31/10/09
----------
Tonight, I decided to unscrew the top and bottom panels that had been effectively used as jigs whilst the side panels cured onto the baffle, so I could re-use the eight screws to glue and screw the centre brace.
I wanted to fit the brace as it would add stability to the structure and allow me to offer up the driver and drill holes on the sub-baffle to fit it. I decided this would be easier as I could stand the box on end and still have two open sides through which I could reach to position and hold things in place whilst holding the t-nuts until they had fully bitten into the MDF.
I may yet apply some adhesive to the the t-nuts, but they seem to have bitten tight exactly where they need to be. I used the driver (sans gasket) as a jig to drill 2mm pilot holes, then after the driver was removed I re-drilled with a 5mm bit. This is a tight fit for the 5mm bolts, but I see no reason to remove more material from the baffle than is necessary. Once drilled, I replaced the driver, pushed the bolts though (with a bit of force) and finger tightened the t-nuts as far as I could. Then, whilst gripping the t-nuts with pliers, I torqued the bolts up until the t-nut was nearly flush with the rear of the sub baffle. The t-nuts seem remarkably secure, but paranoia means I may yet apply a dab of saBesto to each anyway.
So, Pictures. First a couple of views of the secured driver and brace from behind:
A view of a t-nut screwed up tight with the 35mm bolt. 35mm seems about right given the 18mm of MDF, thickness of the driver basket and gasket:
Driver porn:
View from the front:
And finally, a close up showing the slight extra depth allowed to recess the the driver a little deeper than the gasket. The idea is that after applying an extra millimetre ot two of veneer to the baffle, I'll use a round-over bit on the router to radius the edge to 'exactly the front edge of the gasket for a really neat finish, even if it'll make no difference what-so-ever to the sound of the sub:
I must remind myself to let some magnets into the baffle before I veneer, so I can have invisible fixings for the grill which I've yet to figure out.
03/11/09
----------
Tonight, I spent some time playing with stuffing the 65L box and the relative quantities of it.
I'd seen the various 'How much?/Where to buy?' threads and thought that the prices being charged for what amounted to cushion stuffing, seemed a little excessive.
I dropped into the nearest shop I could think of to where I work (that supplied the curtain blackout cloth and black velvet for my original DIY projector screen) and they were flogging the long fibre, hollow core polyester stuffing for £2.50 for 0.25kg. That doesn't sound like much, but the bag it came in was about the size of a well plumped pillow, so I bought four.
The theory goes that stuffing the box absorbs the rearward radiation of the driver by vibrating the threads of the stuffing, that whilst rubbing against each other dissipate the energy as heat. This tricks the driver into thinking it's housed in a larger volume of air than the one it is in. Basically, you're replacing that larger volume of air and it's greater number of molecules to absorb the energy, with a medium of greater density that is more efficient at absorbing energy and hence, 'tricking' the driver into thinking it's in a larger box and therefore giving a flatter, deeper response.
What follows below is the measured responses generated from adding each 0.25kg bag to the box in turn and measuring the response of the sub in the near field (1cm from the centre of the cone) in order to render the room's contribution irrelevant.
The purple line is the empty box, the blue line is with 0.5kg, the green line with 0.75kg, the yellow line is with the full 1.0kg added. I missed out the 0.25kg line as it was near im possible to spread it out thinly enough to occupy the entire box volume. It simply collapsed under it's own minimal weight into occupying about 2/3 of the box. I've never opened a speaker of any sort to see stuffing of that style, so I ignored it.
The graphs with the axis scales exaggerated (compared to the usual 45-105dB, 15-120Hz scales) to emphasize the differences between the resultant curves:
Now, I should explain that I've normalized the graphs to the same output at 15Hz in order to emphasize the changes, but in reality, the lines all cross at around 50Hz which is about the F3 (-3dB point of the sub's roll-off) of the sub. The reality therefore is that the peak in output above F3 is reduced, with the response below gaining a slight boost, rather than a simple flattening of the graph as it would appear. The reality is that relative to 15Hz, the sub's response is flattened by about 3dB.
