Less than you would hope for. If the load is fairly static, like a preamp, those tricks can work pretty well. But a load with huge dynamics like an Class AB, B G or D amp they mostly just add heat and loss to the supply.
Carver took a different approach with a very high voltage raw supply that was allowed to sag a lot under load. However that brings the question of modulation of the audio signal as the supply rails change. I don't see a lot of diy efforts toward duplicating a Phase Linear amp so maybe its not the way to go.
At the other extreme is the large cap bank. Biggest challenge is charging it on turn on without blowing breakers. But with 5 and 10 degree conduction angle the 1A RMS from the wall can become 10-20A peaks. The AC power source has a finite impedance, say .5 Ohm (500' loop of 10 AWG to the pole and assuming a "perfect" transformer to make things easier) that is a 5V drop during the charge period. This explains the 3% THD on the power line (flattened peaks). And the peak current will really stress the power connections way beyond what seems reasonable. I have seen 15A power inlet modules melt under these conditions.
I would focus on a balance based on the amps sensitivity to supply changes and ripple, regulate the parts that can be without a huge power premium (figure a linear supply will probably be less than 50% efficient) and then struggle with the radiated fields, the line noise coupling through the transformer and supply circuits etc. which are much knottier problems to solve.
An array of small caps can have a much lower esr than a single large cap of the same value, but it can all be blown if not implemented well. Start by following the current paths. If any end up in series with a voltage or feedback reference you have a problem. Its easy to overlook a hidden negative or positive feedback node from the currents. They don't show on schematics.
Well, then, we're agreeing. There is no "rule," you have to actually engineer competently to get a desired performance.
If in the 30's/40's the Internet had existed, in this discussion, the values of C would be several orders of magnitude lower.IMHO
Some mJ more, some less mJ due to chokes, E = (1/2) L i^2
Ofcourse, just at that time were designed valves that we use today, and the people who did was stupid enough to suggest, at least, a rule of Dumb.
BTW, hi-end audio was not a major issue in those days.
Some mJ more, some less mJ due to chokes, E = (1/2) L i^2
Ofcourse, just at that time were designed valves that we use today, and the people who did was stupid enough to suggest, at least, a rule of Dumb.
BTW, hi-end audio was not a major issue in those days.
Like anything else over the last decade and a half, bigger and bigger in the top end.
Most of which is marketing, where there's bundles of free cash, there's demand, supply will follow.
Not necessarily proves any technical arguement, as little as me telling you that it sounds better.
Either larger size transformers, increased capacitance, or both, has been a trend for decades.
Capacitance multipliers for SS output stages have gained popularity since the Metaxas Soliloquy of the mid '80s, there's the sporadic 3-phase powersupply, triple "boost" powersupplies of an MBL, and for a number of years SMPS's, as e.g. Levinson.
Another cap multiplier example, Orpheus monaural : http://2.bp.blogspot.com/_FbSDy9bT5...YI/3CPzvya0axo/s1600/__Privilege_Mono_PCB.jpg
> 3J/10W (4Ohm)
Perhaps someone can help me out here.
Capacitance multipliers are good at reducing ripple, but what you want in a power amp is not just low ripple, but also a supply that can deliver large currents at short notice. To that end, you need to store energy that can't be instanteneously delivered by the transformer. A capacitance multiplier does not store much energy.
From the perhaps misguided idea that a power supply for an end stage needs to store an immediately available amount of energy. Isn't it so that a capacitance multiplier would loose steam as soon as the stage that feeds the capacitance multiplier can no longer keep up with the instantaneous current demand? If so, you would need to store the required joules of energy in the stage preceeding the capacitance multiplier, requiring pretty much the same capacitance as you might without a capacitance multiplier in the loop.
This would change if that preceeding capacitance were run at a higher voltage than the rails, with the capacitor multiplier dissipating the delta. But that, in effect, would be pretty much the same as a regulator.
Just some free thinking on the late Sunday night, questioning what the role of a capacitance multiplier in an end stage PS might possibly be.
vac
Capacitance multipliers are good at reducing ripple, but what you want in a power amp is not just low ripple, but also a supply that can deliver large currents at short notice. To that end, you need to store energy that can't be instanteneously delivered by the transformer. A capacitance multiplier does not store much energy.
From the perhaps misguided idea that a power supply for an end stage needs to store an immediately available amount of energy. Isn't it so that a capacitance multiplier would loose steam as soon as the stage that feeds the capacitance multiplier can no longer keep up with the instantaneous current demand? If so, you would need to store the required joules of energy in the stage preceeding the capacitance multiplier, requiring pretty much the same capacitance as you might without a capacitance multiplier in the loop.
