In this blog I test a new cabinet design for the Fostex FE108NS 4" full range driver. This design responds well to a variety of other drivers as well.
- Small to medium sized listening rooms
- Nearfield listening or within 2.5m listening distance
- Low power tube amplifiers (ie. 300B, 45 tube)
- Average listening level of 75dB (low to medium)
- Wide Baffle
- Flatter frequency response through critical midrange
- Clean step step response within first few milliseconds
- Stronger output in the lower midrange (baffle gain)
- Back Loaded Horn
- Maximize sensitivity in bass and mid-bass region
- Load driver across broader frequency spectrum vs bass reflex
- Reduces diaphragm movement and therefore distortion
- Basic Construction
- 12mm thick baltic birch plywood
While visiting Germany I heard the Aries Cerat Symphonia back loaded horn at Klang Loft shown below.
I have heard a variety of back loaded horns but none of them sounded as good as the Aries Cerat Symphonia. The bass was nuanced, palpable, and deep.
So the back loaded horn configuration was back on my radar for a future design of my own.
How it Works
A back loaded horn works by horn loaded the back wave of the driver. The goal is to amplify bass and mid-bass frequencies over a 2.5 octave bandwidth. A backloaded horn is the most difficult type of enclosures to design successfully. As with any horn, there are many pitfalls along the way. I equate the design to walking a knife edge, where balancing all the variables is critical.
BLH Design Goals
- Amplify bass and mid-bass frequencies by +9dB
- Requires exponential flare rate
- Create torturous passage to diffuse midrange frequencies
- Must consist of random passage lengths to diffuse a variety of sound wave lengths and thereby frequencies
- Sharp bends breakup midrange energy preventing escape out horn mouth
- Create path length that is long enough to produce the desired horn cutoff frequency (Fc) but not too long as to introduce too much delay
- Too much delay will introduce a noticeable separation between the direct sound from the driver and the horn output
- Making the horn path too short will raise the horn cutoff frequency and not achieve the required bass extension.
- Pick a target Fc that matches the driver's natural rising response.
- You must blend the horn's output to match the natural rising response of the direct radiator's output.
- A low QTS full range driver is going to have a rising response, so the horn's output should fill in bass and mid-bass frequencies to sum to a flat in-room frequency response
- Minimize Cancelation Dip
- There will always be a cancellation dip in the 150Hz region where the horn output cancels with the driver's direct output.
- There are a variety of techniques to minimize the severity of the cancelation dip which I will touch on later.
Horn Simulation Software
While not perfect, horn simulation software such as Hornresp can aid in the design process. I've come to know what hornresp can do, and what it cannot do. Below is my BLH simulation for my Fostex FE108NS. By making Vrc and Lrc both zero, the rear chamber disappears and you essentially have a BLH.
The amplitude response is shown below. We have good loading down to 50Hz. We can ignore the rise at 125Hz since this will be attenuated by our torturous path mentioned earlier. It is expected that the horn will provide loading across 2.5 octaves which is from 50Hz to 300Hz.
Below is the schematic diagram showing the diaphragm in red on the left, horn throat starting out at 20cm^2 growing to 400cm^2 at the horn mouth, with a path length of 336cm. The overall horn volume is 40.64 liters as indicated in the top right of the diagram.
I needed the horn to have a wide baffle for reasons that I can't get into very deeply in this blog post. But I will summarize by providing a short story. A few weeks ago I had build a small bass reflex enclosure for a 5" woofer. When I tested the frequency response of the woofer mounted in the enclosure, I was shocked at how irregular the response was in the 1kHz region. I decided to place large blocks of hardwood around the baffle to widen the baffle. Each time I extended the baffle by 10cm the response dramatically improved. Once the driver was placed in the middle of a 30cm x 30cm baffle, the response was almost perfectly flat. The reason for this is the proximity of the baffle edges to the driver itself. By widening the baffle the energy from the driver has a chance to dissipate before encountering the baffle edge. This reduces the effect of edge diffraction and smooths out the on-axis frequency response.
In addition to a smoother frequency response, there is also baffle step gain where we see a modest gain in sensitivity on-axis through the lower midrange. This helps blend with the output of the BLH output for even greater overall sensitivity.
Horn Mouth Termination Considerations
There are two schools of thought on how the horn mouth should be terminated. Some propose putting the horn mouth as close to the driver as possible so that the outputs can be more closely perceived as point source. The other school of thought is to place the horn mouth near the floor so that the floor boundary helps reinforce bass frequencies. If done correctly the floor can actually aid in transitioning where it becomes part of the final horn expansion. Even the rear and side walls can physically form the final horn flare geometry. This has a number of distinct advantages if attempting to maximize bass extension as well as keeping the overall cabinet size to a minimum.
