Subwoofer Project - Page Three
Performance
Ok, this is what everyone came for. How does it sound? Before I answer that, I need to preface this with a couple of disclaimers. First off, this sub, and my previous Sunfire, are the only subs I've ever heard in my house. I can't provide subjective comparisons to a wide range of high quality subs. Also, I don't have access to any test equipment beyond a simple Radio Shack sound level meter. And with subwoofers, the room and placement plays a HUGE role in their overall performance. My den is about 13.5' x 14.5', but there is an opening into my living room, which is probably another 15x20' or so, which then opens into the 6x20' or so hallway, which opens to a decent sized kitchen, and dining room. So it's immediate space is fairly small, but it's not closed off from the rest of the roughly 1400 square feet of house, all with 10' ceilings. It's all about cubic feet, but it's somewhat unclear how the space should be counted.
That being said, at this point in time, I've been listening to it for about a week, and subjectively, I'm VERY happy with the results. Unlike the Sunfire, there is no noticeable noise or distortion when playing ultra low frequencies. It also handles quick transients much MUCH better than the Sunfire. Everything I read suggested that the Rythmik servos are very "musical", which when most people describe, I assume they mean accurate and articulate, while not necessarily implying anything about its brute output. And I completely agree. In this respect, it is a very big improvement over the Sunfire. I still have plenty of fine tuning to do as far as equalization goes, but so far, it's appearing to be a very worthwhile improvement.
While I was expecting the extra accuracy over the Sunfire, I was pleasantly surprised by the extra output I'm getting. I had assumed that I'd need two kits to ensure that I wasn't taking a step backwards in output, and it seems I was very wrong. This is the area that I'm at least able to offer SOME objective data on. I did some very quick spot measurements before taking down the Sunfire, then again on the new sub when I put it up. I don't have any way of taking measurements based on a THD limit, so I just ran them as hard as it sounded like they could be pushed (be very careful with sine wave testing - I used only very short durations, with pauses in between each tone). At 20 and 25hz , I'm getting 10db more output than the Sunfire. At 15hz, that margin widens to 30db, and at 10hz, a comparison couldn't even be made, because the Sunfire didn't even register on my meter. With the Rythmik servo, I'm getting over 97db at 10hz! Obviously, this brings movie performance to a whole new level. Below are some frequency plots I've taken at a few different levels:
There are a couple of mistakes I made when making those plots: First, I use a BFD for equalizing my sub response in my room. I had intended on bypassing it while taking the measurements, but I forgot. So far, I've only implemented a couple of filters: a 6db cut @ 40hz to tame a large peak I have in my room at that frequency, and also a 2db boost at 25hz I used to flatten the response some. Both of these filters are less than 1/3 octave wide. Obviously, the 40hz cut was a bit conservative, leaving some of the natural peak there. As you can see, I've not yet done anything to work on the erratic response above that point. Once we get to 40hz, I start running in to some nasty room peaks/valleys. I'm considering a second sub in a different location, and/or some room treatment, to help tame those before I resort to brute force equalization. I also have some erratic response below 17hz. I doubt I'll be able to do anything about those. I suspect that perhaps the small room dimensions are causing the peaks/valleys in the upper range, and possibly the peaks valleys in the lower range are influenced by the larger dimensions of the entire house, which the den is effectively open to through it's doorless opening.
Another correction I forgot to make when taking these measurements: I forgot that the last time I used my DVD player was to play an SACD, so my receiver was using the multi-channel input. This means the crossover was determined by the DVD player, not my receiver. I use an Onkyo DV-SP502, which is based on the Pioneer 578/588, so I THINK its (non-adjustable) built in crossover is at (a too high) 120hz anyway. But it should be noted that this response does differ in the upper end from a previous plot I had done using the receiver's bass management. Ideally, I'd connect the DVD player's left or right output directly to the receivers sub input, and disable bass management, to ensure that no crossovers are affecting the curves, but I'm too lazy to pull my armoire out and temporarily re-route the wiring.