This would change if that preceeding capacitance were run at a higher voltage than the rails, with the capacitor multiplier dissipating the delta. But that, in effect, would be pretty much the same as a regulator.
Just some free thinking on the late Sunday night, questioning what the role of a capacitance multiplier in an end stage PS might possibly be.
vac
Last edited:
Its role is to eat the ripple , in some way.
Of course a big cap is also needed at its output ,
let say 20000uF before and 10000uF after...
Exactly.
My rule of dumbs is to design power supply according to sag and ripples that I can tolerate on max RMS power infinitely long. The same rule applies to power dissipation, both in power supply and output components, passive and active. Hi-fi designers usually assume short period of time for peak power levels, so they downsize power supplies, heatsinks, output components, also they assume that all channels can't work at once on full power. However, it is wrong, that's why amps with bigger power supplies tend to sound better. But let me repeat, it is not because bigger is better. It is because not enough is worse than enough.
It is design criterion. One of many other criteria. No rules of dumbs here. Well, almost no. You know that we deal with art translated to design by science.
Last edited:
Thank you very much to both of you, now I'm just dumb, but a little less ignorant.
From this side of the world, I propose nominees for "the gold electron prize" of the week.
I have some logic issues here.
I was led to believe that NOBODY in their right mind listens to steady state sine waves at full power for hours on end. Therefore, a full power operating mode is by deafult intermittent, unless our objective is to fry our speakers with the first few transients which come along.
After 40+ years of listening to music, I have never encountered a situation in which there were several large peaks inside of say 1 second, or 2 seconds.
Therefore, it seems to me that my amp will in fact be working at least 6 dB below its maximum power in any steady state mode. And even that is cutting it close, but never mind. We use very powerful amps in order to have the first say 25 Watts of power as clean as possible, the rest is just reserve for peaks.
I may have been misled in my beliefs by the logic of DIN 45500 standards, which defined "music power" as power capability with suddent peaks, 1 in 16 cycles, wihtout distorting more than 3% of THD. Meaning that you run your amp at nominal rated power and see what it does when 1 in 16 time intervals requires more. Hardly a perfect test, but far from being stupid either.
We tend to shy away from the fact that our AVERAGE DISSIPATED CONTINUOUS POWER is way below our absolute power capabilities probably because we have no way of knowing what might that be in any "average" situation and because we know all of the factors which come into play. It's effectively impossible to know in general terms, but I think it's safe to assume most folks are at around -10 dB mark or below, unless they purchased an amp initially too weak for their needs (in which case it's not our problem any more).
If so, then we need to design our power supplies so that they can respond to our real world needs, not theoretical models from a lab, and we can design these at will, the sky's the limit.
In that respect, industry designers are better informed than we are, even if they generally still tend to undersize for cost reasons only. Not defendinding anyone here, mind you, just noticing.
I think this discussion has moved on to theoretical grounds a little too much - we keep talkning as if we are about to run our amps into a 2 Ohm load with a -60 degree phase shift steady state all of the time, and for hours on end. All the time knowing that this will never happen.
Let's get down to planet Earth once again, shall we?
Nigel already has. I've followed him through evolution from a steady state 1 Ohm load driving capability to impulse capability - a reasonable move, as I see it, since beside Wayne's 1 Ohm speakers, I have never even heard of anything going that low, I mean, 1 Ohm is practically a short circuit. Sure, you can design for it, no doubt, but be ready not to faint when you work out the price of that baby. It's not going to be gentle.
I was led to believe that NOBODY in their right mind listens to steady state sine waves at full power for hours on end. Therefore, a full power operating mode is by deafult intermittent, unless our objective is to fry our speakers with the first few transients which come along.
After 40+ years of listening to music, I have never encountered a situation in which there were several large peaks inside of say 1 second, or 2 seconds.
Therefore, it seems to me that my amp will in fact be working at least 6 dB below its maximum power in any steady state mode. And even that is cutting it close, but never mind. We use very powerful amps in order to have the first say 25 Watts of power as clean as possible, the rest is just reserve for peaks.
I may have been misled in my beliefs by the logic of DIN 45500 standards, which defined "music power" as power capability with suddent peaks, 1 in 16 cycles, wihtout distorting more than 3% of THD. Meaning that you run your amp at nominal rated power and see what it does when 1 in 16 time intervals requires more. Hardly a perfect test, but far from being stupid either.
