• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

How do companies decide what sort of power supply ~topography?~ to use?

Tks

Major Contributor
Joined
Apr 1, 2019
Messages
3,221
Likes
5,494
I don't know jack about this, but I always see people complaining about:

"oh linear" "oh switching" "oh DC-DC" "oh AC" "oh external" "oh internal" "oh 5v" "oh 12v" "oh even higher volt" "oh it's a massive wall wart" "oh it's USB powered and small" "oh can't be used in the US"

and all sorts of other things about power supplies. What is actually some of the benefits or downsides of choosing one over another (don't really care too much about "sonic commentary" seeing as how all sorts of power supplies are being used in all sorts of high performance products. I'm more interested in what's harder to work with, or which costs less, which approach makes more sense where. Etc...

If anyone has any noob reading material about this, and why there is much varience, I'd really love to learn about it (can't stress the noob part enough, this stuff goes way over my head).

EDIT: One other thing, are power supplies impossible to build on your own that rival OEM's? I don't see even companies doing this in the same way I don't see companies building their own IC's, instead of simply buying stuff off the shelf.
 

win

Senior Member
Joined
Jun 17, 2020
Messages
430
Likes
432
Location
Irvine CA
If you are appealing to the boomer-audiophool demographic then you need linear and toroidal etc. they eat that shit up.
 

Vini darko

Major Contributor
Joined
Jun 1, 2020
Messages
2,280
Likes
3,395
Location
Dorset England
Personally I like double insulated (not earthed) power supplies. Using E core transformers (The square ones). Prefer linear for amps and switch mode for sources.
Be intresting to hear from some manufacturers about this topic.
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,348
Likes
3,462
Location
San Diego
I don't know jack about this, but I always see people complaining about:

"oh linear" "oh switching" "oh DC-DC" "oh AC" "oh external" "oh internal" "oh 5v" "oh 12v" "oh even higher volt" "oh it's a massive wall wart" "oh it's USB powered and small" "oh can't be used in the US"

and all sorts of other things about power supplies. What is actually some of the benefits or downsides of choosing one over another (don't really care too much about "sonic commentary" seeing as how all sorts of power supplies are being used in all sorts of high performance products. I'm more interested in what's harder to work with, or which costs less, which approach makes more sense where. Etc...

If anyone has any noob reading material about this, and why there is much varience, I'd really love to learn about it (can't stress the noob part enough, this stuff goes way over my head).

EDIT: One other thing, are power supplies impossible to build on your own that rival OEM's? I don't see even companies doing this in the same way I don't see companies building their own IC's, instead of simply buying stuff off the shelf.

Here is link to great article on linear power supplies https://sound-au.com/power-supplies.htm .
 

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
3,334
Likes
3,278
Location
.de
That's actually a very interesting topic. There are ties between the power supply and the connections used on a device that you wouldn't initially expect, for example.

There are a lot of considerations that go into a choice of power supply. What sort of wattage and voltages do you need? What other operational requirements exist? What are acceptable efficiency levels / losses? How are you going to pass safety regulations and electromagnetic compliance (EMC)? Can you do it in-house, do you want to buy a pre-assembled board or even something wall wart style?

You will find mains power supplies in two fundamental classes: linear and switch-mode.

Linear power supplies are based on transformers operating at mains frequency. Since that is not particularly high at 50 or 60 Hz, those tend to be a tad big and heavy, a bit more so in 50 Hz territory (airplanes are using 400 Hz AC power for this reason). Efficiency typically ranges between about 50% and a good 80%, depending on load level, transformer construction and what's going on secondary-side in terms of rectification and potential further regulation. Being linear, they tend to pose few EMC problems (once you get your rectifiers sorted out), along with being fairly robust and long-lived. Their generally low complexity makes them fairly easy to design overall.
These factors make them attractive to small-scale operations, although you may have difficulty obtaining good transformers if you need shield windings to reduce primary-secondary capacitive coupling (important for power amplifiers with unbalanced inputs, even before electrical safety concerns) or extra-low mechanical hum levels (transformer hum is related to lots of things, not the least transformer core construction and how much more power you're squeezing out by pushing them closer to saturation, so price and size are definitely relevant factors).
The downside of a linear supply is that it usually requires mains voltage to remain in a fairly small tolerance band (different voltages being accommodated by a split / tapped primary winding if need be), and a transformer from 60 Hz land may be undersized and running too far into saturation in 50 Hz land.

