Filtering the HF hash of a SMPS

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Help me with designing a filter to get rid of some of the HF noise from a brick 12VDC SMPS and to learn something.

I decided to use an external SMPS based on what I have around (a 6A 12V brick or a 12V AC trafo for halogen lighting). The linear supply would require quite a bit of voltage dropping which with installing the trafo would require a bigger case.

I'm powering 4 shunt regulators drawing a total of around 3A. They will work off 10V so I have 2V spare to drop in the filter. I was thinking to use a CRC or a CRCRC, each C being a 10uF tantal paralell with 100nF ceramic. I chose this based solely by feeling so please educate me.
Also I'd like to know how to simulate the effects of different filter on the HF noise once I know what filter to use. Unfortunately I have no scope so I can't measure the HF content of my SMPS to give you some figures. I'll have design the filter blindly, expecting the worst and do the best filtering witouth breaking the bank. Please no top of the line "audio" components", i want engineering.
And a side question, what HF characteristics do PTFE/AL foil capacitors have? There is cheap russian old stock around.
Thanx,
Marko
 
Firstly you need to measure the HF hash if any coming out of the SMPS, then the filtering can be targeted, as you cant its all guess work... Not really recommended.
Also look at Henry Ott's sites...
home page
and this...

http://www.compliance-club.com/pdf/EMCTestingPart1.pdf
http://www.compliance-club.com/pdf/EMCTestingPart2.pdf
http://www.compliance-club.com/pdf/EMCTestingPart3.pdf
http://www.compliance-club.com/pdf/EMCTestingPart4.pdf
http://www.compliance-club.com/pdf/EMCTestingPart5.pdf
http://www.compliance-club.com/pdf/EMCTestingPart6.pdf
http://www.compliance-club.com/pdf/EMCTestingPart7.pdf
EMC Design
Design Techniques for EMC
Design Techniques for EMC: cables and connectors
http://www.compliance-club.com/currentissue/print_article.asp?artid=500.
Design Techniques for EMC - part 4
PCB DESIGN TECHNIQUES FOR LOWEST-COST EMC COMPLIANCE
Design for EMC Part 6: ESD, dips and dropouts, etc.

My self, I would be look at common mode and differential mode chokes as well as Pi filters with caps and ferrites...... without something to aim for though...:)

Leaded devices are not the best for this, you have to use RF design, that means ground planes careful layout etc.
 
You can't design a filter by "feeling". Neither can we. You can't do "engineering" without a spec.

To filter out HF you need some vague idea of what frequencies are involved and how much filtering you need. Otherwise you don't know whether you need a filter at all, or if you need an impossible-to-practically-implement filter.

Having said that, HF can be filtered using LC - if the F is H enough. This is especially true when you need low DC impedance (you need low DC impedance to get less than 2V drop on up to 3A current).

Personally, I would not 'bypass' the caps with smaller caps - this could make things worse. I might 'bypass' the inductors with resistors, but lowish Q inductors might not need this.

You are probably dealing with frequencies from the SMPS switching rate (some 100's of kHz?) plus harmonics up to a few MHz. You can assume that almost any capacitor will act as a capacitor over this range, except highish value electrolytics - even they will maintain a lowish impedance.
 
You can't design a filter by "feeling". Neither can we. You can't do "engineering" without a spec.

You are probably dealing with frequencies from the SMPS switching rate (some 100's of kHz?) plus harmonics up to a few MHz. You can assume that almost any capacitor will act as a capacitor over this range, except highish value electrolytics - even they will maintain a lowish impedance.

So I'll borrow a scope next week and see what Fs are we talking about.
Until I do let's do a practical design that would attenuate the probable switching F and harmonics. What filter would you use to get 50 kHz to 5 (10?) MHz down 60 dB @ 3A and drop two Volts.

I'd like us to come to a practical solution with this example so I can re-do the design myself when I get the measurments.

I'm powering a DCX2496 digital crossover with 4 ReflectorD shunt power supplys from this forum's member Salas. They do just fine without the HF filter for now but I'll measure if some hash does get through once I borrow the scope. And I'd like to know how to design a HF filter anyway.

And hey, thanks for your help!
 
Disabled Account
Joined 2010
So I'll borrow a scope next week and see what Fs are we talking about.
Until I do let's do a practical design that would attenuate the probable switching F and harmonics. What filter would you use to get 50 kHz to 5 (10?) MHz down 60 dB @ 3A and drop two Volts.

I'd like us to come to a practical solution with this example so I can re-do the design myself when I get the measurments.

I'm powering a DCX2496 digital crossover with 4 ReflectorD shunt power supplys from this forum's member Salas. They do just fine without the HF filter for now but I'll measure if some hash does get through once I borrow the scope. And I'd like to know how to design a HF filter anyway.

