LM3875 PCB Layout

[Sci]

Any comments on this PCB layout? I've read what I've been able to find on PCB design for chip amplifiers. Can't hurt to get the opinion of others though.

I've placed the output ground to close as possible to the center between the rail grounds. I've also tried to keep the signal ground separated from the power ground.

I'm not sure though if the voltage and ground planes are wide enough? (100 mils)

Erik

[peranders]

You won't probabaly fuse the 100 mils traces but make them as wide as you can.

Only small error though: The tiny trace from the "SG" should go DIRECTLY to "OG".

[Sci]

Is it OK to connect the chassis ground at the PSU, or should I squeeze in another pad?

Erik

[Franz G]

Do you want to build monoblocks or a stereoamp?

For stereoamps, it is better to remove the trace between SG and OG.

Then you connect both boards with a thick copperwire between the two OG points.

Exactly in the middle is your central ground. Here you connect both SG with exactly same length wire. Here you can connect the psu ground and the chassis and the speaker ground.

Works great!
Franz

[peranders]

Originally posted by Sci
Is it OK to connect the chassis ground at the PSU, or should I squeeze in another pad?

Erik

This is OK now but I had drawn a trace directly from SG to OG. This has no big importanance, only feels better. The main thing is that you don't mix currents from the caps and speaker current together with signal ground. If you do that you will get a small increase of distortion beacuse the output won't get symmetrical due to voltage drops. You can also get hum. This is even more important if you have 35 um pcb.

I still think that you should make all high current traces wider. Make them as wide as you can.

[Sci]

[OT] Just started my studies at Chalmers University of Tech., not to much time for the amp right now... [/OT]

Per-Anders: I've widened the voltage and ground traces to 150 mils. Not sure if the chip and insulator (I use the T variant) are thick enough to reach outside the PCB edge though.

Not really sure how thick the copper is on the boards I have. Might just buy a 70um or even 105um board.
Would 100 mil traces be enough with 70um or 105um boards?

Does anyone know how thick the copper is on BrianGT's boards?

Franz: I'll be building a stereo amp, I will probably leave the signal ground as it is though, to keep the number of wires down.

Erik

[BrianGT]

Originally posted by Sci
[OT] Just started my studies at Chalmers University of Tech., not to much time for the amp right now... [/OT]

Not really sure how thick the copper is on the boards I have. Might just buy a 70um or even 105um board.
Would 100 mil traces be enough with 70um or 105um boards?

Does anyone know how thick the copper is on BrianGT's boards?

Boards look good to me. I am using 2oz copper for my boards.

--
Brian

[Sci]

I've decided to add a Zobel network to the board. This has left me with a bit longer lines for pretty much everything, but I can't imagine it would make that much a difference. The thing I'm not to sure about is placing the Zobel resistor as close to the feedback resistor as I have. On the other hand, none of them are wirewound, so they should have low inductance...

Erik

PS. If I've understood things right, the Zobel acts as a low pass filter to remove amplified HF distortions. Assuming this is correct; what is the formula for the cutoff frequency. DS.

PS2. How much does the value of the input cap matter? Does it reflect the DC offset it is able to protect from? I'm kind of on a budget, and I am considering changing it for a 2u2 one... DS2.

Edit: Value of the input cap...

[Mr Evil]

The Zobel network ensures the amp sees a low impedance at high frequencies. The transition can be considered to occur where Rz = 1/(2 * pi *f * Cz).

Ideally you want that frequency to be well above the audio band so it doesn't have an audible effect. You could experiment to see if you can get away with a smaller capacitance (try testing the with a very long cable and particularly inductive speaker and see if you can spot any distortion or RF on the output). It's probably best not to alter the resistance from that suggested in the datasheet.



If the input capacitor is too small you will lose some bass. Again, it depends on the same formula, Rin = 1/(2 * pi *f * Cin). 2.2uF should be acceptable. You could even go down to 1uF with only a minor loss.

[Sci]

I can't find any information regarding the Zobel network in the datasheet, could anyone point me in the right direction?

I've moved things a bit. Is it OK to route the output through the pins of the input cap as I have? (See the attached image)

Erik