As I know it switched mode power supplies are made to produce high power without the need of a huge/heavy transformer. By increasing the frequency they need a much smaller transformer (less iron) to pass through the same power as a very heavy transformer at 50 Hz.
^ That is true and as a result they are significantly more efficient than 'normal' power supplies. However the increased switching of large voltages at significant levels of current generates masses of switching noise. This is then the 'new' problem: elminating all that switching noise. This I guess is where Linn have succeeded with their specialized supplies. Designing SMPS is FAR more complicated too.
So you'd take 240V AC/50Hz and feed it straight into a device to multiply its frequency to 100KHz - some sort of oscillator - and you must pass the power required too plus the expected losses further down - then you simply feed that into a small transformer and rectify as usual.
I have not done the calcs but the computer boxes (PSUs) at 250W, 300W and recently 600W or better are quite large, I think I could fit 3 * 300VA toroidals in one of them.
So for practical purposes the switched mode requires a huge fan, coolers and so on, so may be you do not save in size all that much plus you are replacing a piece of iron with very little to go wrong with dozens of active devices that can and do go wrong.
I have not done the calcs but the computer boxes (PSUs) at 250W, 300W and recently 600W or better are quite large, I think I could fit 3 * 300VA toroidals in one of them.
So for practical purposes the switched mode requires a huge fan, coolers and so on, so may be you do not save in size all that much plus you are replacing a piece of iron with very little to go wrong with dozens of active devices that can and do go wrong.
Kind of.
First you filter then immediately rectify mains and smooth it to get a lovely 400V DC supply. Then you accidentally touch it and get the shock of your life!
From that 400V supply you create another little 18V supply to drive an oscillator at about 100KHz. This is PWM modulated depending on the output voltage / current of the supply (fed back via an opto isolator).
the 400V DC is fed to a little transformer via a couple of MOSFETs that are driven by the oscillator output. The low voltage output side of the transformer is then rectified and smoothed to produce required DC output.
Adequate heatsinking of the MOSFETs are required but I doubt a fan is needed (you don't want a fan buzzing around when you want to listen to music really).
At a mere guess, off the top of my head I'd say a SMPS would generally be about a quarter the size (weight) of a normal power supply. This is mainly due to the increased efficiency of SMPS which results in a vastly smaller transformer and smaller heatsinks.
Difficulties in 'cooking your own' SMPS: well basically designing those damn transformers and expect high voltage DC shocks and RF burns. (Guess what I used to do when I was younger).
Andy
First you filter then immediately rectify mains and smooth it to get a lovely 400V DC supply. Then you accidentally touch it and get the shock of your life!
From that 400V supply you create another little 18V supply to drive an oscillator at about 100KHz. This is PWM modulated depending on the output voltage / current of the supply (fed back via an opto isolator).
the 400V DC is fed to a little transformer via a couple of MOSFETs that are driven by the oscillator output. The low voltage output side of the transformer is then rectified and smoothed to produce required DC output.
Adequate heatsinking of the MOSFETs are required but I doubt a fan is needed (you don't want a fan buzzing around when you want to listen to music really).
At a mere guess, off the top of my head I'd say a SMPS would generally be about a quarter the size (weight) of a normal power supply. This is mainly due to the increased efficiency of SMPS which results in a vastly smaller transformer and smaller heatsinks.
Difficulties in 'cooking your own' SMPS: well basically designing those damn transformers and expect high voltage DC shocks and RF burns. (Guess what I used to do when I was younger).
Andy
Thanks for the explanation on switched mode supplies.
However, I have not yet seen a PC PSU without a fan, or one that does not get very hot even with a fan. What I have seen is PSU fans increasing in size (from 8cm to 12cm or more ) and spinning lower thus making less noise - noise parameters are all important nowdays. These are 5 and 12 VDC supplies from 250 watts (old times) to 400, 500, and 600 or more watts, recently needing to power CPUs drawing about 100 watts, plus VGA cards drawing another 200 Watts - plus a plethora of other components.
But I can fit two 300VA transformers inside these PSUs and maybe a few capacitors, I cannot see major space savings, only weight savings as the 300VA transformer is quite heavy. So I would rather go with slightly larger and simple than slightly smaller and extremely complex which gets very hot and needs a fan.
Another thing is ripple. The equation V ripple = I / 2 * f * C -- therefore the 100 KHz switched supply will effectively have no ripple for the same currents drawn - I suppose the final filter capacitors will be tiny. That surely is a huge benefit of switched mode supplies?
