advertisement


Rectifier/smoothing PCB for power amps

H

hacker

Delicious and nutritious
I've started a new thread for the second revision of the generic PCB for rectification/smoothing (see old thread here). I'd like to keep the two threads separate, one for the low-power version, one for the high-power.

Without further ado, here's the design as it stands after much tweaking. Red tracks are the top copper, blue tracks are the bottom copper:

Edit: I've removed the pictures of the original board, the actual production design can be found at the bottom of this post.

  • The PCB is exactly 120mm x 85mm. This will apparently allow it to fit in NAP140/NAP110 power amps, although I've not confirmed this. Another PFMer (can't remember who) kindly made the measurements.Edit: this was confirmed and there are pictures somewhere on PFM of folks who've done this.

  • The diodes are TO-220 types, however axials would could be used if stood on end.

  • Headsinking of the diodes is made possible by aligning the rear of the TO-220 devices with edge of the board. A long strip of aluminium or copper would do the trick nicely, and I'm sure other PFMers will come up with cool alternative ideas. If another run of these boards is done I'll adjust the diode spacing slightly to avoid the possibility of them touching. This was one of the design defects in the first run.

  • The arrangement of the components and the silkscreen are all aligned under the assumption that most users of this board will require +/- power rails, not 2 + power rails. The board is perfectly capable of being used in configurations other than +/-, but the silkscreen won't match.

  • The PCB allows for 6 capacitors of up to 35mm diameter. It's designed for snap-in capacitors with 10mm lead pitch, such as this 5600uF/100V: http://www.mouser.com/Search/Refine.aspx?N=11669746&Keyword=5600uf+smh&FS=True Edit: The board will take 6800uF/63V Kendeils

  • There are spaces for 3 0V connectors at the noisy end of the board and 4 0V connectors at the other end. There are 2 output connectors on each power rail. Edit: there are a lot more 0V connections now!

  • The inductor spaces have been placed to allow maximum room for axial parts standing on end. The pads have been enlarged and the through-holes are 1.67mm diameter to allow for very thick wire to be used in hand-made air core inductors. I couldn't use hand-wound inductors on the previous PCBs because my wire was too chunky to fit in the holes.

  • The tracks are really, really thick and will be manufactured with 2oz copper, although I'd like to see how much extra it'd cost for 4oz.
All comments, suggestions, etc welcome. I know this isn't as flexible as the previous board, but it's far more suited to high-power applications and is constrained by the desire for a board that will fit in small Naim amps.

I don't have prices yet, but they're likely to be about the same as last time at around $5 each (about 3 quid), depending on quantities. If we get over 200 orders then things change for the better :)

Cheers,
Carl

The most recent design will always be provided here. It'll save trawling through the pages looking for it!

power-amp-rectifier10.png
 
Carl

Don't go for 4oz. 2oz will be fine.

Also the AC input tracking looks well close the 0V. Could the diodes be moved outwards and the gaps widened between AC and ground. By adjusting the angle of the tracking from the ground you could also open up the gaps.

Also by moving the output connections outwards to the board long edge (mounting holes inwards up the board, remove the lands and make non plated through) you could move all of the caps down and reduce the tight tracking at the input end.

Otherwise I'll have 4 !
 
I think you want to provide multiple independent return paths to the star earth point.

It is hard to generalise, but typically you will need at least:
1) Transformer CT.
2) Loudspeaker return -
3,4) 2 x Output stage bypass return(s)
5) Input signal earth
6) Feedback earth

Many amp circuits may have another couple of low current connections to earth, perhaps from internal bypasses.

For many of these, the actual resistance in the path matters less than keeping it clean, hence isolated from other current paths.

Arguably, the output voltage rails should also be split up in a star manner from the -ve terminals of C6 and positive of C3, so that noisy current gives minimum modulation of the rail for the other stages. This is typically a secondary effect; in many circuits it is 40-60dB less important than what happens in the earth paths.
 
#


# There are spaces for 3 0V connectors at the noisy end of the board and 4 0V connectors at the other end. There are 2 output connectors on each power rail.
Click to expand...

Is the 'other' end supposed to be quiet as opposed to the noisy end? I ask because all the 0v connections are going back to the noisy end due to the star arrangement at the AC side 0v point.

Otherwise looking really good, I might be getting a few of these :)
 
PigletsDad said:
I think you want to provide multiple independent return paths to the star earth point.

It is hard to generalise, but typically you will need at least:
1) Transformer CT.
2) Loudspeaker return -
3,4) 2 x Output stage bypass return(s)
5) Input signal earth
6) Feedback earth

Many amp circuits may have another couple of low current connections to earth, perhaps from internal bypasses.

For many of these, the actual resistance in the path matters less than keeping it clean, hence isolated from other current paths.

Arguably, the output voltage rails should also be split up in a star manner from the -ve terminals of C6 and positive of C3, so that noisy current gives minimum modulation of the rail for the other stages. This is typically a secondary effect; in many circuits it is 40-60dB less important than what happens in the earth paths.
Click to expand...

Excellent points. I've tweaked the design to encompass your ideas, specifically:

- 4 connectors at the star: transformer 0V, loudspeaker 0V, 2x output transistor decoupling connections.
- 4 connectors at the other end, independently linked to the star: signal 0V, zobel, and 2x front-end decoupling returns.

If the zobel wire can be crammed into the same connector as, say, the loudspeaker return connector, then that would free up space for a dedicated feedback return.

I also split the +/- outputs and ran the LED 0Vs back to the star (diminishing returns kick in at this point, but it costs us nothing).

It looks like this:

power-amp-rectifier6.png


power-amp-rectifier5.png


You can see the connectors will be very snug at the output end, but they'll fit - just.

Carl
 
Another point, off Neil Mcbride's site:

"The true 0 V point for the amplifier is the link between the smoothing caps. It is tempting to sit the 2 caps (assuming you have just a pair of caps) side by side, and run a 'bar' between the terminals, and hang the 0 V wires from this. This is how Naim do it. However Doug Self points out that there are actually large currents running back and forth along this connection, and it is better to connect an additional short bar to make a 'T' shape, and connect the 0 V wires to that (ie, the 'sticky out' bit). "

Now, this is more challenging to get right, but involves segregating the dirty and clean earth paths, in the same way that, for example, the Starfish pre-amp board does.

More or less, we want all the dirty currents to flow into one start point, connected through a neck to the clean side; failing that we at least want to organise the tracking so that there is a distinct clean and dirty side.

Very Dirty tracks:
Transformers CT
Capacitor returns

Medium Dirty:
Output stage bypasses

Clean:
Loudspeaker return
Input earth
Feedback earth
Spares

Not sure how that can be accomodated in the layout, but that is the ideal.
 
You must log in or register to reply here.


advertisement


Back
Top