Enclosed SMPS versus open frame regarding noise

[RogerAF]

Thanks for the suggestions but it is a very specific application for a product that was sold in the thousands around 30 years ago and needs to be a simple 'Plug and play' solution for those who are seemingly incapable of just repairing the old, simple linear supplies. Replacing a few dried out electrolytics is SO tough!

What about just having the supply do a slower, current-limited ramp up?

 

[Matt Syson]

I have made a 'solution' but not been able to actually install and 'prove' it yet. There are a number of other constraints that I have managed to work around. stopping 'RF' getting into audio circuitry is a good game or at least it makes you think a bit.
Fitting a switchmode into equipment that was never designed for such powering is like bringing in a dog with fleas. It can be done but needs care.

 

[thor.zmt]

Thanks for the suggestions but it is a very specific application for a product that was sold in the thousands around 30 years ago and needs to be a simple 'Plug and play' solution for those who are seemingly incapable of just repairing the old, simple linear supplies. Replacing a few dried out electrolytics is SO tough!

Increase SMPS current. Rating. Start with 5A.

Not able to change electrolytic Caps? Probably through hole as well... Peshaw.

Thor

 

[thor.zmt]

What about just having the supply do a slower, current-limited ramp up?

Most SMPS do not do this. If they have any kind of soft-start it will be measured in milliseconds.

For an "Audio SMPS" I tend to include a slow start monotonic ramp-up as well as a hold-off delay for hic-up protection. But these need to be designed in at point of initial design, it is next to impossible to fit it aftermarket.

I also tend to design in a C/L-L/C/L/C output filter with a section that has equal value lower value inductors with low DCR in the first section and often an added common mode inductor (also low DCR) and a single relatively high DCR inductor in the positive line of the second L/C cell and the DC portion of the feedback from the output after the final LC cell to give good regulation.

I also use a special transformer design that eliminates (or at least dramatically reduces) the need for the so-called "Y" capacitor which tends to leak 1/2 mains voltage into the PSU output Ground.

This generally gives a PSU that is comparable or better than a standard LM317 etc. regulated linear PSU. This can be done at fairly notional extra cost with any "commodity" style of SMPS that is already decently designed.


This is a 15VA plug top type PSU design. The Switching IC is a 67kHz Quasi-Resonant converter which is in effect uses not quite "zero switching" but valley switching, which offers a similar benefit in terms of switching noise as zero switching. CY1 is 100pF due to custom switching transformer design to reduce noise coupling.

Key visible differences to "commodity" types of the same power level are 33uF instead of 22uF input capacitor and 3 inductors added to the output capacitors which are high ripple current rating solid electrolytic types, instead of generic low impedance types.

The LCLC filtering with ~6kHz & ~600Hz corner frequencies drops the 100mV Peak-Peak switching noise at rated output current (this is actually the common design goal) of the commodity PSU to theoretically 100nV (120dB total attenuation), in reality a little less due to parasitics, say sub 1uV, due to the quasi resonant nature the switching frequency is subject to a natural jittering that actually further lowers observable noise.

Together with the small size components not visible under the PCB that adjust the generic circuit this makes a very much "linear PSU stand-in" with comparable noise.

OEM cost in 5kU is not materially larger than a "commodity type", as items like casework, cables, plugs, main IC, Transformer and other core electronics remain the same. It would be easily for many large SMPS OEM's to offer "Audio Grade" or "general purpose low noise" SMPS's at a modest price premium by simply engineering then correctly. It is just that (almost) nobody perceives a market need for such a design it seems.

Thor

 

[user 133392]

I made some measurements to show a customer who had questions about using switching supplies with the Flat Moving Coil Preamp. Thought I would share the results here...

"Comparative FFTs of the Meanwell P25A, Meanwell GP25A and the TDK CUT-35 to a linear benchtop supply.

The differences, with one exception, were underwhelming...

