Insight request: Parallel 4780

[Crash&Burn]

I created a nice pair of LM3875 boards by myself, nothing spectacular but IMHO nice for the first time in 11 years since high school. Dang I'm old
It worked inspite of my lack of experiance, used a 4.7uF cap on the input (great way to kill bass dead), and it is quite nice even with an underpowered transformer. After blowing it up and re-etching the boards I am happy with it.

Anyway it was a lead up to building the real deal: a 4 channel LM4780 120W version to eventually drive the speakers I am designing.

So I was just doodling at work and made the attached PCB, and I think it should work but would like some experianced eyes to go over it to see if I made any blunders.

Also is it possible to make a regulated powersupply for this? My gut say no but maybe I missed something (cost is some what of an object). otherwise I will be using a 48V centertapped transformer MUR860's and about 44,000uF per rail.... anything else I should consider?

[Leolabs]

Is possible to make a regulated power supply for this,just cost is high and the size is bulky.

U need not a fast rectifier for the power supply.44000uF per rail is too enough to me.But still up to you

[Upupa Epops]

Make power traces wider, there " goes " big current .

[Leolabs]

And add solder mask too

[Alcaid]

Try to get Cs5 and Cs6 closer to the chip.

[digi01]

add holes for pcb stand.

[jackinnj]

the power ground on the PCB is inviting some problems. consider placing a nipple on the output ground and bring all of your signal grounds to this one point.

the decoupling capacitors like to have their feet planted in a groundplane. I would start by placing all of the power traces on the bottom, all of the signal traces on the top. and (Net_0), the power groundplane on the bottom.

On the bridged amp boards which I had burned I identified the signal ground plane as a separate net, brought all the signals together to this one point, and bridged the signal and power ground net (Net_0)s together. there are no problems with ground loops this way.

[Crash&Burn]

Well again doodles at work, I basicaly recreated it from scratch. Ironically the issue I was trying to minimize by making it squareish (size) is actually improved with this more rectangular design.

So I guess this is a whole new set of issues that I may have missed, got the moutning post holes this time tho, moved the caps and tried to increase the power traces. I am working this completely by hand in PCBExpress so I am somewhat limited by choice not to use netlists right now until I have the basics down.
But Jack the power ground issues you describe with the first layout of the board, can you explain it in a little more detail for me?
Will the solution I have with the grounds being tied together right at the output ground suffice?

[Alcaid]

Your output traces are wrong. You have the zobel resistors (Rsn1 and Rsn2) in series with the output signal. Also your output traces are to narrow. This thing can make some serious current!

http://www.geocities.com/CapeCanaver...TraceWidth.htm

[jackinnj]

Originally posted by Crash&Burn
Well again doodles at work, I basicaly recreated it from scratch. Ironically the issue I was trying to minimize by making it squareish (size) is actually improved with this more rectangular design.

So I guess this is a whole new set of issues that I may have missed, got the moutning post holes this time tho, moved the caps and tried to increase the power traces. I am working this completely by hand in PCBExpress so I am somewhat limited by choice not to use netlists right now until I have the basics down.
But Jack the power ground issues you describe with the first layout of the board, can you explain it in a little more detail for me?
Will the solution I have with the grounds being tied together right at the output ground suffice?

Originally posted by Alcaid
Your output traces are wrong. You have the zobel resistors (Rsn1 and Rsn2) in series with the output signal. Also your output traces are to narrow. This thing can make some serious current!

http://www.geocities.com/CapeCanaver...TraceWidth.htm

TH PCB traces for the output should be 0.080 to 0.100 mils (thousands of an inch). They may handle up to 3 amps. If you have the RC Snubber Network you may not need the RL output network.

I found that the snubber resistor should be rated at 1 watt -- when I used smaller ones they discolored indicating some stress.

With regard to the mounting holes for the chip -- you can always place the screw on the fin side of the heat sink (rather than the chip-side) . I prefer to "tap" the heatsink and not use a screw, however -- just saves me from having to handle one more thing.