SMPS Power Supplies vs Linear Power Supplies

sunbeamgls

Wammer
Wammer
Mar 19, 2011
5,088
2,334
193
North Wales
HiFi Trade?
  1. Yes
Below is an extract (complete other than removing the greetings at the beginning and end) from the Linn forum by DanielE, a Linn engineer, about why they chose SMPS and why they think its better than linear.

I'm posting this here for a couple of reasons - first, because I read lots of posts about Linn doing stuff that's not based on sound engineering principles where this post seems to suggest they are using sound engineering principles and secondly because I'd like to see the counter argument please.

If you're a power supply engineer or are fully conversant with power supply design, please post up your thoughts on this piece. If you're not, by all means give an opinion, but please don't spout pseudo science in a way that appears to be genuine knowledge.

Thanks

"I'd like to answer your question firstly with a bit of background information about the larger power amplifier supplies and why a SMPS is benefitial in that case, as I think it will aid the understanding overall, and then move onto the specifics about SMPS for line level products.

The nature of a musical signal means it doesn't present a constant load, particularly in a power amplifier where large amounts of current are provided to move the drive units. A linear power supply relies on the smoothing capacitors on its output to provide the power that the circuits use, which works well enough for a constant load, however as the load draws power from these capacitors they are only getting topped up at mains frequency, 50 or 60 times a second depending on the local mains supply, so as the load increases the capacitors are unable to keep up with large demands at higher frequencies than 50 or 60Hz. The signal created by a kick drum typically used in rock/pop music is a damped impulse, containing large amounts of energy at significantly higher frequencies than 50Hz which causes the smoothing capacitors to start discharging, and the power rails to droop, increasing the likelyhood of power amp clipping.

A switching power supply not only tops up the capacitors at its switching frequency, in the case of our power amplifiers that is in the region of 50,000 times a second, but it also has a feedback network that is monitoring the power rails being delivered. Meaning that the power rails are significantly less prone to mains related pumping (which will eventually occur with a switching supply if you increase the load on it far enough, in the case of our power amp supply that is when delivering over a kW continuously) and the power rails are significantly more stable.

In the case of a line level product, the loads are more constant so they are less susceptible to pumping in the fasion described above, but the circuits are still susceptible to the mains frequency and the harmonics thereof, all of which are audible frequencies, getting onto the power rails and signal ground. The devices within the audio circuits whether it be digital or analogue need to rely on thier own power supply rejection ratio and bulk storage capacitors to prevent these frequencies from interfering with the audio.

A well designed switching supply won't introduce the mains frequency to either the power rails or the signal ground and will filter the rails such that the impact of the inaudible switching frequency is minimised (our switching frequency for line level products is 132kHz). It is much more practical to filter out high frequency noise than it is to filter low frequency due to the physical sizes of the components required to perform the task in hand.

The very nature of any switching supply means that it needs an input filter to prevent emissions getting onto the mains, this filter also doubles up to protect the power supply itself from noise that is already on the mains. I can't speak for the Naim designs as I don't know exactly what is in them, but in the majority of cases a linear power supply doesn't feature any form of filtering against RF noise which will then get through the mains transformer and into the connected circuitry."

 

SergeAuckland

Certified Measurist
Wammer
May 6, 2008
18,607
1,915
173
Bury St Edmunds, UK
AKA
Serge
HiFi Trade?
  1. No
From my perspective, I completely agree with the Linn view. Well designed (i.e. properly filtered) SMPSs have all the above advantages. Conventional sagging supplies achieve their performance by brute-force, meaning large reservoir capacitors and/or large inductors. SMPSs do it by using servo electronics.

It's possible to avoid a sagging supply by using conventional power stabilisation techniques, but that adds complexity and heat, so a SMPS is, in engineering terms, a better solution.

What SMPSs don't do is give an amplifier any dynamic headroom, the maximum output power is the same whether under continuous sine-wave duty or speech and music. If an amplifier with a conventional supply is used for speech and music, its maximum output before clipping can be something like 1-2dB greater than it's output under sine wave conditions, so a 50 watt amplifier provides 60-70 watts before clipping. If an amplifier is used close to its maximum power, a 50 watt amp with a conventional stabilised or SMPS will clip before the same amplifier circuit but with a conventional sagging supply. This has resulted in some people claiming that an amp with SMPS lacks dynamics, when it's just them overdriving a comventional amp.

Tests done in the 1970s (before SMPS existed) compared amplifiers at different powers with and without stabilised supplies, and it was found that an amplifier of X watts continuous sounded better if it had a sagging supply, than one of the same X watts continuous with a stabilised supply, but with two amplifiers of the same dynamic power, one achieved with a sagging supply and the other with a stabilised supply, the one with the stabilised supply sounded better.

S.

 

SergeAuckland

Certified Measurist
Wammer
May 6, 2008
18,607
1,915
173
Bury St Edmunds, UK
AKA
Serge
HiFi Trade?
  1. No
Thanks Serge - do any hifi manufacturers of note use stabilised linear supplies?
Quad did in the 303, then went back to an unstabilised supply for the 405 and later amps, although some power supply stabilisation for earlier stages is used, just not for the main output.

