Power Supply Noise Reduction Methods

I would try adding some smaller higher frequency caps - 0.001 to 0.1uF, good ceramic (X7R) dielectrics. A 1uF cap is not really much good at reducing noise, it's better for supplying current when the load changes. Mount the caps as close to the regulator as possible.

Also, check your input for noise. If the same noise is there, it may be coupling across the regulator.

Make sure your circuit has a good low impedance ground plane. And a good low impedance connection to the ground plane.

I would also zoom in on the noise on the scope and see if it has any periodic components. If so, it may be coming from something else in the area.

Dear Bhankii

I added 0.1uF cap and 10uF cap in parallel and then also added a series resistor of 100 ohms. This did some reduction in noise. However, not much is achieved by putting filter o 0.1uF or lower at the output of the regulator.

Perhaps noise is coming from bad quality transformer or mains is too noisy. The repeat pattern seen in top picture show 50Hz mains frequency source with lots of sharp noises.

I added huge inductance also but did no great good.

With that amount of noise on the input, I would try a power supply choke on your input. I'll bet that the noise is coupling through the regulators on the ground, so a good common mode choke on your power and ground input lines might be what you need.

I added inductor to the output of the regulator, then those ripples were seen. Perhaps the regulator fails to pull the inductor. I will change the inductor to the input of the regulator tomorrow. RC filter removed the ripples a bit and also reduced the noise. However the noise is not fully gone. These traces are on TDS2024 Tektronix oscilloscope with 200MHz probe and 2GSPS recording. No extra filter was used.

I placed another transformer secondary in series to the regulator output to see if noise reduces. This did not do much good. Perhaps I have to look at the input strange of the regulator where I have 1000uF capacitor but no ceramic capacitors or inductors. Small ferrite beads did nothing serious to the noise.

I actually needed a high isolation transformers, but these may be leaking the 230V badly. I am really worried about those high frequency noises and do not know, where from they are coming so strongly as if my place is a radio station or RF oven. Funny.

I am going to get to the bottom of all this and won't leave it.

I am going to connect some 6kV isolated DC-Dc converters also to see if they reduce the common mode. I don't expect them to give me clean signal as they have their own high frequency noises.

Are there not any power supplies made for highly noisy power sources. Perhaps the mains transformers may be over loaded and they are radiating harmonics. My concern is all those sharp pulses perhaps coming from bush type motors and now due to hot weather here, every house runs those water coolers around. I wonder if this noise goes down some time of the day or not.

Any one of you solving the problem can have dinner at Gwalior in five star hotel at your wish.

The solution for power supplies run from noisy mains is to filter the main with a common mode choke:

http://www.butlerwinding.com/inductors/common_mode.html

I'm not aware of any other method.

I am also getting this 20MHz and 2Mhz noise in the form of packets which get reduced by inductance by a factor to 1/6th. I do not know the source of the noise as it is only power supply that I am testing with no load. Initially noise is 300mV pk-pk and remains 50mV pk-pk after insertion of inductance. There is another 2MHz similat noise packet coming from somewhere. We have airforce station within 2km so some of it may be from their radar systems or may be something else.

This 2MHz signal packet in noise is also similar to 20MHz noise source and comes as a burst of signal with added 20Mhz on top of it. Inductor removes the 20MHz and 2MHz is then seen more clearly in above waveform.

Funny thing is that these noise parts are linked to 50Hz very clearly and these packets come as many burts of signals within 50Hz synchronous to mains power source.

One uses a regulator to REGULATE the voltage at some point in a circuit. The regulator will then also provide some "smoothing" function, which is similar to what we use capacitors to do. However, regulators will be limited in their high frequency capability by their gain-bandwidth limits of the amplifiers used internally (AND NO HIGHER). So one of the contributors to the noise on a regulator is to use a capacitor on the OUTPUT that has a HIGHER value than recommended by the regulator manufacturer. The reason why the noise can be worsened is that the regulator is working its heart out trying to meet its internal error amplifier requirements to keep the output voltage steady, by charging and discharging the output capacitor excessively. Now these high currents are dropped across the higher ESR (Effective Series Resistance) that the large capacitor has... NO WONDER THE OUTPUT VOLTAGE IS NOISY. So the moral of the story, if you want a regulator to regulate its output voltage, then don't think you can outsmart the regulator by giving it a larger capacitor on its output than the manufacturer recommends. Rather get the INPUT to the regulator to be squeaky clean with the large capacitors (especially at frequencies up to the gain-bandwidth product that the regulator can handle)

Neil

It just happens that on actual board, a lot of capacitance keep adding with each IC on board. Hence, lots of capacitance any way hangs around the regulated power source. With greater tank capacitor at the output, the charge used is same so voltage drop should also be small and regulator need to pump only when voltage drops too much. While regulator uses voltage function, the circuit uses the charge stored. Small capacitor means frequent change in regulator status and larger capacitor means lower frequency control cycle with perhaps multiple bursts of power. I think this has something to do with the packet I am seeing in the control mechanism of the regulator. The peak error with minimum capacitor is 300mV and with capacitors inductor etc 50mV.