That this would happen to one degree or another, was never in doubt. After all, this is a long established principle by minds far greater than my peanut. What was interesting was to see the point at which you could stuff the box too much, or to put it another way; The fibres would become so tightly packed that they could no longer vibrate in order to dissipate energy and thus the box would start to become (effectively) smaller again.
As can be seen above the first 0.5kg makes a huge difference, with the jump to 0.75kg making less than an additional 50% difference. This shows the tighter packing is starting to offset the desired energy absorption. Adding another 0.25kg, to total 1kg, actually makes things worse, so the line has clearly been crossed at this point.
What does this mean? Well it means that the greatest difference is made with 0.5kg or more and above 0.75kg starts to tend toward worse. If you want a seat of the pants suggestion, then I reckon that 0.75kg may already be slightly more than optimal and that 0.65kg maybe about the mark. I may be making a bit of a leap, but dividing 0.65kg by 65L means that 100g of material per 10 litres of box volume may be a good place to start for optimal stuffing.
I shall try and extract stuffing to measure 500g, 650g and 750g when I get the time and the quiet necessary to do it. It will also be interesting as whatever I discover will be used as the starting point for the 115L box.
05/11/09
----------
More graphs.
This time, the graph below shows the 65L box with and without stuffing and the 115L box with and without stuffing. Any guesses which is which?
The yellow and blue lines are the 65L box with and without stuffing in that order. The pink and purple lines are the 115L box in the same order. The stuffed 65L box was the slightly reduced 650g (10g/L as suggested earlier), whilst the stuffed 115L box used the full 1kg of stuffing I had available. All measurements are normalized to 200Hz where all versions gave the same SPL at the same input level.
I've now realized that the yellow (65L stuffed trace) shows a bit of an unnatural lift below about 21Hz, probably because of a passing car or some such. By placing the mic 1cm from the cone to remove the room's influence, achieving 75dB involves playing at very low levels. So low, you can't actually hear most of the test tone sweep and equally inaudible background noise can therefore affect the plot in a way that it wouldn't with a normal in room, mic-in-the-hot-seat measurement. I'll be more careful in future.
Still, there's some interesting info to be gleaned. Although none of it is exactly 'news' as it's all predicted by programs like WinISD, or my favourite Unibox, but it's nice to see it in practice.
Stuffing clearly flattens and extends the response of both boxes, but it's effects are more marked in a smaller box. The boost at 20Hz in the larger box is about 1dB whilst the small box musters more like 1.5dB, whilst the peak is flattened by 1dB and 2dB respectively.
The bigger box gives an increasing boost in efficiency with descending frequency. From precisely no difference at 200Hz, there is a healthy 2-3dB gain by the bottom end although the increased effectiveness of stuffing the smaller box means there's less of a difference between it and the unstuffed larger box. If I had to guess, the stuffing has given the 65L box an effective volume of closer to 90L and the 115L box an effective volume closer to 135L. That's a 30%(ish) increase against 15%(ish). I'm left wondering at what point stuffing becomes irrelevant. When Vb=Vas?
Although the bigger box only extends about 4-5Hz deeper, once EQd in room it feels and sounds like a
lot more. It doesn't sound quite as dry or overtly damped as the 65L box (although stuffing did help here) but either with or without stuffing, it simply sounds a lot deeper and a lot more powerful, without relinquishing any of the stop-on-a-sixpence control of the smaller box.
The way bass transients appear out of nowhere is absolutely bloody awesome and as such, it delivers that "Is the sub on? Holy crap! Yes it is." moments in spades. Equally, the super clean top end I'd hoped for is as good as anything I've heard or rather, not heard, because unless there is deep bass present it simply doesn't sound like the sub is on. The crossover region really is seamless. I haven't managed to try out any serious reference level movie stuff yet, but Mrs. Williams is already "hearing noises" outside the house, so she's convinced by the low down deep ambient effects the sub is delivering, even from mundane rubbish like Eastenders.
I see no reason not to expect the figures modeling in Unibox suggests, simply because every other box has been ticked so far. 6-7dB headroom over my Monolith at 20Hz, without the roll-off beneath is starting to become a mouthwatering prospect when I can dig out the movie demo favourites.