I don't know of any current designs that do, but then manufacturers are reluctant to publish circuit diagrams.

One trick that was used by a couple of manufacturers was to invert one amplifier channel, so that for a heavy bass signal, which is largely mono, one amplifier went + when the other went -ve and so pulled power out the different sides of the power supply, so it sagged a little less. This mean that the loudspeaker output socket colours on the inverted channel were the other way round, but it also had the advantage that the two amps were operating in bridge, so making the amp a mono bridged amp was very easy indeed.

S.

 

tonerei

Wammer
Wammer
Jan 14, 2012
453
5
33
Dublin,Ireland
AKA
Tony
HiFi Trade?
  1. No
No technical knowledge but saw the insides of a small Teddy Pardo 3 or 5v linear supply. He had an smps inside the box with some other boxes of tricks beside it. It would seem a well engineered smps or linear supply will deliver the goods. I know build a computer audio pc a lot of us have used maplin linear supplies because they are cheap and appear to give a decent lift over the usual cheapo smps supplies. No doubt a well done smps will bring similar if not better benefits. Problem I assume is the cost.

 

Tenson

Wammer
Wammer
Mar 2, 2006
407
2
0
I've got to dash out in a minute but..

as the load draws power from these capacitors they are only getting topped up at mains frequency, 50 or 60 times a second depending on the local mains supply, so as the load increases the capacitors are unable to keep up with large demands at higher frequencies than 50 or 60Hz.
It's not really the point. It would only be a problem if your transformer could not deliver enough current in the time allowed, or from the other perspective if your load was drawing too much current. So use bigger caps or a bigger TX such that the problem doesn't occur. The frequency of mains, or the SMPS is not the point, the current delivery with each cycle is.

A well designed switching supply won't introduce the mains frequency to either the power rails or the signal ground and will filter the rails such that the impact of the inaudible switching frequency is minimised (our switching frequency for line level products is 132kHz). It is much more practical to filter out high frequency noise than it is to filter low frequency due to the physical sizes of the components required to perform the task in hand.
It's not too easy to filter out stuff at very high frequencies because it is RF and air-born from passing through the transformer as a magnetic field and travelling along wires that act like an aerial. That's not to say it isn't possible to make a clean SMPS, but it isn't trivial like that post suggests. Another important point is that any RF signals on the power lines will not be well rejected by the devices (like op-amps) internal regulation and it can demodulate into the audio band very easily, causing spurious noise. At least 50Hz mains hum is coped with well by devices and it stays in the low frequency range where it doesn't sound too disturbing if it is at all audible.

I'm just presenting an opposing argument here, not one method is better than the other as with most things it depends how you do it.

 

SergeAuckland

Certified Measurist
Wammer
May 6, 2008
18,607
1,915
173
Bury St Edmunds, UK
AKA
Serge
HiFi Trade?
  1. No
Tenson's comments are also correct. Linear and SMPS both require decent engineering, both have their problems and have their benefits. I would rather have a well engineered SNPS for its performance, whikst I would prefer a decently sized conventional supply for its ease of service, given how rarely circuit details are provided by manufacturers these days. Pretty much any one can repair a conventional supply whilst with a SMPS, repair without having a circuit diagram and voltage/frequency information is difficult.

Perhaps that's one reason why many SMPSs are modular, and one replaces the module rather than repair.

S

 

toprepairman

Wammer
Wammer
Dec 6, 2010
1,190
573
158
Surrey
AKA
Henry
Bang on there Serge.

When the SMPS goes bang in 10 to 20 years' time then the chips required to repair it won't be available anymore.

H.

 

Clubsport911

Wammer
Wammer
Jan 19, 2013
1,814
241
108
Cheltenham
AKA
Steve
HiFi Trade?
  1. No
SMPSU's have been used for yonks - specifically in (CRT) TV's - in conjunction with special transformers to drive the beam currents required for scanning. The HF of the switching enabled them to be small and of course "efficient" - unless you remember that nothing is for free. The noise is shifted into the HF requiring extensive filtering if used on Audio.

FW BR designs are inherently NON linear [but predicatable] in how they manage mains harmonics - throwing loads back onto the supplyup to the 11th harmonic - all of which corrupts the supply if it's impedance is poor. However.. SMPSU's can also throw out heaps of mains borne noise - much of which these days has to be filtered to meet EN61800-3. Older designs were nasty ++

I've designed hundred of power supply circuits and it's (I'm afraid) horses for courses. Linear PSU's can be even more effective with 3 phase diode arrangements PSU's (hard in a household !) using 6 / 12 or even 18 pulse designs and ST / D / ST transformers for phase shifting - these reduce ripple current to very low levels indeed... and of course the diode conduction angle is also much reduced. But as stated earlier:-

use bigger caps or a bigger TX such that the problem doesn't occur. The frequency of mains, or the SMPS is not the point, the current delivery with each cycle is.
This is key. The depletion of the Vdc by the load current is what determine what happens "before" the cap.