Each IC has its own small value decoupling capacitor with low ESR (high frequency), and when you parallel a whole lot of them across the board, the total ESR REDUCES; but if you use one higher value capacitor at the regulator (depending of course on the type used) your total ESR could be very high. The PCB feed lines (positive, negative and ground) also have inductance (etc) which can buffer load from regulator. Have you tried REMOVING ALL high value capacitors on the output of the regulator?

Another thing to consider is the current characteristics demanded from the regulator (eg. its spectrum). Are your noise spikes that you observe related to this current demanded from the regulator, or does the noise come from the source?

Dear Neil

Noise with 100 Ohms load in parallel with 10uF capacitor is shown in top curve 20mV/div and 2.5ms/div.

Bottom curve is after load 100 ohms and 10uF|| 0.1uF EC low pass filter output with 20mV/div and 2.5ms/div.

I did not use low value ceramic filter at load point. Only the low pass has 104 ceramic parallel to 10uF.

Yes, there was significant reduction in noise and it went a factor of 5 or more down. You can still see those 50Hz packets but they are down in size now with 50mA load. Perhaps I should check this out for different loads.

Your observation is excellent.

Noise at 100mA load has shown increased noise than at 50mA load. Perhaps there is some critical point where noise may be the least from regulator and it may be near 50mA. I am using TO-220 package which is meant for 500mA load current.

hi there,

Can you lay your hands on somthing like the National Semiconductor book? I remember years ago, I read in it that those filter elements mentioned at both input and output got to be linked with no more than 4" wires in relation to the Reg. (I from memory, but you better check!). It is also recomended, instead of a 10micro electrolytic cap use a 2.2micro tantalum cap, and ofcourse the typical 100nf cap for spikes (that's hardly the solution in your case).

The main idea here now is to first get rid of the input noise and that is passing through regulator. This is the first priority. I have noticed that noise is a function of time or you can say that there is a lot of load change on mains power which is also the source of noise. Input filters sure going to help and that is the one I have to use with proper selection. Other things I have to do is make somewhat better PCB and use only those parts clearly can reach the lowest noise. Right now noise is about 4mV pk-pk, which becomes 300mV pk-pk at times. Hence, we can say that mains at times get worst. I do not know the reason of having 20MHz noise packets but that is something keeps coming. Which type of domestic load can do that? Perhaps there are few Microwave ovens without noise filters in use around. These may be in good numbers. I have one in my home. I am not sure if this really passed the certification but is from a well known company. Certifications are done on one test model and they are loosen with payment of kickback. It is not easy to know if those can be trusted unless some one does some serious survey.

Now one question comes to my mind. Can we have a huge damping filter or fast response noise cancellation device to get better results? Some of you can put some light of this subject like input mains filter design criteria etc for the rated power.

Since all you want is the 60Hz (or is it 50Hz in India) power-line frequency, just make a low pass filter at around that frequency. That should keep out the 20MHz well enough since it's so much higher than 60Hz. Just keep the filter in a well shielded box. I don't know if using a shielded power cable will help but everything that could prevent your power supply from acting like an antenna might be good.

Most of the good quality instruments come with shielded cables and may be one reason to not allow the signal to get in even at microvolts level. TV and Radio signals and now mobile phone signals should in that range when picked. Perhaps mobile signal is more powerful in the zone with local transmission power around. I do not know the strength of the broadband in telephone line and that is also around.

Yes the idea of local shielding is good one. I am using a plastic casing and not yet using a metal box. I think I need to switch to metal casing and then see what happens. I will also test with Toroidal transformer if that works any better. It is going to be very bulky.

I've had good results with aluminum spray paint. I spray the inside of the plastic box with it. The mounting bolts serve to connect the PCB's ground to the now shielded chassis. I double-check with a multimeter to confirm if the ground to sprayed-aluminum shield is good.

Good Idea. Let me find out if this Aluminum paint is available in the city. I have only known some plastic paints and metal mixed epoxies.