Using an off-the-shelf wall wart has pretty simple reasons - its manufacturer has already sorted out the whole compliance business for you, so you can concentrate on your product instead of having to worry about electrical safety and all. EMC might still be a concern but definitely less so. This generally is only a option for low-power devices.
As for DC vs. AC, I think you should be able to find some details in the NwAvGuy O2 articles. Basically he needed a typical split supply (with both positive and negative voltages), but you can't buy a supply like that off the shelf. So either you would've had to integrate a split secondary mains transformer into the amp, with all the usual electrical safety concerns (not the greatest idea for a DIY design that's supposed to be easy to replicate), or just get an external transformer of 1x12-15VAC with a voltage doubler kind of arrangement.

Switch-mode power supplies are generally rectifying and chopping up the mains at high frequency well in the kHz (usually ultrasonic), which allows transformer sizes to be greatly reduced at the same power. Modern PC power supplies can output 1000 W while being the size of a small shoe box only, something downright impossible with linear supplies. This approach also allows them to run on just about anything within a certain voltage range, DC or AC, and with some tricks applied it is not unusual to see them being suitable over basically all mains voltages used worldwide (90...240 VAC), all that while keeping output voltages fairly tightly regulated. They can be very efficient, from around 70-80% to >95%.
Alas, where there is switching at high frequency and current, there is lots of high-frequency junk being generated as both electric and magnetic fields, which may upset all kinds of radio applications in particular. Mains filtering is important for this reason, you are also well-advised to keep voltage spikes on the mains away using protection elements (spark gaps, MOVs and varistors), and circuit layouts need to be up to par. Secondary-side capacitors also have to handle lots of high-frequency ripply current but can't be too large either, so their ESR becomes critical for both voltage stability and longevity. Due to all the mains filtering including some capacitors, larger supplies generally need to be grounded via protective earth (PE), and they are often completely PE-referenced even on the secondary side. (I actually wish more wall warts were using a primary-side PE connection instead of dumping their mains filter leakage currents into secondary-side ground.) This makes them problematic or unsuitable for use with unbalanced audio connections.
With such a major can of worms being opened, it should be no surprise that relatively few people are willing to design their own SMPS, and if so it's a major effort. They are often bought as finished boards or modules, or external wall warts. If all you need is a basic wall wart, I've read that you can buy them in bulk for 2 or 3 bucks a piece, so they are often the cheapest, too.

If a wall wart SMPS is being used, you will not uncommonly find that its output isn't actually being used directly, but instead feeds a number of internal switching DC/DC converters (much the same as a full-blown SMPS except for the galvanic isolation bit) that provide all the internal supplies. The RME DACs are one example (so I would take any claims of improvements when fed with 15 V instead of 12 V with a massive grain of salt), and everything that can output >1.7 Vrms on USB power will also have them.
 
Last edited:

gene_stl

Addicted to Fun and Learning
Forum Donor
Joined
Jun 14, 2019
Messages
867
Likes
1,198
Location
St.Louis , Missouri , U.S.A.
I have been repairing electronic stuff since the 1960s and I watched initially with interest and then growing horror as switching mode power supplies
began to replace linear power supplies, which took place much earlier outside of audio than in audio. I am sure the presence of the switching frequency in the ripple and also the complexity of SMPS delayed their use in audio.

Probably switch mode power supplies are quieter because the switching frequency is higher than any tweeter can reproduce. However I have some objections to them based on experience. They aren't really strong objections and I do now have some class AB power amps with SMPS in them.

First of all, some SMPS don't isolate from the AC line. A big ole clunky transformer can do a lot to stop a fast rise pulse from a nearby lightning strike.
NOTHING is lightning proof but linears are more resistant. SMPS also inject noise into the AC line.

