In this blog post I feature a new product segment starting with Horn No.1928.This horn is distinguished by it's shallow design.
Rationale
About a year ago I was designing a 3-way floorstanding speaker which used a 4" midrange and waveguide loaded dome tweeter. I recessed the midrange 40mm into the baffle. In order to do this I had to create a shallow waveguide for the midrange. This project is featured here.
I noticed that the Scanspeak 12MU worked well as a fullrange driver. I also noticed that it had a nice forward presence through the vocal region common to horns. I suspected that the shallow waveguide was helping provide some clarity compared to simply flush mounting the driver to the baffle.
In addition to this, I recall having a discussion with Victor Kung (VK Music) who is the North American distributor for AER fullrange drivers. He mentioned to me that the AER Jazz speaker provides the best sound for his AER drivers. He provided the physical dimensions of the horn which revealed that it was not very deep at all.
It's almost impossible to tell from the photos but the horn is actually quite shallow. I found this image on the web that gives an idea about the shape. I should also note that the horn is non-concentric between the throat and mouth. This would help reduce and distribute edge diffraction. I could possibly do this in a future iteration. But I find that most of the trouble comes from immediately inside the throat area.
So I decided to make my horn square with my own ES horn flare curvature which wraps around to the back of the horn. I was initially going to make my prototype from stacked layers of birch plywood, but later decided to make the horn from solid walnut. I decided to make two pair as well. Just in case somebody wanted a pair.
I purchased three full range drivers to test in the waveguide.
This is a relatively new driver which I had not tested yet. The published response looks good with the exception of the surround resonance at 2kHz. So I figured I would give it a try.
Fostex FF85WK
The published response on the FF85WK looks trouble free with the exception of a small peak at 10kHz. This appears to be an on-axis issue only. The upper treble extends to almost 30kHz which is a good indication that the driver will have very fast transients. Also, this driver is one of Dave's favorites at Planet Ten Audio here in Canada. I also have experience with the bigger FF165WK. I also like the appearance of this series. I hope Fostex continues with this series for many more years.
Markaudio Alpair5.3
The Alpair5.3 is a unique driver in that it has no spider. I used this woofer in my 1900 project and so I decided to try it in this project as well. From published it almost extends to 30kHz as well.
Rear Chamber
I decided to have a sealed rear chamber 3D printed so that I could get accurate off-axis measurements. The rear chamber will also help to control xmax and raise the FS to just below the cutoff frequency of the horn (Fc). This will provide the optimal transient response for the driver. I also designed the rear chamber so that it has no parallel walls. This will eliminate standing waves inside the enclosure. I also included a deflector at the back of the chamber to breakup the rear wave preventing it from reflecting back through the driver.
The rear chamber volume is sized for a QTC of 0.70 which is optimal for driver damping and control.
Throat Transition
The biggest aspect to achieving a flat response is creating a smooth transition near the surround. If there is a cavity or abrupt physical transition at the surround then there will be severe frequency response anomalies likely in the upper treble. To mitigate this I had a ring 3D printed that follows the shape of the surround and partially covers it. This allow the wave front coming off the driver cone to transition into the horn smoothly.
Context
We are not going to achieve the same frequency response flatness as a compression driver, so its important to have realistic expectations when horn loading a cone type driver. The diaphragm is already a very irregular shape and so we are going to see irregularities in the response.
Why such a big horn?
From my reading of Floyd Tool's book on Small Room Acoustics, I read that there is a region in time that the brain has a hard time processing secondary images. This region is in the first 1 millisecond upon arrival of the initial direct sound image. If other images arrive within one millisecond then we perceive this as a blurring of transient detail. So in my design, the goal is to have a very clean initial step response for the first millisecond. This means that the sound wave propagates undisturbed for 30cm or more. This requires that the horn be of a certain size, regardless of it's intended bandwidth or loading. I settled on a 40cm x 40cm horn. If the horn is placed at least 30cm away from any other adjacent objects, then the goal is maintained. This includes the bass enclosure below it, which is why I've elevated the horn above the bass cabinet in my testing. This defies the golden rule of keeping driver center to center distances to a minimum, but I've prioritized other design elements instead.
