Enclosed SMPS versus open frame regarding noise

[Potato Cakes]

Hello, Everyone,

I've been working on a particular project where I thought I had sorted out my PSU noise issues using an enclosed SMPS. For the lastest version of this project where added the final feature set, I used an open frame SMPS to save space. This generated hum centered around 60Hz for the class A sections and raised the noise floor significantly for the ICs. I tried moving the PSU out of the case, filtering, isolating the various amp circuits, and changing grounding schemes with no luck. I went back to an enclosed SMPS and the noise went away, even without any sort of filtering. I'd still like to use an open frame SMPS for the space saving benefit, but I don't know what specs I would need when considering using one in an audio circuit. I have used the medical grade and cost effective versions with the same results. The enclosed SMPS' all seem to work with audio circuits no matter the brand.

Has this happened to anyone else? Has anyone here used a particular open frame SMPS in the same chassis as the audio circuit with great success?

Thanks!

Paul

 

[scott2000]

Apogee Rosetta200 has an open frame SMPS .
There is some kind of metal wall around it though... Like an L-shape iirc...
Maybe there is added filtering circuitry around it though....idk..

edit...knew I had a pic from when it was open...

 

[efinque]

I used a Mean Well open frame SMPS in a mixer build and didn't notice any hum.

It was rectified and stepped-down using a Velleman DC module though.

 

[JohnRoberts]

it may be related the switcher electronic design, not just the packaging.

JR

 

[Bo Deadly]

it may be related the switcher electronic design, not just the packaging.

JR

Agreed. It doesn't matter that much if it's enclosed or open. What matters is the design. Some SMPS are hopelessly noisy and there's nothing you can do about it. Some are quiet from the start and work without any filtering at all in some cases. Like many things in audio, it all depends on the implementation.

My guess would be that the enclosed designs tend to be LED drivers and the powers at be decided that the LED drivers need to be quiet. Probably justified because they are ultimately delivering power to a long unshielded strings of diode junctions that could make for a very bad source of EMI.

 

[Potato Cakes]

Thanks for the response, everyone. I suppose just did a great job picking the wrong open frame SMPS models. I'll check out the Mean Well versions and compare to the ones I have.

Thanks!

Paul

 

[LeeharveyOsmond]

have a look at the Studer 961/62 smps they seemed to dial it in along time ago

 

[thor.zmt]

I think with "enclosed" vs "open frame" you are barking up the wrong tree.

A few basic trouble-shooting tips for SMPS noise.

1) All SMPS use full wave/bridge/Voltage doubler rectification, meaning any ripple will be 100/120Hz depending on local mains frequency.

There is literally no way by which 50Hz/60Hz can make it "through" a SMPS, so if you get 50/60Hz it goes "around' the SMPS in some way.

2) Due to a number of reasons, many SMPS will go unstable and/or into what is called burst mode with light load. Despite switching at 67 or 132kHz for most commodity of-main SMPS this can lead to audio band interference.

If you get this - you will need to either preload the SMPS until it becomes stable or change the design.

3) SMPS high frequency noise can fold back into the audio band due to IMD. It is always a good Idea to add additional passive filtering and/or regulation that that eliminates switching noise to the audio noise floor.

The way an SMPS is mounted in the case can make a big difference.

4) Even using pre-certified SMPS in your design, if you sell it commercially you are liable to require any and all applicable agency approvals, UNLESS the SMPS is external and self contained with an output voltage classed as SELV (Safe Extra Low Voltage).

Medical grade SMPS are recommended for audio use, as they will have much lower leakage currents from the mains into the ground compared to "cooking" SMPS.

Thor

 

[Newmarket]

I think with "enclosed" vs "open frame" you are barking up the wrong tree.

A few basic trouble-shooting tips for SMPS noise.

1) All SMPS use full wave/bridge/Voltage doubler rectification, meaning any ripple will be 100/120Hz depending on local mains frequency.

There is literally no way by which 50Hz/60Hz can make it "through" a SMPS, so if you get 50/60Hz it goes "around' the SMPS in some way.

2) Due to a number of reasons, many SMPS will go unstable and/or into what is called burst mode with light load. Despite switching at 67 or 132kHz for most commodity of-main SMPS this can lead to audio band interference.

If you get this - you will need to either preload the SMPS until it becomes stable or change the design.

3) SMPS high frequency noise can fold back into the audio band due to IMD. It is always a good Idea to add additional passive filtering and/or regulation that that eliminates switching noise to the audio noise floor.

The way an SMPS is mounted in the case can make a big difference.

4) Even using pre-certified SMPS in your design, if you sell it commercially you are liable to require any and all applicable agency approvals, UNLESS the SMPS is external and self contained with an output voltage classed as SELV (Safe Extra Low Voltage).

Medical grade SMPS are recommended for audio use, as they will have much lower leakage currents from the mains into the ground compared to "cooking" SMPS.

