Member
Joined 2009
Paid Member
And that was my point. There is no magic formula (+1 to MiiB) - depends on many variables.
Nico,
Actually, I don't know. I guess it could work, if we used superconductors near zero degrees kelvin. <grin>
The 7240 uF minimum might work, for that squarewave, especially if it were a decoupling cap, right across a power transistor circuit. But what kind of risetime could it provide for that squarewave? In the ideal case it would have to have been around 2 nanoseconds (i.e. it wasn't accounted for). The answer is it could not provide the rising edge at all. It could only maintain the flatnss of the top of the squarewave. If it had used up any charge making the rising edge then it would need to have been a larger capacitor to still be able to also maintain the level of the top part for the full 20 ms. (More on that below.)
WAIT! If we only need 3.54 Amps max (100 W, 8 Ohms), then my earlier calculation of the reservoir capacitance requirement was done WRONG, in the last step, at the end of post # 106!!
It should have been:
C = [(0.6036)(3.54)] / [(60)(2.52)] = 14132 uF
For a 40V/5A power supply, that gives 2800 uF/A.
Or, if you use the maximum output current that can give 100 Watts across the 8 Ohm load, 3.54A, then the rule would be 3992uF/A, or about
"4000 uF per amp of output current".
OK, it kind of makes sense to double your value from 2000 to 4000 uF/A, because the 7240 uF would barely be enough to maintain the TOP of the square wave. But what about the rising edge? Another 7240 uF could provide just enough charge to make a current to raise the voltage across the 8 Ohms to 28.32 Volts, while lowering the rail by no more than 2.52 Volts, in 5.154 ms (equivalent to a being able to handle a worst-case frequency component down to 61.76 Hz).
That should be almost slow-enough, for the main reservoir caps, since the charging pulses will come more often, for lower frequencies (anything below mains frequency). Higher frequencies or faster-rising edges are easier, in one sense, since the same current rise in less time can be done by a smaller cap.
Alternatively, you could keep your "2000 uF/A output max" rule, with the understanding that maybe at least as much capacitance would also need to be used for decoupling.
I would probably rather see it that way, since the capacitance is worth more when used for decoupling anyway, because the rail inductances are no longer in the way.
Cheers,
Tom
Begin,
By what you insinuate you throw all the variables in a bottle, shake them up good and the way the elements fall will determine what you have. Sounds like a shot in the dark. Imagine if Steven Jobs did the with cellphone keys, some phones may lack the #9 while having two #2s and so on. Everyone is happy because they perceive by dialing the wrong number they may make new friends.
By what you insinuate you throw all the variables in a bottle, shake them up good and the way the elements fall will determine what you have. Sounds like a shot in the dark. Imagine if Steven Jobs did the with cellphone keys, some phones may lack the #9 while having two #2s and so on. Everyone is happy because they perceive by dialing the wrong number they may make new friends.
Yes. Your point is well-taken.
There is a magic formula. But it is large and difficult to apply correctly.
We would probably need one of those expensive full-up electromagnetic simulation software suites that solves Maxwell's equations for everything as a whole.
But we can still use what we know, and try to determine what is most important, and maybe find some rules of thumb. Even just expanding what we know and trying to become aware of more of the important considerations is a worthy goal.
Last edited:
"the sheer force and bass slam generated by the Macro,"
"the Krell KSA-250 couldn't supply the Macro Reference's enormous deep-bass volume,"
"The Crown Macro Reference is very possibly the best sound-reinforcement amplifier ever built. I've heard these monster amps drive multiple-woofer towers at extremely high spls at outdoor concerts, without clipping or distortion. "
"the Crown is, generally, clean and powerful. It has a tight, punchy bottom end. In the latter respect it bettered the Krell KSA-250—no slouch in this area. "
"an immensely powerful, classically designed amplifier with a fine, deep, tight bottom-end. "
"If all you want out of your audio system is "lots of bass," the Crown Macro Reference may be just the ticket. But if you're more interested in dynamic shading, harmonic accuracy, transparency, and the more introspective aspects of music-making, look elsewhere."
1200W/4Ω (x2)
While Stereophile didn't like the overall performance, they were impressed with the force and slam in the bass.
How big are the filter caps in this 1200W/4Ω x2 amplifier?
It has a single pair of 6,300µF: that's it.
Regulate the low current for the voltage gain stages, that's enough.
The bypassing of the main filter caps, and the ground-current paths are far more important.
"the Krell KSA-250 couldn't supply the Macro Reference's enormous deep-bass volume,"
"The Crown Macro Reference is very possibly the best sound-reinforcement amplifier ever built. I've heard these monster amps drive multiple-woofer towers at extremely high spls at outdoor concerts, without clipping or distortion. "
"the Crown is, generally, clean and powerful. It has a tight, punchy bottom end. In the latter respect it bettered the Krell KSA-250—no slouch in this area. "
"an immensely powerful, classically designed amplifier with a fine, deep, tight bottom-end. "
"If all you want out of your audio system is "lots of bass," the Crown Macro Reference may be just the ticket. But if you're more interested in dynamic shading, harmonic accuracy, transparency, and the more introspective aspects of music-making, look elsewhere."