The load was the Flat Moving Coil Preamp with the gain set at 61.5 dB.
Flat Balanced Input Moving Coil Phono Preamp Construction Information - Pro Audio Design Forum
The source impedance was the internal 10Ω termination.
The Ein is around -140 dBu in all cases. (20-20 kHz measurement BW).
The ambient hum field of the room was nulled in all measurements by positioning the board.
The FFT is the Left and Right Outputs with the display open to 48 kHz. (FS=96K 24B FFT 65Kpt 5 sample avg.)

The Meanwell P25A is quieter when the 5V supply is preloaded. It is not loaded in this measurement.

The Meanwell GP25A, which replaces the original, is a very electrically noisy supply and no amount of preloading helps.
The GP25A's spread spectrum "hump" above 20 kHz pokes out of the noise floor.

Benchtop Linear Supply

Meanwell P25A (discontinued)

TDK CUT35

Meanwell GP25A ("Green Power")

The GP25A is an electrically noisy supply."

In response to the OP I've found that fields from the primary side of the switcher do radiate but not necessarily more so than a linear supplies' transformer would. If the supply is co-located in the chassis distance is your friend. Call it "range anxiety."

In the measurements above all the supplies were at least a foot from the DUT.
I have found in testing the TDK CUT35 with the Flat Preamp is that as long as the input section of the board is at least 6 inches from the primary side it is relatively quiet.
Though I have both the enclosed and open CUT35 to test I used the open frame version.

https://www.proaudiodesignforum.com/forum/php/viewtopic.php?t=1123

 

[RogerAF]

In response to the OP I've found that fields from the primary side of the switcher do radiate but not necessarily more so than a linear supplies' transformer would.

Does that data include a Toroidal Xfmr for linear supply? That shouldn't radiate much I would think.

 

[user 133392]

Does that data include a Toroidal Xfmr for linear supply? That shouldn't radiate much I would think.

I didn't have one close by to comparatively test today.

But, I do have a client who has asked me to look at another designer's work where he placed the device's mix amp right near the toroid.
It hums far more than it should - the mix amp is within a couple of inches and its definitely from the transformer's field.
With the metalwork and PCs already done the only available option is shielding.
They were going to buy some mu-metal off eBay and I haven't heard back from them.
So toroids are not perfect and distance is still your friend.

I did find an interesting issue however regarding magnetic radiation from transformer core saturation in linear supplies due to DC in the secondary.
The particular example was a "flat pack" type transformer with unbalanced positive and negative loads.
The problem child was a Texar Audio Prism. Texar Audio Prism Schematics and Information - Page 3 - Pro Audio Design Forum
And also: Bipolar Power Supply Load Current Imbalance And Transformer Core Saturation - Pro Audio Design Forum

The clue came when I added load to the opposing rail to lower the hum field.
The DC in the windings was saturating the core in one polarity.

What I learned that day is that a conventional split supply with a single four diode full-wave bridge may look like a full-wave bridge but its not.
With regard to each individual polarity with load current returning to ground it's two full-wave-center-tapped bridges.
Loads supplied by half-wave bridges or full-wave-center-tapped (two diode) bridges setup DC in windings.
Unbalanced loads on a split supply with load currents returning to ground setup center-tap DC current.
Enough imbalance can push the core into non-linear saturation and the harmonic content of the hum field blooms.
And it runs hotter.

I've never tested it but it looks like dual winding secondaries and dual full-wave bridges would eliminate the problem.
The solution, in the Texar, was a TDK switcher.
The infernal thing ran as hot as a furnace, now its cool as a cucumber and far quieter.

 

[thor.zmt]

Does that data include a Toroidal Xfmr for linear supply? That shouldn't radiate much I would think.

Toroid's have a field that is circular around the core.strongest field is along the axis of the doughnut hole.

Mind you, the field on the 90 degree axis from that is still not that low.

Thor

 

[user 133392]

I redacted the designer's and client's name in this pic.
The mix amp is right below the transformer.
I found I could reduce the hum a few dB simply by rotating the transformer.
IIRC the minima wasn't with the lead exit farthest away but I may be wrong.
The circular field around the core may have nodes.
I recall this being the "before" shot.



