17
\$\begingroup\$

I have an audio streaming device I bought that has analog audio output that is way too low (it’s supposedly line level, and I’m not the only one who’s complained about this).

I’ve been using this as an excuse to learn about TOSLINK and DACs, but that isn’t the question here. As an alternative to an external DAC I decided to see if a simple pre-amp would work.

schematic

And it does… for a little while. But after a short time (minutes or hours), some sort of bias voltage starts to appear on the positive input pin of both amps (this is inside the DC blocking cap - on the positive input pin of the chip itself). This, of course, gets amplified and biases the output so much that there's no room for the audio. This happens with or without input or load and seems to degenerate over time.

The output load of this preamp is 10 kΩ - it's a power amp I also built that has been in service for many years. Its input starts with a log taper 10 kΩ pot across the input with the wiper going to the input of the amplifier circuit itself (one each with a common shaft for each channel).

At no point do I see any anomalies in either the positive or negative supply rails. So far as I can tell, the circuit draws negligible supply current.

All of the caps are 0805 ceramic X5R with a 10V rating. Could that have something to do with it?

Again, right after construction, the circuit sounds just fine.

What's going on?

\$\endgroup\$

3 Answers 3

26
\$\begingroup\$

The input bias currents are charging up C4 and C5.

Add resistors between the inputs and GND.

\$\endgroup\$
5
  • 6
    \$\begingroup\$ Note that this also means you didn't destroy the op amps. They're still perfectly functional unless something else happened. \$\endgroup\$
    – Hearth
    Commented Oct 10, 2021 at 17:06
  • 1
    \$\begingroup\$ All 3 of the answers (so far) are the same and all are correct. I did some research in the meantime and discovered this as well. Replacing the caps with 0Ω resistors (removing the AC coupling) worked around the problem and the lack of AC coupling - for the time being - is acceptable. \$\endgroup\$
    – nsayer
    Commented Oct 10, 2021 at 17:19
  • 1
    \$\begingroup\$ To the OP: if you add 20 k resistors from pin 3 and pin 5 of IC1 to ground, you can keep your AC coupling caps. \$\endgroup\$
    – user57037
    Commented Oct 10, 2021 at 18:07
  • \$\begingroup\$ @mkeith I get that, but that means reworking the board, and it's just a lot easier to remove the AC coupling for this prototype. \$\endgroup\$
    – nsayer
    Commented Oct 10, 2021 at 23:14
  • 3
    \$\begingroup\$ To the OP: I understand. I will point out, though, that in your schematic, pin 4 is GND, and pins 3 and 5 are the two pins you need to add resistors to. It seems like it would be really easy to find an SMT package that could fit between pins 3 and 4, and also between 5 and 4. But I won't bug you about it anymore. \$\endgroup\$
    – user57037
    Commented Oct 11, 2021 at 1:58
15
\$\begingroup\$

Op-amps require a DC path on each input to supply the tiny current drawn by the inputs. Without it, the DC voltage of the input is undefined. In practice, that means that the input will drift either up (until it reaches V+) or down (until it reaches V-) depending on the internal circuit of the op-amp.

Fix that by adding a 10 kOhm resistor between the "+" input and ground of each op-amp.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
15
\$\begingroup\$

You're missing the input bias resistors on the positive inputs. The bias current, around -20nA for the LM358, on the input pin will charge up the input capacitor and after some time will cause the op-amp to stop working as intended.

The cure, you need a resistor going from the positive input to ground. Something in the 10k to 100k ohm range.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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