Below is my proposed horn path which places the path in spiral pattern to start and then proceeds to make a few more bends before exiting out the bottom of the cabinet. Now I had originally had the mouth terminate out at the bottom of the front baffle, however it occurred to me that if I simply removed the bottom portion of the front and rear baffle, then this could form a much large mouth area. So essentially the horn mouth splits directions between front and rear radiation.
Horn mouth split into front and rear radiation. The wavelength at 50Hz is 7m long, so this abrupt physical geometry is entirely invisible to the sound at this frequency. The abrupt change only serves to aid in my goal of diffusion of attenuation of midrange frequencies. This is counter intuitive to most BLH designs that I see where an attempt to create a smooth horn mouth in done in error.
So does it work? I don't know, let's build it.
With a fresh sheet of 5' x 5' 12mm (1/2") thick baltic birch plywood, I set off to cut up my test enclosure.
Ah, lets build another one for a stereo pair.
Doing measurements is actually a little bit tricky. Normally I would do a ground plane measurement where I place the speaker on the ground with the mic placed on the ground at a specified distance. However this does not work with this particular type of speaker for the following reasons.
- The BLH output is a full +9dB higher output than the driver's direct sound for the bass and mid-bass frequencies. So there are essentially two sound sources. Where do I place the mic?
- My specific design uses the floor and rear wall boundary to complete the horn mouth geometry. So my cabinet design would not even be able to function properly if the measurement was done outdoors or in an anechoic chamber!
In the end, I decided to conduct two measurements. The first is with the speaker placed in the listening position 30cm away from the rear wall. The second set of measurements was in the middle of my studio so that I can compare and see the influence of room boundary reinforcement.
Below is a frequency response measurement at 1m with the speaker placed 30cm from the rear wall and 100cm from the side wall. My room is quite large at 5m wide x 7m long (not shown).
- Very high sensitivity at 92.5dB for 70Hz region
- Bass extension to 40Hz
- Cancellation dip as expected -12dB at 145Hz
- Mid-bass sensitivity approaching 100dB (200-600Hz region)
Considering that these are in-room measurements, I was not expecting the response to appear flat. But for the specific use case, we are seeing very high sensitivity numbers compared to other types of cabinet designs. To directly compare I replaced the BLH horn under test with a variety of other bass cabinets I had on hand and measured the output at 70Hz calibrated to 2.83v (1w) so that I could measure the output SPL at 1m. Below is a summary of 70Hz output SPL for some various speakers in the same physical position as my BLH cabinet.
Sensitivity (2.83v/1w) for various speakers:
- Markaudio CHR120 in 1cu. ft. bass reflex cabinet --- 75.5dB
- Signet SL16 2-way Stand Mount with SEAS 6.5" woofer --- 76.0dB
- Fostex FF165WK in transmission line enclosure --- 90.0dB
- Speaker No.1630 with dual 5" Scanspeak 4ohm woofers wired in series --- 89.5dB
As you can see, the small 10cm Fostex FE108NS is providing a full 2.5dB higher sensitivity in-room compared to other much larger drivers. Clearly the back loaded horn is having a great effect on increasing the sensitivity in-room compared to the bass reflex and transmission line cabinets.
Following this simple test, I then realized that perhaps the Fostex FE108SS-HP might have even higher sensitivity since it has a massive dual magnet motor. I quickly swapped out the FE108NS for the FE108SS-HP and re-tested. Sensitivity came in at a whopping 94dB for the 70Hz test tone! Talk about suitable for low power tube amps!
Frequency Response of Horn Output
Below is the output of the horn mouth for reference. You can see that it is largely made up of bass and mid-bas frequencies, and then quickly dies at 500Hz.
This contrasts sharply with the prediction in hornresp which shows continued midrange energy well into the 2kHz region. The frequency response below is the horn mouth output simulated in hornresp. This is one example of how hornresp is limited and the information needs to be carefully dissected.
This particular BLH is designed to interface with the room in order to finalize the horn mouth. The room is part of the speaker system in order to provide substantially more sensitivity than a traditional bass reflex cabinet. To highlight this I placed the speaker in the middle of my large workshop which lacks the nearby boundary support. (see blue below) The red response is with rear wall boundary support.
Although the response is more irregular with the rear wall boundary support (which is to be expected), the output gain is at least +5dB through the bass and mid-bass. Additionally, the bass extension hits extends to 35Hz vs 50Hz when positioned in the middle of the room. We also see +3dB gain even at 2kH which is equivalent to doubling your amplifier power.