Hi,

Good solid 0v rail.

Ferrite beads on the input, this sort of noise goes straight through a 7805.

Go right back to the mains input...have you got decent filtering there?

Separate 5v rail for digital and analogue?

Is 0v connect to ground? sometimes this is good , sometimes bad...is the whole thing in a screened enclosure?

What is the electrical environment...? I've worked in medical and military but the noisiest electrical environment I have met was a commercial laundry!

Be careful with your measurement techniques as it can make a huge difference where you connect the probe gnd clip. (sorry if I am stating the obvious...but measuring, and removing this sort of noise is notoriously difficult!)

Your are right, but these things are taken care of and I use good tools even for bad parts to test them. Right now I am testing bad transformers and power supplies made from such bad transformers. I have no load but it is still so bad.

If at all you need a bad transformer then come to India. You will get truck load every where. I paid 10% extra for this transformer only to get the worst one.

Hi Shyam,

Have you tried disconnecting the transformer and putting just a load resistor on the secondary. Then take your 'scope and look at the waveform. If the noise is still there, then it's coming from the mains or the transformer itself. If you can connect your scope directly to the mains, you should try that too. Noise from the mains can be very hard to eliminate.

The high frequency (20MHz did you say) may be coming from radios or other devices that operate at these frequencies. Some remote controls even use the ordinary home electrical wiring as their signal cables! If you can identify the offending equipment, you can try fixing it. Otherwise, you get a good quality low-pass filter choke and connect it to the primary of your transformer. Ferrite cores/sleeves/rings might also work.

Try RS Components-India. We use them a lot for our needs.

Another idea...to experiment with.

Don't connect the mains, connect a decent lab powers supply as the input to you 7805 and then have a look, it may help you see if the noise is mains bourne, or coming from your circuitry. (Or picked up by your circuit).

If it is comming in from the mains, use a good quality commercial filter to see if that cures it, if it does then try building your own filter or using simpler cheaper parts to find what is most cost effective.

I had a problem with mains bourne spikes reseting my microcontroller on a cost sensitive application. running Live and Neutral three times through a very small ferrite sleeve solved it.

Back in an earlier career I worked with the inventor of the C.U.K powersupply. This extremly bright guy had his powersupply operating with five feedback loops to control the voltage, current, and noise. I have a memory of a "Snubber Resistor" in the central part of it that provided some noise control.

Or I am just getting old and it never happened that way, I dunknow?

its normal display when you open circuit.

7805 or 7905 is very normal devices in small power suppy.

add a load and 103 capacitor.

if still get large noise, change one new.

If you want low noise what the heck are you using 78xx and 79xx regulators for??

They are really good noise generators as well as power oscillators if not closely surrounded by capacitors of every type...

John.

Hello John

These regulators are listed and used as freely as burgers and coke at the McDonald's.

It is a very specific problem having a noisy power supply using very local transformer made on street shop and common parts to see, how bad the result could be. I am analyzing just that problem and looking at the signals, I am trying to locate the source of noise and its removal process.

If you have any better design then put up and I will record the noise of that design. Most of the power supplies you may be using may be worst than this as people really don't look at noise in depth. I am doing so right down to millivolts and using 2GSPS oscilloscope to record the data.

It is very import to bring out these facts to public and I am an educator. Here every one is putting their idea to something they do not know and I am doing experiment for them to see the results. You can also join with list of capacitors if you like. Each fellow is coming up with list of items to be added or subtracted to make things work.

The next I am going to try out 6kV isolated power supplies after I get to the bottom of this one.

Each person with idea is very important here. I am hearing all and my students watching all this in the class room. It is an exercise that need to come out with some good results.

Shyam, didn't you just author this thread?

http://cr4.globalspec.com/thread/8620/Contract-Electronics-Manufacturing-Engineering?frmtrk=CR4digest

Well I think you've just answered your own question.

I agree and here I can see how well engineers are interacting. There are ideas and no exact solution to the problem. each one has something to say and no one has the final word.

I am reading every line and it is very interesting one. I wish I can call all of your to dinner party.

Did you get others to do your homework for you in college too? Or didn't that start until you entered the 'work force'?

Looks to me that people have offered suggestions based on thier experience and are leaving it up to you to do your own work. Perhaps you should consider outsourcing the problem to an engineering firm.

Hi JJ

I am doing all the home work that other can see on the screen.

If you are looking for outsourcing work then talk in private and you have my email link. This part is for education and some brain triggers to all those are interested. If this is not fun then go to other topics or create your own. I think, there are good people who are responding well and once a while I find some one not so interested in small things. For such people I have top of the world things.