The best bit? This is with only one sub. The bad bit is that I've got to build the other one yet so I'm going to have to wait a bit longer. Still, at least I can start looking at veneers and materials for the legs and table top.
07/11/09
----------
Okay. I'm sure those with a BFD can probably work it out for themselves, but for the benefit of those wondering how you make a DIY sealed sub deliver a flat response given the graphs above, here you are:
This is the natural response of the 115L sub at it's maximum SPL of 124dB @ 68Hz:
It looks like it's dropping off like a stone, but it's worth noting the figures involved. This is a sealed sub, so 106dB at 20Hz isn't to be shirked at, but 124dB at 50Hz and up is in all practical terms, useless. Note that the 75dB reference line is way down at the bottom of the graph. This is because I want to illustrate the maximum output after EQ. This frequency response is a result of output limited by the driver's capability within its X
max ie, it's excursion within -10% of the motors force limits. It's mechanical limits are considerably higher, but the point of this game is low distortion bass, right?
We need to EQ the response into something useful. A BFD has ten filters. I shall use one to approximate to the 80Hz roll-off that a receiver/processor will employ, so from here on in, remember that out of the ten filters available on a BFD, there will still be eight left, not seven. So, after the receiver/processor does it's bit, the response (the dotted line) looks like this:
Note how the theoretical +/-3dB points have already shifted downward as the 80Hz crossover bites from about 35Hz up. This is a point worth noting because the filter employed by the receiver/processor is not a brickwall filter, but starts acting much lower than most people will suspect.
Now, what we need to do is EQ the response flat within the maximum output (the solid purple line) so the sub delivers a flatter response. First, we need to suppress the response peak with a wide 16dB cut, which results in this:
It's looking flatter already, but clearly were not taxing the limits at the bottom end as described by the solid purple line, so we need to boost in order to compensate for the broad cut at higher frequencies that clearly has an impact far lower down the frequency range. The result is a broad 6dB boost at 20Hz and the resulting response looks like this:
So, because you cant exceed the low frequency capabilities of the driver alone (there's no port here) we've maintained the 106dB @ 20Hz capability, but cut the upper bass to compensate and so the maximum mid-bass output is now circa 108dB.
Remember, the benefits of using sealed subs are not measured in pure SPLs alone. If they were, we'd all be building ported subs, but the interesting point is that the AV15-X driver in a 115L sealed box is delivering 8dB more than an SVS PB13U in sealed tune at 20Hz which is not to be sniffed at. The response bears greater comparison to a 13U in 15Hz tune, but it has a much slower roll-off than the ported box so it punches lower too.
For those of a technical bent, here's the EQ curve relative to the 75dB reference that REW allows:
The eagle eyed will note that it looks much like a Linkwitz Transform and indeed, were it not for the BFD's 20Hz bottom limit, it would be possible to EQ an even greater approximation.
So basically, measure your DIY sub in the near field, use two filters to shape the response to flat and use the remaining eight to tune out any room induced effects over and above the subs response.
15/11/09
----------
Right. Tonight was the night that I finally made up cables, dragged them all under the carpet, drilled a hole in the wall, installed the EP4000 in another room where it can't be heard and took the time to EQ the subs as a a pair.
I can assure you there was more than a bit of trepidation involved because with very little control over global phase, never mind the individual phase of each sub any massive phase cancellations would be there for ever more. That means I had my fingers crossed that not only would there be no phase cancellations between the subs, but between the subs and the speakers too. I'd noted during the ancient 'Dual Subs' thread that I needed to not only use the phase controls on the subs, but also a global phase control available on the Velodyne SMS-1 EQ that I was using at the time.
During the time I was building these subs, the fact that I had long since flogged the SMS-1 and it's phase adjusting abilities was not lost on me. C'est la vie; This is supposed to be a no cost build funded by the things I've sold, so this was a risk to be run.
So, the results. First the left sub:
The usual room lumps and bumps are in evidence. Next, the right sub:
Much the same lumps and bumps below 40Hz, but a big ole dip just above and clearly shown, above that dip and up to 80Hz the response is quite different.
