2
\$\begingroup\$

NOTE: It turns out, the ground connections I wonder about below actually are connected (thanks Sim Son in comments), although I still cannot see how they could be connected (additional closeups). The reason I initially concluded they weren't connected was a combination of being unable to see connections on the PCB (still can't) along with, as it turns out, a broken probe on the meter I was using for continuity checks. So the main premise of this question appears to be incorrect; although the questions about the effect of said ground still stand.


I am looking at the audio amplifier section from a mainboard scavanged from an LCD TV (Sony BU Board, P/N 1-876-561-13, TV appears to be a KDL-40V4150). The amp circuit is built around a TI TPA3100D2PHPR chip (20W stereo class-D amp). The PCB appears to be a 2-layer board.

I have been studying it, as I made a project out of cutting this section out of the mainboard and using it as a standalone amp (mostly worked). However, while closely examining it I noticed what appears to be an unexpectedly absent ground connection in the output section, and I'm not sure what to make of it.

Schematics

Here is part of the schematic from the TV's service manual and the corresponding part of the example circuit in the chip's datasheet (click for high res). The only difference that might be relevant (?) is the Sony circuit doesn't have the 1nF + 20Ω decoupling bits off the outputs before the inductors. I've highlighted the suspicious ground connections in blue:

Service Manual (Sony) Datasheet (TI)
enter image description here enter image description here

Notes:

  • Sony uses this board (or slight variants of it) in a bunch of TVs. I'm confident this is the correct service manual as I found the model # of the TV printed on the chassis.
  • I assume that Sony uses this amp circuit in other products as well, since it is relatively self contained and I see lots of unused pads that I assume are for parts in other variants.

PCB

That's all fine, but the reason for my question is the PCB appears to not include the ground connections highlighted above.

I manually followed all the ground traces on the PCB. Descriptions are with the images below (click to enlarge), the general components of interest are the capacitors after each inductor (C2096 thru C2099):

Description Back Front
TI chip in center. enter image description here enter image description here
Ground trace is green. Caps and the traces between them are purple. enter image description here enter image description here
Back image also shows front traces. Front image is service diagram instead of photo. enter image description here enter image description here

The confusing bit is here:

enter image description here

The grounds are green and the caps and traces between them are purple. Unless I'm missing something, there is no connection on the PCB between the purple and the green. This would mean that the connections that I highlighted in the schematics aren't present.

Question

So that was a lot of pictures but my question is:

Why might those connections be missing? Is it some other common configuration for a class-D amp's output that I don't know about? Is it related to the lack of decoupling capacitors at the chip's outputs before the coils?

Also, how does that ground affect the amp's behavior?

Basically: What's going on here?

Note: The outputs of this circuit go to two places: directly to the TV's internal speakers, and also to some other amp that drives the line-out jacks. There's some resistors and capacitors there too, not sure if they're relevant, but here is that part of the circuit.


There is strong evidence that the PCB is only 2 layers:
  1. I cut this from the main board with a bandsaw and only saw two layers on that border.
  2. I am able to account for every connection on the schematic with the visible PCB traces except for this ground (I checked carefully, every component, in the amp section as well as the main audio processor and aux amp).
  3. The service manual's PCB layouts (page 54) show two layers only, and have no indications of blind vias or traces.
  4. It would be trivial to route those to the ground trace. The purple and green areas (from my images) are directly adjacent on the front side of the board.

So under the assumption that cost reduction is important, it seems to me like it would be strange for the PCB to have another layer just for a few connections that could just as easily be made on the existing layers.

\$\endgroup\$
6
  • 1
    \$\begingroup\$ Following copper traces can be tricky, did you check for continuity between the suspicious points to be sure they really aren't connected? \$\endgroup\$
    – Sim Son
    Dec 30, 2021 at 1:21
  • 2
    \$\begingroup\$ Any questions Jason? \$\endgroup\$ Dec 30, 2021 at 3:02
  • \$\begingroup\$ @SimSon I swear I did. I just did again to double check and... they are connected to ground. And I absolutely can't see how. I took some more closeups. Also this completely changes the question.... I'm 100% baffled how they are connected. \$\endgroup\$
    – Jason C
    Dec 30, 2021 at 4:12
  • \$\begingroup\$ @TonyStewartEE75 I have so many questions, heh.... \$\endgroup\$
    – Jason C
    Dec 30, 2021 at 4:12
  • \$\begingroup\$ @SimSon Oh my; I just now figured out why my continuity tests failed the first time... looks like one of the probes on my meter has a broken internal connection, making it sensitive to the angle it's held at... bad combo when considered with the seemingly invisible connection that's actually on the board. Argh. Time for video games and cats, I'll come back to this tomorrow. I'm fried. \$\endgroup\$
    – Jason C
    Dec 30, 2021 at 5:00

2 Answers 2

1
\$\begingroup\$

Thanks for a well presented question.

