Why are most audio power supplies unregulated?

[fubar.gr]

Many (most ?) audio amplifier power supplies are of the unregulated type (transformer, rectifier bridge and capacitors). Since there's no voltage regulator, they use a ton of capacitors to keep ripple low.

I did a google search on the topic and found some claims that unregulated PSUs sound better, accompanied with other audiophoolery claims, so I can't take them seriously.


So what is the real reason for using unregulated power supplies?

[SeanB]

It is cheaper, and if the amplifier PSRR is good enough it will not have any difference on what it provides as drive to the speakers. As well it reduces power dissipated in regulators that have to have a higher power rail than the amplifier output stage in order to do regulation, and you save on having the high power regulators as well.

If for example your amplifier will operate on 30v rails then the unregulated rail only needs to give 30V unloaded. If you add regulators you will need a supply to them of possibly 35-40V to give the required regulated supply, and if you are driving an 8R load with 75W ( doable with a 30V supply), you will need to only have a transformer that can deliver around 100VA to the bridge rectifier. If you have regulators the power needed goes up to around 120VA, which is a bigger transformer, plus an extra 30W or so of heat in a heatsink as well on the power devices.



Shoot, just thinking Deja Vu again with this topic.

[German_EE]

The short answer, because they don't need to be regulated.

Most audio equipment (and I'm also thinking of battery operated equipment here) works without regulation because nothing has been designed to tight tolerances, amplifiers designed for 9V will work just as well at 8V or 7V but might start to distort once you reach 6V or so. Mains powered equipment is the same and a power amplifier with a 25V supply will happily still work at 20V but at slightly lowered output.

The same applies at RF. Look at a schematic diagram of a tube power amplifier and there will be an HT supply of 2000-3000V which is unregulated, mainly because regulation at these voltages can be difficult. Any amplifier will keep going as the supply continues to reduce until the bias voltage falls out of specification, after which it will start to distort.

[Mark Hennessy]

There is a whole bunch of sane reasons why unregulated power supplies are the best bet in a sensibly designed product. Rather than regurgitate it all here, check out this book:

http://www.unge.gq/ftp/biblioteca%20digital/Amplificadores/Audio%20Power%20Amplifier%20Design%20Handbook.pdf

Chapter 8 deals with this issue. But it's worth digesting the whole thing - it's a fantastic resource.

[bigsky]

Do remember that larger amplifiers require substantial power supplies - up to a few kw rating - the cost of regulating these would be substantial.

An old sound engineer once told me that sonic differences between different audio amplifiers only really become noticeable when they are running flat out. I think this was a wise statement as that is what stresses the design. In such circumstances, there could be some merit in the claim that unregulated supplies result in better sound. I expect that an unregulated supply would have a better transient response than the equivalent regulated supply.

(Amplifier amplifier design invariably involves engineering compromises, especially at higher powers. Audio signals are highly dynamic and it is seldom appropriate to design an amplifier, including the PSU, based on a constant sine wave signal).

However, note that a lot of modern amplifiers have switch-mode PSUs which will invariably be regulated.

I second Mark's recommendation of Douglas Self's book, and anything else by him. At the other end of the scale, his "Small Signal Audio Design" also is a masterpiece.

[dfmischler]

An old sound engineer once told me that sonic differences between different audio amplifiers only really become noticeable when they are running flat out. I think this was a wise statement as that is what stresses the design. In such circumstances, there could be some merit in the claim that unregulated supplies result in better sound. I expect that an unregulated supply would have a better transient response than the equivalent regulated supply.

To the point that part of the sound of the classic tube (valve) guitar amplifiers (e.g. Fender, Vox) is caused by the inability of the power supply to keep up when the amp is running flat out.  Audiophiles don't value clipping and compression, but it can be just the right sound, and quite musical, in the right context.  I'm sure the designers at the time weren't looking for that effect; they were just doing things the way they were done then.

[T3sl4co1l]

If you think of it this way:

The amplifier is a bidirectional voltage regulator.

Then it should be apparent why: pre-regulating it would be fairly redundant!

Now, there are a lot of amp designs out there that suck too much, and require regulation... no one said the audio community was smart.

