• 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!

What is so tricky with SMPS in audio?

jdav

Member
Joined
Oct 28, 2021
Messages
48
Likes
14
It took a long time to shift from linear power supply to SMPS in high end audio producs. Nowadays, SMPS seem to have completely replace linear ones except in some niche areas.

I've often heard the advice to beginners to not touch SMPS and stick to linear supply for DIY projects, because designing for SMPS is too difficult. What is it so tricky? Why couldn't we just put the switching frequency well outside the audible range and be done with it? I'm sure there will be some problem somewhere but I don't know exactly which one.

For comparison, I tore down my RME UFX II interface to find a XP Power VCT60US12 inside. It's not a particularly quite one (max 1% ripple p-p), switching at 60kHz, and I didn't see any massive filtering downstream. It seems they are quite ok leaving a reasonably high amount of noise in their power supply rail. The negative power supply is created using an other switch-mode inverter IC. I'm aware this is quite low power (max 36W) device, but it's also a low signal one, with mic preamps inside. How is it possible on such a high end interface?

Thanks
 

egellings

Major Contributor
Joined
Feb 6, 2020
Messages
4,004
Likes
3,244
PSRR works when the amplifier in question is able to track the PS noise so that the noise can be removed by action of inverse feedback. With the high frequency garbage present on switch-mode supplies, the too-slow amplifier cannot possibly follow that and then cancel it out using negative feedback. The HF garbage also just radiates as RF and, like chicken man, it's everywhere! It's everywhere! Bedawk!
 

MaxwellsEq

Major Contributor
Joined
Aug 18, 2020
Messages
1,627
Likes
2,421
Negative feedback (i.e. in any gain stage) is not the only way to prevent power supply noise exiting the amplifier.
 

AnalogSteph

Major Contributor
Joined
Nov 6, 2018
Messages
3,334
Likes
3,278
Location
.de
In an IEC Class I device it is fairly easy to get away with a SMPS. (Case in point, PC power supplies.) I suspect RME are using an IEC inlet with integrated mains filter? (Those are pretty good for bonding the case to PE, even before considering the internals.) Besides, it's not like noise well in the ultrasonic range requires any truly massive filtering to begin with, you can probably hit it with some (R)LC action and let PSRR take care of the rest.

It's when you require minimum mains coupling for unbalanced I/O that SMPS get tricky, since it's hard to get EMI in check without the odd Y capacitor, and that's several nF right there. There's a reason why (primarily) unbalanced input headphone amps like Topping L30 II, L50, JDS Labs Atom Amp(+) still ship with transformer plug packs - well, arguably more than one, as AC input makes generating dual rails much less of a hassle as well. A small transformer up to about 20 VA might have pri-sec coupling in the mid-hundreds of pF at most, that's generally negligible enough to not warrant any further measures. Transformers with shield windings tend to be necessary past 100 VA only.
 
OP
J

jdav

Member
Joined
Oct 28, 2021
Messages
48
Likes
14
What about the fact switching noise is inaudible? Why would you need to filter it in the first place, beside EMI compliance (which is usually of secondary importance in the amateur world)
 

MaxwellsEq

Major Contributor
Joined
Aug 18, 2020
Messages
1,627
Likes
2,421
What about the fact switching noise is inaudible? Why would you need to filter it in the first place, beside EMI compliance (which is usually of secondary importance in the amateur world)
Just because you can't hear it doesn't mean it's not messing things up. High frequency noise can cause overload problems with downstream items that have wider frequency responses; there can be intermodulation with artefacts back in the audio band.
 
OP
J

jdav

Member
Joined
Oct 28, 2021
Messages
48
Likes
14
Just because you can't hear it doesn't mean it's not messing things up. High frequency noise can cause overload problems with downstream items that have wider frequency responses; there can be intermodulation with artefacts back in the audio band.
What do you mean by overload problem with downstream items?

For the other point, are there realistic situations where intermodulation would create audible distortion? I guess it will depend on the amount of switching noise in the first place and on PSSR, but how is that "tricky" to deal with? Isn't filtering switching noise down to low enough level quite easy?
 

