Electronics > Beginners
DIY Power supply for amplifier testing and more, lowest ripple?
FriedMule:
I want to build my own power supply and have, of cause, already consumed Dave's videos, several times, but I am still uncertain on how to get started and how to reach as low a ripple as possible. I have found a ripple filter circuit that goes down to 20nV but only if you allready have below 300nV, I was not aware that it was even possible to go that low in ripple.
First some simple calculation that shows how I have arrived at my target power:
I was first thinking of building a power supply, capable of driving a 50W 8 Ohm = 20V (400W 2 Ohm) power amplifier but then thought that it is fairly limited and maybe not enough for a later project, so I ended up thinking that 200W 8 Ohm power amplifier will be max I ever will need: 200W 8 Ohm = 40V (1600W 2 Ohm).
1600W / 40V = 20A but that do not leave any headroom at all, so why not 50V 25Amp = 1250VA and call it a day? :-)
Now is the question if linear or SMPS is the way to go, some keep saying that SMPS is a wary hard road, wary noisy and something most wont touch! While Dave says that they are almost the same, a few extra components, but nothing scary.
On the other hand, do I think that 95% effective transformers is out there and if we are only talking about 98% effective SMPS vs 95% effective linear, then, let's go linear. :-)
My problem is the filtering, Dave did make a fantastic video about the Cap-multiplier, but I don't think that it can handle 50V at 25Amp, so where do I please start?
Calvin:
Hi,
the chase for lowest ripple appears to me like the chase for the white rabbit. :o
The Q is ... where is the point of diminishing returns?
That vastly depends on the PSRR of the chosen amplifier.
If its a typical power OPAmp design it may tolerate several volts of ripple on its supply rails without it becoming audible. A Singleended no nfb design might have almost no PSRR at all, placing most of the supply rejection burden on the power supply itself.
Since the ripple amplitude increases with the signal amplitude, hence the amount of current the load draws, it may surpass the audibility treshold at levels where its already fully masked by the music signal.
So, take care that You don't design chasing numbers on a parameter which might have low priority.
regards
Calvin
FriedMule:
--- Quote from: Calvin on December 20, 2019, 05:21:53 am ---Hi,
the chase for lowest ripple appears to me like the chase for the white rabbit. :o
The Q is ... where is the point of diminishing returns?
That vastly depends on the PSRR of the chosen amplifier.
If its a typical power OPAmp design it may tolerate several volts of ripple on its supply rails without it becoming audible. A Singleended no nfb design might have almost no PSRR at all, placing most of the supply rejection burden on the power supply itself.
Since the ripple amplitude increases with the signal amplitude, hence the amount of current the load draws, it may surpass the audibility treshold at levels where its already fully masked by the music signal.
So, take care that You don't design chasing numbers on a parameter which might have low priority.
regards
Calvin
--- End quote ---
Yes I can imagine that one can fill a whole room with components, just to chase the last pico-volt, but I'd like to reach <10mV pk/pk with full load, why... because the psu will also be used for other things then just amplifiers. A steep slew rate wold also be nice:-)
exe:
Usually loads that consume so much power don't care much about ripple.
I'd suggest have two power supplies: one linear for "precision" work (For me 2 channel 15V 1A was enough for everything except very few cases), at least two channels, three of four is better. And a another one ("small welder") when you need a lot of juice, like charging a car battery.
Precise current limit is a good plus when testing semiconductors, etc.
David Hess:
Part of a good audio amplifier design is high power supply rejection so ripple becomes largely irrelevant.
Navigation
[0] Message Index
[#] Next page
Go to full version