Interference in a 386 audio amp

[Oznog]

I have a high voltage NJM386 in the funky SIP-8 pkg on the same board as a 3.3v dsPIC. It's not being supplied from the Vdd of the dsPIC.

There's a considerable amount of "buzz", more towards a hum, in the background. It's during certain dsPIC operations so I know it's from there and not oscillation. I have the basic setup, 100ohm R + 0.047uF cap on the output, and a lot of capacitance on the Vsupply. I even added a choke between the +12v and the pin with the capacitor.

I found the noise is not coming from the digital system output. I disconnected the input wire to the 386 and the noise is basically unchanged.

I tried to add some more ceramic capacitance on the dsPIC's Vdd and the 386's Vsupply. It didn't help at all. If Vsupply ripple and noise on the digital system's output are ruled out it seems like radiated EMI is the only explanation.

Any magical solution other than shielding, shortening the wires, or moving it further away? Is the 386 or the linear SIP pkg particularly susceptible?

 

[williB]

Oznog , you still doing the noise cancellation?
What is a NJM386?

 

[Oznog]

NJM386 is simply NJM's LM386. It's supposed to be equivalent. I liked it in that the SIP8 pkg dissipates a lot of heat, has a small footprint, and is very protoboard-friendly.

It's the Noise Reduction library from Microchip, which reduces the noise in a line signal rather than active noise cancellation to create antinoise in space.

The code is mostly working except I hear periodic pops that I don't know the origin of, it seems somewhere in the dsPIC. It does reduce noise, though I am unclear on just how much. I feel I need a stronger output than the Si3000 can do at 3.3v to really judge the difference.

 

[Roff]

Can you post a schematic?

 

[mister_e]

i'm guessing here... do you share the same GND line or you have splitted it? Usually you should split your analog and digital ground. Once done all of them should be connected to the same end-point.

It's a good practice to use bead-core on the Analog gnd too.

Some will also suggest to have different voltage regulator for the analog and for the digital section. Maybe good or not depending of your layout. On a protoboard it's reaaaaallly hard to have something good so far.

Ground plane may help in some case when properly design... unless it may ruine your life...

What kind of hum do you get 60Hz, 120Hz or totaly random frequency?

 

[audioguru]

The RC network at the output of an LM386 uses a 10 ohm resistor, not 100 ohms. It provides a load when the inductance of a speaker is a high impedance at high frequencies.

New Japan Radio made mistakes sometimes. They made their copy of Texas Instruments' TL07x better, they made it oscillate severely and had to recall it and make a normal "B" version. Maybe they also messed up with their LM386 copy.