Just to echo what abraxalito said ... . The rails are badly modulated by the output stage current draw, unless major efforts are made to to counter this. So then you need regulation everywhere else, or very well engineered PSRR characteristics ...
It really is, a 'how long is a piece of string' question ...
It really is, a 'how long is a piece of string' question ...
1% ripple on rect-C type supply is impractically low, inefficient use of transformer winding, gives too high peak rectifier currents in most applications
you can add passive filtering, active regulation or just build in the psrr needed
sub regulating/filtering front ends, cascoding in the amp itself usually easily makes the psrr numbers as Self writes - can't say about mysterious audio guru calls for extreme regulation as a cure-all
you can add passive filtering, active regulation or just build in the psrr needed
sub regulating/filtering front ends, cascoding in the amp itself usually easily makes the psrr numbers as Self writes - can't say about mysterious audio guru calls for extreme regulation as a cure-all
One of the big problems with conventional supplies is that the charge period, as a percentage of the mains waveform cycle, through the rectifier is limited, unless there is a degree of sag. The bigger the sag, the longer the period of charging of the smoothing caps, more energy in. So, in every situation a balance point is reached for a constant power out, there will a corresponding voltage sag - and there is nothing you do about this - except having ginormous transformers and caps -- Krell, etc ...
The key IMO: what is the actual sensitivity of the amplifier to the power supply? Of course, as as function of frequency. Simulations and measurements!
So many circuits can be rendered much less sensitive to power supply fluctuations via a few circuit "tricks". Thereafter, if one wishes to do so, by all means do more filtering/regulation. There is an undeniable hard limit as you approach clipping that can only be alleviated by heroic measures --- but more appropriately, turn the level down! You really don't need those couple of decibels, although high crest factor material may necessitate limiting average power to much lower levels than might be expected. Beware of asymmetries!
So many circuits can be rendered much less sensitive to power supply fluctuations via a few circuit "tricks". Thereafter, if one wishes to do so, by all means do more filtering/regulation. There is an undeniable hard limit as you approach clipping that can only be alleviated by heroic measures --- but more appropriately, turn the level down! You really don't need those couple of decibels, although high crest factor material may necessitate limiting average power to much lower levels than might be expected. Beware of asymmetries!
Of course there is - use a switcher. No-brainer really.
<edit> Oh I see you excluded switchers right at the start. As you were
I can't speak for all Krells, but the one I have seen the schematics of, from the late 90ies, some 250W per side amp if memory serves, actually had smallish filter caps, if memory serves, like 3*6,800uF per each PSU line, BUT it used one hell of a voltage/current regulator.
BTW, its input stage and VAS per side used 118 very common, off-the-shelf Motorola MPS and MPSA devices, probably costing less than $10 per side ...
KISS is for surfers.
None arm waving question ....
What is considered acceptable for ripple at full power, .55 Vripple is the target.. ..
Regards..
And at 10 watts ?
Anyone wanting to make a modest amplifier sound better should upgrade the PSU . In theory they are wasting their time and money . In reality they are not . The simple reason I suspect is that the measurements mask the ripple . When building a budget amplifier not to have hum with no music playing is as good as it gets . With chip amps it might even be close to state of the art distortion . Theoretically as long as real life ripple current requirements are met there is no reason to go beyond 2 x 10 000 uF in most cases . Reality is that the majority wouldn't choose that . There is a strange alternative . Even smaller capacitors are possible if the majority of the amplifier is fed from a clean supply . This could be right up to the outputs . Going slightly against the grain the VAS supply being lower than the dumpers is best . This allows the dumpers to become regulators in their own right . As long as the ripple bottom isn't reached the amplifier should remain tight sounding . Ripple ratings might just be possible with 4700 uF . The big deal here is the 4700 uF especially if one notch of voltage above requirements ( 100 V running 80V ) might be a better device at HF . That is if it's role was swapped to being an output capacitor it would measure as a nicer component . Personally I see no difference in a Kirchhoff arrangement where a capacitor is the output or a PSU cap. The ultimate thing is the clean amplifier can at the flick of a switch become a dirty amp . Raise the VAS by 15 V +/- and get the best measured result for power output . Still less dirty a than typical amp where no separation is attempted ( filthy amps ) . My guess is not many will do this as capacitors fill the case , like it has 8 cylinders firing . In valve designs the CV values are low . One guy I read tried to convince us that the CV value was a square ratio . That is as voltage climbs the value of a capacitor reduces by a square . I didn't give it any thought and assumed it to be wrong . Valve designs force one to do it right . I was very upset I only got - 85 dB reference 1 watt from my SE valve amp . I learned that is not bad . DC heaters made - 2 db difference on a good day and the noise type was no better at the more audible 100/120 Hz . The CV value of the amp 170 000 at 500 V ( 4700 @ 80 V = 376000 ) . The PSU has a FET capacitance multiplier ( 50 uF ) to get the ripple required . As it is running 100 mA at 8 V loss it almost needs no heat sink .
The punch line is . Forgetting looks did we throw 100 000 uF at a problem to allow the VAS 8 mA of better power ? Down side is our ground point is less well defined ? My big amp I hear has 2 x 47 000 63 V Mundorf per channel ( it has been on holiday for a year or so in Cologne ) . It would be 100 watts 8R and will drive 2 ohms reasonably well .
Last edited:
Excellent point . In PA amps it seldom does more than surf . My friend John uses SM was he doesn't want the hassle of going through certification . How does it sound John ? No comment , it is a non question as he doesn't have the money nor inclination to do it differently . He thinks a class H amp with switch mode is best of breed for PA . Twice the size of class D . Still acceptable weight ( less than 40kg ) . D is becoming popular as it is nicer to handle at knock down .
I guess Carver was right about an energy storage PSU . I fitted one to my dads car about 1978 for the sparks . It didn't make a scrap of difference to consumption . Kept the contacts like new . Morris 1300 countryman . The Mini factory should clone it as the Japanese loved the original . Little RR with Citroen clone suspension . Here is a famous cheaper version . BTW they never gave trouble in real life , rust got them after < 12 years usually . The Lancia Integrali I went around in was sort of similar , I bet no one ever said that before ? Austin Cambridge and Peugeot 404 are clones and that I think is the origin . I suspect both had the same designer although separated by time . Sorry car talk , must resist .
Basil Fawlty beats his car - Fawlty Towers - BBC - YouTube
Hmmm ... the second sentence in the Wikipedia article about such, Engineering - Wikipedia, the free encyclopedia, is:
If you believe the above should be exorcised from the article, feel free to do so ...
- Status
- Not open for further replies.