Power Supply Question

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Speaking from the perspective of good design, the amount of capacitance required is not dependent on the transformer VA; rather, it is dependent on load current, which in turn, is dependent on amplifier power.

You would want to use enough capacitance, that at peak current draw, the ripple voltage is low enough to be suppressed by your amplifier's Power Supply Ripple Rejection (PSRR) capability.

General rule of thumb is something like 1,000uF for every 10W of amplifier power. However, for DIY, it means nothing. Some folks like the GC folks here prefer very little. Some folks like the Zen/Aleph folks prefer massive banks of capacitors.

If you ask a National applications engineer, he'd say the following:
Assuming a single LM3886 with a dual 24V, 250VA transformer, 4700uF would be amply sufficient per rail. Ensure cap is rated to handle peak voltage and ripple current.
 
the last few days i've been playing with capacitance in the psu and it does make a difference, albeit not that big and -i guess- very much speaker/impedance dependent.

I've tried adding 1x 4700uF per rail to my (physically separated from the amp) psu and 2x 4700uF per rail. i'm using 2x 15V secondaries and 1000uF directly on the chip. On 8 ohm speakers the difference was very small, only just audible. On 4ohm speakers the difference was muchg better audible.

Adding 1x 4700uF to the rail wasn;t audible on the 8ohm speakers. On the 4ohm speakers it made the lower bass more detailed and stable.

2x 4700uF improved the bass maybe marginally on the 8ohm speakers. I couldn't hear the difference but a friend of mine (said he) could.

Adding 2x 4700uF per rail made the 4ohm (pretty old) KEF's bass more stable and more in control. Before, it tended to 'wobble' a bit, especially when playing bass-rich lounge music at high volumes.

Today i think that the soundstage also seems more stable (less 'moving instruments'), you could really point your finger at the hihats and bass.

i think i'll stick to my current setup for now: 2x4700uF (Philips) per rail at the PSU, bypassed with small WIMA's i had laying around, and 1000uF back gate standard per rail directly at the chip.

On to the definitive version of my buffered pre!

my EUR 0.02

edit: oh yeah, i'm using an 3875, not an 3886
 
National seems to recommend about 2,500ish uf per rail per chip for lm3886 as per AN-1192.pdf.
That could be a bit high as per the 1000Uf general recommendation for LM3875. I've got a pretty large 3886 project in the works (4 IC per channel) and I've been thinking of using about 1500 to 2000 in that (depending on what can find for good cheap capacitors)

Of course, very few will agree with this, since it seems that one thing people in here can't agree on is amount of capacitance an amp needs.:rolleyes:
 
Just like the responses before me, I too am beginning a large LM3886 project. With one chip per channel accross 5 channels and a bridge configuration for the sub channel, I've found that 20,000uF per rail is sufficient IF you don't push it too hard. As for your question, I had done some initial testing with single chip circuits prior to building and found that National's recommendation of 4700uF/rail to be quite sufficient.

I must say, I do feel a bit like I'm cheating as I go through this project. National has made it almost too easy to do project work with these integrated amplifier chips. Anyone hear me?

~AAW
 
I have not followed this thread, but I might offer some advice. Use as much power supply capacitance as is reasonable for you. It is not because of hum or noise, but the impedance of the return path of the loudspeaker to the power supply. We have found that 20,000 uF and more is necessary, per channel, to make the bass sound right.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.