Page 19 of the referenced TI datasheet suggests the output filtering is optional. Depending on the target market and the exact application, Sony have determined it was not necessary. They may have put a common mode choke on the speaker wires. TVs are a high volume product and heavily cost engineered to specific markets. If they can eliminate a component, they will.

The filtering is for EMC purposes and does not affect the normal operation of the chip.

[edit] if the circuit was perfectly balanced, the voltage between the two capacitors would be 0. In a real circuit with real components that might not be the case, so there would be currents flowing through that point. That may have caused an EMC issue, so they deleted the track. EMC is a bit of a black art - it is all based on some fundamental physics but due to the number of variables, it is not a simple task to model everything. Each pcb track has inductance, capacitance, resistance as does each wire and component. A small change can cause a significant shift in the EMC performance. Experienced people can look at the spectrum analyzer results and can guestimate what might be causing a peak at a particular frequency.

\$\endgroup\$
1
  • \$\begingroup\$ I'm marking this as correct. Even though there's something wrong with the premise of my question, you did a good job describing the effects of that ground and purpose of that part of the circuit, which was the other big part of my question. Thanks! \$\endgroup\$
    – Jason C
    Dec 31, 2021 at 14:33
2
\$\begingroup\$

I would like to add that the missing grounded C connections are much smaller than TI's datasheet, but exist with the close proximity of the aluminum backplate attached to PCB ground for EMI suppression and therefore has an equivalent circuit, if you also used a similar grounded backplate.

If the circuit is balanced by good component and parasitic ESL matching, then that node is 0V and therefore has little current flowing at audio frequencies and only the common mode switching f & harmonics only need a small C to ground plane.

Nice pictures. Analysis needs Test Experience for parasitic ESL (~8 nH/cm) layout balance and coupling to gnd-plane ~1 pF per mm^2 <1 mm gap.

Please verify my assumptions.

\$\endgroup\$
14
  • 1
    \$\begingroup\$ These are normally seen in power circuits to decrease the resistance of the track. It might also be to create a number of ‘lumps’ to lessen the possibility of the track resonating. I can’t say I’ve resorted to doing such things myself. \$\endgroup\$
    – Kartman
    Dec 30, 2021 at 5:03
  • 1
    \$\begingroup\$ @JasonC the Blobs are a normal result of solder stencil with paste applied to whole board "screen printed" but parts excluded from the build then reflowed into a blob. no issue. Faulty Probes are a fart of nature, always calibrate your probes before use with any DC you know exists or square pulse for DSO \$\endgroup\$ Dec 30, 2021 at 14:06
  • 1
    \$\begingroup\$ The cost of EMI CM chokes and plastic filter caps needs to be compared with the cost of copper must be considered for high EMI generation of the class D output. \$\endgroup\$ Dec 31, 2021 at 16:00
  • 1
    \$\begingroup\$ Good . 2L . I forgot was there a technical problem with sound or EMI or just curiosity. My theory is that the 1uF caps are mismatched worse than the traces and inductance so a floating ground plane becomes a radiator, if caps are gnded, while an earth ground suppresses it from low impedance.🍸🧘🛴🚲🚁🛩⏱💵💸💴💰🛠⚙️🔫🧲💊🚬⚰️🎉🎉🎉🎉🎉🎉🎉🎉 Happy '22 \$\endgroup\$ Dec 31, 2021 at 19:46
  • 1
    \$\begingroup\$ If you have FG nearby, wire it with a plastic cap to the subplate . Your 100 pF finger has a 100nF body antenna injecting Efield at high CM impedance. A frame ground ,FG cap will shunt that 100 pF easily like 10 to 100 nF or so. Of course you can also direct connect PE gnd too but so safety in case of some inner fault, a series cap to shunt line E field noise is what you need due to unbalanced source impedance. So add this your your backplate. \$\endgroup\$ Jan 1 at 21:03

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.