3886 power supply cap considerations

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

I'm about to build another multichannel 3886 amp, and thinking hard about how I want to design and build the power supply.

Right now I'm looking at filter caps. What type of caps should I use, what brands would you recommend, and how much capacitance per channel is A) just barely enough B) just right and C) probably overkill? I expect not to exceed the 35-40 watt rating into 8 ohms but I do want the amp to be generally overdesigned with good quality parts from reputable manufacturers so that I can have peace of mind.

I'm shooting for good quality but not exotic/expensive. My first guess was something on this page (1 pair per channel is my estimation of "more than enough"):

Nichicon Radial 10000 uF 50 VDC Aluminum Electrolytic Capacitors - Leaded | Mouser

... After seeing the UKW, UFW, etc. series this is where my knowledge hits a gray zone. I'm not sure what the difference is between these series and whether I should choose one over the other, or am I looking in the wrong place entirely?

Thanks for your feedback
 
Member
Joined 2009
Paid Member
If you haven't already, I suggest reading through the Supply Decoupling section of my Taming the LM3886 article series. Getting the decoupling right is critical if you want the amp to deliver the advertised performance.

For the supply electrolytic caps, I'd go with some 105 ºC rated ones if you can find them. Or 85 ºC ones with a lifespan over 5000 hours. As the previous poster noted 10000 uF is plenty. I use 22000 uF per rail in my 4-channel amp (for bi-amping).

Tom
 
Hello experts,

I'm about to build another multichannel 3886 amp, and thinking hard about how I want to design and build the power supply.

Right now I'm looking at filter caps. What type of caps should I use, what brands would you recommend, and how much capacitance per channel is A) just barely enough B) just right and C) probably overkill? I expect not to exceed the 35-40 watt rating into 8 ohms but I do want the amp to be generally overdesigned with good quality parts from reputable manufacturers so that I can have peace of mind.

I'm shooting for good quality but not exotic/expensive. My first guess was something on this page (1 pair per channel is my estimation of "more than enough"):

Nichicon Radial 10000 uF 50 VDC Aluminum Electrolytic Capacitors - Leaded | Mouser

... After seeing the UKW, UFW, etc. series this is where my knowledge hits a gray zone. I'm not sure what the difference is between these series and whether I should choose one over the other, or am I looking in the wrong place entirely?

Thanks for your feedback
There are two issues here.
One is decoupling.
The other is storage.
Are you intending a conventional transformer, a linear regulated supply or a switchmode power supply?
With a conventional transformer, you need storage to reduce line frequency ripple to whatever level you consider acceptable. Also in this case, think about the ripple current rating of the caps.
 
Hello experts,

I'm about to build another multichannel 3886 amp, and thinking hard about how I want to design and build the power supply...............
All your questions should have been resolved when you built your first amplifier.
What worked for that build?
Repeat for this build.
What did not work? What needed changing to get proper operation?
Repeat the same solutions this time.
 
My first amp came out from the (toroid) transformer, through the rectifier, into some caps, then into the amp boards. I did not consider things like regulation, RC filtering, snubbing, or any other complications as I just wanted to see if I could get something together that worked and didn't sound terrible. Something to play with and learn from.

Now I want to take it seriously and clean up and improve the shortcomings of something so overly simple.

What I've read so far is that 85C caps with 2000 hrs might not be the quality level I'm shooting for. 10,000 uF per channel, 105C and 5000 hr caps would get me into the "i could spend more but I won't hear the difference" territory. I'll also be using the power supply boards for sale on this website.

Next topic I'm interested in is snubbing, which is as I understand it putting small caps after the rectifier which will absorb and smooth out noise coming after rectification if the right values are calculated. Is that about right?

Thanks for the tips Tom, that link is definitely of interest to me and I will be studying it during my free time. I will say however that if I understand the design of your modulus 86, I need not worry much about power supply noise so long as the basics are covered.
 
..................
Next topic I'm interested in is snubbing, which is as I understand it putting small caps after the rectifier which will absorb and smooth out noise coming after rectification if the right values are calculated. Is that about right?..........
The transformer inductance rings with the rectifier capacitance when the current suddenly stops or starts or step changes.

It's the transformer interaction with the rectifier that needs snubbing.
A snubber is a resistance that absorbs energy. It is NOT a capacitor.
You need to place a snubbing resistor across the transformer winding so that the ringing energy is absorbed.
A resistor alone will draw too much current, so instead we place a capacitor in series with the resistor to attenuate all the low frequency voltage across the resistor.
The snubber is an R+C across the transformer winding.
 
Thanks Andrew. I've only been looking into that topic briefly and haven't read much about it yet.

So I have a question... when you said "...rings... when current suddenly starts or stops..." How does the rectifier cause this problem? Is it because the slope of the current at transition from one polarity to the other creates a sharp angle (relative to 0V/gnd) and therefore a lot of high frequency harmonics at that point in time? In my mind I'm looking at what I imagine to be a nice smooth sinewave on an oscilloscope, except it's chopped off by a straight line by the diode. Then I imagine looking at it as a spectrum and I would see that sharp point showing up as peaks where we don't want them. And these frequencies reflecting around between the various LCRs of the power supply components need to be damped. Correct?

I'm still learning...
 
Last edited:

PRR

Member
Joined 2003
Paid Member
Get an old-fashioned spark-coil from a car. Run 12V into it for a second. Now break the circuit.

The place you put the 12V will KICK up to 400V then ring -390V +380V -370V +360v.... on the internal inductance and coil+stray capacitance. (The top-cap will kick 100 times more voltage to give the 40KV needed to spark a dense mix.)

With mains AC and diodes, the same turn-off and kick happen 120 times a second (though typically no kick-UP). The ring frequency is much higher than 60/120Hz, but there is always both L and C in everything so there is a ring. It fades because everything also has R, but the basic function of a power supply leads to small R, so it may ring a while. There's methods to guess, calculate, or measure suitable added parts to make the ring damping near "critical", hardly-any.
 
Just as a no-Math but experience based suggestion:
1000uF per channel: indequate
2200uF per channel: minimum acceptable
4700uF per channel: *good*
10000uF per channel: "audiophile"/overkill ... which does not mean it´s wrong, quite the contrary.

Each step halves ripple, which is a good thing.

No ripple will appear at the output while clean but if playing loud, a few peaks will clip now and then, simply because of Music performance dynamics, and whan that happens, ripple will appear in full force mixed with your audio signal, even if for milliseconds.

If your amplifier has Led peak indicators, you´ll notice they briefly blink now and then; it´s in those instants when ripple reaches the speaker.

Besides above suggestions about decoupling, short thick ground paths, etc.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.