If you are smart and wanna do something serious then I have few projects in the range of $50k to $200K. These are not easy and I am working on these but can also out source to smart fellows. You deliver the product you get the money. that is the way I also work.

These chats are for brain sharpening but also real things. After all you want to see real results. It is impossible to satisfy each and every person so it is worth less to try that.

If you are serious engineer and want to do business in India, then you are most welcome. India is just starting up from nothing in electronics so you can find almost everything salable. Try out and have fun here. I am living in a place where market is in billion pockets. Each pocket is small here but total number is very large.

My email sst (at) sensorstechnology.com is always open for great fellows.

How about using a battery with a low ripple ??

Hi Shyam,

I have seen a situation where noise in the local electrical environment actually came through the power supply of the scope itself. We could connect the scope probe and its ground clip across a 10K resistor and see the noise. In that particular case, the line-locked noise was coming from overhead fluorescent lights.

Yes noise from the lighting system, I remember that. The electronic engineers mantra "What I would give for a really good grounding system"

If your ground is floating up there in the air, you will always be measuring the stuff that comes in on that antenna system. Very good feedback circuits can cancel out much of that but your measurements will be more accurate with that good ground.

To get there a water pipe is not enough, go outside and dig a bit deep. place a copper grid if you can, or just some copper rods if you can't. Soak it a bit with Copper Sulfate, Illegal at you place? just about any salt helps. re-burry the thing, connect your inside ground to it at all electrical outlets. Your actual noise measurement will now be accurate and you can figure out whether to use feedback circuits or filter out the noise.

Works every time I'm telling you.

I fully agree with you. See the difference in #28 due to ground problem.

Dear Mike

To look into what you said, I recorded two waveforms above. Top with power OFF but devices cable inserted into mains socket with neutral connected to transformer. I noticed that noise was higher to about 2mV average and 6mV pk-pk. There is 50 Ohms from regulator output point to GND as load but no power to the regulator. RC filter of 100 ohms and 10uF||0.1uF did not great difference.

However when I removed the connection to the regulator output and GND, I find less than 1mV noise which is perhaps the limit of the TDS2024 DSO. See the traces one below to each high noise. All traces are at 2mV/div and 2.5ms/div 20MHz bandwidth. My oscilloscope reduces bandwidth to 20MHz from 200MHz for 2mV range.

That 50Hz ripple is somehow missing from all traces. Looks like it is radio noise from sources like CFL and other noise sources around.

Hello Shyam et al,

Do you guys know what POO is? It means Point Of Observation (not what you dirty fellows thought it may mean!) So if you get stuck in it, then watch out... you could sink in agony!!!

Unless you take many DIFFERENT Points Of Observation to a problem, then you could get stuck! So Shyam, does your fancy scope have an FFT capability to see what frequency components are coming up in each trace you take? This will tell a much more compelling story!!! Don't stare yourself blind, rather use the tools at your disposal to extract the FREQUENCY components of the noise to get another point of view... don't guess!

Neil

Hi Neil

I have to ask my son to look into FFT part as that must be somewhere in the software. This DSO downloads data to PC so using software on PC must be easier.

If I remove those probes and use shielded box with BNC cables then sure the noise will be down. My hardware are build in 3mm thick Aluminum casing and I/O are through BNC connectors and good 50-ohms cables. Some how this experiment was conducted in open and power supply is in plastic casing. It was not intended to be very great.

Perhaps if I use DC-DC converters that are isolated type, then differential noise in the power sources may get cancelled. I am hoping that but not very sure as DC-Dc also have lot of noise ratings. I wonder where I can buy calm DC-DC.

I am right now evaluating few types of DC-DC and also looking for high isolation low leakage transformers. I do not get good transformers in India for instrumentation except for those big toroidal transformers.

Hello Shyam,

Your download data from your DSO could even be sucked into EXCEL, and use its FFT function, and display the graphs on the spreadsheet. Dedicated PC software for FFT would probably be better in long run.

Reasonably priced and good DC-DC converters can be obtained from Traco Electronic ag www.tracopower.com which are from Switzerland. I have used them OK.

If you don't want noise to get through a transformer, then simply specify the manufacturer to separate the primary and secondary with a piece of brass shim or aluminium foil, which must then be EARTHED. All capacitively coupled noise from the source will then be shunted to ground, and your light current engineers will be happy chappies!

Neil

Watch out!!!

