Here is a picture of the front stage in my living room as of August 2016:
And here it is with the TV moved off the wall and the new center channel installed:
Here are some pictures of the rear of the room as of September 2016:
Regrettably, I did not take a picture of the front before I replaced my old cabinet with the racks or the front left/right speakers Hsus with my SEOS-15 TD12M prototypes.
In case anyone is wondering, yes, this space is very acoustically challenging, but it has its upsides.
Speaker / Room Calibration
Through the years I have experimented with a variety of methods of speaker and room EQ. In my current approach, I mostly ignore the room, except for the subs, and I attempt to primarily correct the first arrival of sound. I aim to make the first arrival *mostly* flat. To analyze only the first arrival, I use the frequency dependent window feature REW with a crude 1/3rd octave resolution. This is roughly consistent with the bandwidths of the ear/brains filters. It's long enough to allow for delays at the crossovers but not so long as to allow inclusion of room reflections that the ear/brain could theoretically isolate. Note that this particular FDW length may not work well with other speakers or other rooms. YMMV.
I say I aim for "mostly flat" because I use a bit of sub bass boost (boost of the bass in the first arrival, actually) and because I use configurable high frequency adjustments to optimize playback for different content. Because the HF adjustments are configurable, I leave them disabled (flat in the treble) in the measurements below, but these measurements *do* include the sub bass boost. A flat-ish first arrival in a listening room that's not acoustically dead tends to result in a frequency response that is sloped up toward the bottom end. Below is my frequency response for left, right, and center with 1/3rd octave-smoothing:
Note that the responses look rather messy and exhibit about 8 dB difference between bass and treble. Note that with high frequency adjustments, the difference tends to come in closer to -10 dB @ 10 kHz vs. 20 Hz. Note also that the center channel response appears very hot in the mid/upper bass. This is because because of speaker/room interface issues that I currently have. The speaker is very close to the bare wall behind it, and this causes most of the sound of those frequencies to disperse diagonally and to the sides instead of forward. As such, the first arrival is relative lean. Now look at the first arrival responses, using FDW in REW with 1/3rd octave resolution:
These responses appear rolled-off at the ends because these frequencies are somewhat delayed (ends of the bandwidth) and escape the influence of the filter. Apart from this, they are very smooth and flat, except for the aforementioned sub bass bump. The center is a bit messier in the mid/upper bass again. I could have made it ruler flat at the MLP here, but I opt instead to compromise a bit for off-center seats. Indeed, think this may sound more balanced, to the extent that the brain uses information from early reflections to ascertain the content of these lower frequencies. I'm not exactly sure at this point.
The configurable high frequency adjustments are of three different kinds: distance compensation, upper mid-range tilt, and X curve. The measurements below depict the effects on the high frequency response of my center channel from using -0.75 dB of upper-mid tilt (blue line) and several degrees of distance compensation (red lines) on the high frequency response of my center channel:
Distance compensation mostly affects frequencies above 8 kHz or so. Almost all recordings need this compensation, in part because high frequency sound naturally rolls off with distance. I chose to model this roll-off on actual physical data of distance roll-off, and it seems to be working out quite well. There is no standardized distance for monitoring music or theatrical recordings, and furthermore, a lot of music is monitored on near-field monitors that implement some kind of roll-off of their own. The consequence is that the optimal roll-off varies substantially by content.
For mid range tilt, I currently use a Q 0.5 @ 2 kHz high shelf filter with variable gain. I'm not sure this is really optimal, but it works pretty well for me. Most stuff sounds good either flat or with a tiny bit of treble reduction. Rarely do I need more than -1 dB. I'm not sure why this is necessary other than perhaps precedence. Monitors in days past likely had just a bit of tilt, and the precedent stuck. A lot of monitors include a HF trim switch that engages a tilt like this. Of course, every design is slightly different, and many include additional > 8 kHz roll-off here as well.
