If you had understood rules of electronics and thinked a little, you would realize that two amplifiers with the same topology and the same load impedance needs exactly the same puffer capacitance on supply rails! The more the supply voltage the more the current, but at more voltage the acceptable voltage variation increases too with the same percentage, so C=I*t/deltaV remains perfectly constant. Capacitance depends only on topology, number of channels, load impedance, and a "quality factor" wich is subjective. Power is absolutely irrelevant, falls out from equations, and there is no need for stupid rule of thumbs! Switching freq? Don't be ridiculous! Why don't you tell something about the color of the power cable? This would have the same relevance.
You could tell with the same way, that a ClassD amp needs 4700 uF regardless of everything.
It's a crying shame! For a biginner this kind of lack of knowldge would be acceptable.
Pafi said:
Pafi
Maybe I was not paying proper attention, but are you saying that the required buffer capacitance is NOT dependent upon the switching frequency?
That would imply that for the same load, a "regular" power supply and a SMPS would have the same requirement as far as output capacitors, right?
If so, I attended the wrong electrical engineering school... or the alcohol has melted my last brain cell.
Please advise/clarify. Thanks!
Hi Pafi
For a class D amplifier it's the energy stored compared to the power output that matters, not simply the capacitance.
Stored energy is 0.5*C*V^2, so a 2000uf capacitor charged to 100V stores 10 joules, exactly the same as a 500uF capacitor charged to 200V or a 125uF capacitor charged to 400V. You'll also find that all three capacitors will be roughly the same size and cost for the same amount of stored energy
If the amplifier is putting out 100W into the load then this will make the supply drop by the same percentage over time in both cases -- it would take 0.1 seconds to completely discharge the capacitor if the SMPS wasn't topping it back up.
Cheers
Ian
For a class D amplifier it's the energy stored compared to the power output that matters, not simply the capacitance.
Stored energy is 0.5*C*V^2, so a 2000uf capacitor charged to 100V stores 10 joules, exactly the same as a 500uF capacitor charged to 200V or a 125uF capacitor charged to 400V. You'll also find that all three capacitors will be roughly the same size and cost for the same amount of stored energy
If the amplifier is putting out 100W into the load then this will make the supply drop by the same percentage over time in both cases -- it would take 0.1 seconds to completely discharge the capacitor if the SMPS wasn't topping it back up.
Cheers
Ian
dudaindc!
Not exactly, because between a regular PSU and an SMPS there are more difference then simply the freq, they work with different waveform, differrent filter topology, etc..., but basically they needs similar cap value, because we are talking about amplifiers! The cap is not only for PSU, but for amplifier, and this is the strongest requirement! SMPS needs 10...100 uF, depending on freq, but who cares, when amp needs more then 1000 uF?
Also we were talking about real SMPSes (for amplifiers), so audible switching freq is out of question already!
Are you satisfied by this answer?
iand!
You are absolutely right, and this is exactly the situation with my calculation (and this is what is failed with fredos's calculation). If you increase the supply voltage for twice, max. output power will grow to 4 times of original, and energy in capacitor will also grow 4 times. You can calculate capacitor in many different ways, this is one of them, but all ways you will find the same figure, if you do it correctly, considering real facts.
This is the theory, based on volume of dielectricum, but in reality an elco have a bigger amount of conductor (electrode) in it then dielectricum, and in a small voltage elco there is a bigger electrode for same energy, hence they are bigger (and more expenisve). Check in a price list! At my favourite part supplier 2 pcs of 100V 2200 uF costs 4 euro, while a 200V 1000 uF (90% energy, 60% size) costs only 1 euro! (OK, this is extreme, at other suppliers there are smaller differences in price.)
Not exactly, because between a regular PSU and an SMPS there are more difference then simply the freq, they work with different waveform, differrent filter topology, etc..., but basically they needs similar cap value, because we are talking about amplifiers! The cap is not only for PSU, but for amplifier, and this is the strongest requirement! SMPS needs 10...100 uF, depending on freq, but who cares, when amp needs more then 1000 uF?
