Speaker Return location choices

Status
Not open for further replies.
I am working on a LM1875 amp build from scratch. This is my first amplifier build, but not my first electronics project. Here is an interior view.


20180824_102504.jpg


My question is about star grounding and the speaker return. From the jack on the right, I can attach the speaker ground to (at least) three different places. #1 returns to the 0V point on the amplifier module. #2 returns to the 0V point at the filter caps. #3 returns to the earth ground on the chassis.


I have seen arguments in favor of #1, specifically in "More Guidelines for Minimizing Amplifier Hum" by Andrew C. Russell. He says if you do #1 to "ensure there is adequate on board decoupling and that it is wide-band". Following Rod Elliott (and TI's) schematic, I have 100nF and 100uF on the amplifier board, but that doesn't match the requirements Russell gives to "use 1uF and 0.1uF localized decoupling around the output stages along with 220uF to 1000uF bulk decoupling on the amplifier module(s)". Does that mean that #2 is a better choice?

I think that Elliott argues in favor of #2 when he says "Note that the speaker must return to the central 'star' earth (ground) point. If connected to the amplifier's earth bus, you will get oscillation and/or poor distortion performance." I'm not clear if the "central 'star' earth (ground) point" might instead refer to #3. And is the reference to "the amplifier's earth bus" specifically arguing against #1? or against #3?

My apologies for a newbie question, but some of the terms being used aren't completely clear to me. I hope that the image makes my particular situation intelligible.

-Neil N0FN
 
Hi. It's a good question and it'll be interesting to here the different answers. Forget #3. Generally it is best practice to keep a flow and return close-coupled, therefore #1 is best.


Agreed on #3. It was a bit of a straw man. I think I just got confused by Elliott's "central 'star' earth (ground) point". That point isn't very central 😉


So far, one vote for #1 and one for #2. I imagine that I will be "okay" either way, but I am very interested to see all the different perspectives.


-Neil N0FN
 
If everything else is done good than #1 and #2 can give similar result.
Theoretically #1 is the best way, but practically #2 can be easyer to do, and can give +- the same result (use twisted pairs and "twisted" triple-wires in both power and signal paths).
 
Hi. It's a good question and it'll be interesting to here the different answers. Forget #3. Generally it is best practice to keep a flow and return close-coupled, therefore #1 is best.

#2 definitely.
This minimises the current path of the return signal.
The power comes via the caps so should go back to them.
Taking the return to the pcb modulates the ground line with audio between the pcb and smoothing caps.
 
You might want to have a look at this post (and the thread in general) http://www.diyaudio.com/forums/chip-amps/252436-lm3886-pcb-vs-data.html#post3846783


Thanks Tony, that is indeed a monumental thread. Tom Christiansen sees a clear advantage in terms of THD when doing #1 instead of #2. As it is explained there, the current to feed the speaker is being sourced from the local bypass caps and should return to the 0V of those local bypass caps.


I've implemented #2 on-board currently since I don't have much confidence in my layout prowess which would be required for #1 to be equally good. I ended doing much of what was recommended in that thread by coincidence. There is a monolithic ground plane on the board and the power ground is on one edge and the signal ground is on the other edge (because that is how the pins on the LM1875 are arranged).



I did not separate the two grounds, letting the full ground plane "connect" the sides. In another thread, it is suggested that a 15R resistor between the two would help killing ground loop noise, but I didn't take that into account when building the "board".



Thanks for the high-quality discussions and links. I've learned a lot here.



-Neil N0FN
 
Status
Not open for further replies.