A Monster Box Construction Methods Project
8/10/2020 Initial post. PLEASE DONATE TO MY GOFUNDME!
9/26/2020 added results of Rounds 1 & 2
11/15/2020 added results of Round 3
11/28/2020 added results of Round 4
6/7/2021 added results of Round 5
9/4/2021 added results of Round 6
Which wood to use? Which fill? What does bracing do? Is CLD best? I'm starting this project to start quantifying some answers to these questions. A lot of work has been done before in fits in starts by others (BBC, Kef, etc.) but they were often narrow investigations and you couldn't compare results because measurement conditions were different among the different studies. Some have cited manufacturer's damping specs, but I have no idea how that translates to the real world. Ditto on accelerometer data. So here I will be doing SPL measurements, that is after all, what we actually hear.
The cabinet is ordinary, large enough for the panels to have some radiating area and magnify differences between changes. I will be using a Peerless 830970 firing into the cabinet using MLS signals. When I do the fill/lining I will mount the SB15 as a dummy woofer and measure what radiates through it. Mic placement is 1/8" from the SB15 dustcap when doing lining/fill testing, and 1/2" from either the side or rear panel when doing wall radiation testing. I use MLS signals and have settled on a 50ms gate and 1/24 octave smoothed.
To discuss the results, I've started at a thread at diyaudio.com
In the plots where "bare driver" is shown, this is the far field measurement of the Peerless 830970 (levels are adjusted so relative SPL is correct, if you see a resonant peak near the bare driver level, it really is that loud outside the box) as shown here:
First up, tests of 6 cabinet materials. Note: given the relatively low SPL of the sound radiating from the cabinet, the ambient noise floor can start to show up below 200hz, so consider this the cutoff for useful information.
1/2" JUNK PLYWOOD
Got this stuff from Home Depot. Looks like 3 layers plus thin veneers.
1/2" HIGH QUALITY PLYWOOD
Good quality 7-ply birch plywood
Note: I started harmonic distortion testing after I had cut a hole for the SB15 dummy woofer in the 1/2" MDF and 3/4" mediocre plywood, so I capped the hole for HD testing. Because of this, I don't really think it is strictly comparable to the other boxes that never had a big hole cut in them.
3/4" MEDICORE PLYWOOD
AC Radiata pine plywood from Home Depot. 7 plys, occasional small void. Note: I started harmonic distortion testing after I had cut a hole for the SB15 dummy woofer in the 1/2" MDF and 3/4" mediocre plywood, so I capped the hole for HD testing. Because of this, I don't really think it is strictly comparable to the other boxes that never had a big hole cut in them.
3/4" HIGH QUALITY PLYWOOD
Very high quality Russian birch plywood, 13 plys
Following is a comparison of 1/2" and 3/4" materials respectively:
Fill and Lining testing
Next we test filling and lining, and measure how much sound radiates through a woofer cone to the outside. The baseline is an empty box, no lining, no bracing. Dummy woofer is an SB Acoustics SB15, terminals shorted. The wood going forward will be the 3/4" mediocre plywood.
Monacor MDM3 wool batting
Looks like this isn't available anymore. About 1/2" thick batting of wool and IIRC poly. Given the performance of 2 layers (1" nominal) I would like to find a new source of wool batting, since my big complaint with the denim insulation is the amount of fuzz and fiber it gives off. Makes me worry it will get into the voicecoil of a driver. Some lining will be multiple layers, so note the description in the plots
1" Sonic Barrier
From Parts Express
1.8" Ultratouch Denim insulation
Available at HD (and other places), reasonably priced, performs well. No wonder the popularity. Now if would just stop shedding so much stuff!
Meniscus Audio bonded dacron batting
This is interesting stuff. Not as dense as wool or denim insulation, but for exactly that reason it used by the transmission line designers to completely fill the first half of the line behind the woofer with a consistent, know density material. The plot labeled "stuffed" is filling 1/2 of the box, just past the top of the dummy woofer.
Polyester pillow stuffing
1/3 of a pillow I bought from Walmart
Following are comparisons I find informative:
Round 2 results! Couple issues make these results not exactly comparable to the first round. First, repeatedly disconnecting the source driver ended up with a terminal coming off, lead and all. I bought a couple more drivers, and of course the response isn't exactly the same. So I've taken a new baseline empty box measurement with the 3/4" mediocre ply identified before (AC Radiata pine plywood). Second, I already cut holes in that box for the previous fill testing, so built a new box. So new baseline is a new driver in a new 3/4" mediocre ply box.
