Linear or switching power supply in Sound Process

[steven2583]

I recently had a mod(LR channels new op amps, PGA2310 and voltage raised to 12v) done where an additional switching power supply was added to my sound processor for the +/-12v. The unit had to go back for something and mod company found out the electronics needed was overloading the switching power supply. So they added another one. So one switching power supply runs the +12v and one runs the -12v. Waiting for the until to come back to hear it. Wondering if I should just change the power supply to a larger linear or a bigger switching. What usually sounds better in a pre-amp?

[jitter]

Switch mode power supplies (SMPSs) will probably be frowned upon by most audio enthousiasts. The reason for that is that a SMPS produces switching noise which, if inadequately filtered, might affect sound quality.

By the choice of my words you'll see that I'm not taking a fixed standpoint here, I'm trying not to be prejudiced as I haven't experimented with linear vs switch mode myself. But I do think that cheap off the shelf SMPSs might not belong to the "well designed" category and might therefore harm sound quality. I also believe that it's harder to design a good SMPS than a good linear power supply.

I know of at least one manufacturer of professional and high end audio equipment that uses SMPSs in their products, where appropriate. They actually make a point of mentioning it in their product brochures.
With that in mind, I don't think that an SMPS is by definition worse than a linear power supply. You'll just have to experiment to find out...

I work for an industrial electronics OEM. In that kind of electronics, linear power supplies are only still used in some very specific cases (e.g. sensitive measuring equipment or intrinsically safe power supplies). If SMPSs were incapable of producing clean power, they wouldn't have found their way in some of the products that we make, e.g. TV studio equipment or medical ultra sound equipment.

[steven2583]

The power supply is from LC AUDIO which I would think would be decent since they are a mod company. That is not who I received the mod from. It just seems odd that I have a different one for the positive and negative DC. Never heard of doing that before.

[jitter]

Originally Posted by steven2583

[COLOR=Black]The power supply is from LC AUDIO which I would think would be decent since they are a mod company.

You would certainly think that they chose that particular SMPS for a good reason.

That is not who I received the mod from. It just seems odd that I have a different one for the positive and negative DC. Never heard of doing that before.

Aha, the plot thickens. I certainly find it strange to use only half of one symmetrical power supply and the other half of the other. I don't know too much about the peculiarities of SMPSs, but do they allow for such unsymmetrical loading of the outputs?

If not, the only right thing that should have been done was yank the underpowered one out and replace it by a chunkier item.

[twest820]

Originally Posted by jitter

The reason for that is that a SMPS produces switching noise which, if inadequately filtered, might affect sound quality.

If the switch frequency's sufficiently far above audio its subharmonics lie above the audio band I wouldn't be particularly worried about this. Unfortunately the switch frequency is among several parameters not specified by the OP, such as the converter type, feedback type, compensation type, LC selection, load, and so on. So the answer to the question, as posed, is indeed just "It depends."

[gmarsh]

Linear power supplies have switching noise too - it's produced by rectifiers, and sits at 60Hz/120Hz plus harmonics. Plus there's all sorts of resonances, stray inductances, etc. that can exist in a linear power supply and ring during diode conduction.

In my experience, it's often easier to clean up a SMPS output than a linear power supply output.

[jitter]

Originally Posted by gmarsh

Linear power supplies have switching noise too - it's produced by rectifiers, and sits at 60Hz/120Hz plus harmonics. Plus there's all sorts of resonances, stray inductances, etc. that can exist in a linear power supply and ring during diode conduction.

In my experience, it's often easier to clean up a SMPS output than a linear power supply output.

Several questions arise:
- Doesn't a SMPS also use rectifier diodes?
- Isn't the switching noise of a SMPS far "louder" than that of rectifier diodes?
- Won't the little bit of rectifier diode noise in a LPS get suppressed by voltage regulation?

[zjaakco]

Hi Guys,
I walk around for long time to create a smps. As I looking around on the web, for the best cd players, the linn cd 12 pops up. A long time the best cd player in the world.
And wen I read the review of stereophile again, the first time was long ago, a few things remark me. The 8x oversampling. digital filter, and a switch mode power supply. I think if one will a real good , clean power supply, a smps is necesarry.
And here I am on a point, I dont no how to start. I prefer to make a diy, and as good as possible. I will a smps in a dac.
Do you think this is realizable?
best regards
zjaakco

[steven2583]

What's better to have on a power supply for the regulator output side 1 large capacitor or 2 smaller ones?

[twest820]

Depends on the caps, the regulator, and your purpose. A meaningful answer requires a specific context. Typically it doesn't matter if the caps are selected appropriately and the layout's done well. Personally I tend to prefer a single larger cap as it makes the layout more compact and favors lower ESR parts selection.

Originally Posted by jitter

Doesn't a SMPS also use rectifier diodes?

Depends on what exactly one considers an SMPS and what exactly one means by rectifier diode. Broadly speaking, non-isolated topologies follow a linear supply with a switchmode regulator involving some sort of Schottky catch diode which you'll find in the rectifier diode section of a parts catalog. Isolated topologies typically involve a second bridge on the output. Pretty much any reference will include schematics. Wikipedia, for example.

Originally Posted by jitter

Won't the little bit of rectifier diode noise in a LPS get suppressed by voltage regulation?

The diodes themselves will be quieter than the regulator but I think you're probably asking about some combination of power supply snubbing, EMI, and ripple. For a given load regulated switch topologies usually pull less current through the bridge and trafo than a regulated linear supply so, given good bypassing and layout in a hard switched topology (or resonant conversion), there's less somewhat less demand on the snubbers and typically a little less EMI. Switchmode output ripple is usually larger than that of linear regulation, but it depends on the regulators, load behavior, parts selection, phase margins, and so on. However, the switching frequency can be selected so the ripple is above the audio band whereas with linear supplies the ripple is in the audio band. For any given design you have to work through the supply coupling mechanisms, GBPs, slew rates, PSRR as a function of frequency, intermodulation opportunities, and so on to figure out the tradeoffs. With current generation parts usually the control loops have sufficient PSRR, feedback, and slew rate the nature of the supply doesn't matter much if the layout and bypassing is reasonably well done.

There are also a number of cases where the comparison of interest is regulated switchmode versus unregulated linear or even unregulated switchmode to unregulated linear. Typically switchmode has lower ripple and lower output impedance but, as always, there are exceptions. For example, with a voltage feedback buck converter with type III compensation and a few amps of output it can be difficult to select the output LC to admit a truly low ESR cap in the vicinity of 100uF without clobbering the phase margin. This leads to a supply impedance peak that tends to be around 150 milliohms at 150kHz with a typical 500kHz switcher. In comparison a linear supply can hand off from a 10,000+uF to a low ESR ~100uF in the same frequency range, resulting in something like a 15 milliohm impedance in the same frequency range. Is the difference meaningful in an audio design? Not usually.

Since the OP asked about solid state preamps: in those applications I'd default to linear supply and regulation. The efficiency advantage to switchers is limited and linear regulator thermals aren't much fuss due to the relatively low supply rails and current draws. The audio interface I use uses a 3x7 pair with the board as their heatsink (about 70 cents per regulator in DIY quantities plus the cost of the board area) with op amps that are 37 cents each (NJM4565) and hits an IMD floor of -110dBFS. Good enough for most things.