I am seeing limited experience here with real loads and the capacitance needs of power amps. An important factor is 'ground return impedance' so that you keep your damping factor high where you need it.
Of course, PEAK current reserve. You know: 20A-60A for a short time.
Ripple is a distant third.
Of course, PEAK current reserve. You know: 20A-60A for a short time.
Ripple is a distant third.
Plenty Class A power, a whole lot more Class AB, and way over 1kW regulated.
Left 2 boards are the regulation stages for each rail, 16 Sanken 200W in MT200 package per rail.
Amp boards on the right, balanced setup, so that rail supply voltage can stay low to enable maximum output current level.
The key phrase being "to save on ...".
I refer to a different kind of beast, to regulators specifically made for quality, not savings. There are NO savings in my world, I will use a regulator and still precede it with lots of capacitance which might just as well be standalone.
Later on, I'll dig up the schematics of both the shunt and the main regulator and post them here, so you can make up your own mind.
Completely agreed.
But we do need to put some things in some reasonable proportion. The expectations from a say 100 WPC and a 300 WPC amp are not quite the same. Nor is their very design likely to be quite the same.
John, let me ask you straight - disregarding everything else at the moment, what would you consider to be sufficient capacitance per channel for what you might consider to be a decent design offering nominally 100 W/8 Ohms, yet expected to be load tolerant?
In uF, please.
Err, yes , my, not too much details now ..
Right, not too much details, only one bridge.
Cheap and easy. However, longer wires than in case of 4 separate diodes.
Diode in this looks useful and cheap . If it took 5A it would be useful . Who knows it might be 600V like the device ? Ultra fast / soft has great appeal .
Might be OK for PSU's also .
Well, it says 20A in the lower right corner ...
Thank you, John.
It may please you to learn that this is exactly what I installed in my Marantz Brothers amps (1152 DC, 170 DC) - frankly, I spat blood before it was done, a VERY tight fit, but I did it. Their supply lines are 51 and 53V respectively, the caps are not 15,000uF/56V like Marantz did, but a nice 22,000uF/63V, power ratings 76 and 85 WRMS respectively.
Of course, that's retrofitting, so you have to deal with various imposed constraints which are simply not there when you design for it straight away.
For my own amp, I was thinking 22,000uF in parallel with 10,000 uF, plus the housekeping little 'uns in parallel, ending with a 1 Ohm/2W resistor in series with 470 nF to the ground, which is intended to get rid of any residual inductance left over by the capacitor bank. The 470 nF value is tentative, it may be a step down or a step up in real life. I know you can measure it, but that particular one I traditionally do by ear.
Not meaning to minimize your experience, but I think there is a simple way of determining when is enough enough. You can stop adding caps once your amp start to sound like it is a bottomless pit of power under your regular conditions of use; it's what I truly treasure in an amp, this feeling that it has all the power in the world on tap, and what it normally does in your room is child's play, almost an insult to it overall capabilities.
This trait is where American amps are usually best at. Very, very few European amps manage this feeling, Electrocompaniet being one of them, and in Japanese amps, it has been decades it seems when I heard it last, and it was a Yamaha B4 amp (not the general "R" version, but an Asian version, which has double the capacitance of the "R" version). Late 70ies.
If at the last moment I should decide to go down the dual mono way and use full electronic regulation for all stages, I will still use the the same capacitance.
I said to a friend when he wanted a " new Idea " . Why don't you arrange the circuit with Fibonacci values where possible . Being that he made his money designing the fundamentals of computers I didn't have to say why as maths is his passion . He came up with a Fibonacci PSU . Approximately the caps go like this . 22000 , 13600 , 8400 , 5200 ,3200 , 2000 , 1200 , .................................................100 , 68 , 39 , 22 ,15 , 10 , 6.8 , 3.3 , 2.2 .
If a maker of caps wanted to do the correct values ( plus / minus the usual ) what harm would it do ? Fit as many of the sequence as you like . Looking at it, the values might be better than what we have even as practical values ?
Gary and I suspect that feedback systems are best analyzed with a Fibonacci sequence . That would be for statistics also . We mean any feedback system of which amplifiers is a distant cousin .
I was in a pub talking about this when a man came up and insisted we tell him the best mathematical curve to describe the stock market . We said if we knew would we be drinking in this pub ? This seemed to make him very unhappy like we were holding out on him . He wouldn't accept it so said we said Fibonacci . You would have thought all of his birthdays came at once , he was delighted . I suspect it was about the best we could do ? He did show us all the curves he knew of . I must say it looked exactly like the problems we discuss here . Apparently he was on a forum just like here and was trying to get somewhere .
The very gifted amplifier designers must sit back and say none of what is discussed goes anywhere because fundamental problems exist . However the courageous often make small contributions . Stupid things like what VAS cap to use for example ( quality , value , reliability ) . Not stupid in reality . If sending a space craft on a journey such things matter and yet cost pennies to get right .
