Now, it's time to pick a size of the sealed box.
Sealed boxes have the happy case that you can change the alignment
of a sealed box continuously simply by changing the box size. Vented
boxes are much tougher to handle this way - a vented box works best
at discrete sizes based on the alignment you want.
Further, you can change the apparent size of a sealed
box by the amount of stuffing that goes inside of it. You need some
stuffing to reduce internal higher frequency reflections and too
much stuffing will dampen the driver noticeably so there is a tradeoff
but around 50% stuffing works well. This increases the apparent
box size by as much as 20-25% and improves the sound quality as
well.
Here's a chart of the three most common sealed box
alignments.

The black lines are the chebychev (Q=1) alignment,
the red lines are butterworth (Q=0.7) and the blue lines are bessel
(Q=0.57). As the Q increases you can see that the frequency response
get a bigger hump around resonance, the resonance increases, and
the low bass decreases. Also the low-end cone excursion decreases,
the transient response worsens and the group delay worsens.
Many people thus prefer a low Q alignment sonically
at the expense of the bottom end (not deep deep bass but the low
end of the bass spectrum). For myself, I'm worried about box size
and low end more than anything else. At these frequencies a lot
of the other effects are inaudible.
For the NHT woofer we find that a bessel alignment
is 35 cu ft, the butterworth alignment is 12.5 cu ft, and the chebychev
alignment is 3.8 cu ft. I think chebychev is too high a Q (I don't
like the effect on transient response, in particular) so I'm planning
on a Q of closer to .9. This produces a box size of 5 cu ft and
a resonance of 32 Hz with a 3db down point at 26 Hz. This all sounds
good and I'm actually planning a box of around 4 cu ft, stuffed
to produce a Q near 0.9. Spec shows that an empty box of 4 cu ft
will produce Q of .97 with a resonance around 35 Hz and a 3db down
point of about 28Hz so if my stuffing is less than expected the
results will still be pretty good. |