TPA3116 strange output wave form

[Michaelaudio]

I designed a tpa3116 board exactly as the typical application says. There 3 tpa's on the board (high, mid, bass) high and mid works just fine but I have a strange behavior on the output for the bass. At 20 HZ it starts to get a strange form at around +/- 10V ptp. At 60hz it gets better but still looks weird when its close to clipping. Strange thing is that the input signal looks the same as the output signal, but ONLY under load. Without load it looks just fine, its like the chicken or egg problem and i dont know what to do (already replaced the tpa) In the pics u can see the output at 20 and 60hz, and the input at 20hz  (1V ptp)

[Michaelaudio]

Okay I found out that when I connect my Phone directly to the input cap of the ic its a clear sinus on the output. So it must be a Relation between the amp and preamp but only under load.
Has no one an idea?

[Michaelaudio]

Update: I dont think its something to do with the tpa. Had a TPA board layin around and connected the output of my filter to it, so it must be my filter but only when the tpa is load es enough.. wtf?
The signal comes clean out of IC1 but is distorted at the output of IC5. Its not the ic itself, already replaced it. I dont know what to do

[SMdude]

Is it something with the ground or power rails?
Can you separate the ground and feed it from an external wire. Also are you able to power the preamp and TPA separately?
It looks like clipping, but if it doesn't happen with the input from the Iphone, then it must come back to power/ground disturbances.

Also I note that the setup for input/output is different for base vs Mid/high.

[Michaelaudio]

Ive tested it with another amp board justconnect Ing the output of the filter in the other board, but it was with the same power supply. Ill do that again but with a separate psu for the amp board.
What do u mean by that the Setup is different?

[SMdude]

As in the outPr and outNr are shorted and the L outputs are shorted too.
The inputs for L are shorted to gnd.

That may be intentional and normal. I just noted that it was setup different to the mid and high.

[Michaelaudio]

Ahh yea, thats the Setup for bringe Mode.

So I tried it out, separate psu for the other amp board in bridge Mode. Competely clean output wave till its clipping at +/-24v! 
I kinda knew it was something with ground, this shit can get u so much in teouble.  But how do I solve the problem now?? I have absolutley no clue. I mesaured every ground connectIon around the IC 5 and 6, everything good.

[Renate]

As in the outPr and outNr are shorted and the L outputs are shorted too.
The inputs for L are shorted to gnd.

Yes, as counter intuitive as that is, that's how the "Parallel Bridge Tied Load" (PBTL) works.
The TPA3116 specially detects hard grounded left inputs and alters its mode.

Those scope photos are certainly hard to decipher.

[Michaelaudio]

Sadly I cant get better pics. But I guess u can tell what happens (with my description too) when the input Voltage is about .5V ptp. But only under load, looks wonderfull without any load.

[Renate]

Have you tried bypassing (shorting) your power MOSFET?

[Michaelaudio]

Didnt change anything. Oh and if I turn the volume up too much, the amp goes crazy and the speakers go up and down really heavy. Its like the amp is "pumping" and the current goes verry high and the voltage down, its like up and down up and dow with a awfull noisy crackling  sound, till I turn the volume down or disconnect the speakers.

[Renate]

It sounds like your power supply is not sufficient.

The other question is whether the bootstrap voltage is sufficient at low frequencies.

[Michaelaudio]

Thought that too but thats not the problem.

It has to be, when I connect my Phone directly it works just fine

[SMdude]

Is the ground for the preamp section(audio ground) in the current path of the tpa?
You need a clean ground for the preamp section that doesn't see the current consumed by the TPA.

You might be able to separate the ground and run a wire from a common point for the preamp/audio ground.

[Michaelaudio]

Thats the pcb. U6 is tpa for the Bass, I guess the ground shouldnt be a problem?

Edit: soldered a wire on C172 and connected it to the psu clamp (ground) didnt change anything, but I guess the ground on the pcb should be fine anyway

[Michaelaudio]

Weird thing: when I switch the bass boost on (which is at 40hz)
The input (and o the output) is going crazy at ~0.6V ptp. Without bass boost it looks just a lil out output shape.
Settings for the pics:

Input 0.5V/div  20hz signal
Output 10V/div
Tested w a 8ohm driver

When I plug my 2 4ohm driver  (parallel) in, then things r going crazy earlier, so yea its obv that its something with current and ground maybe. 