Don't let the transformer manufacturer short the SINGLE TURN OF FOIL, BECAUSE THE TRANSFORMER WILL BE LOADED WITH A SHORTED WINDING. THE SINGLE TURN OF FOIL SHOULD BE OVERLAPPING AND THE START INSULATED FROM THE END.

Neil

Good. You want to keep the induced voltage at near ground voltage reference for the foil without making it to lose power.

In Fixed or Variable 3/4 pin V_regulators you have to connect cmall mica capacitors in I/P & O/P as hi-freq filters.

See Rs coponents App note supplied on order with components.

I will see if I can trace somewhere their values.

Is someone close by using power line networking product? What kind of communication use 200MHz and 20MHz signal? The power cable is a big antenna picking up lots of signal.

Pineapple

Broadband communication is used in power line as well as in the telephone lines and both lines are more or less exposed. The net is through board band and perhaps the power line systems monitor the power line equipments and power problems through power lines. These monitoring should be on main back bone of power line and not at distribution point as power metering has not yet reached to my zone. It looks like a collective noise generated by the equipments used in about few hundreds of users. There is no regulation on power quality so each person hooks whatever one can hook on the like. India is in very primitive stage and both equipment industries and electricity departments are equally bad. Some electricity department fellows are so bad that they even sell the power lines to thieves and they themselves are thieves. Public is somewhat mixed type so some are some what good and some are little bad and some are worst. People put hooks to steal power and electricity department fellows do not ever disturb them as those fellows have little to lose and they may beat these people for disconnecting their free power tapping points. I can put picture if you like to see these. Right from my house window I can see dozens wires hooked.

Perhaps in developing countries this stage is not considered all that bad. Here people who can not put on shirt properly lay fiber optic cables and their children become top software engineers. It is all like diamonds in coal mine. I watch all this and it is really very interesting to be here after I have been all over the world. It is very funny life here. I am using the net right now having no shirt as it is very hot so I removed it. I am running a water cooler to increase the moisture as it is too dry out here. Power is survival here but it is all types of power that I get from 100V to 500V in same wires.

I haven't read the whole thread, so apologies if this is repeating others' contributions:

According to National Semiconductor's data sheet, the supply rejection of these regulators is only about 60-dB at (probably) the optimum output current of 40-mA, and this degrades above 10kHz. So you should anticipate almost no intrinsic rejection above 10-MHz, particualrly at low output current. Referring to the graph of output impedance, a good quality 1-uF tantalum load capacitor is likely to provide about 26-dB suppression when combined with the regulator. If you cannot quieten your input supply directly (it sounds to be a bit of a dog, with a 20MHz internal resonance), you need to place a filter on the input to the regulator. This should be a well-damped LRC; a 0.1-uH inductor in series with 0.1 Ohms* feeding a 10-uF tantalum load could be worth trying as a first step. (The self-resonant frequency of the inductor needs to be above 20-MHz). If the problem includes radiation from the supply, you may find a toroidal inductor beneficial.
You also need to pay attention to board layout - don't let the power-supply ground current pass through the operating circuitry, keep the ground substantial around the capacitor, and avoid multiple ground paths that can bypass this route.

*Although the resistor will only dissipate in the region of mW on average, it needs to be able to withstand an input inrush of a few mJ (=C.V^2/2 depends on input supply Voltage)

Dear Physicist

That is pretty good information. Let me try out this filter at the input stage. Perhaps supply noise rejection is not working well.

I been having similar problems designing my own shunt regulator for audio use.

Does your scope have FFT options? if so, you should be able to sample the waveform and pick out any harmonics and identify their frequencies.

Make a list of these harmonics & then remove the output of your transformer & find out it's resonant frequency.

It could be that the reverse recovery pulses of your Bridge Rectifier diodes is exciting the LC Resonant Circuit of the secondary winding of the mains transformer & giving problems...Change them for Schottky Types & Use LCR Circuits to trap the resonance frequency before & after the bridge

I've managed to identify 49Khz harmonics in a 3-4mV signal that relate to the resonant frequency of the transformer.

Otherwise it could be mains grunge getting into the system from a number of sources & filters are the only means of preventing them getting into your design

Linear regulators are renowned for letting hf through. Their internal amps are compensated to make them stable.

Hope this helps

Brian

That is good information. Perhaps, the diodes are not so good. I am using 1N4007 diodes so will try to change these Schottky diodes I have from On Semi. I will also try to remove the resonance frequencies of the secondary windings. In thin case frequencies in the spectrum are at 2MHz and 20MHz and these can be cut down by various means.

Good thinking.