Would I be right in thinking 22 000 uF is good for 60 A transient ? That's not a rating of the cap . It is a guess as to the current needed into 1 ohm on a big amp . My instinct is that it will be right . One has to say what sort of pulse if so . Bass guitar bottom D seems about right and a genuine 8 speakers in paralell . That's what I mean by 1 ohm . Do we have a Spice model for a Fender ?
If a maker of caps wanted to do the correct values ( plus / minus the usual ) what harm would it do ? Fit as many of the sequence as you like . Looking at it, the values might be better than what we have even as practical values ?
Gary and I suspect that feedback systems are best analyzed with a Fibonacci sequence . That would be for statistics also . We mean any feedback system of which amplifiers is a distant cousin .
I was in a pub talking about this when a man came up and insisted we tell him the best mathematical curve to describe the stock market . We said if we knew would we be drinking in this pub ? This seemed to make him very unhappy like we were holding out on him . He wouldn't accept it so said we said Fibonacci . You would have thought all of his birthdays came at once , he was delighted . I suspect it was about the best we could do ? He did show us all the curves he knew of . I must say it looked exactly like the problems we discuss here . Apparently he was on a forum just like here and was trying to get somewhere .
The very gifted amplifier designers must sit back and say none of what is discussed goes anywhere because fundamental problems exist . However the courageous often make small contributions . Stupid things like what VAS cap to use for example ( quality , value , reliability ) . Not stupid in reality . If sending a space craft on a journey such things matter and yet cost pennies to get right .
Would I be right in thinking 22 000 uF is good for 60 A transient ? That's not a rating of the cap . It is a guess as to the current needed into 1 ohm on a big amp . My instinct is that it will be right . One has to say what sort of pulse if so . Bass guitar bottom D seems about right and a genuine 8 speakers in paralell . That's what I mean by 1 ohm . Do we have a Spice model for a Fender ?
Last edited:
All this cycling back to sound quality. Well I replaced my e-mu box with an m-audio and it does not hang my system. Got the long firewire cable today, built the test box where I can mount the Seas reed woofer and both the metal and cloth version of the same tweeter so I can A/B them for my wife. Their specs are so close ( identical motor) I can use the same crossover, so it is really just the dome itself switching.
Maybe in the next couple of weeks I can determine for sure if the preference she has for the Rotel amp is because it does not excite the tweeter breakup as much. Just maybe, the entire problem is the tweeter all along. If by some chance they both sound OK, then that means I just needed to build better speakers all along. If the both still fail, then I am going to build a 3 way with a wide range for a mid guessing the problem is trying to get 2K out of a 1 inch dome. That's the plan anyway.
Maybe in the next couple of weeks I can determine for sure if the preference she has for the Rotel amp is because it does not excite the tweeter breakup as much. Just maybe, the entire problem is the tweeter all along. If by some chance they both sound OK, then that means I just needed to build better speakers all along. If the both still fail, then I am going to build a 3 way with a wide range for a mid guessing the problem is trying to get 2K out of a 1 inch dome. That's the plan anyway.
Could some type of Zoble help the tweeter ? Perhaps an LC one with resistor damper to the choke . It would change the Q of the problem ? I am seeing this as a mechanical problem with an elector-reactive solution . Analogy is how suspension of a car works ( or doesn't very often ) . Think oil , spring , gas in the oil . oil flow restriction hole , bump stop rubber . I know Derek Hugues ( Spendor ) said the Audax tweeters needed care if used that low . His Preludes worked so much better with those tweeters than other people managed . I wish I had asked him how he did it . I know he struggled with it . Interestingly he would have preferred to use Celestion HF1300 which only went to 15 kHz . Derek said a for him 15 kHz was enough . Falcon Acoustics in the UK know about these things .
I tried zobels, notch filters, LP filters, and even built a twin T. Best any can do is about 3 dB. Going through my DCX which brick walls at 20K and the Rotel 840, I can get about 6 dB. One has to understand not everyone is sensitive due to the very high frequencies. (27K), so I am for sure not condemning metal domes categorically. Of course, many soft domes do things just as ugly. Exactly what that does to the frequencies we can hear is TBD. I am working only with measurements as I am not sure I can hear the problem my wife does. I think so, but not clearly.
With some kind prodding, I am ordering some Fostex to do the midrange test sooner than later. Logically, 2K or 2.5 with an energetic Jazz band has just got to be too tough for a 1" driver. I am betting at least one of these two problems is the source of her pain.
My only experience with Celestion was the T-2000's in my DItton 44's. They were not the outstanding part of that speaker, and when one came unglued, I replaced them with Seas. Pretty good for the late 70's.
With some kind prodding, I am ordering some Fostex to do the midrange test sooner than later. Logically, 2K or 2.5 with an energetic Jazz band has just got to be too tough for a 1" driver. I am betting at least one of these two problems is the source of her pain.
My only experience with Celestion was the T-2000's in my DItton 44's. They were not the outstanding part of that speaker, and when one came unglued, I replaced them with Seas. Pretty good for the late 70's.
- Status
- Not open for further replies.