Recently I bought a couple of BA3121s to fight the annoying ground loop problems when using several/separated modules (MCUs, BlueTooth, USB etc) in class-d amplifier audio projects and all the modules share the same power source.

For every solution provided, I made a test board (made it for isolation transformers, isolated DC/DC converters) but there always seem some downside to using it (see also footnote below). Not a fan of breadboards, like solid connections and mostly solder a demo board to try a new component.

Like this one, followed the application example mentioned in the datasheet and works pretty well however very thin sound. With thin sound I mean there is no bass. It sounds clear, stereo and noise free however no bass. It sound like a high pass filter. Checked the connections (ok), soldering, and wiring several times and is exactly the same like the example.

example schematic used demo test board

I don't think my result is the intention of the example, so what could be wrong? First thought is the small output decoupling capacitors (4.7uf) in the example which is pretty low and by using a bigger cap in serie (for example 100uf), on the output terminals there is more bass. However, by doing this, I change the characteristics of the circuit.


My question is, why do the manufacturer provide an example that produce low quality sound or do I miss something inhere? Or is there an error in the schematic? Can I replace the output caps and/or input caps with a larger capacity caps without changing the ground loop characteristics? What minimal value should be used?


When you want to take a look at the datasheet, you can find it here and here.

footnote (not a part of the question):

Downsides found by using several different/other ground loop solutions:

  • Isolation transformers on the output (input amplifier) is cheap and easy however can be big and bulky, reduces output capacity (volume), can reduce bandwidth especially lower frequencies;
  • Isolated DC/DC converters, expensive and limited, can still introduce some whine, USB doesn't function when there is no common ground connection

UPDATE 08 may 2018

I'm very confused, the datasheet explain to increase the caps values on the Vm pins however when I do this, nothing happens, same harsh rich treble sound. They talk about CMMR and doesn't effect the sound quality but the noise suppression.

However when I change the output caps, tried 10uf, 22uf, 47uf, 100uf and 220uf. When using 100uf and 220uf there is no difference in sound quality and provides rich bass, so I think 100uf could be right. However there the harsh treble still exists, it is not the same as the original source, the treble is amplified.

There is even more confusion when I found the "evaluation board manual", it is practacly the same as the example in the datasheet however there are some resistors in serie on the input path with jumper pads to be able to bypass the resistor. Especially the 50k(!) resistor, which is enabled by default, really don't understand why you want to do this because it kills the signal completly (or am I wrong?).

Evaluation board schematic

Because it is practacily the same as the datasheet example, why I get thin sound using the example? I don't get it. Why do I need to change the output caps and why is the sound quality not the same?

You can find the "evaluation board manual" here

  • \$\begingroup\$ Bass cutoff -3dB depends on f=0.16/RC Do you know Rin? \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 May 6 '18 at 22:20
  • \$\begingroup\$ Sorry, don't notice your comment. Rin can be different because there are different audio sources. \$\endgroup\$ – Codebeat May 7 '18 at 22:51
  • \$\begingroup\$ Then your load is non-std or too low for chip circuit design .It is not intended to drive a speaker but a power amp with a reasonably high and input impedance \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 May 7 '18 at 23:25
  • \$\begingroup\$ @TonyStewartolderthandirt : It doesn't drive a speaker, it acts like a pre-amplifier for a power amplifier, in this case isolated to suppress/avoid ground-loop noise on the input and finally the output of course. \$\endgroup\$ – Codebeat May 8 '18 at 0:28
  • \$\begingroup\$ If you can code you can compute RC something is wrong with your assumptions for bass response. For the tinny sound , show us a picture of your cables and system connected since this is a high game noninverting amplifier that Attenuates ground noise it’s prone to inductive feedback which will cause your results. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 May 8 '18 at 16:00

The 4.7uF output capacitors that feed the power amplifier pass frequencies down to 20Hz when the input impedance of the power amplifier is 1.7k ohms and feeds lower frequencies when the input impedance of the power amplifier is higher. BUT maybe these coupling capacitors have their polarity backwards? Measure the DC across them to see.

  • \$\begingroup\$ Hi thanks for the answer, interesting. No, the caps are mounted like the example, positives to the pins and negatives to the input/output. Measuring Out:+13.6mV In: +6.4mV \$\endgroup\$ – Codebeat May 7 '18 at 11:58
  • \$\begingroup\$ See my update at the question. \$\endgroup\$ – Codebeat May 7 '18 at 22:52

The schematic is not complete. There is no source of DC_bias for pin_2, pin_6.

As an aside, I'd increase the size of the caps on pin_2 or pin_5, by 5:1, so the HPF high-pass-filter behavior of the biasing drops by 2 octaves.

Perform some causal exploring.

  • \$\begingroup\$ Hai thanks for the answer. Well, this schematic is copied from the datasheet. Measuring Out:+13.6mV In: +6.4mV I already discovred that changing the caps increase sound quality however there is a reason why they provide this example. Will increasing the caps effect the isolation? \$\endgroup\$ – Codebeat May 7 '18 at 12:03
  • \$\begingroup\$ See my update at the question. \$\endgroup\$ – Codebeat May 7 '18 at 22:52

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.