In both cases we see a rising response into the upper treble. This is expected and normal due to the increasing directivity of the full range driver. Anytime you see a narrowing of directivity in terms of the off-axis response, you will see a corresponding rise in on-axis output. Simply put, the energy from the driver has to go somewhere, so if the driver's high frequency energy is being concentrated into a narrow "beam" (if you will) then the on-axis frequency response is going to increase for the frequency region of concern. In other words, if we ran an RTA measurement taking into account the full radiated power of the full range driver, it would be a flat. This is another area of full range drivers that is counter intuitive. If we attempt to flatten the rising response with EQ, then the speaker will sound dull and lifeless, and the total radiated power of the driver will have a lack of energy in the treble region.
Some other measurements
It did not make sense to me to measure distortion at my usual 85dB test signal level for the following reasons.
- The average listening level for this use case is 75dB 1m
- The listening distance and room size is typically smaller than with my larger systems, requiring less output SPL
- Typical use case makes use of low power tube amplifiers that operate in the sub-one-watt category
- The typical musical genre is often quieter with less energy
So I decided to test at 75dB 1m instead.
Below is the IMD performance for bass, mid-bass, and midrange. We can see that with a complex and difficult test signal we have IMD at -55dB for the 100Hz region improving to -60dB for the mid-bass and midrange.
If we high pass filter (HPF) above 100Hz we can see the mid-bass and midrange improve to -65dB.
Below is the IMD performance for treble frequencies with modulating tones extending down to 50Hz. Again, we see IMD hovering around the -60dB mark for the treble. We do see some Doppler distortion coming into play in the upper treble frequencies however its audibility would be in question since it hugs so close to the fundamental tones.
If we high pass filter above 100Hz we see IMD improve to -62.5dB for the treble.
Some Design Renders
I listening evaluation was done in my small listening room. My upstream components consisted of the following:
- Wiim Pro music streamer using coaxial digital out
- Hypex FA123 plate amp, channel 3 (lowest power amp module)
- No DSP or filtering I mwas used.
- Music source is Network Attached Storage, FLAC files ripped from CD
The above setup is an extremely low distortion solution based on my measurements. I measured distortion on all three channels of the Hypex and found that the third channel had the lowest distortion.
China Hok-man Yim, Poem of Thunder, Poem Of Chinese Drum
On the Hok-man Yim track the drums came through like a good 8" woofer, perhaps even more forcefully than the Fostex FF208HS woofer. Bass is palpable and nuanced, with excellent transient attack. The snap of the wood sticks held proper dynamics. Overall it is quite shocking that a 10cm driver can provide this level of bass. Even during the quiet passages where the drums are barely touched, you can perfectly hear and feel the modulations of the drum.
Spiritchaser (Remastered), Dead Can Dance, Song of the Stars
The opening sounds on this track is a low continuous bass note which comes through strong and impressive with the BLH. This is a complex track but the BLH keeps it together at medium listening level. But if I try to push it to 90dB at my listening chair 2m away then things start to sound a bit stressed in the midrange and treble.
Drums for Modern Dance Class > Coccyx Balance
The opening bass percussion is right where the cancellation dip occurs in the mid-bass and so there was little percussion energy with this track. I mention this to highlight that there are some tracks, such as this one, where the cancellation dip becomes noticeable. However this is the only one that I found noticeable after testing dozens of tracks.
I listened to a number of other albums from Ray Lamontagne, Cowboy Junkies, and The Great Lake Swimmers. All of my listening confirmed that there is excellent soundstage depth and clarity, with vocals coming out of the mix with excellent separation and intelligibility.
I had my twin brother come over to listen to the setup. He noted that the mid-bass was perfect especially with male vocals, noting that this is where many speakers fail to provide the proper resolution and energy.
Expressiveness and Spaciousness
As I listened to these BLH (No.2314) I kept thinking about the term used by Fostex to describe many of their products. The term "Expressive" is used quite frequently. But what does this mean? In the context of small speakers in a small listening room, there is nothing that can come close to the dynamics provided by these BLH speakers, especially through the bass and mid-bass. The only area falling short is in the treble, where I feel dynamics could be improved with a high frequency horn. The fullrange driver sounds clear, detailed, with excellent transient detail, however the sound in the treble region does not have a very spacious sound character. The sound is more forward and "in your face" which is directly a result of the narrowing directivity in the treble region. In contrast, the midrange has an excellent spatial quality. This further emphasizes the need to go 2-way and perhaps even integrate my ultra-wide horn into the baffle. A small horn could be accommodated directly underneath the wide band driver by shifting the spiral horn path down and around the HF driver chamber. So stay tuned for more of that in the near future!
I hope to include some sound samples of the 2314 playing some of the tracks mentioned above. My sound demo will include a direct comparison with the Markaudio CHR120 where I change nothing on the volume dial to highlight the increased sensitivity of the BLH compared to a much larger 6.5" bass reflex. The Markaudio represents a pretty common configuration since the driver has a very popular QTS (Motor strength). Consider how many 6.5" stand-mount 2-way speakers exist, especially from company's like Klipsch that claim high sensitivity from such a configuration.