Thor

50 / 60 Hz can still find its way around to the output. Particularly in compact designs.

 

[Matt Syson]

On way mains frequency or doubled can get 'around' and through a switcher is due top the similarity to AM radio where the switching frequency is 'modulated' by the mains 'ripple', as frequently the reservoirs after the rectifier are intentionally small. Switchers that have Power Factor Correction are even more 'exciting' as they are a curious switcher ahead of the main switcher design.
Bear in mind the commercial Linear supplies easily manage to have ripple voltages of around 1 millivolt or less, even up to 10 Amp units and as long as it's rectifiers are properly snubbed, there are few further mains ripple components above a few Hundred Hz which in turn is unlikely to capacitively couple into audio circuitry as long as basic wiring hygiene' is observed. The noise specification for many switchers class 100 millivolts as 'good' which is of course 40dB 'noisier' than any competent linear design AND with the disadvantage that noise, common or differential mode is high frequency and can couple capacitively quite easily. One conole manufacturer posted an article in Studio Sound magazine around 20 years back concerning their findings when their desks which previously used linear supplies went head to head with a switcher unit they were attempting to use where the sound engineer noticed 'increased distortion' of high frequency audio. Resolved by decent filtering after the switcher modules. When switchers are used you also need to consider what will result when other switcher interference is 'summed' (Sum and difference frequencies) if your audio cabling picks up 'radiation' from phone chargers, LED studio lighting or other HF sources, all carefully recorded on 192K convertors. Whether a switcher is good for any project depends on strict attention to detail.

 

[thor.zmt]

50 / 60 Hz can still find its way around to the output. Particularly in compact designs.

Precisely, around. By magnetic (highly unlikely) or capacitive (highly likely) coupling.

Compact or not is not relevant. No matter how compact, I never really worry about 50/60Hz leakage in my SMPS designs (commercial) compared to 100Hz.

Here the output of a compact (plug top type) SMPS I designed. This is a nominally 15W design with universal input and full agency approval delivering 1/2 rated power. Observe that the 50Hz component is around 20dB down on 100Hz.



Thor

 

[thor.zmt]

On way mains frequency or doubled can get 'around' and through a switcher is due top the similarity to AM radio where the switching frequency is 'modulated' by the mains 'ripple', as frequently the reservoirs after the rectifier are intentionally small.

That ripple will be 100/120Hz, not 50/60Hz.

Typically 50/60Hz are the result of the stupid version of measures to ensure compliance with EMC regulation (e.g. FCC)

Switchers that have Power Factor Correction are even more 'exciting' as they are a curious switcher ahead of the main switcher design.

Normally these days switching frequencies are synchronised. As PFC is a Boost Converter operating from a bridge rectified but not capacitor smoothed mains voltage, the ripple passing through it will be double the mains frequency.

The noise specification for many switchers class 100 millivolts as 'good' which is of course 40dB 'noisier' than any competent linear design AND with the disadvantage that noise, common or differential mode is high frequency and can couple capacitively quite easily.

I would call that very low grade switchers not suitable for audio, without additional circuity. Mind you, 1mV 100Hz is also not suitable for Audio MNSHO.

One conole manufacturer posted an article in Studio Sound magazine around 20 years back concerning their findings when their desks which previously used linear supplies went head to head with a switcher unit they were attempting to use where the sound engineer noticed 'increased distortion' of high frequency audio. Resolved by decent filtering after the switcher modules. When switchers are used you also need to consider what will result when other switcher interference is 'summed' (Sum and difference frequencies) if your audio cabling picks up 'radiation' from phone chargers, LED studio lighting or other HF sources, all carefully recorded on 192K convertors. Whether a switcher is good for any project depends on strict attention to detail.

Correctly designed switchers with full agency approval and EMC regulation compliance tend to have fairly decent behaviour in that.

Many Chinese SMPS makers have two versions of their models, one domestic and one "export" (EU/USA).

They often cannot be told apart externally except a suffix on model number (and weight). Internally mains side filtering is minimal, capacitor values are lower etc. They often also do not match the power spec on the full rated input voltage range, some times not even on 230V, while the export versions do match specs.

If buying from Aliexpress or E-Bay it is likely what you get is is the chinese domestic production.

Thor

 

[Matt Syson]

All in all you can't just 'throw' a switcher module into any gear without proper consideration of likely noise issues, inrush surge current etc. The final implementation has to be fully tested and 'proven' to be of equivalent quality if it is replacing previously supplied linear designs with noise artifacts properly evaluated. It has been shown (IIRC) that human hearing can perceive single 'tones' when even 6dB below the 'average' (white) noise level which I think was addressed in the various weighting curves that were applied to test results where a 12dB 'boost' at around 7KHz would give interesting and revealing 'data' about a piece of audio gear. Using two (or more) switcher modules to power 'split rail' gear invites sum and difference 'noise' as two switchers with non synchronised switch frequencies may well be 'out' by a few kHz which is of course well within hearing range.
Having only been installing mixing desks in professional studios for about 45 years it is interesting to see the cavalier attitude to some 'recent' designs. I was being generous with my comment about a maxi of 1 millivolt at mains frequency from a linear supply. The Coutant and Kingshill supplies I used to work with specify around 200 microvolts peak to peak (referred to as low frequency noise as I assume they weren't really looking a lot beyond 20KHz .