On measurement conditions, remember that the amount of signal escaping the box is relatively low, so low frequency noise can start to show up <200hz. So take that as the lower limit. The one exception is the fill testing where the amount of bass leaving the dummy woofer is actually pretty high, so you see a typical low frequency roll off there.
For the CLD enclosure tests, I used 1/4" MDF as the primary material. Why so thin? Well it seems to me that a proper CLD should be doing more with less - design smarter, not harder. I built box-within-a-box, minus the baffle, which was a solid piece of MDF left for last. I used a fine tooth trowel that made 1/16" beads to try to keep a consistent application of adhesive. See pics below to get an idea. For the Nidacore and 1/2" XPS foam I used Loctite PL300 adhesive, since it is made to use with foam.
First plot is the new baseline of an empty box for CLD testing, using the new driver and 3/4" mediocre plywood used before. Second plot is a baseline with the new driver of the denim insulation, identified in round 1 as a top performer.
1" Melamine foam
.80" 100% wool for quilts from homeofwool.com.
This looks the same as other wool batting for quilts, so I think any 100% wool batting for quilts you can find will behave similarly as long as the nominal thickness is the same:
Comparison of the top 3 performers:
Next is bracing. I used 3/4" x 1 1/2" oak braces from Home Depot. The control is just a simple cross brace. Then test CLD braces using Sikaflex 292i or Weicon Flex 310M Classic as the adhesive. There is a 95% overlap between the braces.
Weicon Flex 310M Classic
A construction adhesive made from MS polymer, a new-ish family of adhesives that supposedly has vibration dampening also. Very creamy and easy to work with:
This is a construction adhesive with claims of vibration dampening. Rather thick and tacky:
Following are some comparisons of the results. I was surprised that the CLD braces didn't really seem to do much, at least in my implementation. I'm guessing with the 95% overlap, the bracing is effectively solid, again, in this implementation. Next round I will try 50% overlap and see what that does.
And now the full CLD construction! The first two boxes are simply 1/4" MDF joined by either the Sikaflex or Weicon. The last pair use either 1/4" Nidacore or 1/2" XPS foam as an interior layer, glued with Loctite PL300, and again using 1/4" MDF as the inner and outer skin. I built box-within-a-box style, but the baffle was a single piece of 1/2" MDF.
And last, Resonix CLD squares applied to the 3/4" mediocre ply box above. The story of these is interesting. A hobbyist at diymobileaudio.com had done a very popular CLD test and one product rose to the top in overall performance and price: Sundown Solutions CLD tiles. The surprise was that there were other similarly constructed CLD tiles, but performance was surprisingly variable. Some years later SDS went out of business the hobbyist started his own line with the aim of being the same (or better) as the SDS tiles. I write all of this to alert you that just any butyl rubber and aluminum tile may not behave this well.
I'll be rolling out results over the next week, but the first batch is the port performance. As mentioned before, the inspiration for this experiment was the Kef LS50 white paper and a Harman study by Salvatti, Devantier, Button (attached). The Kef port's innovation is using a foam center section to dampen the higher port resonances. The Harman paper looked at different port terminations and their impact on sensitivity, harmonic distortion, and power compression. Their own designs in the test were ports that were not straight walled, but a constant radius. They related this radius to the overall length via an equation, but the one labeled N=0.5 was the best balanced and the one that I selected to copy. Simply, the radius of the wall curvature is equal to the port length:
There is also a small 1/2" roundover on either end that isn't shown in that pic. What was interesting about Harman's results were slight sensitivity gain at low power levels vs the competition, and very good harmonic distortion and chuffing onset performance. I did not see the sensitivity gain, probably due to length, which I will address below.
For the Kef port I used common foam pipe insulation and also decided to print a version using TPU filament. I used Hatchbox TPU with a Shore 95A rating. I used 2 perimeter layers and 18% fill, which is about as low as I think a person could go. I copied the roundovers of the Precision port so it would just be the flexible center sections that were the variable.
The Precision port is just a 3D-printed copy of a 2" Precision port.
The "typical DIY" port is a simple 2" ID pipe with a 3/4" roundover at the baffle, as a DIYer would typically do. This would be the reference port.
The square port is similar to the typical port, except rectangle (it just now dawned me I labeled all the pics square, oops) with a 1:4 aspect ratio. The cross-section is equal to a 2" round port at 3.14 sq.inches. This also has 3/4" roundover at the baffle.
Speaking of diameter all of these are 2" nominal ID ports. For the Harman this would mean the very center of the port where it is skinniest.