At this point, the penny dropped and I dropped my original plan. I have wired the room (and the one the amp is in) to allow the subs to run as separate stereo subs. The idea being that if you're going to run one cable, you may as well run two, thus future proofing the Blu-ray promise of 7.2 soundtracks and the ever present possibility that I get sick of this multichannel madness and regress to vanilla stereo. As such, I have cabled up for each sub to be run off and be EQ'd independently from each channel of the BFD I'm using for EQ. It is, after all, a stereo device.
However previous experience has shown that the individual responses of two subs in the same room does not match the averaged response (that REW allows you to predict) as the combined response is not a mere average of the individual SPLs. REW cannot predict the combined response PLUS the phase cancellations which cannot be EQ'd, so as it currently stands with mono LFE and bass managed mono bass from stereo, the only sensible option is to EQ the combined response as if it were a mono sub.
So, with that out of the way, best get the speakers involved as EQing a sub is pretty pointless unless you involve the speakers it (they) is going to combine with. The combined sub plus speaker response:
And lo and behold, the one obvious dip is ameliorated to a large degree. It should also be noted that (and maybe I should have done a 'before' graph to illustrate the point) that in this graph I have reduced the overall level of the subs by 6dB. I had hoped for a 3dB gain by using two subs, but below 40Hz and prior to EQ, I am getting the best part of a 6dB gain.
This is excellent because it virtually removes the need for the 20Hz boost I speculated on earlier in the thread and Nimby correctly identified as a potential factor. I was speculating based on the need to apply a Linkwitz Transform in order to flatten the sub's response based purely on near field measurements.
The reality is has less to do with boosting the bottom end and flattening the upper bass and much more to do with simply equalizing the response in much the same way as my previous ported sub. Sure, there's less cutting low down and a fair bit more cutting of the room modes higher up, but there are no wide and deep/high boosts/cuts employed to shape the response. Result. Further more, it means no need to waste two of the available ten filters in response shaping. In the end, I only used nine anyway.
So the combined subs plus speakers response:
As I am now accustomed to, I've EQ'd as much for a short response as a flat response and I think I can safely say that the ringing sub 60Hz is massively reduced. Incidentally, I listen to a couple of reliable 'yard stick' music tracks to fine tune the results and hence what looks like a dip in the power response between 30 to 40Hz, is as a result of
shortening the response.
There is nothing I can do about the response above 60Hz as the speakers are contributing additional energy in this area but it's actually pretty much in line with the rest of the speakers power response above this point so it's not quite the issue that it looks to be.
How does it sound? With more than a hint of relief, I can honestly say pant wettingly good. Not only is it delivering a seamless musical integration with the speakers I cherish above all else, but the few movie clips I've had the chance to try simply blow my wildest hopes out of the water.
There is something staggering sudden about big sealed subs that even really good ported boxes only hint at. It's a bit like the description you give to people when they go from their fart box to their first real sub and they then wonder where the 'bass' went. Only in this case it's not distortion that's missing, it's overhang.
Unibox (I really don't like WinISD anymore) allows you to tune your box to the drivers parametres and then allows you, if you take the time to optimize each alignment (sealed, ported, bandpass, etc) to compare the impulse response of each box. That is, how quickly the sub stops after the signal has passed and for the case of a sealed sub, it's at least twice as quick as the next best option.
In use, the initial impression is of a much quieter sub, but when the big bass hits happen, they hit far more suddenly and then leave again even faster, which initially makes them sound a touch reduced in impact. However, sustained effects seem just as mental but have far, far more texture and a more impressive sense of grip.
Whilst listening to Adams dual driver sub, I was put much more in mind of the Velodyne style of bass as opposed to the fatter sounding SVS bass, but in my room at least, it doesn't suffer the Velodyne limitations at the bottom end as the volume goes north. To put it another way, it sounds like I have the HUGE Velodyne that Velodyne haven't made. Effortless, very deep, clean and fast. Everything I had dared hope for and a bit more besides at the bottom end. That makes me one very happy bunny.
I shall have to wait for the rare occurrence of the wife
and son being absent and then I shall try and measure some absolute figures for the in room repsonse as regards SPL and extension at those SPLs.