Questions for thor.zmt:

1) Have you looked at the DIYAudio/Jan Didden Silent Switcher? What are your thoughts? Jan has said he has no interest in manufacturing it now.

2) What's going on in the Meanwell GP25A?

It has a spread-spectrum "hump" (which may be an alias with FS@96K) that narrows as the 5V rail is loaded. (See: Meanwell Single, Bipolar and Triple Output Switching Power Supplies for Audio Projects - Pro Audio Design Forum )

The +15V rail has what appears to be a 1/2 line frequency synchronous tri-wave ripple even when the 5V loading spec is met. (See: Meanwell Single, Bipolar and Triple Output Switching Power Supplies for Audio Projects - Pro Audio Design Forum )

The top trace is +15; bottom -15V about 50 mA load on the rails. 500 mA min load on +5V.

The original P25A didn't have this ripple: Meanwell applications engineers didn't seem to think the GP25A "green power" sample I had was defective.

 

[thor.zmt]

1) Have you looked at the DIYAudio/Jan Didden Silent Switcher?

Yes. Not DIY friendly, uses difficult (Chipaggeddon) to get TI Chip's that are expensive. And the design is wholly unimaginative.

It would be trivial to do better:



Single Chip (U700) single inductor (below PCB, 6 X 6 X 4.5mm) with +/- 15V @ 0.3A continuous from a 3.7V/4,000mAh LiPo Battery, U703 is the battery charger/controller, output voltage is MCU adjustable from +/-6V to +/-15V continuously (PWM Pin), plus 5.2V from the Battery Charger IC when not charging.

There is also a -24V zener regulated, low current to bias J113 & J109 JFet audio switches and it would be trivial to add 48V and/or boost voltages.

All outputs are additionally LC filtered, switching frequency is 1.2MHz nominally and can be synchronised with an external clock, e.g. 1.411MHz or 1.536MHz or with the battery charger which cannot be synchronised externally.

The IC under the U700 label (Taiwan company made in china) is good for around 10W output from around 12W input power with 3.7V as source. More powerful options exist.

In principle, as long as the input voltage remains notably below the main +/- Rail voltages we can allow greater output power and higher output voltages are also possible if the input voltage is boosted.

So 12V would allow +/-24V @ 0.6A and with the next larger IC in the lineup we are looking at +/-24V @ 1.2A needing 12V/5A (continuous - so a 10A rated china adapter is recommended). At absolute maximum +/-48V @ 0.6A are possible, though that gets awfully close to absolute max stress values for all parts.

But I have no intent to produce something like this. Just too much hassle for little pay. I pay upfront to make 500, sell a few each month, even if I were to majorly gauge on the price it would not pay for my time just to pack this stuff and send it out.

2) What's going on in the Meanwell GP25A?

Honestly, there are so many options to loblox up (as we say on planet anagramia) a simple switcher, never mind one with multiple outputs, that I'd have to analyse the the internal design in detail.

I prefer to talk about how I would go about things based on my experience delivering switching power systems for audio that don't have these problems.

The original P25A didn't have this ripple: Meanwell applications engineers didn't seem to think the GP25A "green power" sample I had was defective.

They are only concerned if the unit meets factory spec, which I am pretty sure it does. Past that they don't give a hoot.

Thor

 

[RogerAF]

Toroid's have a field that is circular around the core.strongest field is along the axis of the doughnut hole.

I wonder what the effect would be of adding a big ferrite cylinder down the center of the doughnut? And/or just MuMetal shields top and bottom?
Just blue sky curiosity.
Not planning to build one anytime soon. And If/when I do, it will probably have the power supply as an outboard unit with power and sense lines to the audio device. That way I can have a UL licensed mains power in and my own low voltage supplies out. Even a wall-wart if it has enough current.

