Someone sent me a link this morning to this thread, suggesting I chime in.
First, thanks for the feedback. I agree this behavior is unexpected. Sorry in advance for the long message.
The relays are produced by Kemet, rated for 2A. We have used the exact same line of relays in other products for years in various circuits (gain control, muting, and system power). I was just as surprised as everyone else when the behavior appeared while prototyping Atom Amp, so I spent considerable time studying the circuit, trying different relays (all with the same behavior), reviewing application notes, and reaching out to the relay manufacturers themselves.
The circuit is simple: a constant 12VDC regulator feeds the SPST switch of the Alps potentiometer to make or break relay coil current. A diode protects the coil.
If you're reading this message board, I do not need to explain to you that relays are an old mechanical technology. We chose to use a relay to toggle amplifier power because doing so permits a front mounted switch. The same level of performance without a relay would have required placing a 2A DPDT switch on the rear side of the amp. Customers over the years have complained about rear mounted power switches. Thus, Atom Amp's power is relay toggled.
After a relay stuck once in prototyping, I handed the board to a coworker and asked him to switch the amp on/off repeatedly. He came back after 100+ tries, convinced I was crazy. Nevertheless, I spent weeks seeking to understand the behavior. Relay contacts can stick under strong magnetic fields, or due to micro-arcing from high currents through the contacts, or from a sudden bump (i.e., relay dropped or hit). I was already aware of these basic possibilities--we've observed gain and muting relays stuck in the wrong position in various boards over the years, always resolved by cycling the relay once.
Anyway, the relay controlling power in Atom Amp was not being hit or dropped, and there was no extreme magnetic field, so this left micro-arcing as a possibility. Micro-arcing happens when strong inrush currents cause the relay contacts to temporarily stick. Thus, I delayed release of Atom Amp by two months to implement soft-start circuits in the AC/DC regulation circuits. Atom Amp's +/-15VDC rails slowly ramp up from 0V to their final value, minimizing inrush currents.
Still, we've observed about 1 in 200 Atom Amp relays stick in final Q/C. The vast majority of Atom Amps never exhibit a stuck relay contact. For those that do, allowing the contacts to stick, then power cycling several times is generally all that is required to ensure that the power switch works as intended there on out. Once the contacts break-in, they tend to be reliable. I placed a tip in the Instructions PDF in case more break-in was required.
The Atom Amp on my desk was one with a stuck relay when first powered on. I've been trying all week to replicate the behavior. It only works as expected. The good news is, relay sticking is limited to initial use of a small number of amps.
To be clear, I do not intend to downplay or otherwise justify this behavior. Yes, the power relay can stick when Atom Amp is new. When a design has Achilles' heel, people focus on that.
Although 99.5% of customers never experience a stuck power switch from Atom Amps, I am actively seeking advice from outside engineers. It's a challenge, as a single prototype is unlikely to prove consistency of relay operation. Even 100 prototypes is not definitive. And such an enhancement must also not compromise Atom Amp's leading performance and FCC/CE compliance.
Thank you all for the support!