Also, I took these measurements starting from the highest frequencies, then working down, knowing that at the higher volume levels and lower frequencies, I'd hit the excursion limits of the drivers once I got down to a certain frequency. I reached that point when measuring the top curve shown here when I hit 16.8hz. So for the measurements at 16.8hz and down, I started adjusting the volume downwards to prevent bottoming the drivers. The Rythmik amps have a 2nd order high pass filter at 14hz anyway, so I didn't have to do that TOO much, but the response below that point is not linear due to that. But for 16.8hz and below, the yellow line represents maximum output, not linear output. For that reason, I chose to not do a fourth curve at a higher level.
Because of the issues mentioned above, this graph should only be used to analyze compression. In my room, I can maintain 105db down to around the high pass filter point of 14hz. While I haven't done full plots at this level, some quick tests suggest I can maintain 110db and greater above 25hz.
UPDATE 1-28-06
I've done many more measurements, as well as some more calibrating and EQ'ing. Below is a screenshot of a spreadsheet I used to predict some values for EQ'ing some of the room issues I have:
The dark line is the response I measured in my room. The pinkish line shows the projected response after adjusting the filters, while the green line shows the equalization curve. At this point, I only have three filters entered. Sine that time, I've fine tuned them further, and also created some new test tones and spread sheets to get better measurements. I'm now testing with 1/12 octave test tones from 5-320hz, to allow me to see the response above and below the sub's typical range. I then did some tests with the sub essentially bypassing any crossovers inherent to my DVD player or receiver. I set all the speakers to Large on the DVD player, and turned off it's Sub output. This will pass the full signal from my test CD out the L & R outputs, one of which I connected to the sub input on my receiver. While the signal is still passed through my receiver and BFD, the crossovers should be bypassed, and the BFD can also be put in bypass mode. Following is a comparison of measured response with all BFD filters bypassed (raw sub performance in my room), as well as with my EQ settings implemented. Those first two measurements are done JUST with the sub. Note that the sub amplifiers have a built in crossover that I can't bypass, so there is still high end roll-off. I have it set as high as possible (which is 160hz, if I remember correctly). After taking those two measurements, I then connected my DVD player back up to the proper inputs, and took too more: one in conjunction with the center channel (standard DPL-II mode), and another in pure stereo mode (sub and L+R). Both of those measurements were taken with the BFD filters in place:
I re-adjusted my volume setting for the DPL-II (center channel) test run to match the the runs where the DVD L output was going straight to my sub out, but after that run, I did not re-adjust for the stereo mode (since no change was made to the input). Strangely, stereo mode ran with much more output, presumably due to some level calibration inherent to the Dolby processing. Strangely, the Dolby processing also seemed to influence the sub performance below 10hz, as the curves for all the other runs were essentially similar (adjusting for the higher gain of stereo mode). The blue curve is the only one that bypasses the BFD, so the effects of the equalization can be seen. The BFD does a good job of smoothing out the peaks/valleys I have from 40-55hz. Above that, I'm going to need room treatment. The bulk of the problems I have above that are within the realm of main speakers, so the BFD can't help. Furthermore, we can see the peaks/valleys vary greatly depending on what speakers are being used (center vs. L+R). After seeing how much the response varies in the upper frequencies, I decided it would be wise to compare the response from various seats in my room. Following is a picture of the "seating end" of the room:
Obviously that was taken long before I got a good programmable remote. Now there is only one. :-) The end table in the corner is exactly opposite the corner the armoire and sub are in. Therefore, the two "best seats" are the two closest to the end table. I've taken measurements for five locations: three for the sofa on the left, and two for the loveseat on the right. From left to right, the seat numbers are: 5, 3, 1, 2, 4 (numbered from the "best" seat to the worst, with the dog obviously taking the best seat, #1). Here are the results from each seat:
Each seat has its own color in the graph. Up until this point, all measurements were taken from a point in between seat #1 and #2. The above test made it apparent that the response varies greatly from seat to seat, and EQ'ing based on that one spot wouldn't be the best compromise. I therefore added the white line, which is the average of all five seats. Some changes would have to be made in my filters for a better compromise between all seats, but It's also obvious that seat #5 was the most unlike the other seats, and skewing the average curve. I therefore decided to add another average that excluded seat #5 (the black line), since that is the "worst" seat anyway. It began to appear that the average of seats 1-4 most closely followed the curve of seat number 1, so that is the seat I'm now using for further fine tuning of my filters (rather than re-running full tests of each seat every time).
