Power supplies

[trl]

Well, it still has linear regulators in the end.
http://www.electronics-lab.com/silentswitcher-quiet-mains-free-power-supply/

​

However, it's a good thing that more and more low-noise SMPS regulators appear on the market (mostly from XLSEMI and TI) and I'm confident in few years the 10 Hz - 100 KHz output noise will be low enough to be used successfully in audio equipment, without needing the use linear regs as followers. I also hope to see such SMPS regs fully protected internally too.

 

[SIY]

Best of both worlds?

Jan likes to say, "Specify the design and accept the performance, or specify the performance and accept the design."

 

[trl]

Just measured today the AC ripple & noise for the well-known XL4016 boost-converter and compared it with a regular linear supply.

SMPS boost-converter powered from a DELL laptops SMPS power adapter (AC RMS: 178 mV)

Linear PSU with powered from an isolation transformer (AC RMS: 0.4 mV)

​

Pure resistive load, 12V/3A, boost-converter, DIY linear PSU​

DIY linear PSU (the two paralleled BD249 output transistors are placed behind, not shown here)

​

Pure resistive load, 12V/3A, boost-converter

​

The 4 bucks Alibaba SMPS boost-converter (https://www.aliexpress.com/item/32788530654.html)
​

In this case the difference in AC ripple & noise between the two power supplies was a bit over 350 times, 0.43 mV vs. 177 mV, so quite a difference, but probably normal for such a cheap SMPS converter device. Although, I'm sure with a bit of care and engineering this can be improved a lot.

 

[DonH56]

I am not sure this matters, you are comparing apples and oranges... The Dell supply feeds a battery (which acts like a big capacitor) and then a bunch more SMPS' on the motherboard to power everything else. So it goes through additional regulators before anything sees the output from the external brick. Including any audio circuitry.

If you are driving a DAC with poor PSRR, then you wouldn't (shouldn't) be using the brick for a laptop anyway. You can design an SMPS for very low ripple (ones I use are in the <100 uV range then filtered after that) or not, depends upon the end application. The 12 V rail in a PC can be fairly ugly because it is sent to a bunch of other regulators. The place to compare to a linear supply would be at the end point. In our case, an SMPS delivers far more power with far less noise and wasted energy than a linear supply. But we also specify higher-tier caps and all that jazz to meet our 10-15 year MTBF specs.

 

[trl]

I am trying to point out measurement differences between a regular SMPS vs. a regular LPSU. Like you said, without linear regulators as followers to a SMPS the ripple can't get rid of. Benchmark guys did the same, they used linear regulators after the included SMPS, which is great, of course.

It would be interesting to see a really low-noise SMPS "in action", under a load, if you have the time for. Thanks!

P.S.: I do have a very low-noise 150W SMPS; I might try to feed that boost-converter and see how the ripple looks like.

 

[Speedskater]

Measuring the external bulk power supplies is of little interest. The important supple is the internal one in the component.

Try measuring the signal-to-noise ratio differences at the audio output.

 

[DonH56]

The regulators after the SMPS wall wart in your notebook are also SMPS', just designed to a different target spec.

I can't share internal data from work but there is nothing to see. With a 20~30 A load the only noise on a DSO is its own noise, and a spectrum analyzer shows a noise floor in the 10 to 100 uV or so region. Note these are low-voltage supplies, ~1 V to 5 V, not power amp or even audio line amp voltage rails.

There are a myriad of design trades in design a regulator, SMPS or linear, so comparison is really sort of pointless. A notebook brick does not have to be clean because it is just the rough power source for all the internal regulators. It does not need to be clean. As @Speedskater said, what matters is the noise at the audio output. SMPS have a lot of advantages over linear supplies for audio, but as others have noted need attention to detail in their design and implementation for reliability and long life.

What you have to love are the companies selling high-dollar linear wall warts/bricks to replace the SMPS versions, only to have them drive an SMPS inside the box...

 

[trl]

I was able to power a BURSON PLAY (board rev. V2.2) from an external DIY linear power supply and found little to none differences with ARTA:

PLAY V2.2 with the original SMPS

PLAY V2.2 with the DIY linear PSU

​

I've also attached two pics with background noises only where a bit of mains hum can be seen on the LPSU image, although completely inaudible with my most sensitive 16 Ohms IEMs.

I can see that 3rd harmonic on the LPSU graph disappears and I have no idea why, but the rest of the spikes are all still there.

Note: The AC RMS + noise of the LPSU was about 400uV, per attached picture; test done with 12V/3A, resistive load. LPSU is based on the standard schematic with a 24V/6.67A - 160VA isolation transformer, a 5A bridge rectifier, a LM723 chip followed by a BD139 and 2 x BD240 with heatsink and fan, so nothing special. Current drawn by the PLAY during the test was 0.6A while playing.

L.E.: Worth mentioning that inside PLAY there are SMPS boost-converters that might not care much if the main power supply is switching or linear design.

 

[trl]

The Dell supply feeds a battery (which acts like a big capacitor) and then a bunch more SMPS' on the motherboard to power everything else. So it goes through additional regulators before anything sees the output from the external brick. Including any audio circuitry.

Actually the battery inside a laptop doesn't acts like a capacitor at all, because it's separated by the built-in regulator and it's not even active in feeding power to the laptop, unless you have a gaming laptop or a desktop replacement laptop that has this feature: for short-term loads/peaks the power is loaded from both PSU and internal battery.

If a poorly designed SMPS emits lot of RFI around, injects DC into the mains and perhaps outputs more than 1% of ripple into the output circuit, maybe the audio USB devices connected might not get affected, but laptop's internal regulators will struggle in their operation, same applies to the caps inside the laptop, hence will lower their lifespan. A low ripple and low RFI is desirable to any SMPS, even for the SMPS used in light appliances (mostly to the DC induced to the mains).

 

[DonH56]

The battery still looks like a capacitor but the point of my post was that it is effectively isolated from the actual power supplies inside the computer. Read again all the words after ..(...capacitor).

Perhaps the most critical part of SMPS design, aside from choosing quality components, is the board layout to ensure the current loops on the input side and output side are constrained. That is something every manufacturer I have worked with emphasize and something that will cause all sorts of noise issues if done poorly.

 

[Wes]

When I talked to them at AXPONA a couple years ago, they made a big deal about how all the wires from the inlet to the plugs were the same length. "This insures coherence." God knows you don't want incoherent power.

or incoherent engineering analysis...