The box is a 4th order bandpass tuned to 52hz. The driver is the SB15NAC-8. Front volume is 15L, rear volume is 10L (net, but see below). Front was lined with wool batting, rear was lined and somewhat stuffed with the same wool batting. I adjusted all the port lengths so the Fb was the same, as verified by impedance:
Length can give you an idea of how restrictive a port is. The less restrictive it is, the longer the port must be to maintain a given Fb. The lengths were: Harman 5.4", PP and Kef 4.25", and typical and rectangle port 3.75". This is where I deviated from the Harman paper. There they equalized the ports by length, which doesn't make much sense to me, in the real world you would want to know relative performance for a given tuning. For this reason, you might notice unexpected results vs the Harman paper. For example sensitivity, which was worst for the Harman in my tests, instead of the best in their paper.
This brings me to caveats. Since these are all different lengths, I had to add a little closed cell foam so the port volume was *roughly* equal. It wasn't exact, but I'm satisfied with my mitigation. Next caveat is that all of these are more or less close to the driver. A problem? I don't know. See pic for an idea:
Another caveat is surface finish. The Harman paper showed surface finish makes a difference and smooth is best. These are 3D printed and so not as smooth as PVC pipe you might typically use. A problem? I'm not sure, but I would say a necessary evil to keep the playing field level. Otherwise I would have just used the actual 2" Precision port I had. Another benefit was that I could make the mounting flange the same for all of them for quick attachment/removal.
Test setup> MLS measurements for the SPL: 500ms gate, 1/24 smoothing, mic at 1/4" from the baffle. I tested a number of distances and settled on 1/4". I had to change things up for the harmonic distortion measurements to not overload the mic and soundcard, here it is 20.5" from the baffle. On a related note, I could not use the MLS function for these high SPL measurements, so to judge power compression look at the level of the fundamental F1 in the HD measurements. Also when trying to get an idea of the onset of chuffing, you can see the harmonics starting to cluster. Why this is so is what you can't see in the plot: the entire spectrum is lit up with something almost like white noise, with other concerning sounds mixed in. You'll also notice the fundamental F1 goes from smooth to jaggy. I know the plots are limited, but I'll try to add subjective commentary to give clearer perspective. I added the drive voltage to the HD plots, if you are bored you can look at the impedance and calculate power and SPL.
TYPICAL DIY PORT:
KEF PORT W/FOAM:
KEF PORT W/TPU:
I need to delve into the results much deeper but here are some initial thoughts. The typical DIY and rectangle DIY ports were surprisingly similar in frequency response and sensitivity. I'll need to dig into the HD results to pic a winner.
The Kef ports didn't show their stuff as much as hoped. Foam looks better than what I printed, so chasing down high performance flexible materials will be the key to making the concept work. One thing I should mention, since the Harman port was the largest and not length adjustable, everything had to match that for Fb. This resulted in the Kef center section only being about 1 5/8" of exposed pipe for flexing. Perhaps if it was longer you would see a bigger impact on damping those port resonances.
The Harman was the most interesting to me. I was let down by the sensitivity being lower than all the others. But as soon as I did the HD measurements it clearly pulled ahead. I've only just glanced at the HD plots to confirm what I heard, but I can tell you sitting here running the test the difference was huge. Most of the others were having difficulty by 8.75V, and are downright scary sounding by 10-11V. The Harman was just so much cleaner. I remember thinking at 13.52V, yeah that's chuffing. But lower than that and you kinda had to listen closer to pick out audible distress, it just wasn't that noticeable, certainly not if I were actually playing music or movie to mask it. I need to compare it against the actual Precision port with smooth walls. If the 3D printed surface is not an issue, I would encourage anyone to try the Harman for any fullrange or subwoofer duty where ultimate SPL performance is important.
Looking at the HD results closer for the first time, I see the rectangle and typical round port are basically the same in my implementation, with maybe a slight edge to the rectangle. The Precision port steps up performance with noticeably better HD and compression performance than the previous two DIY ports. But notice at the highest SPL it falls apart hard and actually falls a bit behind the the typical ports. Harman noticed this same behavior in their study where the straight port with no roundovers lagged all the competition until the point where they were all chuffing, where it actually was a bit better than the others. Still, they were chuffing, so sort of a pointless win. The Harman OTOH is just a beast. It takes another step up in compression performance from the Precision port, and the HD is just great until finally starts chuffing pretty good (to my ears) at 13.52V.