Second if someone presents me with a piece of equipment with a sick power supply if its a linear I can almost always fix it if the transformer is not shorted or burned. In contrast with switchers the only guy who really understands how a given smps works is the guy what designed it. He probably doesn't even work for that company anymore. If its a generic circuit board they are as easy and cheap to replace as wall warts.

When they became widespread semiconductor companies started offering power supply controllers on a chip. You added external components to suit your application. But now its later enough that some of these chips are obsolete and unobtainium. Even if they were a package that was unsolderable.


SMPS , like LED lighting are an idea whose time has come. But like anything else they have advantages and dis advantages. I became aware of them
because a brilliant engineer I know was an early adopter and he thought they were the bomb. He didn't use them for audio for which he built gigantic linear supplies. (1970s) Same on his kool kilowatt ham gear.
 

gattaca

Member
Joined
Nov 29, 2019
Messages
97
Likes
87
SMPS are difficult to design correctly and impossible to repair as stated above. Most SMPS are made in China and typically, they are made by bidding/ pricing them at "cents per watt". All SMPS have down-sides. For instance, they inject huge amounts of electrical noise into the lines they are connected into. Depending on the supply filtering, that can make a mess of your home's or businesses wiring if you use things certain things which may be sensitive to electrical noise. How SMPS are built + the quality of the components, in particular the capacitors, really determine the longevity and stability of the units. There are some very good units Made-in-the-USA. In fact right outside NYC in Flushing, NY. I've toured their facility a few years ago. I now only use Kepco's in my AV setup (you can find'm on eBay if you are patient).

So when someone calls BS on it's just a bit of "electrical noise" that SMPS can generate or I see SMPS being used in supposedly high-end AV equipment... I just shake my head, say well.. BS and usually point them to this article. There's a huge difference between 3mV peak-to-peak noise and 180mV peak-to-peak noise even from good SMPS. Do the math! Exceptional engineering + attention to details makes the difference!

For instance, did you know that Kepco, a USA company in NY had a major hand in in the LIGO project? They had been searching for gravitational waves for years... and it turns out the SMPS being used earlier in the project were way too noisy and had been masking the gravitational waves all along. Once LIGO switched to the ultra quiet low-noise linear units, boom, they suddenly found the gravitional waves they had been searching in vain for nearly a decade earlier! (I'm sure some engineer or supply chain purchaser is really kicking themselves for not paying more attention.... )

https://physicsworld.com/a/the-firm-thats-made-no-noise-over-gravitational-waves/


"LIGO started taking data in 2002 and Kepco’s equipment – along with other switched-mode power supplies – was installed in both detectors. The devices were used to convert standard 150 V mains AC into DC that was used to power the lasers, electronics and other equipment that make up the giant 4 km-long detectors.

When LIGO was upgraded in the early 2010s to form Advanced LIGO, however, the project’s scientists decided to replace all the switched-mode power supplies with linear power supplies, with Kepco supplying 600 power units – most of those used – to both the Hanford and Livingston sites. Eliminating RF noise is vital with Advanced LIGO, which has to detect vibrations 10,000 times smaller than a proton’s diameter.

Without the company’s equipment, the gravitational-wave signal would have been swamped. Sure, LIGO also uses many other vital bits of equipment, but it’s fair to say that without Kepco, we might have had to wait a whole lot longer for this year’s ground-breaking discovery of the waves that Einstein first predicted."

Stay safe, stay alive! Peace.
 
Last edited:

Vini darko

Major Contributor
Joined
Jun 1, 2020
Messages
2,280
Likes
3,395
Location
Dorset England
SMPS are difficult to design correctly and impossible to repair as stated above. Most SMPS are made in China and typically, they are made by bidding/ pricing them at "cents per watt". All SMPS have down-sides. For instance, they inject huge amounts of electrical noise into the lines they are connected into. Depending on the supply filtering, that can make a mess of your home's or businesses wiring if you use things certain things which may be sensitive to electrical noise. How SMPS are built + the quality of the components, in particular the capacitors, really determine the longevity and stability of the units. There are some very good units Made-in-the-USA. In fact right outside NYC in Flushing, NY. I've toured their facility a few years ago. I now only use Kepco's in my AV setup (you can find'm on eBay if you are patient).