Lastly, I have X curve correction, which is used only rarely for movies that need it. Movies mixed on a dub-stage tend to use speakers behind an acoustically transparent screen, and this almost always introduces high frequency roll-off of some kind. Instead of compensating by making adjustments to the monitoring system, tended to boost treble in the mix to compensate. The X curve was intended to standardize the high frequency roll-off in the playback system so that mixes would translate, and it was based on the roll-off that was commonly observed at the time. (IMO, it largely failed because it measured pink noise energy response instead of first arrival SPL.) AIUI, dub stages largely ignore the X curve these days opting instead to simply accept the faults of the screen but also avoid hyping the high frequencies too much with the understanding that doing so may make the mix way too bright in some settings. In any case many movies, particularly stuff from the 90s and older, do benefit from some additional high frequency reduction, and the X curve shape is probably the most generically useful.
Click this image. It is an animation of the box assembly sequence:
The cabinets are separate, and I'm building a left and right pair. The other bass box is visible in the background, surrounded by a bunch of clamps.
On the bottom is an AE TD12M-4A in a 1.5ish cuft box with an aggressively flared 4 inch port. The cabinet is 1/2" B/BB plywood with double thick front baffle and pairs of shelf style braces running horizontally and vertically (4 braces total). The design is somewhat noteworthy for being under 9" deep. On the top is an SEOS-15 horn with a Denovo DNA-360 mounted on it. The horn cabinet is 1/2" B/BB and is not braced. It is slightly wider and deeper than the bass cabinet, partly because I started them before I started the bass cabinets and changed my mind and opted to make the bass cabinets smaller.
These are currently up and running with fully active crossovers and EQ provided by a prototype version of my custom DSP software, running on a Linux PC with a Motu A16 interface connected via USB. Each mains driver is powered by a channel from an Emotiva XPA-5. The amp is rated for 200W/300W into 8/4 ohm with all channels driven, but I expect to "effectively" get 500-700W peaks for the 60-120 Hz range due to the resonance above the tuning frequency, the very low DC resistance of 3.2 ohm, and the amp's beefy power supply that'll be hardly tapped to run the horns.
So far, I am very pleased with the performance of these speakers and active crossover. Here is the a recent measurement of the response (no gate, 1/6th octave smoothing) of each driver playing separately and in unison along with the subs.
(Note that I'm tweaking the response all the time. I'll try to keep this image current, but it probably won't work out that way.)
This measurement was taken with a -30 dBFS RMS sine sweep and a calibrated mic. The pink noise calibration signal (-30 dBFS RMS pink noise band-limited to 500-2000 Hz) reads at about 73.75 dB. I zoomed in on the data a lot, which makes the response look ugly, but in reality I'm +/-4 dB or so with 1/6th octave smoothing everywhere except for the big dip at 165 Hz. Even though my room has some treatments, I still have some significant acoustical problems involving the floor and ceiling, which accounts for the unevenness. I could have EQed much of that out, but for the time being, I'm concentrating on EQing the speaker response and avoiding acoustics, and I have purposely slanted the response through the mid-range to obtain a flatter power response.
The crossover filters, 2nd order HPF for the horns and 4th order LPF for the woofers, are centered at 950 Hz. Note that the woofer and horn aren't completely in-phase here. This is an experiment, in which I have purposely mis-aligned the drivers in time to try to steer the lobe above or below the MLP, putting the MLP closer to the edge of the lobe. Then, the response must be boosted to avoid a dip, and this increases power response too. That's a good thing because directivity typically goes through a maximum where two drivers are crossing over. The idea here is to avoid that directivity peak to achieve both smooth direct response and smooth power response through the crossover region. My measurements suggests that the strategy works, but it's not clear how this affects the sound at off-axis locations that may be at a slightly different vertical angle.
The cross to the subs is at 110 Hz for now because that works best with the filters on the subs as they are now. In the long run, they will be blended with the subs over a wide frequency range to optimize multi-seat response. This means, I am also planning to build subs that can player higher than the traditional 80-100 Hz as well.
This is the first part of a larger project to upgrade all the speakers and DSP in my system to fully DIY-built solutions. I have started building a center channel of identical design, and I have drivers for new subs and surrounds drivers on order as well.
Edit: Added details about speaker/room EQ configuration and in-room response.
Edited by SME, 26 January 2017 - 11:16 AM.