Also we were talking about real SMPSes (for amplifiers), so audible switching freq is out of question already!
Are you satisfied by this answer?
iand!
You are absolutely right, and this is exactly the situation with my calculation (and this is what is failed with fredos's calculation). If you increase the supply voltage for twice, max. output power will grow to 4 times of original, and energy in capacitor will also grow 4 times. You can calculate capacitor in many different ways, this is one of them, but all ways you will find the same figure, if you do it correctly, considering real facts.
This is the theory, based on volume of dielectricum, but in reality an elco have a bigger amount of conductor (electrode) in it then dielectricum, and in a small voltage elco there is a bigger electrode for same energy, hence they are bigger (and more expenisve). Check in a price list! At my favourite part supplier 2 pcs of 100V 2200 uF costs 4 euro, while a 200V 1000 uF (90% energy, 60% size) costs only 1 euro! (OK, this is extreme, at other suppliers there are smaller differences in price.)
Pafi...Just to be sure...We are talking about unregulated switching PSU, like the one you can find everywhere in the QSC website? What the output capacitance have to do with amplifier (at certain level....)??? Output transformer waveform is pur squared wave, that full wave rectified, then go directly to output capacitor...What's the job of these capacitor? They are there just to damp the peak power draw by amplifier. Ripple current have nothing to do there, they are feeded by nearly pur DC, and with low impedance source (the transfo, except if you do something stupid like 100 turn on secondary...). The capacitor that have the major effect are the one at the input of the switching PSU. But they did'nt see large peak because the current draw by the amplifier are damped, first by the output capacitor ( the are feed by nearly DC low impedance voltage...) and they operate at high voltage ( low ripple current). So did you still convict that the output capacitor have a so big influence on amplifier performance?
For sure they need to be there, and they need to damp the power supply, but not as critical as in linear power supply. And yes frequency have something to do, more it's hight, lower is the impedance of the switching power supply, less these damping capcitor will have to do there job...
Anyway, we can talk about that like to talk about speaker cables that change the sound, what's important is that you do your amp in the way you think is better and you find is sound better for you..
Have a nice day!
For sure they need to be there, and they need to damp the power supply, but not as critical as in linear power supply. And yes frequency have something to do, more it's hight, lower is the impedance of the switching power supply, less these damping capcitor will have to do there job...
Anyway, we can talk about that like to talk about speaker cables that change the sound, what's important is that you do your amp in the way you think is better and you find is sound better for you..
Have a nice day!
Pafi
Thanks for your answer - I now have a better understanding of your point of view.
I do understand the basics of the power supply topologies (both linear and switch mode), but I have a different perspective regarding their use to feed power amplifiers (particularly class D ones).
I am not a manufacturer or an electronics systems designer, but I have a "bit more than basic" knowledge of analog/digital electronics and many years of experience with both professional equipment (as a consultant) and hi-fi/diy gear (as an "inquisitive tweaker").
Based on that I can say I am more in line with what Fredos wrote above.
I do see the value of "rules of thumb" although they have to be used carefully. I see them as "good staring points".
So, when you design a PSU to feed an audio power amplifier (single channel, 4-8 ohm load) what is the output buffer capacitance (uF/100W) do you typically use (per rail) for the following scenarios:
a) Class A // linear power supply
b) Class AB // linear power supply
c) Class D // linear power supply
d) Class AB // SMPS
e) Class D // SMPS
Just like I am learning here, other people will benefit from your (and Fredos') experience/perspective and may develop their own "rules of thumb".
Cheers!
(Any more pictures of that nice SMPS and amplifier?
)
Thanks for your answer - I now have a better understanding of your point of view.
I do understand the basics of the power supply topologies (both linear and switch mode), but I have a different perspective regarding their use to feed power amplifiers (particularly class D ones).
I am not a manufacturer or an electronics systems designer, but I have a "bit more than basic" knowledge of analog/digital electronics and many years of experience with both professional equipment (as a consultant) and hi-fi/diy gear (as an "inquisitive tweaker").