So when someone calls BS on it's just a bit of "electrical noise" that SMPS can generate or I see SMPS being used in supposedly high-end AV equipment... I just shake my head, say well.. BS and usually point them to this article. There's a huge difference between 3mV peak-to-peak noise and 180mV peak-to-peak noise even from good SMPS. Do the math! Exceptional engineering + attention to details makes the difference!

For instance, did you know that Kepco, a USA company in NY had a major hand in in the LIGO project? They had been searching for gravitational waves for years... and it turns out the SMPS being used earlier in the project were way too noisy and had been masking the gravitational waves all along. Once LIGO switched to the ultra quiet low-noise linear units, boom, they suddenly found the gravitional waves they had been searching in vain for nearly a decade earlier! (I'm sure some engineer or supply chain purchaser is really kicking themselves for not paying more attention.... )

https://physicsworld.com/a/the-firm-thats-made-no-noise-over-gravitational-waves/


"LIGO started taking data in 2002 and Kepco’s equipment – along with other switched-mode power supplies – was installed in both detectors. The devices were used to convert standard 150 V mains AC into DC that was used to power the lasers, electronics and other equipment that make up the giant 4 km-long detectors.

When LIGO was upgraded in the early 2010s to form Advanced LIGO, however, the project’s scientists decided to replace all the switched-mode power supplies with linear power supplies, with Kepco supplying 600 power units – most of those used – to both the Hanford and Livingston sites. Eliminating RF noise is vital with Advanced LIGO, which has to detect vibrations 10,000 times smaller than a proton’s diameter.

Without the company’s equipment, the gravitational-wave signal would have been swamped. Sure, LIGO also uses many other vital bits of equipment, but it’s fair to say that without Kepco, we might have had to wait a whole lot longer for this year’s ground-breaking discovery of the waves that Einstein first predicted."

Stay safe, stay alive! Peace.
Top quality post m8 will go watch that video thanks.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
You can build SMPS that have less 'garbage' going back into the mains as a linear power supply.
You can build SMPS that have lower noise than a linear power supply.
These, however, are nothing like the bricks and wallwarts you can buy for a few $ or get for free with a device.
Those can even be fed from a range of AC and/or DC voltages (wide range)

You can build massive (heavy and big) impressive linear power supplies with excellent regulation and stability but these cost way more than the cheap transformer+rectifier+78xx regulator found in some gear.
These are either around 115 or 230V AC only.
The more expensive ones can be select-able with a switch (or jumpers inside).

Manufacturers often buy 'bricks' that are already available (for cheap in bulk) on the market that are suitable for the job.

Power supplies are mostly chosen based on price, weight, size, wide-range or not, onboard or external. Sometimes for specific reasons like leakage currents, isolation or because customers may kind of 'demand' it.

Its mostly audiophiles that worry about power supplies based on facts and nonsense.
 

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
3,334
Likes
3,278
Location
.de
impossible to repair
Definitely not. While a repair may be uneconomical in a lot of cases (few people are going to put a lot of time into repairing a module that costs 30$ brand new), a lot of common faults actually aren't rocket science - dead (through-hole) electrolytics on the secondary side or in the auxiliary supply (sometimes a resistor there as well), shorted switching MOSFETs, bad rectifier diode, and did I mention the odd bad solder joint? Things can get tedious when you have to crack open an ultrasonically welded wall wart first or need to fix a little soldered-in sub board, and finding a replacement for a controller chip may prove difficult, so repairs are more demanding much like the circuitry itself.
You can build SMPS that have less 'garbage' going back into the mains as a linear power supply.
You can build SMPS that have lower noise than a linear power supply.
These, however, are nothing like the bricks and wallwarts you can buy for a few $ or get for free with a device.
Exactly. I do wonder how big a "SOTA performance" wall wart might turn out to be, it certainly wouldn't be short of chokes of all kinds on the input and output side...
 