1200W/4Ω (x2)
While Stereophile didn't like the overall performance, they were impressed with the force and slam in the bass.
How big are the filter caps in this 1200W/4Ω x2 amplifier?
It has a single pair of 6,300µF: that's it.
Regulate the low current for the voltage gain stages, that's enough.
The bypassing of the main filter caps, and the ground-current paths are far more important.
I think it is a lot simpler than we think. The current discharge for a cap is described by:
The energy is 100 watt the low voltage is 28V and current is 3.5Amp these are given..To store 100 Joule for 1 second would require a cap with applied voltage V so that E = 0.5CV^2 If the rail voltage was Vrms x 1.414 plus 6V drop accross the lateral mosfet then Vsupp = 45.5V thus 100=0.5 x C x 45.5^2 or 0.01 = C or 100 000uF, but the requirement calls for 20 mS, thus 100 000uF x 0.02 = 1.930 mF or 1930uF. According to my simple arithmetic a 1930 Capacitor is the minimum requirement for this application. Also knowingly that the requirement was that the transformer has no influence at any time, but if it was a 50Hz line it will and as the capacitor is discharged the transformer will resume charging it so this capacitor is in fact just a reservoir of energy.
Okay so can we say that for a 100 watt amp into 8 ohm the minimum reservoir cap should be 2200uF/rail. Into 4 Ohm obviously it would have to be doubled to say 4700uF.
DJK what are you saying that minimum is what we should target. Earlier on we said that the front end being regulated is a given we are only concentrating on the back business end.
The energy is 100 watt the low voltage is 28V and current is 3.5Amp these are given..To store 100 Joule for 1 second would require a cap with applied voltage V so that E = 0.5CV^2 If the rail voltage was Vrms x 1.414 plus 6V drop accross the lateral mosfet then Vsupp = 45.5V thus 100=0.5 x C x 45.5^2 or 0.01 = C or 100 000uF, but the requirement calls for 20 mS, thus 100 000uF x 0.02 = 1.930 mF or 1930uF. According to my simple arithmetic a 1930 Capacitor is the minimum requirement for this application. Also knowingly that the requirement was that the transformer has no influence at any time, but if it was a 50Hz line it will and as the capacitor is discharged the transformer will resume charging it so this capacitor is in fact just a reservoir of energy.
Okay so can we say that for a 100 watt amp into 8 ohm the minimum reservoir cap should be 2200uF/rail. Into 4 Ohm obviously it would have to be doubled to say 4700uF.
DJK what are you saying that minimum is what we should target. Earlier on we said that the front end being regulated is a given we are only concentrating on the back business end.
Last edited:
In the end, for a power supply, ALL that matters is what happens right at the load's power and ground pins.
So maybe the magic formula is THERE.
One important thing is that it doesn't get us too far to talk in generalities. And maybe that's what you guys meant about there being no magic formula.
The only way to really get it right is to develop and define actual numerical requirements for what must be able to happen AT the power and ground pins of the loads, and then make that always be able to happen, or fail partially and from that define what the actual capabilities are (and why they can't be better).
That's why I started looking into decoupling requirements. How do you determine how much capacitance is needed (for example)? You either have to use the worst-case (i.e. best) capabilities of the active device, from the data sheet, or use the overall system's specs if they are not as strict. Then you have to consider everything upstream as one system or network, while remembering that the final result AT the pins of each device is the only place that really matters.
So maybe the magic formula is THERE.
One important thing is that it doesn't get us too far to talk in generalities. And maybe that's what you guys meant about there being no magic formula.
The only way to really get it right is to develop and define actual numerical requirements for what must be able to happen AT the power and ground pins of the loads, and then make that always be able to happen, or fail partially and from that define what the actual capabilities are (and why they can't be better).
That's why I started looking into decoupling requirements. How do you determine how much capacitance is needed (for example)? You either have to use the worst-case (i.e. best) capabilities of the active device, from the data sheet, or use the overall system's specs if they are not as strict. Then you have to consider everything upstream as one system or network, while remembering that the final result AT the pins of each device is the only place that really matters.
It creates a 'free' LC with the wire / trace, along with providing the very first bit of the charge to the demanding output. As the cap delivers charge, the big caps from the main supply start to provide current to both recharge the local caps and continue providing power to the output. It lessens ringing and thus removes sources for oscillation, especially noticable with very fast output stages that can draw big currents in an instant; it's what keeps it fast and stable
At least, that's my way of thinking.
exercise 1
Thank you for looking at it. We have to do 1/10 of 20Hz comes to 2Hz , Thus comes to 266uF, The available cap size being 270uF. Isnt it just cute
. Can you workout transformer specs using #98 ?
