1
\$\begingroup\$

I have designed a 4 layer board for a DIY DJ mixer I'm building. The 4 layers are:

  1. Connectors. Some traces going to/from components to Molex connectors.
  2. Power plane.
  3. Ground Plane.
  4. Signal plane.

The circuit is designed using dual power supply essentially creating +12V, virtual ground, -12V. The virtual ground is all connected through the ground plane.

I have noticed that the part of the circuit that uses op-amps, specifically the TL072, seems to be ok.

The headphone part of the circuit which uses 2 LM386, is susceptible to oscillations and distortion. I.e: When I touch my case or event pots or components on the PCB, the op-amp circuit seems fine, no distortion, the sound plays clean. The headphone section though goes crazy.

I have attached my KiCad PCB project. Just wondering if I should continue to use ground plane, switch to star ground or use both, ground plane for the op-amps and star ground for the headphones amplifiers?

Or do I have serious design issue?

https://www.dropbox.com/s/shabj2at0fffa25/main-v5.zip?dl=0

The images are in the order that they where described above.

Connector Plane Power Plane Ground Plane Signal Plane Schematic

\$\endgroup\$
4
  • \$\begingroup\$ Post pictures not projects that are inaccessible to those not using KiCad. \$\endgroup\$ – Andy aka Apr 6 '20 at 15:44
  • \$\begingroup\$ Ok cool I figure it would be harder to see... But I will. Thanks for the tip. \$\endgroup\$ – user432024 Apr 6 '20 at 15:46
  • 1
    \$\begingroup\$ Star GND and GND Planes can/should be used both. It’s not one or the other. \$\endgroup\$ – Stefan Wyss Apr 6 '20 at 16:06
  • \$\begingroup\$ Post schematics too. Otherwise we need to reverse engineer the PCB to find out if something is wrong in the schematics. \$\endgroup\$ – Justme Apr 6 '20 at 16:24
1
\$\begingroup\$

given anough power gain, and enough stray currents from Power Output back to sensitive inputs, you can easily make an oscillator.

You've not diagnosed the cause of the oscillation. People on stackX often suggest the LM386 NOT be used. I have no opinions on that.

I can say the standard ground planes will not eliminate 60Hz (or harmonics, from diode rectifier surge currents) coupling into your audio circuits.

neither will STAR Grounds

The planes will provide excellent electro-static shielding on ONE side of the circuit. One side only. And through-holes will degrade even that shielding.

I suggest you use the planes, and learn to SLIT the planes, so as to steer high currents away from sensitive regions of the PCB.

I recall, on "diyAudio simplistic NJFET RIAA preamp" thread, the achievement of over 100dB isolation between output and input of a PCB, using a NARROW region of Ground strip between the 10volt/10mA output (to 1Kohm input load of a Power Amplifier) and the 100 microVolt input from a Moving Coil cartridge feeding a lownoise JFET.

The need of the Preamp ground design was to prevent ECHOES caused by the dominant 50Hz pole (3 millisecond delay) of RIAA compensation, with output---input crosstalk.

Thus designing the Grounding and the current flows is the better approach, rather than deciding between Planes and Stars.

First --- find out why the headphone drivers are oscillating.

Diagnose? clip a long wire to the HOT pin of the headphone output, and bring that wire near your PCB or input cabling from audio sources.

\$\endgroup\$
2
  • \$\begingroup\$ Thanks. Any tips on how to diagnose the issue? I don't have low frequency hum issues with the op-amps. I can physically touch any part of the case, PCB pots etc... and the sound plays supper clean. Only the headphones there's massive distortion when I touch anything. \$\endgroup\$ – user432024 Apr 6 '20 at 16:17
  • \$\begingroup\$ So what you are saying I should basically "slit" the ground plane by adding a big empty rectangular area and leave a small section connecting the two? \$\endgroup\$ – user432024 Apr 12 '20 at 14:07
1
\$\begingroup\$

The trouble with the LM386 is how power is connected to the device. For instance, if power and input signal are applied like this: -

enter image description here

Then no supply currents that feed the output stage flow through the input circuit. However, if the input circuits are wired like this: -

enter image description here

Output stage power currents flow along the ground reference point (purple) and can "modify" the input signal at the LM386 inputs by several mV. This can cause oscillation, noise and instability.

You need to employ star pointing at pin 4 of the device and your main 0 volt power to anything else needs to start point away from that point. Because you use two LM386's there is some compromise but try and keep both pin 4s as close as possible and your incoming 0 volt power needs to hit that point before any other part of the circuit.

\$\endgroup\$
15
  • \$\begingroup\$ In my schematic 4 is going to ground and 6 is going through the capacitor and ground. So I'm assuming I'm using the second approach two what you describe. Except it's ground plane, not a star ground. \$\endgroup\$ – user432024 Apr 6 '20 at 17:52
  • \$\begingroup\$ If the 0 volts power feed to your LM386 is last in the chain then you'll get problems. Your 0 volt power (from the circuit that generates your DC power) should go nowhere else first other than pin 4 to be absolutely sure. Every other circuit node that uses 0 volts should tee-off from that point to ensure there are no LM386 PA output stage currents upsetting input circuits. \$\endgroup\$ – Andy aka Apr 6 '20 at 18:48
  • \$\begingroup\$ OK so if you look at my pcb above then what you are saying is I would have to move the headphones all the way to the left near the PSU input? \$\endgroup\$ – user432024 Apr 6 '20 at 19:03
  • \$\begingroup\$ You might try this idea out by externally applying power directly to the LM386 circuits first. Make sure this power supply is isolated of course and let the rest of the circuits just ride to whatever level they naturally acquire with this change of emphasis. You have to prove it this way just in case I’ve missed something that didn’t reveal itself when looking at the circuits. \$\endgroup\$ – Andy aka Apr 6 '20 at 22:21
  • \$\begingroup\$ Could I physically saw that part of the board off and just solder two wires directly to pin 4 and 6? Wuld the other half still continue to work? WOuld I need to isolate the edge of the boards since the power plane goes all the way? \$\endgroup\$ – user432024 Apr 11 '20 at 21:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.