Measurement
Fostex FE83NV
The response on the FE83NV did not provide satisfactory results. The loading simply did not recover from the severe surround resonance at 2kHz. We also see severe anomalies in the upper treble. This cancellation is likely due to the inverted surround forming an air gap void.
Markaudio Alpair 5.3
Below is the result for the Markaudio and we can see similar issues in the upper treble that is also a result of the inverted surround. Although it does not suffer from the surround resonance being much more mild at 2.2kHz.
Fostex FF85WK
Below is the response for the FF85WK.
Subjectively I found the FF85WK to sound the best. So I continued further development using this driver.
I flattened the response using DSP which is shown below as an overlay. I will use the flattened response for the time domain, distortion, and off-axis measurements.
If we look at the step response we can see that we have a very clean result after the initial impulse of sound. We don't see anything until 6ms later. This was the design goal mentioned earlier.
Below is the impedance sweep using the sealed rear chamber which confirms the QTC of 0.70 (bottom left of graph).
Time Domain
Distortion
Below is the harmonic distortion for 85dB and 95dB for a 1m distance.
Intermodulation Distortion
Below is the IMD for 75, 85, and 95dB at 1m.
For the 75dB test signal we see -60dB across the spectrum.
At the 85dB test signal we see IMD generally around -55dB. We don't see any trouble spots along the spectrum.
Increasing the test signal to 95dB we see IMD at -50dB IMD.
Gedlee Metric (Gm)
I collected the Gm distortion data for the FF85WK and compiled the data along with two other drivers for comparison.
Generally the THD numbers for the FF85WK are the least performing compared to the other two drivers. However the tables turn when looking at Gedlee Distortion (Gm) where the FF85WK offers the lowest distortion of the three.
I then decided to test Gm for a lower listening level. Below is the values for 75dB at 1m. However I do not have any comparison data for context, I will likely refer to these numbers in future blog posts.
I then decided to measure increase the test tone frequency from 1kHz to 10kHz just simply for future record.
Perhaps I should contact Earl Geddes and ask him if 0.0014 Gm is a good number!
Off-Axis Colored Polar Map
Below is the initial off-axis colored polar map for 5 degree increments out to 80 degrees off-axis. However there are some peaks in the response 15 degrees off-axis which is scaling the colors incorrectly. So I have a few options in terms of changing the graph settings.
If I change the smoothing from the highest seting (1/12dB/octave) to 1/3dB per octave, I get the following results. This embellishes the results a little but gives you a sense of the coverage pattern. We have an 80 degree listening window at 15kHz which is really good for this type of solution.
Changing the graph settings so that we only see up to 11kHz is shown below. I also changed the smoothing back to the more revealing 12dB/octave setting. Generally we can see a very well behaved off-axis with pattern control down to 200Hz.
I decided to show the frequency response results with 15, 30, & 45 degrees off-axis as an overlay.
Honing in on the response at 15 degrees off-axis you can see that things are actually looking pretty good. We see solid extension out past 20kHz down -15dB at 30kHz.
Subjective Listening
When setup as a stereo pair crossed at 200Hz to an 8" bass cabinet, I immediately noticed the wide soundstage. There was certainly no "head in a vice" aspect to the sound.
Soundstage Depth --- 9/10
Soundstage Width --- 9/10
Smoothness --- 9/10
Coherence between midrange and treble --- 9/10
Coherence between mid-bass and midrange --- 7/10
Vocal Clarity Male --- 8/10
Vocal Clarity Female --- 10/10
Accurate Musical Instrument timbre --- 9/10
Sense of Dynamic Range: --- 8/10
Conclusion
Shallow horn No.1928 along with the Fostex FF85WK provides great overall performance. There are some response anomalies in the upper treble when listening directly on-axis, however things clean up slightly off-axis. Perhaps Fostex can work on an elliptical driver?