CLD Box Construction
Loctite PL300 CLD
In Round 2 I was impressed by the simple CLD of 1/4" MDF and Weicon 310M Flex adhesive, but wanted to check cheaper alternatives. I had plenty of Loctite PL300 on hand from the other CLD boxes and liked that it stayed softish when cured (acrylic latex). And it's super cheap. I also grabbed some polyurethane PL Premium adhesive. This cured quite hard, I really can't see it doing anything beneficial so I just built a box with the Loctite PL300. Here are the results:
1/4" cork CLD
Similar to the prior boxes made from 1/2" XPS foam and Nidacore, this cork is sandwiched between 1/4" MDF and glued up with PL300. Sorry forgot the pic, but it looks like the Nidacore build.
I've wanted to try this since a box I built years ago that seen improvement when judged by the old knuckle rap test. Problem with that is things can sound different but not better, and we have no way of knowing until we measure. Application is 3 thick coats to the inside of the box, dabbing with a sponge so the surface finish is rough.
Some comparisons of the above with Weicon 310M Flex Classic:
Last time my CLD braces with 95% overlap behaved pretty much like a solid brace. So this time I used 33% overlap and the Weicon 310M Flex adhesive as before, and also added a version using 3M's VHB tape, which is a vibration dampening adhesive tape.
Weicon 310M Flex Classic at 33% overlap:
3M VHB tape at 33% overlap:
Here are the results of the new port test. Here I wanted to investigate a few things. One was abandoning the Kef method of flexible center sections to reduce pipe resonances, and go to tapping the center of the port and then dampening that energy. I also wanted to investigate whether the performance of the Harman port was inherent, or whether it was simply due to the average port diameter being larger than the others (thus it being longer than all the others also). I did this two ways.
First by downsizing the Harman port so it matched the overall length of my Precision port clone. Port ID ended up at 1.756". Since I hadn't yet determined an end correction factor, I had to make an educated guess and the final port tuning was high by about 2-3hz. Not bad.
Second, I upsized my Precision port clone so that the overall length was the same as the original Harman port. Port ID is 2.28". Tuning was virtually identical.
I also measured an actual 2" Precision port to see if surface finish made a difference. If you recall I 3D printed all of these to remove the surface finish variable, since that is known to have a measurable impact on performance.
Additionally I added a 2.38V harmonic distortion plot to better capture typical listening conditions.
For some reason the F1 was not quite as smooth in the distortion plots this time around so it is not as obvious when chuffing starts when just looking at F1. The higher harmonics are still indicative though. Because of the difference in F1, power compression vs the earlier testing is a little harder to see.
Harman port (shorty):
The Harman port is superior. It's a bit ahead in distortion performance for the original vs jumbo Precision port, and more clearly in the shorty Harman vs the original Precision port.
Precision port (jumbo):
Precision port (factory smooth finish):
Surface finish matters. Quite a bit. The harmonic distortion of the smooth finish Precision port is cleaner than my clone, and is really noticeable at the highest power. Not enough to catch up with the Harman though, but close. Close enough that smoothing the Harman print is well worth the effort. I've seen if you print with ABS you can smooth the finish with acetone.
Tapped port v1
I should note that I heard (and measured) a "flapping" sound during the harmonic distortion tests. No doubt the taped cavity being pumped back and forth. So treat these HD results as preliminary until I can print the dampening cavity, so it is sealed with hard material.
I've designed and measured 7 more versions of the tapped port above, now dubbed "AugerPort"! I don't want to do a huge data dump of the 10,000 ways to make a lightbulb that doesn't work, so I'll just present a quick summary. There are taps at 1/4 and 1/2 the length of the port. All of them used wool batting in the channels seen below, and then plastic caps are glued on to seal them off. Eventually I made great progress in knocking down the pipe resonance, but at the expense that harmonic distortion and chuffing was greatly increased. I've never seen this measured by others who have went down this path, so it was a (bad) surprise. Here are some pics. Versions 1,2, and 3 used square ports like this:
Versions 4 and 5 used .15" wide slits, running vertically on v5, and the other horizontally on v4. Version 5 really knocked down the resonance, but harmonic distortion was still poor. So for versions 6 and 7, I shrunk the slits down to only .06" wide and cut the number down, like this:
Above you can see AugerPort 5 really knocked down the resonance over the Harman port on which it is based. With AugerPort 6 I had hoped to get a best of both worlds, but harmonic distortion and chuffing seem unavoidably poor whenever you put holes in the wall of the port. My next designs will be inspired by the latest Polk port which used a closed-end pipe suspended along the centerline of the port. The pipe is tapped in the center to absorb the 1/2 wavelength pipe resonance. We'll see if it can do that and not chuff.