However, I have not yet seen a PC PSU without a fan, or one that does not get very hot even with a fan. What I have seen is PSU fans increasing in size (from 8cm to 12cm or more ) and spinning lower thus making less noise - noise parameters are all important nowdays. These are 5 and 12 VDC supplies from 250 watts (old times) to 400, 500, and 600 or more watts, recently needing to power CPUs drawing about 100 watts, plus VGA cards drawing another 200 Watts - plus a plethora of other components.
But I can fit two 300VA transformers inside these PSUs and maybe a few capacitors, I cannot see major space savings, only weight savings as the 300VA transformer is quite heavy. So I would rather go with slightly larger and simple than slightly smaller and extremely complex which gets very hot and needs a fan.
Another thing is ripple. The equation V ripple = I / 2 * f * C -- therefore the 100 KHz switched supply will effectively have no ripple for the same currents drawn - I suppose the final filter capacitors will be tiny. That surely is a huge benefit of switched mode supplies?
Resuming the topic could maybe look like this:
Depending on your circuit design, you´ll need clean noiseless power to feed your amplifier. Op-amp based ones thogh can almost work perfectly on even very noisy supplies, as they have very huge PSRR values.
Discrete designs are a lot more sebsitive to this, as designers often design low noise power suplies for them, rather than doing a lot of work making the circuit unsensitive to noise and riple.
One could say, that in discrete design, the problem is often solved where it is, in op-amp design circuits are made unsensitive to unsolved problems.
Depending on your circuit design, you´ll need clean noiseless power to feed your amplifier. Op-amp based ones thogh can almost work perfectly on even very noisy supplies, as they have very huge PSRR values.
Discrete designs are a lot more sebsitive to this, as designers often design low noise power suplies for them, rather than doing a lot of work making the circuit unsensitive to noise and riple.
One could say, that in discrete design, the problem is often solved where it is, in op-amp design circuits are made unsensitive to unsolved problems.
Unregulated powersupplies are the noisiest ones, and will no matter what, always leave some ripple, which can enter the amplifier circuits.
To me the only logical place to use this way of smoothing is in the power stage of poweramplifiers.
IC Series regulators will attenuate noise and ripple by a certain factor, just around 60-70 dB for IC regulators, which are widely used in consumer electronics. Unfortunately they do not regulate much @ higher frequencies, which sometimes might not be necessary though. In my universe these are suitable for preregulations only.
Discrete series regulators are often much faster than IC regs, but they do not contain any protection circuits, so they can pass spikes at start up, and they can burn if overloaded. They can be designed in a lot of ways, voltage reference, zener diodes, Jung regulators and so on.
These regs are mostly pretty good, and are suitable in most preamps, DACs, and the vooltage amplification in poweramps and others.
Shunts are the Ferrari´s of regulators, they regulate at lightening speeds, and the ripple rejection is fantastic, the downside is heat, as they could be said to work like a class A amplifier, but they would be suitable in preamps, DACs, RIAA amps, MC headamps and maybe headphone amps too.
In power amps they are impossible.
Switch mode PSU´s are by far the most effective there is, they also eliminate the problems about ripple. The price is complexity, and it will independent of the load always have the same high frequency ripple.
But they are the only usable way to regulate supplies for poweramps.
It also eliminates the need for overdimensioning, as they actually performs at their best right until they are overloaded. Their performance does not decline as linear supplies do. So there is no need for electrolytic caps as big as beer cans, as the electrolytics are charged i.e. 100.000 times pr. sec.
The downside is noise and the need for power factor correction.
Last edited:
The need for fans in SMPS is only present, if the load is constant, which it mostly is in computers. For amps there is no such need, also the voltage is a lot higher and the current much lower as in PC´s.
Both Linn´s and other brands aproach towards SMPS can be seen as a need directed by design. SMPS is way smaller and also much cheaper to produca as linear supplies.
The noise can be avoided ore made less harmfull by a few means, switching frequency, screening, filtering and power factor correction.
There isn´t really much more to adress, but marketing needs food, hence the impressing names as "silent power", "Dynamik" and others. They still are noisy SMPS´s with both upsides and downsides.
As far as technology reaches today, SMPS could IMHO be a very attractive alternative in power amps, as you then get a really low impedance regulated supply for your output stage.