I can take better pics, but I guess u can tell whats going on

[SMdude]

I would be trying to isolate all of the grounds on the signal path, ie C54, C53, the 10k resistor(d19?), C51, C49, D10, C47, C46, RV5.
I'm actually not an audio nut at all, but do design and layout some low noise circuits from time to time and one needs to be very considerate of current flows through the ground plane.

I see you have a separated area in your ground plane under the bass section.

Is that a 4 layer board or only 2?

[Michaelaudio]

What do u mean by Isolate? I mean ground is ground?
2 layer board

For example this is a board from ebay and it works fine. Only bottom layer is filled w ground!

[tooki]

What do u mean by Isolate? I mean ground is ground?
2 layer board

No, ground isn't ground. (There's a reason there are different symbols for earth ground, chassis ground, signal ground, analog ground, digital ground, etc.: so that they can be used separately within a circuit.) Current flows through ground, like any other wire or trace, and sometimes you want different parts of the circuit on "separate" grounds -- yes, they'll all connect somewhere, but the idea is that almost no current ever flows through that one spot.

Take a look at https://www.onelectrontech.com/pcb-layout-design-tips-grounding-considerations/ -- it explains it quite well.

Your PCB layout really isn't very good. I fear your power traces (VCC) may be much too thin, contributing to the problem. But also, all the caps on VCC are too far away from the ICs -- they should be located literally as close as possible, with as fat traces as possible. On your board, they meander around, creating loops that can induce currents in other traces, potentially resulting in noise and instability.

I also see that many signal traces take very... creative paths. Like the output of RV5 (subwoofer out), which starts on the top layer, goes down to the bottom layer (through a via right next to the through-hole pin it came from!  ), then goes back to the top, then back to the bottom, back to the top, back to the bottom again, and finally back to the top to connect to C153. That literally could have been accomplished with a single via, by having it come out of RV5 on the bottom, where it has a clear shot all the way over to near C153, where it would pop back up to connect. (And with careful layout, it likely could have been routed entirely on the top, with other signals traveling beneath C153, for example.

Was this autorouted? Audio is something where autorouting really isn't wise, in that you often have both sensitive signal paths as well as heavy power paths right near each other.


I'm expecting my hand-routed TPA3126 PCBs from China this week. I'm of course rather curious as to how well they'll perform. I'm sure I've made some layout mistake somewhere, but I think all the important bits got covered.

[Michaelaudio]

I dont think its a problem with the caps, many Chinese Boards are like that and they work. Um yea no.. Not autoruted but my first bcb with so many parts, as u can see.
If its a ground problem, why was the problem still there as I connected another amp board to the same psu, and just feed the signal of this board into it?
On the other Hand as I fed the other board with a different psu und put the signal of this board into it, it just worked fine

[tooki]

We are doing our best to guess what you mean from your unclear descriptions.

I dont think its a problem with the caps, many Chinese Boards are like that and they work. Um yea no.. Not autoruted but my first bcb with so many parts, as u can see.
If its a ground problem, why was the problem still there as I connected another amp board to the same psu, and just feed the signal of this board into it?
On the other Hand as I fed the other board with a different psu und put the signal of this board into it, it just worked fine

Well if a Chinese board happens to work OK (not that it's been tested under any kind of controlled conditions) then I guess TI must be wrong in their recommendations for their own chips. And if you're such an expert already, then why are you here asking questions?

Realistically, doing one thing wrong will likely not cause any trouble. But your board and circuit clearly have multiple things wrong, and that can be enough to cause real problems.

[Michaelaudio]

Who says im an expert🤔 I just try to understand things. When the TPA ground needs to be "far away" from the signal ground, then why was the problem the same with the other board, that was directly connected to psu ground? And why did I worked with a separate psu for the other amp board? -> that would speak for the ground problem ur refering to.

I just wanted to try to get this working, but it seems that I cant.

[Renate]

I just wanted to try to get this working...

Yes, you can!

How about those split rail bias supplies on the bass filter channel?
Those things have a time constant of 5 milliseconds.
Maybe you've got motorboating from the power supply loading being fed back.
Throw a bigger cap on it and see if something changes.

[Michaelaudio]

U mean instead of the 1uf maybe a 10uf parallel to the Resistor?

[Renate]

Yeah, just tack solder on a bigger electrolytic as a test.