The Truth About HiFi Amplifier Power Supplies

[Angsty]

Not sure if this was shared in the earlier pages, but it has appeared on ASR previously. Benchmark has some thoughts about linear versus switching power supplies.

Audio Myth - "Switching Power Supplies are Noisy"

This Myth Goes Something Like This: "Switching supplies are noisy." "Linear power supplies are best for audio." We disagree!

benchmarkmedia.com

 

[fpitas]

I've designed a lot of very sensitive electronics using switchers. It can be done.

 

[Punter]

Switchmode power supplies are fitted to just about every piece of sensitive medical equipment out there. If they were "noisy" and caused electromagnetic interference, they would be useless in that application. The golden ear brigade are so convinced that everything modern is somehow evil and that it is poised like a vulture, ready to steal your enjoyment of music that they miss out on equipment that has been designed and manufactured with modern thinking. How do they square this thinking when it comes to DAC's and streaming? Oh, the answer is, they allow snake oil peddlers to convince them that this equipment has all the same faults as stuff designed in the 70's and therefore needs special cables, routers, switches, etc etc.

 

[TimF]

As a result of comments on ARS, particularly by restorer-john, if I recall correctly, I got curious about ADCOM. I found a used ADCOM GFA 545 ii in very good condition and bought it. I've been listening to it for the last week and enjoying it. I'll keep it so I can occasionally alternate with it and a NAD class D amp. It's not hard to find good clean ADCOMS here in the USA, but then it's also not hard to find beat up worn down ADCOMS too. I know not to go poking around the insides of the ADCOM. In any event, thanks to restorer-john on steering me to the ADCOM.

 

[ahofer]

Somewhat OT, but I'm proud to announce that my son, with only a year at TI under his belt, has had his new power supply design accepted and built. For a consumer device from a very large company. It may well be tested here.

In his words, "everything in it has been done before, but they hadn't been combined in this way".

He sent me pics, but not my place to post 'em. Of course, it has to get through reliability testing....

 

[fpitas]

The toughest part of the design is picking the cord.

 

[solderdude]

Nah... it just needs to have some rope in there.

 

[atmasphere]

 

[Cbdb2]

I simulated current bandwidth ages ago: https://www.whatsbestforum.com/threads/power-supply-bandwidth.15198/#post-275474 -- have not updated and ported over here, think I decided not worth the effort. But when I click on the pictures "over there" to make them large enough to see I get a "not found" warning. I'll have to dig up the originals sometime.

However, the results are in line with what @solderdude showed, natch, so nothing really new or worthwhile. It was to show that power cord bandwidth for current needs to be a couple orders of magnitude greater than just 50/60 Hz, but not a MHz -- at least for a conventional supply. I don't have data for a SMPS; that might be interesting. Be nice to repeat on a Hypex or whatever class-D amp wall input and see.

4 pages in and finally the crux. Current spikes. Thats how a linear supply works. They can be huge in power amps and that can lead to noise issues. The output power and the transformer/filter caps dictate the current spikes, but the output impedance of the AC into the transformer (including the power cable and fuse) can have some effect.
So a resistor in the power line will flatten out the spikes (less current but you can still get almost the same power in, it just takes a little longer) and may make an difference in 50/60hz (and harmonics) noise pick up in an Amp thats not layed out properly, by keeping these large current spikes from inducing into the rest of the circuit.
So that expensive power cord is only connected 10% of the time. As is the power conditioner.

 

[fpitas]

4 pages in and finally the crux. Current spikes. Thats how a linear supply works. They can be huge in power amps and that can lead to noise issues. The output power and the transformer/filter caps dictate the current spikes, but the output impedance of the AC into the transformer (including the power cable and fuse) can have some effect.
So a resistor in the power line will flatten out the spikes (less current but you can still get almost the same power in, it just takes a little longer) and may make an difference in 50/60hz (and harmonics) noise pick up in an Amp thats not layed out properly, by keeping these large current spikes from inducing into the rest of the circuit.
So that expensive power cord is only connected 10% of the time. As is the power conditioner.

Sometimes the rectifier switching transients can be a problem, too. Ironic for a "linear" supply. Generally the solution is an old -fashioned RC snubber on the diodes.

 

[atmasphere]

Sometimes the rectifier switching transients can be a problem, too. Ironic for a "linear" supply. Generally the solution is an old -fashioned RC snubber on the diodes.

Actually the 'diode noise' is a product of the power transformer. To snub it, a capacitor in series with a resistor can be placed across the output of the transformer at the rectifier input. Because its more of a transformer issue, its best to keep the transformer leads as short as possible.

How it works is something called a 'swept resonance'. Its caused by the inductance of the transformer in parallel with the capacitance of the rectifiers and is energized by the commutation of the rectifier. Bypassing the individual junctions of the rectifier will work; its putting out the forest fire rather than blowing out the match.

The correct way to snub the rectifier is still to use a resistor and capacitor in series around each junction, rather than ye olde 0.01uf cap in parallel with the rectifier junction, which does nothing or makes things worse.

 

[fpitas]

The correct way to snub the rectifier is still to use a resistor and capacitor in series around each junction, rather than ye olde 0.01uf cap in parallel with the rectifier junction, which does nothing or makes things worse.

Yeah, never understood why people thought they were absorbing anything with just a capacitor. Maybe it tunes the noise lower in frequency?

 

[SIY]

Actually the 'diode noise' is a product of the power transformer. To snub it, a capacitor in series with a resistor can be placed across the output of the transformer at the rectifier input. Because its more of a transformer issue, its best to keep the transformer leads as short as possible.

How it works is something called a 'swept resonance'. Its caused by the inductance of the transformer in parallel with the capacitance of the rectifiers and is energized by the commutation of the rectifier. Bypassing the individual junctions of the rectifier will work; its putting out the forest fire rather than blowing out the match.

The correct way to snub the rectifier is still to use a resistor and capacitor in series around each junction, rather than ye olde 0.01uf cap in parallel with the rectifier junction, which does nothing or makes things worse.

Have you seen Morgan Jones’s article on snubbing? Nice analysis and supporting measurements.

 

[egellings]

Yeah, never understood why people thought they were absorbing anything with just a capacitor. Maybe it tunes the noise lower in frequency?

If you're gonna 'snub' energy, you need to have a dissipative element to pee it away in, i.e. a resistor. The cap, absent its minuscule ESR, won't quite cut it.

 

[atmasphere]

Have you seen Morgan Jones’s article on snubbing? Nice analysis and supporting measurements.

Yes. I bought the book on that account.

 

[Sal1950]

Have you seen Morgan Jones’s article on snubbing? Nice analysis and supporting measurements.

I've seen it, but I don't understand it.
Guess I have to rely on you to explain it.

 

[fpitas]

I've seen it, but I don't understand it.
Guess I have to rely on you to explain it.

I'll let them tell you exactly what Morgan Jones said. But in broad terms the subject is the response of a solid state diode as it switches OFF from a conducting state. They tend to ring, some quite violently, as the stored charge is emptied. Some are even used to generate microwave signals as they do that.

 

[fpitas]

Step recovery diode - Wikipedia

en.wikipedia.org

 

[fpitas]

The switching noise can be problematic in some cases, even in audio. The solution is a series RC circuit called a snubber, which absorbs the energy in the resistor. The discussion here is where to place the snubber in a power supply for maximum effectiveness.

 

[Sal1950]

Thank you kind sirs for you tutoriage.