I've done countless back to back tests. I always end up with linear supplies. They can be designed in a cinch, tweaked with line or load chokes and sound very good when optimised. You have to realise that a good power supply is the foundation to a great sound. Linn's article is all plain fare - nothing revolutionary in it at all so for me, the big breakthrough will be when the Matrix converter is commercialised for household applications (no DC link !)

And... as Serge mentioned... all the bits for a standard TX / BR / CAP supply can be found (almost) in Maplins. You cannot often say the same for SMPSU's

Rgds

 

Tenson

Wammer
Wammer
Mar 2, 2006
407
2
0
Funny, ai just got in today to find a SMPS giving a funny whistling noise. That will be the second time I've eepaired it. I can't help but thinking the switching frequencies put a lot of strain on the psu caps. Cheaper smps do seem rather unreliable compared to linear supplies here even vheap ones keep going for longr than the other parts of the kit.

 

SergeAuckland

Certified Measurist
Wammer
May 6, 2008
18,607
1,915
173
Bury St Edmunds, UK
AKA
Serge
HiFi Trade?
  1. No
Don't Naim and Krell use stabilised power supply units? Expensive??
Don't know about Naim,I suspect that they have used stabilised supplies, but don't know if that's usualor not. As to Krell, their earlier amplifiers were conventional Class A, and so had a constant current draw, so a stablised supply wouldn't of any great benefit, but their later and larger amplifiers use a sliding-bias Class A which also had switched voltages, so would have a pretty complicated power supply and I would be very surprised if that wasn't stabilised.

As to the reliability of supplies, I think on balance a conventional linear supply is more reliable, in that with the exception of a rare rectifier failure, supplies fade gracefully as capacitors lose performance over many years, (although I've had a fair few small regulator failures in equipment) whilst SMPS fail suddenly and totally. However, that's based on my limited experience in that I don't see as much faulty equipment as someone who repairs and services equipment for a living.

Professionally, more and more equipment is going over to SMPS, because of weight, bulk and autosensing mains voltage that avoids transformer taps that can end up in the wrong place, and also avoids the need for a larger transformer to do 50Hz mains. It's hard to get a decent sized transformer into a 1Uhigh piece of kit whilst an equivalently rated SMPS is a lot smaller and lighter.

S.

 

Cable Monkey

Moderator
Staff member
May 16, 2006
9,069
2,089
158
Birmingham, UK
AKA
Henry
HiFi Trade?
  1. No
A good SMPS can certainly do a good job. I have one in my Aqvox phono stage that, together with a good deal of filtering has made it an extraordinarily quiet stage. The problem is not one vs. the other, it is the fact that an SMPS is often used to save costs over a linear supply or is designed to fit in a smaller space than a linear supply with the compromises that those considerations may bring. Let's be clear that linear is not automatically better and a badly designed one can have its own issues.

 

awkwardbydesign

Perfect, apparently.
Wammer
Mar 5, 2012
10,399
2,293
173
HiFi Trade?
  1. No
all the bits for a standard TX / BR / CAP supply can be found (almost) in Maplins. Rgds
LOL! I can remember when Maplin were an electronic parts supplier. :shock: They killed off all the little suppliers and then turned into toy shops. :grrr:

 

George 47

Subjectivist
Staff member
Jan 1, 2006
8,423
6,070
193
United Kingdom
HiFi Trade?
  1. No
Don't know about Naim,I suspect that they have used stabilised supplies, but don't know if that's usualor not. As to Krell, their earlier amplifiers were conventional Class A, and so had a constant current draw, so a stablised supply wouldn't of any great benefit, but their later and larger amplifiers use a sliding-bias Class A which also had switched voltages, so would have a pretty complicated power supply and I would be very surprised if that wasn't stabilised.S.
You are spot on there with the Krell amps, which also includes some mains filtering in the newer versions, probably to deal with the issue below. Or the less than good quality mains in the US.

I think the real problem is the low cost SMPS that have poor filtering (wall warts) that we use in the same room as the hifi, especially for the computer audio components that can 'leak' RF.

 

awkwardbydesign

Perfect, apparently.
Wammer
Mar 5, 2012
10,399
2,293
173
HiFi Trade?
  1. No
I think the real problem is the low cost SMPS that have poor filtering (wall warts) that we use in the same room as the hifi, especially for the computer audio components that can 'leak' RF.
It's fun to stand in the middle of the room with an oscilloscope and put your finger (for example) on the probe and watch the trace go crazy! One reason I wouldn't live under a high voltage line or pylon. And why I am not a fan of wireless solutions.

static-hair.jpg


 

Tenson

Wammer
Wammer
Mar 2, 2006
407
2
0
The problem is... the fact that an SMPS is often used to save costs over a linear supply or is designed to fit in a smaller space than a linear supply
That us exactly what they are good for though. Why would you use a SMPS if you are not constrained by space or budget?

 

YNWAN

100% Analogue
Wammer
Jul 22, 2007
1,675
52
93
Sheffield
AKA
Mark
HiFi Trade?
  1. No
Naim Nap-135's used a regulated supply as did Mark Levinson (certainly in their older stuff).

 

Forum statistics

Threads
113,444
Messages
2,451,263
Members
70,783
Latest member
reg66

Latest Articles