Next up is another CLD box, this time using Hardie backer board. This is a cement and paper/cellulose composite board used as underlayment. It is similar to the simple Weicon Flex 310M and PL300 constructions, where the inner layer is the 1/4" Hardie backer, and the outer layer is 1/4" MDF, glued together with PL300. Below are the results, followed by a rerun of harmonic distortion tests from earlier designs to compare to, since I forgot how I had my measurement rig setup back then.
Some comparisons with some the better performers earlier in the testing:
Considering the performance above, and the ease of construction, I think I like the simple 1/4" MDF panels glued together with Weicon Flex 310M best. Easy to build, low cost, performs well. I'll be using this going forward. Next I'll add braces and see how that changes things. If nothing unexpected pops up I'll do a new box will all the lessons learned and compare that to a typical MDF box.
Here are tests of two ports and some new box materials. AugerPort 8 is similar to my earlier attempts, but instead of large taps, I use microperforations, defined as being <1mm in diameter. The idea came from this really interesting video. The idea is to damp the pipe resonance but not disturb the flow of air in the port, as this has increased chuffing and harmonic distortion in my prior attempts. AugerPort 9 is inspired by Polk's new X-port, which uses a suspended closed-end pipe along the port axis with taps in the center to again absorb the half-wave resonance of the port. Polk claims chuffing was actually improved.
Damping of the pipe resonances was good, 2nd best behind AugerPort 5. The hissy sort of chuffing was better too than prior attempts, but there was a distinctive "burrrr" sound at all power levels. You can see in the single tone spectrum plots how lit up it is. I don't see how I can improve this situation, so that's disappointing.
As far as damping the port resonance, this design did absolutely nothing. The resonator diameter is 1/4 that of the port, but watching Polk's video again it looks more like 1/3 or even 1/2 the diameter of the port. I'm going to resize to 40% of port diameter and try again.
Some comparisons, including with AugerPort 5, which had great damping, but poor chuffing as all these tapped ports do:
Next are some new brace schemes. One is "matrix" style suggested by others. The other is similar to Kef's CLD brace in the LS50 where they place the damping material between the box and brace. I used the dowel brace here with 3M VHB tape as the constrained layer. But given my muted results vs Kef's simmed results, I wonder if using something like a window frame brace with the VHB tape would be better? More contact across the panel to absorb more energy? Also included are some comparisons with earlier tests: regular oak dowel bracing and CLD bracing where oak braces are overlapped 33% and glued together with Weicon 310M Flex.
Kef-style CLD using 3M VHB tape:
Some comparisons of prior bracing methods:
It might be good to recap where my cabinet testing has led. I started measuring just simple materials: plywood, MDF, in 1/2" and 3/4" to establish a baseline. Moving on to CLD construction, I then used 1/4" MDF "skins" joined by either the adhesive under test, or with an additional layer of foam, cork, etc. in between. From this the Nidacore construction was probably best, but the simple 1/4" MDF glued together with Weicon 310M Flex was so close and so much easier and cheaper to make, I would use that in the future. From my bracing tests (prior to the results just posted above) the CLD with Weicon 310M Flex made a decent improvement over solid braces. The next question is whether adding those CLD braces to the simple Weicon CLD box would retain the benefits if both. The following is just such box, compared to a simple 1/2" MDF box with solid braces, and a 3/4" plywood box with solid braces to represent the typical DIY boxes.
BTW the Weicon 310M Flex can be purchased here.
I would say the rear panel measurement shows the Weicon CLD box to be superior to the others. The side measurement has it about the same as the MDF, with both being ahead of the 3/4" plywood box. I will say the addition of the CLD braces to the CLD box, didn't complement each other quite as I had hoped. Here is that CLD box with and without braces:
I also wanted to try a version of the simple Weicon CLD box (no braces) but using 1/4" MDF inside and 1/2 plywood on the outside. The idea being that the two materials would resonate at different frequencies and therefore damp each other. I also did a version using 1/4" MDF & 1/2" MDF. I also switched from "box-within-a-box" construction to gluing the panels up ahead of time, then cutting to size and constructing the box. See pic to see what I mean. These are not braced and glued with Weicon 310M Flex.
Here is another comparison I find interesting. This is the CLD of 1/4" and 1/2" MDF glued with Weicon, versus solid 3/4" MDF. So same nominal thickness and material, but one is solid and the other CLD: