Are Re1 and Re2 Always Equal?

Good catch — and the honest answer is: conventionally yes, but not strictly required.

Why Re1 = Re2 is the Standard Assumption

In a perfectly complementary push-pull stage, the circuit is topologically symmetric around the output node. Setting Re1 = Re2 ensures:

• Both half-cycles see identical source impedance → no even-order harmonic distortion
• Equal quiescent dissipation between Q1 and Q2 → thermal balance
• No DC offset current at the output node at idle

This is why virtually every textbook and reference design (LM3886, TDA7293 datasheet) specifies identical emitter resistors. The optimizer treats them as a single shared variable for this reason.

When Re1 ≠ Re2 is Actually Used

In practice, real NPN and PNP transistors are never perfectly complementary:

• Their β (current gain) values differ
• Their V_BE temperature coefficients differ slightly
• High-power PNP devices (e.g. MJ15004) often have lower β than their NPN counterparts (MJ15003)

In those cases, a designer might deliberately mismatch Re1 and Re2 by a small amount to compensate and null the DC offset — essentially using Re as a trim. Some high-end audiophile designs do this empirically on the bench with a trimmer pot.

What This Means for the Optimizer

The current A* treats Re as one shared variable, which is the correct starting point. To handle asymmetric pairs properly, you'd extend the state space to two independent dimensions — Re1 ∈ E12 and Re2 ∈ E12 — and add a DC offset cost term to the objective:

cDCoffset = k · (Re1·β_P - Re2·β_N)²

The state space grows from 3,125 to 15,625 states, but A* still handles it efficiently due to the admissible heuristic pruning most of the search tree.