Waiting for exercise 2
sample capacitor
Here is a sample for preamp and mixer
MAL213816102E3 Vishay BC Components | 4219PHBK-ND | DigiKey
Here is a sample for preamp and mixer
MAL213816102E3 Vishay BC Components | 4219PHBK-ND | DigiKey
It creates a 'free' LC with the wire / trace, along with providing the very first bit of the charge to the demanding output. As the cap delivers charge, the big caps from the main supply start to provide current to both recharge the local caps and continue providing power to the output. It lessens ringing and thus removes sources for oscillation, especially noticable with very fast output stages that can draw big currents in an instant; it's what keeps it fast and stable
Maybe the C is free but any L would be more like a cost. Maybe it pays for the C.
Tom,
Your calculations are predicated on max voltage drop at the cap.
This is essentially a power consideration, as you naturally want max rail voltage for high amplitude signals.
What if the power angle differs from the subjective issue, in that the max power will not necessarily deliver the best sound at listening levels? This would considerably reduce the metric from 2000uF/amp......
In this I'm reminded of subjective bass response of amps. If you have an amp with a single output pair, bipolar or mosfet, and relatively high Zout, you can actually improve subjective bass response by interposing a small 0.15R resistor into the speaker output line. This promotes overshoot on the voice coil - that is, loss of VC control - but adds a subjectively stronger bass response. Hardly correct in engineering terms, but quite useful when sculpting bass presentation. Perhaps something similar is operating here?
The other aspect you have touched on is decoupling the amp reservoir cap from the power supply filter cap. I use 0.15R between each cap electrode; 0.3R in all. I have found this very greatly improves sonics presumably by greatly reducing speaker earth return currents from intermodulating with charge currents from the rectifier.
Of course, the rail voltage does sag rather more, yet the sound quality is superior. Should this not be the criteria we should address, and if so, why the hell is this happening?
My experience is empirical. I cannot plausibly explain why. But I suspect the issue is much more complex than baseline power calculations suggest.
That said, I applaud what you are doing. Adding some engineering rigour to this vexing issue has been a long time coming....
Cheers,
Hugh
Your calculations are predicated on max voltage drop at the cap.
This is essentially a power consideration, as you naturally want max rail voltage for high amplitude signals.
What if the power angle differs from the subjective issue, in that the max power will not necessarily deliver the best sound at listening levels? This would considerably reduce the metric from 2000uF/amp......
In this I'm reminded of subjective bass response of amps. If you have an amp with a single output pair, bipolar or mosfet, and relatively high Zout, you can actually improve subjective bass response by interposing a small 0.15R resistor into the speaker output line. This promotes overshoot on the voice coil - that is, loss of VC control - but adds a subjectively stronger bass response. Hardly correct in engineering terms, but quite useful when sculpting bass presentation. Perhaps something similar is operating here?
The other aspect you have touched on is decoupling the amp reservoir cap from the power supply filter cap. I use 0.15R between each cap electrode; 0.3R in all. I have found this very greatly improves sonics presumably by greatly reducing speaker earth return currents from intermodulating with charge currents from the rectifier.
Of course, the rail voltage does sag rather more, yet the sound quality is superior. Should this not be the criteria we should address, and if so, why the hell is this happening?
My experience is empirical. I cannot plausibly explain why. But I suspect the issue is much more complex than baseline power calculations suggest.
That said, I applaud what you are doing. Adding some engineering rigour to this vexing issue has been a long time coming....
Cheers,
Hugh
This here is also closely connected to the grounding scheme of the amplifier. and where the lowest point of impedance for the speaker return currents is formed. I normally have that right at the Center tap of the Transformer. But I have also had good results with a dual-bridge solution where the star point is formed between the rail-caps.
I used this in a Hiraga inspired amplifier some years ago, where I had a dual-bridge- then one large central reservoir followed by a R-C bank for each channel.
Hugh do you also insert a resistor in the gnd....??
I used this in a Hiraga inspired amplifier some years ago, where I had a dual-bridge- then one large central reservoir followed by a R-C bank for each channel.
Hugh do you also insert a resistor in the gnd....??
Member
Joined 2009
Paid Member
People have reported noticeable changes in sound between different power transformers. The power transformer's inductance and resistance are in the circuit whenever the rectifier diodes are conducting. The capacitors are in a circuit, where there are other elements and multiple time constants.
Hugh, you are always on the right path!, perhaps the overshoot behaviour of the voice coil you referred to is another example where the transient response / damping behaviour of the power supply-amplifier is the factor being sought and the value of the capacitor is just one ingredient.
Hugh, you are always on the right path!, perhaps the overshoot behaviour of the voice coil you referred to is another example where the transient response / damping behaviour of the power supply-amplifier is the factor being sought and the value of the capacitor is just one ingredient.
Last edited:
Wouldn't you want the star point at the output of the power supply, or at least after most of the smoothing caps? It seems like the star ground should be kept out of the loop formed by the transformer secondary, rectifier, smoothing caps, and the other end of the secondary or the center tap. Otherwise the voltages induced across the inductance of the smoothing caps' ground conductors by the charging current pulses would get into ALL of the grounds. If the star point is just past the last smoothing cap, then the caps' grounds and the center tap would just be another arm of the star, and none of the charging currents would ever pass through the star ground point, or anything else connected to it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.