4
\$\begingroup\$

At the moment, I'm working on converting a single ended audio signal to a differential one (source is a CS42L55, need a differential input for a CSR8670. The latter's datasheet is indecisive about how to approach this issue). I have a few options, and I'm wondering about the validity of each.

  1. Tie the negative differential input to analog ground
  2. Use an op-amp to create an inverse signal
  3. Use a transformer (I've only just read about this, not sure how it works yet)

This got me thinking - why does a differential signal need an inverse input, if the noise on both cables is the same?

For example (all waves start from 0v):

                   /\
Positive input:   /  \

Negative input (neutral) : ----

If the same waves had noise, like this for example:

                 __
Positive input: /  \
                          _  _  
Negative input (neutral):  \/

Therefore, pos - neg = pos without noise right? (the sum being the function of the converter?) Is there a benefit to using an inverse wave instead of ground (op-amp > just tying to gnd?)

I guess I am missing something here, and I'd like to know what.

\$\endgroup\$
2
  • 2
    \$\begingroup\$ The idea of routing differential signals is to couple them as tightly as possible so the noise will be the same on both signals, so when you subtract them you will cancel the noise out. So your example of noise only on the positive side only doesnt address the main benefit of having differential signals. \$\endgroup\$ – Wesley Lee Mar 3 at 16:56
  • \$\begingroup\$ Transformers for small audio signals are sooo 1960s... one can almost smell the high anode voltage and see the cathode glow. \$\endgroup\$ – fraxinus Mar 4 at 11:15
9
\$\begingroup\$

why does a differential signal need an inverse input

They don't. What makes "differential" signals work is that the impedance of both wires is equal (balanced) which is necessary for noise to be picked up as common mode which can then be rejected. If the impedance of both wires is different, then some of the noise will be picked up as differential signal.

For example if you use a current sense resistor you'll have better noise rejection by routing the sense lines as a differential pair. In this case there is no inverted signal, but routing as a differential (and balanced) pair ensures the noise picked up by both lines is common mode and rejected by the diff amp.

enter image description here

Using a positive and inverted signal just doubles the available signal amplitude, which gives you a bit of extra signal to noise ratio or headroom. It also minimizes some distortion mechanisms in the receiver (ie, odd order distortion).

\$\endgroup\$
3
  • \$\begingroup\$ Thanks for this answer. This also cleared up why the signal wires need to be balanced, which I didn't understand fully before. Do you think that using an inverse input with an audio signal would mean an actual audible (to humans) difference in sound quality? I'm not sure if there is a general answer to that question or if it's more situational (in that case don't worry). Thanks for your help again. \$\endgroup\$ – brazed_blotch Mar 3 at 18:41
  • 1
    \$\begingroup\$ I'd say it depends. If the bluetooth audio chip has an internal differential opamp right at the input, it probably won't matter much. If it's straight into the ADC, then these will probably require a differential signal to get best performance. Note this kind of mixed signal chip is quite noisy, so its internal ground might not be very clean, and a differential input solves this problem by not being referenced to ground. \$\endgroup\$ – bobflux Mar 3 at 18:46
  • \$\begingroup\$ Seems there is an op-amp that takes the differential signal before the ADC so I think I should be OK. Cheers mate \$\endgroup\$ – brazed_blotch Mar 3 at 18:56
3
\$\begingroup\$

The idea behind differential lines is to avoid common mode noise.

When using a single ended line the signal line will have a difference impedance to ground than the ground reference line. Therefore, environment noise will generate a different potential on the line than on the ground line, hence the receiver will see the noise.

in the case of a proper differential (balanced line), you have 2 equal lines, each with an opposite signal, when substracting the signals, since they are opposite you will get a signal twice the amplitude, however since the lines are equal, the noise component in both will be equal, substracting two equals = 0, hence you eliminate the noise that way.

You can see in a typical xlr connector that you have 3 pins, where 1 is shield, which acts as ground reference and provides shielding for the balanced line 1-2.

\$\endgroup\$

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.