Power Supply filter capacitors

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I am working on my first gainclone stereo amp based on the LM3886. It looks like there are two camps on power supply filter caps. One is relatively low values such as 1,000uF to maintain the "magic" of the amp. The other is traditional large say 10,000uF filter caps with a snubberized circuit such as the Carlos Filipe design. What are the considerations and tradeoffs?

Thansk,

George
 
Why would you use small smoothing capacitors?

I have read more than one thread that says to use small capcitor values or you destroy the "magic" of the amp.

One example from Decibel Dungeon/sticky above:

"How much capacitance do I use in the power supply? This subject is hotly debated topic amongst chip amp builders. If you are new to chip amp building I suggest the following arrangement. About 10-100 uF after the bridge rectifier, and 1000 uF on each (power supply) pin of the chip amp. If your speakers are a difficult load, you will need much more capacitance than this. But if you put say 10,000 uF after the rectifier, you will almost certainly notice that the chip amp loses some of its magic. It's another hi-fi compromise, trading some good mid-range, for better bass. There are a couple of ways around the compromise though. Either go for the snubberised supply as shown here, or a regulated supply as described here."

George
 
Last edited:
I have used a separate 1000uf after the larger caps, usually closer to the load.
Best of both worlds, and I can’t believe it isn’t common to do this on diy amps of any variety.
Why does it have to be one or the other?
I learned this from a guy here on another thread who has since passed on, and confirmed it with my own tinkering.

I always used a much larger case than typical, and raise the circuit board on 1”+ standoffs so additional parts can be installed on the underside of the board. Surface mount methods don’t help this approach.
 
I think what the author is getting at there is smoothing and decoupling, two different things which he is confusing with the help of the word magic

I understand decoupling vs smoothing as well as ripple etc. No complete circuit diagram is provided and neither word is used in the tutorial. Also 10,000uF is discussed as an alternative to 1,000uF so I am going to have to do further research, this just is not clear enough.
 
I have used a separate 1000uf after the larger caps, usually closer to the load.
Best of both worlds, and I can’t believe it isn’t common to do this on diy amps of any variety.
Why does it have to be one or the other?
I learned this from a guy here on another thread who has since passed on, and confirmed it with my own tinkering.

I always used a much larger case than typical, and raise the circuit board on 1”+ standoffs so additional parts can be installed on the underside of the board. Surface mount methods don’t help this approach.
It is common practice to use decoupling caps close to the output devices, often in the region of 100uF and 100nF
 
The DD suggestion is quite clear to me:
I suggest the following arrangement. About 10-100 uF after the bridge rectifier, and 1000 uF on each (power supply) pin of the chip amp.
Minimal filtering/storage just at bridge terminals, 2 to 8 inches away from the actual chipamp and 1000uF (which I guess way back then was maximum possible value to allow such connection) straight at Chip pins, no shorter path possible whatsoever.
Brushing away any significant wiring resistance/inductance.
I bet you can nowadays use 2200uF and even 4700uF there, just hanging from their legs and with some handy drop of silicone sealant so body does not move around and fatigue cut legs..

If your speakers are a difficult load, you will need much more capacitance than this. But if you put say 10,000 uF after the rectifier, you will almost certainly notice that the chip amp loses some of its magic.
Chip amps in general *are* quite sensitive to wiring path and length, so doing things the opposite way, large filter caps away and puny "HF/RF only" local decoupling may cause audible problems ... but you can have your cake and eat it; 1000 uF to 2200uF local straight-on-chip-pads plus whatever you feel comfortable with just by the bridge.
Best of both Worlds.

I wouldn´t overthink it.

As of snubberizing, I does not convince me very much because it´s usually done far away from the chipamp, so negating its possible usefulness.

In any case, IF I considered it really useful, I´d connect the snubber straight from chip power pins to chip ground.
 
Last edited:
That's it indeed. It was a trend a few years back, inspired by some audiophile designs.

Thankfully, the trend has shifted back towards more reasonable designs. Have a look at the neurochrome website. Tomchr has a good serie of articles on the lm3886, including power supply design.
 
The big snag with using a small cap after the rectifier and a big cap by the chip is that it significantly increases the area of the charging pulse loop, and potentially makes it harder to get a clean ground reference. Apart from this, it is slightly bonkers advice to use a small cap where science says to use a big cap (PSU reservoir) and to use a big cap where sciences says to ue a smaller cap (local rail decoupling). However, in audio there is no idea which is too daft for someone somewhere to promote it.
 
I assume since there is not a single response in support of less than 1,500uF total capacitance (all functions) between the bridge and the chip amp power pins that this approach is no longer considered for new designs?

I am going to assume that 10,000uF for smoothing is standard and then you decide on decoupling/snubberizing/etc. In this area I will keep reading.

George
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.