This goes for all speaker processing gain structure:
There MUST be headroom between the point that the amp limits (or clips) and the point that the weakest link in the drive electronics clips. This difference is called headroom and is the amplitude that the gain reduction of the limiter acts upon when entering limiting. If you want to insure that you can drive the system 6dB into limiting without the danger of clipping then you must have 6dB of drive headroom from which to subtract gain from. You must also be aware of processing anomolies where drive headroom may be taken away from in terms of algorithems and possible hysteresis in compression, and conversions where displays are not as accurate as you might think (this were 2 specific defects in the DSP I mentioned in post 14 and accounted for an ~4dB MEASURED error.
Also, it's critical to verify the numerical incications that the DSP provides and specifically what it's referenced to (dBu, dBv, dBm, dB full scale, or some arbitrary scale), and what the maximum I/O levels really are (the DR series is jumper selectable between +14 and +30+dBu AND the imput and output scaling are independant).
Generally, I recommend 12dB of drive headroom which allows for adequate limiting compliance as well as accommodating possible errors in level/scaling throughout the chain.
I think the DR product manager(s) got way too focused on maximizing the S/N ratio number and forgetting the importance of headroom under real world operating conditions. I don't know why, the product is plenty quiet to worry about specsmanship IMO.