|16th July 2018, 09:56 PM||#11|
Join Date: Feb 2009
To be honest, I can't see enough of the eBay board to be able to provide any meaningful critique of the layout. If it's a well done 4-layer board, it could be a good layout. I strongly doubt it's anything beyond a connect-the-dots 2-layer layout, however.
From what I can see, it looks like the eBay vendor forgot to add the local decoupling on the LM3886es - both the 2x100 nF and the 2x10-22 uF per LM3886. They could be on the bottom as is the case with Shine7's board, but I doubt they are. You rarely see boards with hacks like that for sale on eBay due to the added assembly cost of such fly-by-wire components.
Another thing is that you can't buy the components alone, even at high quantities, for the total cost of the done-up board with shipping. That should tell you something.
Looking at the Shine7 layout, I see several issues. First off, bringing the supply in from each side is about the worst you can do from a rail-induced distortion point-of-view. Douglas Self writes about this in his Power Amp Design Book.
There is no local decoupling on the LM3886es. The 2x100 nF he solders to the bottom of the board are not enough. You need a pair of small electrolytic or tantalum caps (10-22 uF) there as well. The Thiele network (L||R in series with the output) is missing in the Shine7 circuit, but strangely present in the eBay circuit. Go figure.
The passives around the LM3886 look to be 0805 size, maybe 0603. This means the feedback resistor is rated for 125 mW or maybe 62.5 mW. Granted, Shine7 wants you to use 200k/10k for the feedback network, so you'll be OK from a power dissipation perspective. Sadly, 200k/10k is 20 dB noisier than the more commonly used values of 20k/1k. Change the parts to 20k/1k and you'll bump into the dissipation limit on the 20k and even if you don't, it'll certainly get hot, which means you'll get additional distortion from its temperature coefficient.
I would also add ballast resistors on the MUTE pins of the LM3886es to make sure that they all come all the way out of the MUTE state. That's more of a CYA thing, though.
I do like that the Shine7 board includes options for using either SMD or leaded resistors for the 0.2 Ω ballast resistors. Finding low TCR resistors in SMD can be a challenge at times.
It's interesting to note that the eBay circuit carries forward some of the design flaws (lack of decoupling, for example) of the original circuit. That should tell you something about their technical skill.
If you skimp on the decoupling (i.e. use less than 10 uF || 100 nF), the LM3886 will oscillate at higher output currents (so as the output swings near the rail with a 4 Ω load). Ask me how I know. You can see my take on LM3886 decoupling here: Taming the LM3886 - Supply Decoupling.
Best of luck with your build. I hope I've given you a few search terms to pursue and nudge you in the right direction.
First and foremost: Read the LM3886 data sheet, including the application section in the back.
|17th July 2018, 01:52 AM||#12|
Join Date: Aug 2002
Location: victoria BC
bauderline - I think it’s fair to say that Tom understands at least as much about maximizing performance the performance of the LM3886, and engineering work-arounds to mitigate for its shortcomings as any member frequenting these forums. One only need to read the “Taming” section of his site to realize he’s put some thought into the subject.
you know I can’t quit you, baby
|Thread Tools||Search this Thread|
|Thread||Thread Starter||Forum||Replies||Last Post|
|My take on BPA300||FyrBridd||Chip Amps||2||25th October 2017 02:23 PM|
|WTB : Bpa300||dhole||Swap Meet||1||29th July 2011 06:58 AM|
|Parts for the BPA300||Agramerus||Chip Amps||2||1st July 2009 09:52 PM|
|New To Site?||Need Help?|