Tim

[bigsky]

An old sound engineer once told me that sonic differences between different audio amplifiers only really become noticeable when they are running flat out. I think this was a wise statement as that is what stresses the design. In such circumstances, there could be some merit in the claim that unregulated supplies result in better sound. I expect that an unregulated supply would have a better transient response than the equivalent regulated supply.

To the point that part of the sound of the classic tube (valve) guitar amplifiers (e.g. Fender, Vox) is caused by the inability of the power supply to keep up when the amp is running flat out.  Audiophiles don't value clipping and compression, but it can be just the right sound, and quite musical, in the right context.  I'm sure the designers at the time weren't looking for that effect; they were just doing things the way they were done then.

He was referring to transistor PA amplifiers, not tube guitar amplifiers. For example, I would expect a QSC Powerlight (lightweight switch mode PSU) to be audibly different to a Crown Macrotech (heavyweight linear PSU) at rated power. I'd also expect the differences could be measured - for example different THD figures, and different spectrums for the distortion, and for this to vary with different types of input signal which stress the circuitry in different ways. Unlike audiophools, I take the view that if you can't measure it, you can't hear it!

[dfmischler]

He was referring to transistor PA amplifiers, not tube guitar amplifiers. For example, I would expect a QSC Powerlight (lightweight switch mode PSU) to be audibly different to a Crown Macrotech (heavyweight linear PSU) at rated power. I'd also expect the differences could be measured - for example different THD figures, and different spectrums for the distortion, and for this to vary with different types of input signal which stress the circuitry in different ways. Unlike audiophools, I take the view that if you can't measure it, you can't hear it!

It really doesn't matter: it's only a decade or two between the cherished guitar amp designs and the advent of transistor amplifiers.  When those old guitar amps were designed the folks who did it were just following standard amplifier practices of the time.  They didn't know that the "defects" would become a valued part of the sound.

I remember the Crown DC300A fondly.  Drunk college kids could set it up and run it and pack it up and it survived, night after night.

I also believe that what can't be measured can't possibly be heard.

[coppice]

I also believe that what can't be measured can't possibly be heard.

True, but be careful not to confuse that statement with "what isn't being measured can't be heard". Serious audio engineers have a history of being just as unrealistic as audiophiles, but in the opposite direction. Engineers started seeing fantastic results from audio amps in the late 70s and early 80s, and insisted they were doing a fantastic job. All the while the actual audio quality went down and down and down. Then people started worrying more about dynamic performance (e.g. TID) than static performance (e.g. THD), and amps improved again. In between was a period of the best engineers being called idiots.

As far as regulation in audio amps is concerned, preamps usually regulate their supplies. You don't want lots of 50/60Hz getting into the sensitive circuitry of something like a mic pre-amp, and if the mains has a lot of distortion you might have considerable energy across much of the audio band. A pre-amp regulator is small and cheap and runs cool. Its a no brainer to have one. The PA is another matter entirely. It takes a really badly designed PA to let much of the supply noise appear at the output. It is a natural ripple rejecter. A regulator to achieve a stable, but lower, supply voltage just burns off energy and gets hot (even if its an SMPS it will still be wasting some energy). What you need is enough stored energy to ride over the mains cycles when the largest possible drain in output current occurs. That means having a substantial stack of capacitors.

[bigsky]

He was referring to transistor PA amplifiers, not tube guitar amplifiers...

It really doesn't matter: it's only a decade or two between the cherished guitar amp designs and the advent of transistor amplifiers.  When those old guitar amps were designed the folks who did it were just following standard amplifier practices of the time.  They didn't know that the "defects" would become a valued part of the sound.

I think it does matter.

Valve power amplifiers have low open loop gain and little, if no, negative feedback. There is plenty of scope for non-linearities, and the specifications of individual components will directly affect the amp's performance. What you call the "defects" will always be present  (although they will be increased at higher power levels).

Transistor power amplifiers use op-amp techniques - high open loop gain and lots of negative feedback - to linearise the transfer function. The design overcomes the shortcomings of individual components. At low power levels, a well-designed transistor amplifier will be almost perfect. It's only at higher power levels that "defects" will become evident, due to, as I said before, the necessary compromises required in real-world design.