 

[Newmarket]

Precisely, around. By magnetic (highly unlikely) or capacitive (highly likely) coupling.

Compact or not is not relevant. No matter how compact, I never really worry about 50/60Hz leakage in my SMPS designs (commercial) compared to 100Hz.

Here the output of a compact (plug top type) SMPS I designed. This is a nominally 15W design with universal input and full agency approval delivering 1/2 rated power. Observe that the 50Hz component is around 20dB down on 100Hz.

View attachment 102405

Thor

Well "compact" can be an issue. Given that a more compact design may have greater capacitive coupling. Although I'd agree that it is also very dependent on layout / power planes / PCB stackup etc.
Even -20dB down on the rectified frequency might be an issue in some applications (not necessarily audio).

 

[Bo Deadly]

Observe that the 50Hz component is around 20dB down on 100Hz.

I see a lot of noise there and not just at 50Hz. An off-the-shelf supply will make a much better spectrum than that.

Here's spectrums of a 150W console supply with phantom:

Source: AH ML5000 PSU Replacement

 

[ccaudle]

Here's spectrums of a 150W console supply

What is the 0dB reference there?

 

[thor.zmt]

I see a lot of noise there and not just at 50Hz. An off-the-shelf supply will make a much better spectrum than that.

Here's spectrums of a 150W console supply with phantom:

Source: AH ML5000 PSU Replacemen

Observe the scales.

The noise is absolute. The highest noise component is 100Hz at 2uV or -114dBV which dominates measured overall noise.

As this is a 9V PSU (e.g. for a pedal board) the noise is 0.000022% of Vout, or -133dB below Vout.

Here the same supply from a comparison test, 180mA & 360mA current, respectively.



Here a 9V battery for comparison:



Here a "typical" china made linear powersupply. This one uses R- Core transformer, several big capacitors and 3-pin regulator, rating 9V/2.5A, price ~ USD 100:



Battery is lowest noise, but my design SMPS is also quite low noise. The "cooking tech" linear PSU, less so.

Thor

Images from:
Power Supply (8+) Group Test, LPS and SMPS

 

[Bo Deadly]

What is the 0dB reference there?

Totally arbitrary but perfectly satisfactory for relative comparisons. I think I just eyeballed 0dB to the noise floor. For whatever reason it looks like the mix bus output is 4-5dB lower.

Normally I would use a test tone to find the clipping point into the capture device and make that 0dBFS. That would be a lot more meaningful. But in this case the capture was being performed by someone else who just sent me a .WAV file. So trying to establish a proper 0dB reference was not worth the trouble just to make comparative measurements.

 

[Bo Deadly]

Observe the scales.

The noise is absolute. The highest noise component is 100Hz at 2uV or -114dBV which dominates measured overall noise.

As this is a 9V PSU (e.g. for a pedal board) the noise is 0.000022% of Vout, or -133dB below Vout.

Thor

While your log 10 y axis is a little difficult to visualize I apprecate that the noise level is very small.

However, my USB interface can do 120 dB no problem and I have seen completely flat noise floors with SMPS.

This is a APC-16-350:

Source: Mini SMPS Mic Power Supply

In this particular example the SMPS was under-loaded which is the source of the whistle at 2kHz and it was located physically very close to the mic pre inside a small enclosure so you can see a little mains. But still, it's pretty flat compared to yours so I think my point stands: don't try to roll your own SMPS because the off-the-shelf units are hard to beat.

But respect to you for trying.

 

[thor.zmt]

You say "mine does -120dB" without stating what is 0dB. That means your statement cannot be evaluated.

If it is for example an Audio Interface with 0dBFS at +22dBu or +20dBV, to make things comparable, the top of my chart would be at -60dB and the bottom at -180dB.

In this case the top of my 2uV/100Hz would just pokebabove your -140dB like and most of what visible would be below your charts cutoff, making my SMPS nearly 40dB better than what you show.

Thus stating the reference is vitally important.

If you compare the other measurements I referenced, you can see that my SMPS design is only marginally noisier than a battery. This is for a design based on commodoty parts, mass manufactured with under 3 USD BOM cost including accesseories and packaging.

Note that I do longer have any current association with the company marketing the design and I do not recommend the product as it is seriously overpriced.

Bit you are right, I'm only trying and there is no point to what I do.

Thor