And hey, thanks for your help!

You assume there is some noise between pos and neg output of your smps - this might be true. In that case we talk of symmetrical noise that can be reduced by a symmetrical LC filter or similar. Unfortunately, this is only the first half of the truth.
There exist asymmetrical (or common mode) disturbances as well that require a different type of filter (common mode filter). You need basic knowledge of both types to deal efficiently with filtering.
 
a wise man once said "return the noise from whence it came".
filtering is not always the best way. filtering often will push the noise from one end to another. input to output or visa versa. sometimes you can drive the noise into the grounds to pop up again elsewhere E.g Common mode sourced.
best to optimize or add snubbers to absorb the hash energy from where it comes E.g. switches and rectifiers. once that is done look at filters and decide based on measurements and topologies. bricks for laptops is more line ripple than hash once loading near max.> brick topology is pretty bad generally, you get what you pay for, so upgrade to decent half frame SMPS can save grief, time, and money in the long term.
 
On reflection we are discussing EMC, Henry Ott, Ralph Morrison etc. have written books about it, the EMC club is full of articles as well as notes and papers by many others...
On this or any other thread all we would be able to look at is a specific well defined problem. :)
 
Pure audiophile nonsense, there is a wealth of equipment far more sensitive to audio that runs on SMPS's as well as audio gear, including a lot of professional stuff... you are just generalising...
Linear supplies also produce noise , quite often in the audio band.
In this case there is no indication of noise and no empirical data just a presumption.
Like using ferrite beads, this view that SMPSs are all bad gets a bit boring...
 
Maybe not all are Garbage ;)
Absolutely the 2! chinese SMPS gizmos I bought are.
First one dumps massive energies to the item being 'powered' .
Second one outputs virtually No power.
Yet both read their spec Voltages when metered. Lovely.
Yes SMPS do work at least in ones' desktop computer.
But It's gonna be a frosty day in Hades.. before I buy one that Isn't a Brand Name Computer supply from a reputable store.
 
I'll get a scope in a couple of days and post the result.
In the mean time I found in my junk box a 15V SMPS from a dead NuForce Icon amp. It's rated at 4.5A so that should be good. The quality of the inbuilt filtering should be above that of the no-name 12V one I first wanted to use.

I've read the papers marce posted but aren't much wiser now. Also I've found a site that made things a bit clearer.
KA7OEI's blog: Reducing switching supply racket (RF Interference)
KA7OEI's blog: Quieting high-current switching power supplies used in the ham shack.
 
So I loaded the Nuforce SMPS with 3A and hooked up the scope. It was ugly, 50 mV peak to peak, 40 kHz saw-tooth ripple with plenty of noise riding on that.
The Salas shunt's output looked clean though, as clean as a AC transformer powered one.
Then I disassembled the SMPS, there was one AC cap and a skinny choke and on the output just capacitors.
Sorry, no pictures, I was in a hurry.
I decided to use a transformer PS, with both input and output CLC filtering. Thanks for your help, I did learn something new after all and most importantly I didn't end up using a bad power supply.
 
Got the make of the supply... 40kHz is a bit low for switching frequency in my view for audio to near the audio band and thus hard to filter out without affecting the upper frequencies, though not impossible with a notch filter. Higher frequency switching, not only because the components can be smaller and more compact, but also the higher frequency's are better for layout, the switching loops can be kept tighter and as they follow more the path of least inductance the return path can also be tighter minimising the noise spread outside the switching loops.
For purely digital this would not matter so much as the circuitry is more immune to noise, again though 40kHz is hard to filter out for clean power islands especially with compact digital layout, again higher frequencies can be combatted with simple pi filters to isolated supplies using cap ferrite cap.
 
Thanks for your help, I did learn something new after all and most importantly I didn't end up using a bad power supply.

Not sure how you determined that the power supply was 'bad' for your application - it doesn't look like you used the 'bad' power supply for powering your equipment and then measured a 'bad' change in your equipment's response?
 
I didn't want to use a sub-par switcher that lacks RF filtering and has a low switching frequency, but I didn't measure the change in the response.

I replaced the original SMPS (which also lacks RF filtering in the output) in the DCX digital crossover with shunt regulators. Although the shunts do a good job filtering the hash from the brick SMPS I just don't want any more RF in the crossover case then necessary. And I had the parts for the linear supply on hand.
Also I got my hands on a written-off UPS and took out the AC line filter components and there were some nice low R iron core inductors there. I'm going to use those in the new PS.

Sent from my C6903 using Tapatalk
 
I thought it was just worth checking whether you ever had poorer, or improved, audio performance with the variety of power supply changes you made. It's fine if you want to conjecture a possible link between changing a power supply and getting an audio improvement, but it is also likely that there has been no change to your audio performance with any of the power supply changes you have made.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.