In linelevel gear it is IMHO crazy to consider them, as there is in fact no upsides compared to a nice regulated linar supply to gain from it, only downsides in the form of much more noise @ critical frequencies.
The need for filtering also brings the output impedance of the PSU up @ high frequencies, which calls for carefully filtering of the input signals in the amp, else you might risc uncontrolled behavior of your amp @ high frequencies.
Both Linn´s and other brands aproach towards SMPS can be seen as a need directed by design. SMPS is way smaller and also much cheaper to produca as linear supplies.
The noise can be avoided ore made less harmfull by a few means, switching frequency, screening, filtering and power factor correction.
There isn´t really much more to adress, but marketing needs food, hence the impressing names as "silent power", "Dynamik" and others. They still are noisy SMPS´s with both upsides and downsides.
As far as technology reaches today, SMPS could IMHO be a very attractive alternative in power amps, as you then get a really low impedance regulated supply for your output stage.
In linelevel gear it is IMHO crazy to consider them, as there is in fact no upsides compared to a nice regulated linar supply to gain from it, only downsides in the form of much more noise @ critical frequencies.
The need for filtering also brings the output impedance of the PSU up @ high frequencies, which calls for carefully filtering of the input signals in the amp, else you might risc uncontrolled behavior of your amp @ high frequencies.
SMPS are cheap, noisy and unreliable. More failures than any can be attributed to SMPSs. They stress components, make a lot of RF and audible noise and run at the limit of destruction most of the time. Linn use them because they are cheap. Copper and steel cost money, take up space weigh a lot and last forever- no company wants that.
Note also that the fan included in PC SMPS is inteded for use for the WHOLE system i.e. not just the power supply. In other words, when you buy a PSU for a PC, you're buying a PSU and a fan for the computer system as a whole.
So it is!
But for a power amp, it can be considered an upside, as you actually get much less electrical noise. The RF is still there, but ripple in the audible range is much lower.
The only downside within the audible range I see, is the possibility of the PSU midle moving @ low frequencies and high power.
Of course reliability is a matter, and you are right, SMPS was and is a lot less reliable as linear ones. Actually I hate SMPS, but I think they could have some justification in poweramps.
@ linelevel they are an awfull and brainless choise
made only by
I hope they later they will meet their
as consumers are confused about their short life.SMPS unreliable for 2 reasons in my mind (from those I've repaired in the past):
1) Dodgy caps. The industry had a plethora of dodgy Chinese & Japanese Electrolytics a while back. They got EVERYWHERE.
2) With regards to those PSU SMPS's, use of fans not great as they suck a lot of crap into the electronics ... this results in one of two things - i) components get a nice wooly jumper & overheat. ii) @ 400V switching points the dust & crud & moisture & flakes of skin start to short things out & not surprisingly, things go BANG!
Andy
1) Dodgy caps. The industry had a plethora of dodgy Chinese & Japanese Electrolytics a while back. They got EVERYWHERE.
2) With regards to those PSU SMPS's, use of fans not great as they suck a lot of crap into the electronics ... this results in one of two things - i) components get a nice wooly jumper & overheat. ii) @ 400V switching points the dust & crud & moisture & flakes of skin start to short things out & not surprisingly, things go BANG!
Andy
Not impossible. I know of people having their t-amp powered with all discrete, mosfet based, shunt regulators. Works fantastically, they say.
http://www.diyaudio.com/forums/showpost.php?p=1901286&postcount=987
I think what you mean is that cheap, poorly implemented SMPS as found in consumer electronics are unreliable. SMPS can be made to be reliable in even extreme environments and are used almost universally in military and aerospace equipment.
^ Absolutely.
In fact I admire the SMPS in PCs... How on EARTH they make them for that amount of money is quite frankly, staggering. Lots & lots of cheap Chinese labour is how... but how????
Andy
In fact I admire the SMPS in PCs... How on EARTH they make them for that amount of money is quite frankly, staggering. Lots & lots of cheap Chinese labour is how... but how????
Andy
Reliable PC SMPS that are actually capable of the power output they claim are not cheap. It's the cheap ones that give the rest a bad reputation.
So an idea for a low ripple supply that I've been tossing around for a while is as follows:
Use a PWM regulator (eg MIC4684) first stage to take the voltage down to about 15V. Then use a linear regulator to take the voltage down to 12V. It allows higher efficiency and should give good ripple characteristics.