Last edited:

gene_stl

Addicted to Fun and Learning
Forum Donor
Joined
Jun 14, 2019
Messages
867
Likes
1,198
Location
St.Louis , Missouri , U.S.A.
[QUOTECan't forget Lambda and Power Designs. :D ][/QUOTE]
I just did!:p:facepalm:

A place where I used to work the new guy got stuck on a bench with a shelf full of Wyse monitor/terminals. You opened them up and replaced all the electrolytic caps and a couple of semiconductors. And anything else you noticed that was cooked. But with just the caps and two semis about 60% of them lit right back up. I think I did 30 of them and 19 came right back up. They were from the seventies and eighties.
 

gattaca

Member
Joined
Nov 29, 2019
Messages
97
Likes
87
^^^ I covered the part about those *(@*@ electrolytics being the most common failure point - often because the are crammed in right next to something very hot which just bakes/dries them out - a very poor design - attention to details matters! But of course most SMPS are only designed to last 3-5 years - which says a lot about what they are going into. Most vendors have muc less concern once it an item goes out of warranty. Why? It is simply no longer a liability on their warranty costs. Of course, I see SMPS installed in $80-$150K servers / computers all the time. In these settings, they could care less about "electrical noise!" These are almost always built for $0.10-25 / watt in China and as long as they last 3-5 years they have met their design goal. It's a completely different perspective vs building something designed to last 10-15+ years.

OK "uneconomical" is more accurate than "impossible". ;) My challenge to anyone seeking to repair a SMPS would be locating someone competent and well trained enough now-a-days to actually work on most SMPS. I live in a major metropolitan area and all those techs / guys / shops are long retired and gone! You are right, economically it is easier and more cost effective to recycle whatever the SMPS is in or try to find a spare salvage unit or part on eBay. I've lost count of how many times I've done both. It is however a big PITA to keep the noise off your lines. I ended up installing line noise filters in front of my smaller UPSes b/c of all the noise they interject in my power lines!

That is unless you are looking for gravity waves, then well SMPS are not the deal.

Thanks for mentioning all those other companies. I know those names but I am not as familiar with them as I am Kepco's products. Many of those companies sadly are no longer around and if they are, they are no longer "Made-in-the-USA..." There are only a few left. :( Stay safe, stay alive. Peace.
 
Last edited:

KaMiKaZeCATS

Member
Joined
Jul 26, 2020
Messages
37
Likes
43
Generally speaking.

Linear Power supply are used for highly sensitive electronics. They usually sacrifice size and efficiency for low ripple and noise.

Switch mode power supplies are generally used for electronics used in pretty much everything else. Most consumer goods almost always ship with switch mode psu simply due to thier size and efficiency.

But this is NOT a hard rule at all. It heavily depends on the situation.

For example a Bluetooth speaker will use a switch mode power supply and almost always be class d based. The noise and ripple are a small consideration in the grand scheme of things. Size and efficiency are more important but that doesn't mean that design is "bad" its just trying to achieve different goals and if careful design choices are made they can easily achieve extremaly high performance. If you crack open some "quality" brand studio mointers most (not all) use class D with switch mode power supplies. If you consider that your music was made and mastered on a switch mode based amplifier. Shouldn't we all be using switch mode power supplys? .... No of course not since careful design considerations went into that power supply in order to achieve low noise.

Don't be afraid of switch mode power supplies especially quality ones made from quality components. Meanwell make fantastic switch mode power supplies that are easy to service and are made from quality parts. With good amplifier design you can easily use switch mode power supplies and reach the noise and ripple of transformer based psu. But its definitely not as easy as linear.

Linear power supplys are much easier to get a clean smooth dc supply. Just Google linear power supply circuit and you will find tons of schematics and designs that are free. They are very DIY freindly and much harder to get wrong. (but you still can definitely get them wrong... )
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
Good power supplies can be made in many topologies, bad ones too. I suspect bad ones are more commonly switching than linear because the minimum viable design is much cheaper in the former than the latter. If you're chasing pennies, you're not going to stop halfway. With a bigger budget you can pick the technically best design, which may well be a hybrid incorporating both linear and switching stages.
 
Top Bottom