I'm trying, for MONTHS (obviously not everyday, but I spent at least more than 80 hours on this) to do a 15 W audio amplifier with two TDA 2002's in a bridge configuration, following the circuit provided in the datasheet itself, but it never works.

This is the circuit:enter image description here

(Note: both the TAB and the 3rd pin are the same thing)

The circuit is only stable in one configuration: with the potenciometer P1 completely grounding the non-inverting input of the chip on the right, setting its output to 0 V.

This causes the chip on the left to work like it was alone in the circuit, with a speaker directly on its output with no DC decoupling. It sits around half VCC with no signal, causing a constant ~1.2 A to flow in a 4 ohm speaker.

Other than that, if the P1 is far from grounding (ie, if its on the opposite end), the circuit is wildly unstable, it keeps making loud clicking sounds and making the speaker jump from end to end, like if its output was a square wave going from one extreme to the other.

If it ever stays stable for a while, any input signal makes it go back to that unstable state.

In some of the circuits I made, it was so ridiculuously unstable that clapping my hand or screaming loud near the speaker would make it start again (just because of the small fluctuation in voltage caused by the sound reaching the spaker, making it behave like a microphone).

I've checked every single component, specially the capacitors many times, all of them are working and correctly valued.

I've checked my power source and even used a car battery to ensure it wasn't the cause of the problems.

I've disassembled the circuit and reassembled it TWICE (all components are soldered in place) and checked them to make sure no connection was wrong or failing. I also tested changing the load from my speaker to an equivalent resistor to avoid unstability caused by its inductance.

I even changed the chips 2 times. I even thought they could be fake, so I did a test circuit for the single chip amplifier configuration and all of the ones I bought work correctly on it.

I've come from the "something maybe wrong with the design" to "something is surely wrong with the design".

What's curious is that the bridge configuration for other chips like tda 2003 is almost identical, I doubt they would make the same error twice.

But I can't see any way the problem is in any of my equipment or pieces.

Do you have any idea what the issue could be, and how to solve?


  • \$\begingroup\$ Where did this design originate from? It is not what I normally see for a bridge configuration. P1 is to set the bias voltage, which is normally not needed. I do not think this is a good design at all. There are many bridge designs for that IC alone. Often the mandatory fewest parts works the best. \$\endgroup\$
    – user105652
    Dec 22, 2018 at 1:58
  • \$\begingroup\$ @Sparky256 It came from the datasheet itself, from Fairchild semicondutors. I know this potentiometer is used to set the bias voltage, but it never really manages to do so. The output of the chip on the right never reaches half VCC as it should, and its unstable. Not using this potentiometer doesn't help either, it still is unstable as hell without it. Maybe I'm using fake parts that get to work in the single chip configuration but are too unstable to work in bridge configuration? \$\endgroup\$ Dec 22, 2018 at 2:07
  • \$\begingroup\$ This is certainly a positive feedback problem with many causes as shown. Vcc ought to be filtered 10mH 1000uF for each for isolation. The 220uF need to be Low ESR to get << 1 Ohm \$\endgroup\$ Jan 23, 2021 at 5:29

2 Answers 2


The power supply rejection is not very good (25dB at low frequencies).

Run separate HEAVY wires to each amplifier, and put 10,000uF capacitors on each of the VDD pins.

Also, this chip has 80 dB gain. That means a 0.001 ohm feedback resistor, shared in the GND daisy-chain between PowerRTN and input, with 10 ohm (or smaller) Rload, will cause oscillation.

I suggest you reduce the resistance of the GND wiring at least 10:1.

Remember that ONE MILLIOHM of Ground resistance (or impedance) will be a problem, given your open-loop voltage-gain is 10,000X.


Here is a sketch of the GROUND-as-feedback situation


simulate this circuit – Schematic created using CircuitLab

  • \$\begingroup\$ Although Aol is 80 dB, Acl is 40dB with R3/R1, R8/R5 but I wonder if they are wirewound.and not sure about R8/R6=46 dB, I wonder what layout differences in inductance exist. 1 mΩ ? not sure about that \$\endgroup\$ Dec 22, 2018 at 3:01
  • \$\begingroup\$ @ Tony As you know well, 4" of wire is about 100 nanoHenry. At 1KHz (assume a risetime of 0.5 millisecond, to produce that 1KHz), the Zl is 2 * pi* F * L = 6.3 * 1,000Hz * 0.1uH= 6.3 * 0.0001 = 0.6 milliOhms. Yes, that is my expectation. \$\endgroup\$ Dec 22, 2018 at 8:49
  • \$\begingroup\$ From the OP "... it was so ridiculuously unstable that clapping my hand or screaming loud near the speaker would make it start again..." \$\endgroup\$ Dec 22, 2018 at 8:50
  • 1
    \$\begingroup\$ I was thinking more along the lines of reducing the phase margin to create the instability that draws current with some hysteresis as well to cause the observed bistable instability \$\endgroup\$ Dec 22, 2018 at 8:55
  • 1
    \$\begingroup\$ Look at all the power snubbers with 10ohm RC’s to prevent oscillation with a gain BW product of 100MHz. WW Resistors. could wreak havoc. \$\endgroup\$ Dec 22, 2018 at 16:57

The TDA2002 and the improved TDA2003 are both obsolete and not made anymore. Then you bought counterfeit replacements? Why not use a modern IC instead?

Did you make the unstable amplifier on a breadboard? Amplifiers made on a breadboard usually oscillate.


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.