Using these graphs with test tones higher into the range where the mains take over makes it apparent that my sub is running very hot relative the the upper bass/lower midrange. Over the past week, I've reduced the output of the sub several times, but this imbalance is still there (though obviously not to the extent that it was before). While I can reduce it further to make the graphs look flatter, I've already reduced the bass as much as I can subjectively stand. Testing my levels with the pink noise calibration tones built in to my receiver don't suggest that my sub is hot, yet this graph makes it appear that it is still about 10db hot. A quick test with another test CD I have shows that a 1khz warble tone is pretty much in line with warble tones within the sub's range, so that suggests that the problem is a lull in the upper bass/lower midrange region, and that perhaps it curves back up (if I had made test tones that go higher - maybe another day). I have fine tuned my filters further since the above graph, but haven't created full plots.
Conclusion
At this point, I am extremely pleased with the results of my sub. The subjective quality is a significant improvement over my previous (and more expensive) sub, and the objective tests of extension and output show that I'm a completely different class than I was with the Sunfire. I had considered adding a second one to help tame some of the room response issues I have, but I've actually been able to get a fairly decent response curve within the range that the sub operates. And while you can never have too much power, all the time I've spent doing measurements has shown that the time/money for a second sub would be much better spent towards room treatment to help take care of the range just above where the sub operates. THAT is where the nightmare exists. It seems now my challenge is getting the rest of my response up to par with the sub.
Downloads
If anyone is looking for any of the tools I used to take these measurements... the initial spreadsheet used for BFD calibrating can be found here. This spreadheet is EXTREMELY helpful, as it allows you to actually see what kind of effect your settings will have. Another major benefit over some others I've seen is that it is based on 1/12th octave intervals. If I had used the more common 1/6th octave test tone intervals, most of the peaks and valleys in my graphs would have been invisible (though maybe I'd be happier if I didn't know they existed). However, for some reason, the test frequencies in that spreadsheet are just a little off from perfect 1/12th spacing. Not that it really matters, but because I wanted to test lower than 10hz and higher than 100hz, I decided to make my own test tones and spreadsheets on 1/12th octave spacing. You can download my test_tones.zip file with MP3s all the way from 5-320hz in 1/12th octave increments. Since perfect sine waves compress so easily, the fact that they are MP3s shouldn't affect their quality much. To reduce damage to the sub at high levels, I limit the test tones to four seconds, and burn these to a CD with three seconds of space in between each tone. This gives just enough time to be able to read the meter, log it during the silence, then read the next one. A complete sweep can be logged in under 10 minutes. Four seconds on and three off is only a 57% duty cycle, which should help reduce the chance of overheating your voice coil during loud testing.
To use as a template, you can also download the spreadsheet I used to compare output at the various seating locations. It has all the radio shack sound level meter compensations built in, but NOTE: the compensations are based on the Tony Gomez BFD spreadsheet linked above from 10-100 hz, but outside of that range I had to fill in my own values. There are values available on the internet above 100hz, and in that range, they don't vary much anyway, so it was easy to interpolate. I could not find any calibrations anywhere for below 10hz, so I had to extrapolate those. I was conservative with those values, simply extrapolating the 1db adjustment per 12th octave seen in the range immediately above 10hz, but I have no idea how close that actually is. I expect that at some low frequency, the meter will just stop responding. In my testing, my low frequency results were limited by the point where the output was reduced to the noise floor of my room, and the sensitivity of the meter. The "stereo" mode test, which was about 5db higher than the other tests, gave me measurable output to about 6hz. Higher volume levels would likely extend that further, but I haven't yet tested how far.
As I said before, I'm still early in the process with this sub. I still have plenty of fine tuning to do. If/when I have more pertinent info to add, I'll update this page. If anyone has any questions, comments, or suggestions, please direct them to bass (at) consumes.us