Based on that I can say I am more in line with what Fredos wrote above.
fredos said:
I do see the value of "rules of thumb" although they have to be used carefully. I see them as "good staring points".
So, when you design a PSU to feed an audio power amplifier (single channel, 4-8 ohm load) what is the output buffer capacitance (uF/100W) do you typically use (per rail) for the following scenarios:
a) Class A // linear power supply
b) Class AB // linear power supply
c) Class D // linear power supply
d) Class AB // SMPS
e) Class D // SMPS
Just like I am learning here, other people will benefit from your (and Fredos') experience/perspective and may develop their own "rules of thumb".
Cheers!
(Any more pictures of that nice SMPS and amplifier?
)dudaindc!
OK, you can have a different point of view, but are your conclusions different form mine? Do you have any conclusions anyway?
In fredos description there are some untrue statement, and it doesn't make his "rule" any plausible (not to mention prooving). It's just philosophy without any quantifiable conclusion, so if you find it symphatic, it doesn't mean anything. Actually he didn't show the tiniest sign that he sees the problem I mentioned, this because I can't write more argument to him.
"uF/100W"? Undefinable!!! All I try to make clear during this long and exhausting disputation is that ANY uF/W value is essentially WRONG! I prooved it with an example! Didn't you notice it? Neccessary capacitance is directly related to (nominal) admittance of load, and some other factors, exept power!
Despite of my all efforts you forgot some very important factor: PSU topology (single supply/split supply) and amplifier topology (bridge, single ended, transformer push-pull). Impedance is also important, for 4 ohm you should use 2 times higher value then for 8 ohm, so I can't tell a single value for 4-8 ohm!
a) ClassA: its completely impossible to tell typical value, because of extremely widely varying PSRR. Eg.: Pass Zen needs doubly (or more, CLC, stabilised, active capacitor etc...) filtered PSU with very big (>22000 uF) capacitors, or one giantic one (>100000uF, and this is still a compromise) but Hiraga needs smaller caps without additional filtering.
Actually this is somewhat true for all amplifiers: there are more sensitive and less sensitive circuits, for example a no feedback true digital amp needs completely smooth supply voltage, so for good result you have to use either a huge cap (eg. 47000 uF) or stabilisation. In these cases the cap is not just an energy storage, and this is not related to classes as fredos told (this is his strongly limited experience again).
I will try to make a calculation table later.
OK, you can have a different point of view, but are your conclusions different form mine? Do you have any conclusions anyway?
In fredos description there are some untrue statement, and it doesn't make his "rule" any plausible (not to mention prooving). It's just philosophy without any quantifiable conclusion, so if you find it symphatic, it doesn't mean anything. Actually he didn't show the tiniest sign that he sees the problem I mentioned, this because I can't write more argument to him.
"uF/100W"? Undefinable!!! All I try to make clear during this long and exhausting disputation is that ANY uF/W value is essentially WRONG! I prooved it with an example! Didn't you notice it? Neccessary capacitance is directly related to (nominal) admittance of load, and some other factors, exept power!
Despite of my all efforts you forgot some very important factor: PSU topology (single supply/split supply) and amplifier topology (bridge, single ended, transformer push-pull). Impedance is also important, for 4 ohm you should use 2 times higher value then for 8 ohm, so I can't tell a single value for 4-8 ohm!
a) ClassA: its completely impossible to tell typical value, because of extremely widely varying PSRR. Eg.: Pass Zen needs doubly (or more, CLC, stabilised, active capacitor etc...) filtered PSU with very big (>22000 uF) capacitors, or one giantic one (>100000uF, and this is still a compromise) but Hiraga needs smaller caps without additional filtering.
Actually this is somewhat true for all amplifiers: there are more sensitive and less sensitive circuits, for example a no feedback true digital amp needs completely smooth supply voltage, so for good result you have to use either a huge cap (eg. 47000 uF) or stabilisation. In these cases the cap is not just an energy storage, and this is not related to classes as fredos told (this is his strongly limited experience again).
I will try to make a calculation table later.
fredos!
