Hi DIYers , looong question already discussed BUT many modern / recent amplifiers are using nowadays more and more capacitance for the output rails (symetric PSU). 30 years ago, a twin 8000uF was huge, now 20000 is normal, 40000 is considered as high quality, 80000 is "High end" 😀 Where did you stop stacking caps ? 😉
No more caps when the peak-to-trough ripple on the supply is less than X% of the DC voltage. Measured when the amp is delivering 75% of its rated output power to an 8 ohm load.
For me, X=2% seems like an acceptably conservative choice. Others will have different opinions.
_
For me, X=2% seems like an acceptably conservative choice. Others will have different opinions.
_
As long as they don´t trigger breakers feeding your home and you have big enough pockets, be my guest.
Capacitance for low ripple for ac plus the max integral of peak constant load.
Parallel caps for lower esr and heating, lower noise plus volume discount.
In the end your psu is limiting from the ac waveform.. even with 3 phase..
Parallel caps for lower esr and heating, lower noise plus volume discount.
In the end your psu is limiting from the ac waveform.. even with 3 phase..
Still professional amps (Yamaha) use around 2000uF/Ampere in their A/B amps, the same as in old times.
This seems to be a good compromise between peak and continuous power and cost.
1975 I build an amp with 3055/2955 with 60W and used 4x 4700uF.
This seems to be a good compromise between peak and continuous power and cost.
1975 I build an amp with 3055/2955 with 60W and used 4x 4700uF.
Last edited:
Twice the caps, double the peak charging current. At some point you start to torture everything, including the transformer, and to radiate double mains frequency and harmonics, pulses into your amp. Everything interacts.
80,000uF is worse for everything than 40,000uF
Jan
80,000uF is worse for everything than 40,000uF
Jan
mmmh, not sure ... and 2x80000 ? 😀 BRYSTON 28B SST3 CUBE PRO Amplificateur de reference 1000 watts @ 8 Oh😉
or even 6x 100000 in late Dan d'Agostino creation ?
or even 6x 100000 in late Dan d'Agostino creation ?
Last edited:
Doubling the caps does not double the peak charge current, which is limited by the transformer, bridge and interconnecting wire impedances. The rms current stays exactly the same. In fact if the rms charge currents changed it would break Kirchhoff's circuit laws.
In the 1980s did an experiment with John Dunlavy at Duntech where we took a very good amplifier connected to a pair of his flagship Sovereigns, and then doubled the supply capacitors, and then doubled again, and again until at 160,000µF per rail we ran out of cans.
Each doubling gave a distinct and easily noticeable improvement in sound quality. We were both absolutely gobsmacked at how there didn't seem to be any sensible limit to the improvement in sound with more supply capacitance. John Dunlavy had predicted that there wouldn't be much if any improvement over 40,000µF!
In the 1980s did an experiment with John Dunlavy at Duntech where we took a very good amplifier connected to a pair of his flagship Sovereigns, and then doubled the supply capacitors, and then doubled again, and again until at 160,000µF per rail we ran out of cans.
Each doubling gave a distinct and easily noticeable improvement in sound quality. We were both absolutely gobsmacked at how there didn't seem to be any sensible limit to the improvement in sound with more supply capacitance. John Dunlavy had predicted that there wouldn't be much if any improvement over 40,000µF!
😉
you're absolutely right, there is an evident improvement every doubling ... crazy 🙂
nothing in relation with peak current, peak power (even improved too) , it seems that noise of the zero loudspeaker's line is deeeeeply reduced as ESR is lowered.
i'd build an extent cabinet for each of my Bose 1800 with 200000uF inside ...
you're absolutely right, there is an evident improvement every doubling ... crazy 🙂nothing in relation with peak current, peak power (even improved too) , it seems that noise of the zero loudspeaker's line is deeeeeply reduced as ESR is lowered.
i'd build an extent cabinet for each of my Bose 1800 with 200000uF inside ...
Last edited:
even if the instantaneous peak current doesn't increase when you double the capacitance, the duration of inrush certainly does. You'll probably pop a fuse or your rectifier if you put too much capacitance. If you use a soft start circuit you have to beef up your soft start resistor.
As for how much you 'need', it depends on hum/distortion goals and power supply rejection ratio (PSRR) of your amplifier. A very good amplifier design can get away with much more ripple on the powersupplies than a poor one with non-existent PSRR.
As for how much you 'need', it depends on hum/distortion goals and power supply rejection ratio (PSRR) of your amplifier. A very good amplifier design can get away with much more ripple on the powersupplies than a poor one with non-existent PSRR.
2000uF/Ampere is more than enough.
Instead staking caps, I would look for high quality caps and synchronous rectification.
Regards,
Tibi
Last edited by a moderator:
PS, I pushed rectifiers to 250A continuous , and fuse to 15A slow-blow ... and noticed PSRR improvment too ... Yes i hear a "blooonk" in transformer on START, not much more than before, not worst of my spot-welder 😉
@Tibi : yes theory tells 2000 is more than enough until i did tests wth all laying capacitors i found in basement ..
@Tibi : yes theory tells 2000 is more than enough until i did tests wth all laying capacitors i found in basement ..
Last edited:
yes 🙄 but on scope, residual 100Hz is reduced , and on ear , amp is more silent 😛 Or would you say Marklev, D'agostino, Krell, Bryston, Naim and many others didn't understand anything ? 🙂
Last edited:
The mean current (at the rectifier outputs) stays exactly the same (Required by Kirchhoff), the RMS current (And thus heating) increases as the conduction angle drops.
I would tend to take differences in adding truly stupid amounts of cap as being an indication of a PSRR problem with the circuit design, better to fix that then band aid it by throwing piles of expensive and bulky caps at the problem.
i said, result is an extended infrabass response (lowered reactance ) + "cleared" signal ie. less grainy, more dynamic, more natural ... 🙂