So an idea for a low ripple supply that I've been tossing around for a while is as follows:
Use a PWM regulator (eg MIC4684) first stage to take the voltage down to about 15V. Then use a linear regulator to take the voltage down to 12V. It allows higher efficiency and should give good ripple characteristics.
OK, maybe I was a bit hard on SMPSs in general. I'm sure military and aerospace switching supplies will be reliable, but then again, all systems have to be...except, ironically, the in flight entertainment systems which were (especially the old ones), you guessed it- based on heavily modified consumer equipment!
My attitude is filter the incoming AC for RF/distortion and DC 'before' it gets to the equipment and then do your dammdest not to generate any more- at any frequency. You have to wrap ferrite chockes around every bit of exposed wire just to come close to an EMI regulation. SMPS supplies are like petrol leaf blowers- they clean your driveway really well but spread all the rubbish someplace else.
My attitude is filter the incoming AC for RF/distortion and DC 'before' it gets to the equipment and then do your dammdest not to generate any more- at any frequency. You have to wrap ferrite chockes around every bit of exposed wire just to come close to an EMI regulation. SMPS supplies are like petrol leaf blowers- they clean your driveway really well but spread all the rubbish someplace else.
It is right that designing a reliable and noiseless SMPS is somewhat of a challenge, actualle me and my friend "Hurtig" are trying to do just that. The idea in the first plase came aboutas we tried to measure ripple on a linear supply, andwere quite shocked about the size of it. I have to ad that it is in power amp design these problems are only.
So if it should be a linear one, there would be no way around seperating current amplification and voltage amplification, so that the supply for the voltage amp would be regulated heavily. In the supply for current amplification there is just nothing to do about it away from enormous electrolytic banks, and these, in our experience, do not sound very well. Thus the SMPS came about, but it is also for a class D amp which switches heavily allready.
But as it is with linear supplies, you´ll have to take your precautions of the imperfectnesses of your supplies. In linear it is lowfrequency ripple, in SMPS it is high frequency ripple and noise sent backwards into the wall outlet, as well as airborn noise inside your amp.
It is therefor not advisable to change a linear supply agains some SMPS, as the amp almost surely wont like it. This might even also be true the other way around.
SMPS is surely a devils creation, but cleverly used I think it might bring some good with it after all.
Now I just hope that my house wont burn down, as a class D amp with SMPS should be left on always
So if it should be a linear one, there would be no way around seperating current amplification and voltage amplification, so that the supply for the voltage amp would be regulated heavily. In the supply for current amplification there is just nothing to do about it away from enormous electrolytic banks, and these, in our experience, do not sound very well. Thus the SMPS came about, but it is also for a class D amp which switches heavily allready.
But as it is with linear supplies, you´ll have to take your precautions of the imperfectnesses of your supplies. In linear it is lowfrequency ripple, in SMPS it is high frequency ripple and noise sent backwards into the wall outlet, as well as airborn noise inside your amp.
It is therefor not advisable to change a linear supply agains some SMPS, as the amp almost surely wont like it. This might even also be true the other way around.
SMPS is surely a devils creation, but cleverly used I think it might bring some good with it after all.
Now I just hope that my house wont burn down, as a class D amp with SMPS should be left on always
For power amps perhaps the switched mode PSU may be a good solution.
For pre-amps a cascade of filters/regulators seems to be the concensus or a shunt regulator but at the expense of more complexity. Some practical experiments I did produced some unexpected but interesting side-effects.
When you cascade filters you are really making an oscillator and it may seem to work well at one level of current being drawn, but try to increase the current and it starts oscillating, and I have not drilled down but it must be all the R-C one after the other, creating this ... sounds like a low frequency oscillation, perhaps like 7 Hz or thereabouts. So it seems to me that adding an R in the power supply followed by a C and then other Rs and Cs is not a great idea unless you never plan to increase the current being drawn.
For pre-amps a cascade of filters/regulators seems to be the concensus or a shunt regulator but at the expense of more complexity. Some practical experiments I did produced some unexpected but interesting side-effects.
When you cascade filters you are really making an oscillator and it may seem to work well at one level of current being drawn, but try to increase the current and it starts oscillating, and I have not drilled down but it must be all the R-C one after the other, creating this ... sounds like a low frequency oscillation, perhaps like 7 Hz or thereabouts. So it seems to me that adding an R in the power supply followed by a C and then other Rs and Cs is not a great idea unless you never plan to increase the current being drawn.
- Status
- Not open for further replies.