Actually I don't know why to restrict discussion to unregulated PSU with strongly coupled transformers, there are many other possibilities (QSC is not the world) but I noticed you talk about these. Your calculation is basically wrong in these cases too. You ignored rules of physics, and the result at small power, small impedance made this clear.
Here is your (incomplete, and contrary to your other statements) answer to your question:
And not to forget: damp the current wich is pumped back by amp! This task is absolutely not done by such a little cap as your "rule" tells at low power, low impedance, in half bridge. And needlessly you blamed me, this is not my fault, nor of the amplifier!
You have some other statement contradictionary to reality, but I have no time to correct them (and wait for ridiculous answers), and what is worst, they have no relevance on original statements, so I finish it now.
Actually I don't know why to restrict discussion to unregulated PSU with strongly coupled transformers, there are many other possibilities (QSC is not the world) but I noticed you talk about these. Your calculation is basically wrong in these cases too. You ignored rules of physics, and the result at small power, small impedance made this clear.
Here is your (incomplete, and contrary to your other statements) answer to your question:
And not to forget: damp the current wich is pumped back by amp! This task is absolutely not done by such a little cap as your "rule" tells at low power, low impedance, in half bridge. And needlessly you blamed me, this is not my fault, nor of the amplifier!
You have some other statement contradictionary to reality, but I have no time to correct them (and wait for ridiculous answers), and what is worst, they have no relevance on original statements, so I finish it now.
Pafi said:
Yes and no.
Similar to your approach, I tend to use more buffer caps than what Fredos suggested when using linear power supplies to feed any type of audio amplifiers.
On the other hand, I use significantly less than you suggested when using SMPS to feed any type of amplifier under any load condition (between 4 and 8 ohms).
Since I have been successful (lucky?) with my DIY amps so far, I never went through the exercise of defining better design parameters before.
I guess I was lazy and "got a ride" on the knowledge of well known designers that have more years of experience than my age and yours added together.
I do not believe we are having a "disputation", but rather an informative discussion. You seem passionate about your opinions and you seem to hold them in high regards.
Yes I did notice your efforts to quantify/qualify your point of view. However, the fact that you proved something to yourself (based on your assumptions) does not make it the absolute thruth.
I am not saying you are right or wrong (or even a bit of both) - that does not matter anyway.
I do value your intention to share your knowledge and experience - otherwise this discussion would be a waste of my time and yours.
Yes indeed... my request was apparently neither clear nor comprehensive. That is my fault since English is my 4th language and not my first - I do believe it is not yours either.
In a nutshell, what I am trying to learn from you are your general design rules for amplifier power supplies (not necessarily the theory behind them).
As I understand it, from your perspective load admittance and all the other relevant factors (except the amplifier output power) should be taken into account when determining the amount of buffer caps.
I, on the other hand, believe that the amplifier output power is somewhat relevant.
So, I want to see where our common grounds are (no punt intended) by comparing your "rules of thumb" to mine. (If you do not like the term, please refer to them as practical guidelines or something else.)
The idea of using diferent scenarios (amps and PSUs) was my attempt to bridge your approach and mine.
If you are still game for this exercise, please forget about class-A and concentrate on:
(1) a class AB single amplifier channel connected to a "real world" 4-ohm (nominal) load; the amplifier is being fed by a symmetrical linear power supply; the RMS ouput power is 100W/4 ohms.
(2) a class AB single amplifier channel connected to a "real world" 4-ohm (nominal) load; the amplifier is being fed by a symmetrical switch-mode power supply; the RMS ouput power is 100W/4 ohms.
(3) a class D single amplifier channel connected to a "real world" 4-ohm (nominal) load; the amplifier is being fed by a symmetrical linear power supply; the RMS ouput power is 100W/4 ohms.
(4) a class D single amplifier channel connected to a "real world" 4-ohm (nominal) load; the amplifier is being fed by a symmetrical switch-mode power supply; the RMS ouput power is 100W/4 ohms.
Please assume that these single-channel amps are:
- used for full-range audio
- not the result of bridging 2 amplifier channels
- not transformer push-pull configuration
(I hope I provided better information this time around.)