A simple example is frequency response. Variations in the frequency response of components in the signal path of a valve amp directly affect the response of the overall amplifier. With a transistor amp, it doesn't matter. Provided that the open-loop gain is much higher than the closed loop gain throughout the bandwidth, the negative feedback will compensate for any variations and the overall response will be as the designer intended.

What I am saying is that if you took two valve amplifiers, of different topologies, but the same overall specification, and evaluated them at, say 10% of rated power, you would find measurable differences (eg frequency response and distortion) that were audible under blind testing. If you repeated the exercise with different transistor power amplifiers, the differences would be far less, maybe even unmeasurable, and blind testing would be unlikely to reveal any audible differences.

[vk6zgo]

He was referring to transistor PA amplifiers, not tube guitar amplifiers...

It really doesn't matter: it's only a decade or two between the cherished guitar amp designs and the advent of transistor amplifiers.  When those old guitar amps were designed the folks who did it were just following standard amplifier practices of the time.  They didn't know that the "defects" would become a valued part of the sound.

I think it does matter.

Valve power amplifiers have low open loop gain and little, if no, negative feedback. There is plenty of scope for non-linearities, and the specifications of individual components will directly affect the amp's performance. What you call the "defects" will always be present  (although they will be increased at higher power levels).

Transistor power amplifiers use op-amp techniques - high open loop gain and lots of negative feedback - to linearise the transfer function. The design overcomes the shortcomings of individual components. At low power levels, a well-designed transistor amplifier will be almost perfect. It's only at higher power levels that "defects" will become evident, due to, as I said before, the necessary compromises required in real-world design.

A simple example is frequency response. Variations in the frequency response of components in the signal path of a valve amp directly affect the response of the overall amplifier. With a transistor amp, it doesn't matter. Provided that the open-loop gain is much higher than the closed loop gain throughout the bandwidth, the negative feedback will compensate for any variations and the overall response will be as the designer intended.

What I am saying is that if you took two valve amplifiers, of different topologies, but the same overall specification, and evaluated them at, say 10% of rated power, you would find measurable differences (eg frequency response and distortion) that were audible under blind testing. If you repeated the exercise with different transistor power amplifiers, the differences would be far less, maybe even unmeasurable, and blind testing would be unlikely to reveal any audible differences.

 "Valve power amplifiers have low open loop gain and little, if no, negative feedback"

Sorry,but that is not so,except in really cheap radio output stages.
Valve amplifiers with any pretensions to HIFI status.all used negative feedback.

The limitation on NFB in valve amplifiers is the possibility of phase rotation in output transformers causing oscillation at supersonic frequencies.

I've tested valve amplifiers----distortion figures & frequency response errors were far below anything audible.
Testing the valve amplifiers,back to back with solid state replacements showed no discernible audio difference.

[ciccio]

He was referring to transistor PA amplifiers, not tube guitar amplifiers. For example, I would expect a QSC Powerlight (lightweight switch mode PSU) to be audibly different to a Crown Macrotech (heavyweight linear PSU) at rated power. I'd also expect the differences could be measured - for example different THD figures, and different spectrums for the distortion, and for this to vary with different types of input signal which stress the circuitry in different ways. Unlike audiophools, I take the view that if you can't measure it, you can't hear it!

The difference will be audible to any listener, when reaching the power limit of the amplifier and maintaining that condition for some time: the non regulated power supply (if decently designed) will keep it's available power constant, the switch-mode  PSU will go into some kind of protection, and the amplifier will not sustain it's rated output for more than some time (sometimes seconds), then the amplifier will clip.
I've seen/heard this in any switch-mode powered amplifier I've tested.
I've never done a detailed cost analysis, but I believe that a correct switching PSU will cost more than an unregulated one.
Weight / Bulk and the fact that a classic unregulated PS will not be compliant with the modern requirement for Power Factor Control are obviously not taken in account.
If the amplifier itself is not a linear design, but a switching type, the amp itself will reduce it's power output when stressed, and the reduction of available maximum power versus time will  be emphatised. 
This is the reason I'm still selling conventional power amplifiers: I've customers that ask for them: they do not want the lightweight and cool new models.