage regulator gives the output attached. With spikes, but with a frequency of 13.7kHz. Again I think I should use a low pass filter, but what kind of a capacitor to use?

thank you very much for your input
linear power supply spikes II.jpg
Doros
 

Attachments

Bordodynov

Joined May 20, 2015
2,939
If you want to get rid of spike, then I will give you advice.In stabilizers, use a PI filter consisting of two ceramic capacitors and a bead.A bead is a conductor through a ferrite ring or a ferritic tube (the second is better).Perhaps through the ring will have to take a few turns.
In a linear regulator, this filter must be placed in front of the stabilizer.In the pulse, after the stabilizer.
 

Thread Starter

Doros

Joined Dec 17, 2013
137
Thank you very much Bordodynov for your help.

When you say stabilizer, you mean the voltage regulator? For example the LM317 I am using for the linear supply? So to put the π filter, after the electrolytic capacitors just before the LM317 circuit.

In the linear power supply the spikes have no specific frequency. They appear from time to time with no defined frequency

For the SMPS, you propose to place the π filter at the output? I use an LM338 after the SMPS to regulate the output. So I can put it after the LM338 circuit. I found this SMPS not noisy that's why I intend to use it. This output I posted is at 20V & 1A

When you say bead you mean the beed II photo I inserted, with the cable making some turns through it?

thanks again

doros
 

Attachments

Bordodynov

Joined May 20, 2015
2,939
Yes, speaking stabilizer, I meant the voltage regulator, although this is not the same thing.The voltage regulator is not always a stabilizer.Put the filter before or after electrolytic almost without a difference.Spike on the output of your stabilizer regulator is usually connected with the input spike.But the spike can appear at the output due to a sudden change in the load current.This is treated by shunting the output with a ceramic capacitor.
You can put a filter on the output before the load, but SMPS is the source of the spike.It is more correct to extinguish the impulse noise in the place of its occurrence.But keep in mind that the ceramic capacitor at the output (parallel to the load) is desirable.Especially if the load is an electric motor.
You showed the correct examples of beads. I once used a simple bead to help a lot. I used a voltage regulator on one board and a laser driver on another board. The driver worked very poorly at the same time (laser emissions were observed, which is very bad for a laser). I put a small ferrite ring (toroidal) on the power conductor with an outer diameter of 4mm. Ceramic capacitors on both circuit boards already were. And everything became good!
 

Thread Starter

Doros

Joined Dec 17, 2013
137
Thanks Bordodynov,

I always put, in this case too, ceramic capacitors of 0,1μF at the end of the circuit, parallel to the load.

The sizes of the two capacitors of the π filter should be 0,1μF also?
 

Bordodynov

Joined May 20, 2015
2,939
Probably enough.But the more, the better.I calculate my schemes using LTspice.True beads I use ready.I used capacitors in filters up to 22μF.Sometimes 2μF is enough.
 

Thread Starter

Doros

Joined Dec 17, 2013
137
Hello Dana,

you mean i.e. if the spikes have a frequency of 13kHz, to use a capacitor showing the lower ESR at this frequency so it will charge/discharge easier.

Why then do we always use, (I do the same with good results), as decoupling capacitors next to the power pin of the ICs a 0,1uf ceramic X7R, without even bothering check the pattern of the noise of the power supply?

thanks Dana
 

Thread Starter

Doros

Joined Dec 17, 2013
137
It worked well. I made and used a bead 0,7mH, and used capacitors electrolytic and ceramic in parallel 12μF combined, and no spikes at all.

thanks so much

doros
 

danadak

Joined Mar 10, 2018
4,057
Why then do we always use, (I do the same with good results), as decoupling capacitors next to the power pin of the ICs a 0,1uf ceramic X7R, without even bothering check the pattern of the noise of the power supply?
There was a practice I knew in the community I was a part of that used .01, .1, and bulk, usually
tant. I think these practices were not always analytic, simply breadboard derived. That "art" in our
profession.

The other lesson I learned is not all .1 uF ceramics are equal in performance. Some quite inductive
and almost useless. I started paying a lot more attention to datasheet characterization curves.

Regards, Dana.
 

ebp

Joined Feb 8, 2018
2,332
"Why then do we always use, (I do the same with good results), as decoupling capacitors next to the power pin of the ICs a 0,1uf ceramic X7R, without even bothering check the pattern of the noise of the power supply?"

I'm not sure I fully understand what you are asking, but ...

How much noise on power supply lines can be tolerated depends a great deal on the type of circuitry. Digital circuits are generally, but not always, less susceptible to noise on the supply. Analog circuits vary greatly - some are OK with moderate noise, depending on the frequency, others are very sensitive to noise. It can be very important to select a low-noise power supply for some applications.

Decoupling capacitors next to ICs are not usually there to filter noise that is already on the supply, but to prevent transients due to what the IC itself is doing.
Power distribution tracks on circuit boards have both resistance and inductance. Any change in the current through a such a track will cause a change in voltage. If the rate of change of current is quite low, the inductance may not matter much. If the rate of change of current is high, the inductance can be very important. Most digital ICs cause a fast current spike every time an output switches, even if there is no load connected ("shoot-through" current; the bipolar version of the 555 timer has really high shoot-through current and can be interesting to experiment with to see the effects of lack of decoupling, especially with long wires to a power supply). With a load connected, there will also be capacitance that has to be charged or discharged. The capacitance is usually quite small but with logic devices the very fast switching even a small capacitance can still cause current spikes of tens of milliamperes. In analog circuitry it is often the load that results in current transients. Driving a 50 ohm transmission line, for example, can require a lot of current. Sometimes in analog circuitry you can create what is effectively an unwanted feedback path if you allow the supply voltage to vary with the output current of an amplifier.
A decoupling capacitor close to the device that requires fast changes in current helps to overcome the effects of supply distribution resistance and capacitance. It acts as a local charge reservoir. As you probably know, you should keep the path between the power pins of the IC and the decoupling capacitor as short as possible so that the inductance of that path is minimized.
 

Thread Starter

Doros

Joined Dec 17, 2013
137
Thanks ebp,

I always use a 0,1μF ceramic XR7 capacitor next to the power pins of the ICs. Whenever I checked with my oscilloscope the power line, was highly improved when using these capacitors.

It is because of the above you mentioned at your post, which I liked it a lot due to the highly concentrated, descriptive, and simple to understand content.

regards

doros
 
Top