We tend to shy away from the fact that our AVERAGE DISSIPATED CONTINUOUS POWER is way below our absolute power capabilities probably because we have no way of knowing what might that be in any "average" situation and because we know all of the factors which come into play. It's effectively impossible to know in general terms, but I think it's safe to assume most folks are at around -10 dB mark or below, unless they purchased an amp initially too weak for their needs (in which case it's not our problem any more).
If so, then we need to design our power supplies so that they can respond to our real world needs, not theoretical models from a lab, and we can design these at will, the sky's the limit.
In that respect, industry designers are better informed than we are, even if they generally still tend to undersize for cost reasons only. Not defendinding anyone here, mind you, just noticing.
I think this discussion has moved on to theoretical grounds a little too much - we keep talkning as if we are about to run our amps into a 2 Ohm load with a -60 degree phase shift steady state all of the time, and for hours on end. All the time knowing that this will never happen.
Let's get down to planet Earth once again, shall we?
Nigel already has. I've followed him through evolution from a steady state 1 Ohm load driving capability to impulse capability - a reasonable move, as I see it, since beside Wayne's 1 Ohm speakers, I have never even heard of anything going that low, I mean, 1 Ohm is practically a short circuit. Sure, you can design for it, no doubt, but be ready not to faint when you work out the price of that baby. It's not going to be gentle.
Like most standards, it's a Procrustean bed. Unfortunately, real world source material is not made to DIN standards, and the increased use of compression has made things worse. THD is far less relevant than recovery from overload (e.g., blocking in cap coupled amps, "sticking" to rails in DC amps) and modulation of the signal from the increased ripple components and sag from the power supply.
Lots of words have been used, but we basically still have three criteria, in descending order of uF:
1) use as much capacitance as you can afford, (with some adding "but not too much"),
2) get the ripple sufficiently low on continuous full power signals,
3) with real music you can get away with less than (2).
So far, only (2) and (3) have had engineering explanations and only (2) has the benefit of a calculation. A calculation for (3) would be like that for (2), but with the addition of something relating to duty cycle or music dynamic statistics.
1) use as much capacitance as you can afford, (with some adding "but not too much"),
2) get the ripple sufficiently low on continuous full power signals,
3) with real music you can get away with less than (2).
So far, only (2) and (3) have had engineering explanations and only (2) has the benefit of a calculation. A calculation for (3) would be like that for (2), but with the addition of something relating to duty cycle or music dynamic statistics.
Please follow the link in my sig line.
But it has not, so far in this thread, had the benefit of a precise number. We just say low enough ripple or quote some percentage. How do we determine what that is? How much ripple is good and how much is bad? Is the criterion "Not audible on my speakers at the listening position", or something else?
At some point we have to set how much ripple (or other fault) is acceptable to us, or to those who buy amplifiers.
Berendsen in Germany still offers a 150W stereo power amp type in two versions ; a regular with 60.000uF and a Special Edition with 100.000uF capacitance.
Back in the late '80s, early '90s, i audited their monaural Blue and Red editions on different loudspeakers.
Red or Blue was not distinguishable on an easy loudspeaker, the trickier loudspeaker model favored the pumped up Red brigade.
http://www.diyaudio.com/forums/solid-state/138411-thought-process-diy-er.html#post1742515
Interesting (to me) still are the geographic preferences ;
High VA/W numbers in the US, with a relatively large uF bank, e.g. Parasound amp models.
UK designers who favor soft powersupplies, relatively low size capacitors (and Nigel is unmistakenly British
).Powersupplies as rigid as a tank, with steep electrolytic size, in the Germanic region.
Ultimately it's still the amp circuit that's the key factor.
The Marantz MA24 monaurals (Philips production) and Accuphase P102 from the 80s/90s are both Class A power amps.
Marantz does a nominal 60W, but was conservatively rated, the balanced design Accuphase produces 80W in 4 ohm.
Cost about the same at the time, have close to identical size power supplies, both 2 times 22.000uF and a similar size toroidal per channel (Marantz >3.5 J/10W for 4 ohm).
The Philies beat the Accuphase hands down, doubling the powersupply in the P102 wouldn't make it the winner.
See post 6374.
FWIW, in my own gear, I design to a "lower than the noise floor from 16 bit source material" standard. But that's an individual choice, not a "rule."
I manage, and I'm just a rank amateur.
I should qualify my statement- for a phono stage, I aim for within 1dB of cartridge thermal noise.
- Status
- Not open for further replies.