Speaker frequency response at sitting position

[glynnw]

While not exactly a Bottlehead product topic, I know that at least here I will not immediately be mocked by some "expert" out there in internet land, as seems to happen whenever I post on the other forums.  I recently used a simple 10 band sound level meter to match the output of the subs to my speakers.  I noticed that while the tweeter output level looked fine when the microphone was  just a couple of feet in front of the speaker, when I placed the mic at my sitting position, about 10 feet from the speaker, the tweeter trailed off to about 10 dB below the midrange.  This was on my Basszilla speakers, with ribbon tweeter.  I found the same result with a small Wharfedale speaker with a dome tweeter.  Changing from the Paramounts to a gainclone amp got the same results.  Next I removed most of the sound absorbing panels from the walls and still had the same result.  Boosting the treble with the graphic equalizer in the J River player until it showed a flat response on the spl display resulted in a different, but not a better, sound.  Finally I installed a couple of resistors in an L-pad on the midrange and lowered it a couple of db, incurring a very small change, but to cut the midrange enough to match the tweeter by this method would render the speakers too inefficient for my 8 watts.

So my question is - do almost all speakers trail off the treble with distance?  My experience with the digital equalizer showed me that some aspects of the sound could be improved, but it would take a much better toy than the one in J River to accomplish this.  Any suggestions, comments, or good jokes?

 

[Paul Joppa]

When you are more than a few feet from the speaker, the sound is dominated by the reverberant field. Close enough to the speaker, it's the direct field that dominates. Too close to the speaker, near-field effects distort the measurement. That's probably why you see that difference - the tweeter is more directional (especially vertically, if it's a ribbon) so when it is flat on-axis, the power it puts into the reverberant field is smaller. Plus, almost everything in the room absorbs high-frequency sound better than mid-frequency sound, so the reverberant field is reduced more at high frequencies.

Now you know why we measure speakers outdoors whenever we can - no reverberant field!

In practical terms, the first-arrival sound dominates the perceived timbre, so your near-field measurement is probably the most useful for high-frequency performance. More specialized test gear will take data from a restricted time window, and can isolate direct sound from reflected (i.e. reverberant) sound, but at the price of reduced frequency resolution. It can get to be a real can of worms!

Incidentally, if you do need more reverberant high frequency energy, turning up a directional tweeter is not the way to go - you get bleeding ears on-axis. Some systems use an auxiliary tweeter, pointed away from the listening area.

 

[Grainger49]

I would say give the speakers toe in if they are not already.  The high frequency roll off that PJ mentions is just reality.  Toe in is one way to combat it but the highs fall with the square of the distance regardless. 

This would make you want to listen in near field if you still have a good HF response in your ears.

 

[Doc B]

FWIW when we are stuck with in room measurements instead of outdoors, I usually use a 1 meter distance to adjust the tweeters and then we adjust the mid/woofer balance at 1M and check again at the listening position to see what the room is doing to the bass balance. Sometimes one needs to tweak that mid/bass balance at the listening position by moving the speakers or adjusting the crossover. But you can usually be pretty confident about your 1M measurements giving you a good idea of what you are hearing at the listening position with respect to the mid/treble balance.

 

[glynnw]

Thanks for all the input.  I have ordered a pair of Selenium tweeters (as I used in the Straight 8) and will modify them as per your past instructions and then experiment substituting them for the Aurum Cantus ribbons or even using them as rear firing 2nd tweeters.  With the high efficiency of the Seleniums I can easily match the output of the main Fostex driver in the speaker.  Since the Fostex is used as an open baffle diver in these speakers, the rear firing may be the best answer.  At the very least I now have a new fun experiment going on. Whoever talks about retirement being boring isn't a DIY addict.

Thanks, as always.

 

[ironbut]

One thing that I seldom see mentioned is the "tilt" of the speakers.
If you're finding that the sound at your listening seat is a tad dull, and your speakers baffle extends above the level of your ears, try putting some wedges under the speaker at the front or back. Start out kinda small (1/8-1/4") and increase if necessary. One you find the best angle, you can replace the wedges with something more permanent.
I have Magnapans and tilting them really helps but I would think that an IB like the Basszilla could benefit from it too.

 

[Grainger49]

My suggestion for the rear firing tweeter is to wire it out of phase with the front firing tweeter.  This is a trick that Infinity used in the 1001 to simulate a single diaphragm with an open back.  Like electrostatics and Magneplanars.

If it sounds odd swap the phase on the rear tweeter and see if it sounds better.

 

[richardl]

A trick I used was to measure the speakers at 50", so I could compare it to the data that John Atkinson publishes in Stereophile.  It is convenient to have a similar style graph to compare to. That way the variance in drivers doesn't fool you and you can check to see if your xover is wired correctly.  FWIW, when I balance speaker/subs, I do it at the chair and try to get the mids to match the bass.

For various reasons, in-room treble measurements are not too easy at larger distances.  Treble rolloff is the norm.  You can easily see this in the graphs Atkinson publishes that are taken at the listening chair, again a source of comparison.

Good luck, Richard