That's a good question. I'm not sure about the exact functioning of Dyno but the issue involves a filter jumping to about 14 orders of magnitude(!!!) over its expected amplitude in a extremely brief time. We would need a limiter with a attack of absolutely 0 and a infinity to 1 compression beyond a certain level (not just a very high ratio). Most limiters I know don't have those extremes. Normally it's assumed that input levels will be more or less reasonably set and such extreme response times are undesirable because of the effect they will have on the spectrum. Dyno does have a "ratio" setting but considering the really quite extreme magnitude of the signal to be corrected I would have doubts about our ability to correct it completely. We can't trust the volume will stay below the maximum value of a double-float because Nyquist allows for reconstructed waves that exceed the maximum value of a sample. Dividing that by the maximum float we can express still won't do and may still leave a amplitude that would be painful. 

What is needed here, IMHO, would be a combination of a normal limiter (with a high ratio beyond .9 amplitude or so) and a hard clip at 1. That still won't sound pleasant once things blow, but at least Tomasz's ears will live for another day and his partner and cat won't have to be startled by thrown headphones.

To do that we could use the tutorial that Stephen wrote for adding a hard-clip as a UGen to ChucK, either rushing that in as a patch or by Tomasz recompiling ChucK.

That would buy us some time to look into verifying the filters are stable at the set coefficients, at least for non-modulated filters. This is quite possible, I believe, and the current issue indicates that this isn't -yet- done or not correctly. We would still need to keep a clip at hand because of Perry's notes on modulated filters and their stability. 

I agree with Tomasz that this issue deserves priority because of the very real chance of physically harming users when combining sensitive headphones with the output jacks of MacBooks running OSX. I know we make no warranties but it would still be good to try to keep everyone healthy. I believe it's a good thing that Tomasz keeps addressing this because of that. It may be a rare scenario but I don't want to imagine combining this issue with a big PA and a audience accustomed to subtle sounds listening attentively.

Of course in a live situation a hardware limiter may be used. I'd have one for you to borrow for a while, Tomasz, if that would help.

Yours,

Kas.