Archived from groups: rec.audio.tech (
More info?)
Carlos wrote:
> According to Dick Pierce in a previous thread the maximum
> accuracy is achieved when the test box internal volume is
> close to the Vas of the driver being measured. I'm going
> to build one to measure a specific small driver with a
> spec'd Vas of 10 l, but would like to build something that
> I can also use for bigger drivers in future projects.
>
> So, my question is, how big can I build the test box and
> still get reasonably accurate measurements for a Vas of 10 l?
> By "reasonable accurate" I mean a worst case error of, say,
> 10%.
The reason a test box is used in the first place is that by
combining a known quantity with an unknown and measuring the
resulting quanity, you can derive the unknown. In this case,
we're combining a known compliance (the test box acoustic
compliance, which we'll call Cab) with and unknown acoustic
compliance (that of the driver, call it Cas), and measuring
the resulting total (call it Cat). We know that:
Cat = (Cas * Cab) / (Cas + Cat)
and thus we can derive:
Cas = (Cat * Cab) / (Cab - Cat)
(Remembering, of course, that Cat is derived implicitly from
the resonant frequency measured).
If the measurement procedure is such that the expected error is,
say, 5% of the larger of Cas or Cab, it becomes clear that to
minimize the effect of this error, we must make Cab and the
expected value of Cas as close as possible.
Let's see practically what this means. If Cas=Cab, the resulting
Cat will be half that of either, and the resulting resonant
frequency will be equal to sgrt(2) or 1.414 times that measured
in free air. Let's say you have a built-in +-1 Hz error in
measuring that resonant frequency, and let's say you're measuring
a driver whose resonant frequency is 25 Hz.
With equal volumes (ideally), the new resonant frequency will
be 35.35 Hz, but you might measure anything from 34 to 36, or
about -3% to +2%. Since Cat goes as the square of the resonant
frequency, that translates to an error of -7.5% to +4%.
Now, let's do the same with a box not the same size but 10 times
the size. Now, the expected resonant frequency is not 35.35 Hz,
but sqrt(1.1), or 26.22 Hz, with the error, from 25 to 27. That's
an error of -5% to +3%, resulting in an error in the derived Cat
of -10% to +6%.
Now, this might seem acceptable, but this is just ONE source of
the error and not the biggest one, either (but it is the easiest
to identify and quantify). How big is your test volume? Really?
Are you SURE?
Reasons your test volume MAY NOT be the size you think it is is
that ANY leaks will change how the volume behaves and could
significantly skew the results. Your test colume is acutally the
volume of the box AND the volume included in the side of the
driver facing into the box. How well did you measure that?
The solution is to have a range of test volumes, suited to
different ranges of compliances. And since, for woofers as an
example, the expected compliance is a rough function of diameter,
you build a test box for 4" drivers (with a volume of about 6
liters), for 5" (with a volume of about 12), for 6" (with a volume
of about 25) for 8" (with a volume of about 40) and so on.
Make each on a little on the large size: you can through in a
block of wood to cut it down. Indeed, if you want, make a 100
lieter box with interchangeable mounting panels and have a bunch
of bricks sitting around to take up the volume. This only works
so well, because if you end up with MOST of the volume taken
up by bricks, frictional losses can start affecting accuracy.
Frankly, I found it more reliable and easier just to have
purpose-built boxes with purpose-cut holes on the front.
Facebook
Twitter
Instagram