MaxwellsEq

Major Contributor
Joined
Aug 18, 2020
Messages
1,627
Likes
2,421
What do you mean by overload problem with downstream items?
Some power amplifiers have wideband frequency responses into the 100kHz. I've measured an amplifier clearly drawing a lot of current when no signal was applied. It was badly designed and suffering a resonance when exposed to RF.
For the other point, are there realistic situations where intermodulation would create audible distortion? I guess it will depend on the amount of switching noise in the first place and on PSSR, but how is that "tricky" to deal with? Isn't filtering switching noise down to low enough level quite easy?
Yes it is relatively easy to do with filtering, local regulation and bypassing. But designers need to also pay close attention to circuit layout to minimise parasitic coupling. In the past this was an "artform", but with modern simulations, it's easier.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
It took a long time to shift from linear power supply to SMPS in high end audio producs. Nowadays, SMPS seem to have completely replace linear ones except in some niche areas.

I've often heard the advice to beginners to not touch SMPS and stick to linear supply for DIY projects, because designing for SMPS is too difficult. What is it so tricky? Why couldn't we just put the switching frequency well outside the audible range and be done with it? I'm sure there will be some problem somewhere but I don't know exactly which one.

For comparison, I tore down my RME UFX II interface to find a XP Power VCT60US12 inside. It's not a particularly quite one (max 1% ripple p-p), switching at 60kHz, and I didn't see any massive filtering downstream. It seems they are quite ok leaving a reasonably high amount of noise in their power supply rail. The negative power supply is created using an other switch-mode inverter IC. I'm aware this is quite low power (max 36W) device, but it's also a low signal one, with mic preamps inside. How is it possible on such a high end interface?

Thanks

It is just a matter of proper engineering.

SMPS are cheap, have less heat, are smaller, weigh less and one can easily create various voltages, can be wide range and when properly filtered can be very low leakage and inject allowable 'garbage' back into the mains.

Transformers are big, limited in frequency (50-60Hz), weigh a lot, are not regulated, can still leak currents capacitively but arguably can have much better longevity.

It is the weight and longevity that inspires 'confidence' in such power supplies.
Both can be very low noise.
 
OP
J

jdav

Member
Joined
Oct 28, 2021
Messages
48
Likes
14
SMPS are cheap, have less heat, are smaller, weigh less and one can easily create various voltages, can be wide range and when properly filtered can be very low leakage and inject allowable 'garbage' back into the mains.
Leakage from primary to secondary?
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
Yep and the other way around.
So do transformers but usually less and construction dependent.
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
It could be a problem when 2 devices connected to mains have a different leakage.
Could be low frequency but this will be low. For HF most power supplies are either transparent or attenuate very little between real ground and the power supply (be them balanced or not). This HF could, with some devices, cause a problem. Usually weird sounds/noises.
When you connect those devices using an RCA cable (or 3.5mm TRS cable) then there will be a current flowing through the shield of that cable.
As all cables have a resistance there will be a small voltage added to the signal. Not so with balanced as the audio path and ground are not the same ones.
With RCA they are.

And it's not the RCA shield resistance alone. In a lot of cases it is the ground path inside the device that may be a bit too high making the problem worse.

Grounded power supplies can help here as well but also make it worse depending on the situation.

With SMPS there often is a deliberate leakage between primary and secondary because of regulations regarding HF emission.
In that case the output is 'grounded' for HF using a small capacitor.
For medical devices this may be too high. So often we see SMPS medical grade being used.
 
OP
J

jdav

Member
Joined
Oct 28, 2021
Messages
48
Likes
14
Thank you very much! Why would introducing leakage reduce emission? As the result of putting a low impedance pass for HF between your secondary and your primary?
 

wwenze

Major Contributor
Joined
May 22, 2018
Messages
1,283
Likes
1,827
Everything has a spec, the power supply comes with its spec, the product works when powered by a power supply that meets a spec, what's there to design?

And for the same amount of money spent on transformer and capacitors, an SMPS will have way better ripple than a 60Hz power supply so you even get less noise in that aspect. The only real issue is that instead of 60Hz you have to filter out e.g. 15,000Hz which arguably requires different approaches.

The only real issues are when you're making a new product for example when I tried to make a class AB headphone amplifier with the negative rail generated using a charge pump with a switching frequency inside audible range so the output has an "eeeeeeeeee". If we are talking about complete products today, audio products have no issue being powered by SMPS, including this $2600 headphone amp here.

 
OP
J

jdav

Member
Joined
Oct 28, 2021
Messages
48
Likes
14
Everything has a spec, the power supply comes with its spec, the product works when powered by a power supply that meets a spec, what's there to design?