I look forward to that. Thanks!
You speaks english much better then me, but the way you think is from the far east. This seems to be the root of our understanding problem. You can easily accept the result of two contradictionary theory at once. OK, it's your choice!
Similarity is apparent. I didn't say I use more cap. Actually sometimes I use smaller cap. Try this xls file with 2*50W 2 ohm, in-phase, and 2*1000W 8 ohm, out of phase! In first case very much more cap needed, in second case I suggest a little smaller cap then fredos.
If you don't care reality, and pay much more attention to emotions then arguments, then I can't proove anything to you indeed. Absolute truth doesn't exist, but this doesn't mean that every attempt to prove something can be ignored by saying "this is only your opinion". But this is really only my opinion.
Similarity is apparent. I didn't say I use more cap. Actually sometimes I use smaller cap. Try this xls file with 2*50W 2 ohm, in-phase, and 2*1000W 8 ohm, out of phase! In first case very much more cap needed, in second case I suggest a little smaller cap then fredos.
If you don't care reality, and pay much more attention to emotions then arguments, then I can't proove anything to you indeed. Absolute truth doesn't exist, but this doesn't mean that every attempt to prove something can be ignored by saying "this is only your opinion". But this is really only my opinion.
Pafi said:
East yes, but not quite as far east...
I did not "accept" either theory, but tried to understand both viewpoints without judgement or jumping to conclusions. I am grateful for your help on that.
This is great - for nuckleheads like myself, this illustrates your perspective much better. Thanks!
Originally posted by Pafi
If you don't care reality, and pay much more attention to emotions then arguments, then I can't proove anything to you indeed. Absolute truth doesn't exist, but this doesn't mean that every attempt to prove something can be ignored by saying "this is only your opinion". But this is really only my opinion.
I do care about reality and that is why I am trying to learn more about things that are not clear to me.
I am probably just passionate about stuff than you are, but I do my best to leave emotions out of discussions (not always successfuly, of course).
As far as proving anything, I would not worry too much about that. Just make your point and let other people draw their own conclusions - you have been successful with that.
By no means your opinion was ignored - it was only questioned. As I wrote before, "I do value your intention to share your knowledge and experience".
Time to "play with numbers"...!
Thanks again!
not you Dudaindc.
Anyway,there are some thread in power supply design people trying to make chaper smps.I dont remenber now the name of the thread,but one of them was about modifing an old pc smps fo 250w to unregulated and feed a class d amplifier+-45volts,but this involved modification even primary turns per volt,a lot of capacitance in output...
Other used 2110+3525+toroid+irfp460.
Its not very expensive...
Anyway,there are some thread in power supply design people trying to make chaper smps.I dont remenber now the name of the thread,but one of them was about modifing an old pc smps fo 250w to unregulated and feed a class d amplifier+-45volts,but this involved modification even primary turns per volt,a lot of capacitance in output...
Other used 2110+3525+toroid+irfp460.
Its not very expensive...
Luka!
Aren't you afraid of electrical shock? I think this SMPS is the one with the smallest creepage distance I have ever seen. Official rule says 8 or 10 mm.
odnaizutra!
I'm sorry about correcting false advices, and working for hours with the correct one! I promise it won't happened again!
2*2=4? It's just another point of view. You are still free to think 2*A=4 instead.
dudaindc!
I know the phrase "to be questioned" have different meanings, but I like exact expressions. You didn't have any question about what I tried to tell. I'd rather say you denied it. If you had any question about it, I could answer it, and I had any chance to convince you.
Aren't you afraid of electrical shock? I think this SMPS is the one with the smallest creepage distance I have ever seen. Official rule says 8 or 10 mm.
odnaizutra!
I'm sorry about correcting false advices, and working for hours with the correct one! I promise it won't happened again!
2*2=4? It's just another point of view. You are still free to think 2*A=4 instead.
dudaindc!
I know the phrase "to be questioned" have different meanings, but I like exact expressions. You didn't have any question about what I tried to tell. I'd rather say you denied it. If you had any question about it, I could answer it, and I had any chance to convince you.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.