This video at the end (from EEVblog).

I am kind of lost. Why does that blip in the waveform happen?

 

Could YOU say at what point in time it happens? 49 minutes watching to find your blip?

The end here it is too wide.

 

At 48:30

 

Looks like crossover distortion.

 

The blips are due to the lower input impedance of the 1K, 1oK resistors, causing higher current on the output stage makes the crossover distortion more visible, higher values increase the input resistance and lowers the output current.

 

That's crossover distortion, caused by the opamp's class AB output stage which operates class A for small output signal currents, and class B (to reduce DC supply current) for large output signal currents. So when Dave reduced the circuit resistances, drawing more current from the output, the resulting "gap" between current sourcing and sinking suddenly became visible. The LM324 (the quad version of the LM358 dual) is the same way, and that gap is the reason you don't see either of these opamps used in hi-quality audio designs.

But the crossover distortion can be dealt with, to a degree. From the original National Semiconductor datasheet for the LM124:

To reduce the power supply drain, the amplifiers have a class A output stage for small signal levels which converts to class B in a large signal mode. This allows the amplifiers to both source and sink large output currents. Therefore both NPN and PNP external current boost transistors can be used to extend the power capability of the basic amplifiers. The output voltage needs to raise approximately 1 diode drop above ground to bias the on-chip vertical PNP transistor for output current sinking applications. For ac applications, where the load is capacitively coupled to the output of the amplifier, a resistor should be used, from the output of the amplifier to ground to increase the class A bias current and prevent crossover distortion.

 

IIRC for the LM358 there was a reccommendation to use a resistor in parallel with the output returned to -V, I think. Sorry, I cannot recall properly.

 

That is correct, and a common trick even with high quality "audio" opamps from Burr-Brown, etc. The resistor always goes from the opamp's output to the opamp's negative supply voltage pin, whether it is GND or an actual negative voltage. Figure out the maximum load current at the negative signal peaks, size the resistor so the current through it is about 1 mA greater, and check that the opamp can supply the total load+resistor current when the output is at it's positive signal peak to make sure you're not overheating anything.

ak