 

[Tubetec]

Ive tried a few small switcher type supplies for tube projects , with limited enough success ,

As I mentioned before Ive used the camera flash pcb to generate a few hundred volts HT from a 1.5v AA cell .
I tried several brands of camera , the Kodak pcb with the single transistor had by far the least garbage at the output ,
how can I best filter a circuit like this , the Hammond 15x series chokes have some low current high inductance types , with several kOhms resistance, only 0.25lb weight ,there physically small and dont cost much ,
how would I arrange any CM filtering along with this and what types of inductor are suitable for small currents/high voltage ?
Everything after the 100uf is unused , with extra RC-RC-RC sections it works fine but seeing as the Hammond iron is easy to get maybe its worth trying . I have plenty of reclaimed 100uF 330v caps to throw at it also ,
It would be nice to repurpose the flash PCB in the form of a quiet/stable bench HT supply of a few mA for testing tube mic circuits .




ohms Lbs


The other questions I have is how can I generate a quiet LT supply from 120v DC ?
I found an off the shelf switcher module which can do 120v down to 5 or 12v at 3A ,
It might feed a 6.3v 150ma heater supply , how can I best arrange the extra passive filtering and noise canceling circuits etc?
https://www.ebay.com/itm/225134946575

 

[thor.zmt]

I wonder what the effect would be of adding a big ferrite cylinder down the center of the doughnut? And/or just MuMetal shields top and bottom?

Generally a Mu-Metal wrap around the outside is used in most equipment. The vertical field is only bothering if stacking and the transformer end up below or above sensitive circuitry.

In this case it is better to Mu-Metal shield affected circuitry as EM fields fall off by the square of distance. Even a few cm reduce field strength a fair bit so shielding directly the noise affected circuitry always delivers superior results.

That's why commonly put audio transformers into mu metal cans but not mains transformers.

Thor

 

[thor.zmt]

Ive tried a few small switcher type supplies for tube projects , with limited enough success ,

OTS supplies are crapshoot in the dark and once the model you found to work well is replaced by an "improved version" that now no longer works well, as it was never designed for Audio.

As I mentioned before Ive used the camera flash pcb to generate a few hundred volts HT from a 1.5v AA cell .

I have used an IC version to generate the HT (adjustable too) for electrostatic headphones. 650V dC from 9V input without voltage doubler.



It is extremely quiet even when working. In this setup it charges a film capacitor bank and the headphone and then turns off. To generate HT continous switching would be needed.

how can I best filter a circuit like this ,

Make sure the frequency is high enough (> 500kHz) and use appropriate SMD Chokes and Capacitors.

how would I arrange any CM filtering along with this and what types of inductor are suitable for small currents/high voltage ?

Mouser has a huge catalog of suitable SMD Parts.

The other questions I have is how can I generate a quiet LT supply from 120v DC ?

Step-down converter with 120V DC Input rating.

I suspect this needs a re-purposed off-mains AC->DC adapter switcher IC, or literally the PCB from a 115VAC in to desired out PCB stolen from a plug-top supply. Add LC/RC filtering to eliminate noise on input and output.

I found an off the shelf switcher module which can do 120v down to 5 or 12v at 3A ,
It might feed a 6.3v 150ma heater supply , how can I best arrange the extra passive filtering and noise canceling circuits etc?

That is just standard electronics, no difference to (say) a tube rectified power supply, except frequencies are different and thus parts needed.

For example, if you have (say) 10H/470uF for 100Hz supply, 100mH/4.7uF or 10mH/47uF will do the same at 100kHz and 10mH/470n or 100uH/47uF will do the same at 1MHz.

For high frequencies designers tend to use more C and less L.

For stuff that switches up to ~ 132kHz throughhole parts are usable, but often adding SMD to mix can lower noise.

Anything higher needs SMD parts IMHO, but has the advantage that there is a HUGE range of commodity parts to choose from.

TI has models of some of their switchers for TINA-TI, allowing unusual topologies to be modeled and simulated.

Thor

 

[Tubetec]

Thanks Thor