And for the same amount of money spent on transformer and capacitors, an SMPS will have way better performance than a 60Hz power supply so you even get less noise in that aspect.

The only real issues are when you're making a new product for example when I tried to make a class AB headphone amplifier with the negative rail generated using a charge pump with a switching frequency inside audible range so the output has an "eeeeeeeeee". If we are talking about complete products today, audio products have no issue being powered by SMPS, including this $2600 headphone amp here.

I feel like your answer is slightly out of point. You're trying to prove to me that SMPS can work well, which I never said was false.

I was just wondering why beginners are often discouraged to use SMPS at first because they are "more tricky to design with", and also how a high end audio product could get away with a quite noisy and cheap SMPS (120mVpp, and only 20$ for a 1200$ unit).

I think other have given a reasonable answer to it.

A question remain: why did it took so long to transition from linear to SMPS? Only 20 years ago, it seems like SMPS were not the norm in high end products.
 

fpitas

Master Contributor
Forum Donor
Joined
Jul 7, 2022
Messages
9,885
Likes
14,191
Location
Northern Virginia, USA
I feel like your answer is slightly out of point. You're trying to prove to me that SMPS can work well, which I never said was false.

I was just wondering why beginners are often discouraged to use SMPS at first because they are "more tricky to design with", and also how a high end audio product could get away with a quite noisy and cheap SMPS (120mVpp, and only 20$ for a 1200$ unit).

I think other have given a reasonable answer to it.

A question remain: why did it took so long to transition from linear to SMPS? Only 20 years ago, it seems like SMPS were not the norm in high end products.
Us engineers tend to be very conservative about new stuff. There's nothing quite like discovering a horrible problem in production, or worse yet when 1000s of units are deployed to customers.
 

wwenze

Major Contributor
Joined
May 22, 2018
Messages
1,283
Likes
1,827
I feel like your answer is slightly out of point. You're trying to prove to me that SMPS can work well, which I never said was false.

I was just wondering why beginners are often discouraged to use SMPS at first because they are "more tricky to design with", and also how a high end audio product could get away with a quite noisy and cheap SMPS (120mVpp, and only 20$ for a 1200$ unit).

I think other have given a reasonable answer to it.

A question remain: why did it took so long to transition from linear to SMPS? Only 20 years ago, it seems like SMPS were not the norm in high end products.

That question is multi-faceted...

Well, the first one is, if we are talking about designing an SMPS module from scratch, it *is* hard. Not just buying a module designed and manufactured by some guy in Asia, but making your own SMPS product or putting an SMPS circuit inside your product. The funniest example I can always remember is seeing my senior at work replacing a whole batch of output capacitors with Panasonic FM because the low-cost default one is not good enough. We're not talking about audiophile kind of "not-good-enough", but "dude the whole batch of LED lamps are flickering" bad. And it's funny because we never see Panasonic FM being used in real life products. Rubycons, yes.

And when you're selling a product, having to buy modules from a third party eats into your profit, so you want to build yourself as much as possible.

And at the lowest power level and price level, unregulated / linear regulators are the cheapest. Just slap a $0.20 IC and you're done. This is why even the products today that you see are using SMPS for their main supply, the onboard minor regulators are usually still linear. But as power level increases, the advantage starts to tilt towards SMPS for cost/performance. Originally this tilt happens later and we could still see 1A 12V transformer wall-warts powering my Casio keyboard, but now a 2A 12V SMPS weighs less and costs less.

And, the noise issue, which is not entirely unrelated to the first point so my post is ending up pretty messy. But putting an SMPS on the same PCB as your amplifier takes skill. Heck, today we have speakers that emit noise when its Bluetooth is disconnected or when its Bluetooth is connected, just to illustrate the difficulty.

And finally perhaps the most important reason, it's the high end product buyers who believe SMPS bad. Class D amp bad. Diamond power cable good.

20 years ago SMPS was still expensive at low power levels. Hence I have memories of using those fat adjustable wall-wart to power my Game Gear.
 

fpitas

Master Contributor
Forum Donor
Joined
Jul 7, 2022
Messages
9,885
Likes
14,191
Location
Northern Virginia, USA
The electrolytics for SMPS improve all the time too, if for no other reason that there is now money to be made making good electrolytics.
 
Top Bottom