eWhammy: Last-Minute Refinements

OK, I’m posting this because I’m all for complete transparency.

As we were about to ship, we did a final round of heavy testing and found something we couldn’t ignore: the mass-produced spring plate had a bit more hysteresis than our original prototype. Bummer, yes. But don’t panic. This story has a good ending.

You see, hysteresis is the small return-to-center “gap” when a spring doesn’t land in exactly the same spot after every bend-and-release. It comes from energy lost inside the metal plus tiny stick-slip at contact points (clamps, stacks, edges).

We had to do something. Otherwise, I was inclined to call off the release.

The first fix was obvious: soften the edge. The right geometry, especially around the clamp transition, keeps that gap small. Adding a 3 mm fillet to the top and bottom clamp plates reduced hysteresis by about 67%. Our prototype was already decent at around 3%, while the first production plates came in at a ghastly 10%, which wasn’t acceptable. With the fillet in place, both improved significantly: the prototype dropped to 2.35% (best case), and the production plates came down to 3.28% (best case).

The other fix, of course, is to widen the deadband to cover whatever hysteresis remains. Push it too far, though, and the center starts to feel sticky and unresponsive. My workaround was to bypass the deadband, but only for the “wiggles”: I pass only the higher-frequency movement (about 1 Hz and up) straight through, so small vibrato gestures still come through naturally even while the center stays locked.

This works so well that the deadband control is almost unnecessary. As it stands, the deadband covers up to 5% hysteresis, which gives comfortable headroom over the 3.28% we measured. A little margin never hurts. The high-pass bypass handles the small wiggles, while the deadband circuitry takes care of the deeper dives and aggressive pulls.

The Mod

The Hall-effect preamp output feeds the output amplifier’s non-inverting (+) input through a high-pass filter (C2 and R17). We’ve now fixed the deadband at ±2.5%. For this batch, we’re simply hacking the existing boards to add the HP filter (see image on the left). A bit of SMD hacking goes a long way.

Takeaways:

1. It’s always challenging when exploring uncharted territory. Even a seemingly simple sourcing decision can lead to unexpected results.
2. We’ll have to rework the current production PCBs and clamp plates to incorporate the necessary changes, which is fine for this small batch.
3. We’ll probably have to find a better supplier for future batches, hence more sampling and testing. The annoying part is MOQ: we’ve ordered a lot for this batch, and we may have to scrap the remainder.
4. The spring plate, like traditional trem springs, should be replaceable, and it is. Over time, it may sag or have degraded return-to-center. As soon as we find better spring plates, we’ll send the first batch a replacement plate for free; you just cover shipping (via regular post).

Bottom line: the first batch is covered, and future batches will be tighter.