**UPDATED 11/03/04 with distortion measurements at the bottom of the page. Also shown is a family of FR curves at different drive levels to see if any FR changes occur.**

Introduction

This project was developed to fit a need for small, ceiling mounted satellites to be used in my family room for home theater and general music usage.  Now, "small" is a relative term.  It doesn't bother me to have larger speakers hanging off the ceiling, but most folks think of wall speakers as in-wall or bose sized cubes.  One day, I asked my wife about it, and surprisingly she said it would be ok if the ceiling speakers looked nice and were no bigger than a pair of old Paradigm titans I had lying around. This was fantastic as it's not possible to design a legitimate home theater satellite with a  woofer smaller than 5.5 to 7 inches in diameter.

So there was the genesis of the project. Design as good of a unit as possible given a box size somewhere around the size of the Paradigm atom or titan. Or, to put it another way, "how to cram the largest 7" midwoofer into the smallest sealed box."  Since it was to be used with HT, it didn't need to go lower than 60-80 Hz. I looked a fair number of woofers in both sealed and ported configurations, and the configuration that looked most promising for the midwoofer was the Usher 8945. It modeled with an F3 of about 70Hz in a sealed box of about 7 liters. It also matches well with the Usher 9950 tweeter, which is an exceptional tweeter, especially above 3k, according to my measurements. And, at the time I was beginning this project, one of the first distributors of Usher here in the US, Euphase, was offering an introductory special on Usher. I purchased five woofers and tweeters for around $550. I didn't quite know it at the time-it was a bit of a gamble since Usher was relatively unknown-but that turned out to be a very good deal.  Usher is most prominently carried by PE and Zalytron.  It's still a very good deal, although the price has gone up. 

I don't want to get too far into the debate on whether or not these are "Scan Speak Clones." They are very similar in appearance, but also in performance. And the Usher units are favorably priced, compared with their more well known siblings. In my case, for a small sealed box, the Usher TS parameters were better suited than the SS8545.

Box design

The drivers are on the centerline vertically. The distance from the top edge to the tweeter flange is 1/4 inch.  The distance from the bottom edge to the woofer flange is 1/2 inch. At some point I will start using a different box program. My cabinets have 3/4 mdf on the front and 1/2 on the sides, for weight savings, but you can use 3/4" all around and extend the back to keep the same volume. If you want to know by how much, download the BoxyCAD spreadsheet from www.audiodiycentral.com  For now, you'll have to get out your ruler, but everything you need to know is here. Of note, the diagrams above are from BoxyCAD.

The crossover

The crossover modeling was done with lspCAD. The final schematics, after a couple of iterations and a cabinet redesign, are shown below.

The modeled response and null.

The crossover frequency is 2.2k. The topology is standard 4rth order LR.  Note the crossover is really quite simple simple except for the notch filter in the woofer circuit. I recommend not using an air core. It would be very expensive and is really not necessary.

Measurements

The overall frequency response is shown below.

The overall response is very respectably smooth given that there is no edge treatment to reduce diffraction and the edges are "lit up" quite a bit since the baffle is so small. Under the circumstances +/-2.5 dB is about as good as you'll see. There is about 4 dB of bafflestep compensation, give or take about one dB or so of uncertainty. The highs are not rolled off, despite the well preserved high frequency extension, as the Usher 9950 is a very smooth sounding tweeter and really doesn't need it. (Note the curve extends to 30k, not the usual 20k.) Despite the presence of a notch filter, the pesky hump around 850 Hz still shows up in a muted fashion.

Note the actual sensitivity of the unit is around 83dB/2.83v. Praxis normalizes the measurement to 1v and the measurement distance was about 1.6m, so add approximately 13dB to the above plots.

The horizontal off axis plots are shown below

The vertical axis plots are shown below.

This deserves a bit more comment.  These units were designed with the design axis approximately 10 degrees below horizontal. That is, they should be pointed ABOVE your head by about 10 degrees. This explains  the difference in the nulls in the graph. In a design that has the zero delay plane and the design axis on the horizontal axis, the nulls will occur approximately 30 degrees above and below.  By trying to get on the zero delay axis, which is somewhere around 10 degrees below the axis, everything is shifted down by about 10 degrees.  So, above the horizontal axis the null becomes prominent at around 15 degrees while below the same degree of prominence is not present even at 25 degrees. If you don't catch all this, don't worry. Just remember that optimally, the listening axis should be below the horizontal axis by about 10 degrees, i.e. tilt the speaker up a little.

Here's the impedance graph.

Let's call it a nominal 4 ohm unit.  The F3 is around 65 Hz. The box Q with heavy stuffing is about 0.65, according to the TS curve fitting routine in Praxis.

Distortion measurements

I spent a bit of time doing some distortion measurements to see how loud this speaker could go and to see where was its weakness. 

First, a single 100 Hz tone.  This was adjusted until very minimal distortion was heard by me. Note this is very subtle, just a bit of a change, a bit of higher order products creeping in.

Below is a 3 tone signal at 100 Hz.  Note that the SPL level has dropped, but not by much.

Next a 2 tone signal to look at IM products.  Note the drive level here is notably higher and the corresponding 2nd and 3rd order products from the woofer dominate. There is some IM, but for the most part it's down about 60dB.

Lastly, a 1.5 kHz pure tone followed by an SL style 3 tone test.

I would have liked to do some more tests at additional frequencies, but ran out of time. However, I think these plots capture the flavor of the unit at moderately high SPL's. The tests are done in my garage and so average in some room gain and reverberant soundfield. You can see though, you probably won't get reference levels with -40 dB or better distortion products. In the 80 Hz to low 100 Hz region, you won't get more than 94 dB SPL before getting 10% distortion. As far as I'm concerned, this is audible and really, for music, mid 80 dB SPL level is realistically what you'll get to keep the distortion under 1% in this range. As far as IM and higher up at 1.5k, the unit does better.

These levels are somewhat relative, and it's probably best to compare these numbers with the numbers of similar units. I consider the Usher woofer to be an excellent unit and this unit to have overall good distortion. But you probably can conclude that a sealed 7" unit, even a high quality one crossed at 80 Hz  to a sub cannot be considered truly full range and won't achieve reference levels.

There's something to be said for MTM's, TMM's, and three ways for dynamic range. 

Addendum part two. 

FR changes with drive level. For what it's worth.

And a magnified view around the crossover.

Listening impressions and final comments

These turned out to be great units. They have a very clean, neutral sound. The highs are very clean, but not falsely detailed. Currently, I have 5 of these in my HT setup matched up with a Denon 3803. Prior to this I was using some older Paradigm CC and Titans. The difference is astonishing. I no longer have any dialogue problems and they can be played very close to reference levels cleanly. No 7" based MT can truly play at reference levels, but these come very close. The cost for five of these is around $1000 currently. That may seem steep; this in not a budget system. On the flip side, there is nothing available commercially for anywhere close to this.

Now there are some caveats to this project.  The units are not shielded since they are on my ceiling far away from the television, which is an RPTV anyway. However, with some effort and crossover reworking, the 9950C and 8836AC2 could probably be used.  The truncated flange is a bit of a hassle, but otherwise this looks promising.

They really are best as a high quality alternative to in-wall speakers or the plethora of tiny plastic satellites available at most local audio-video retailers.

Feel free to email me with any questions or comments

mark k