In this blog post I test the Celestion Axi2050 compression driver. I test it on my 300Hz Biradial (ES290 No.1670).
I began my measuring the impedance sweep which is shown below.
The raw on-axis frequency response is shown below. Sensitivity is excellent at 110dB shelving down -5dB at around 5kHz. Treble extension is good until 17kHz. There appears to be mild breakup starting at 7kHz.
In preparation for further measurements I padded the driver down by -10dB with the circuit shown below. This is just a starting point and I may change this later depending on how well it matches up with my bass cabinet.
I then implemented a contour circuit to flatten the response in the midrange. The resulting response is shown below. (purple response)
Off-axis which is shown below at 0,15,30, & 45 degrees. The ES-290 holds good pattern control off-axis across its operating bandwidth.
I then conducted harmonic distortion which is shown below at 85, 95, and 105dB SPL at 1m. The microphone was placed 20cm from the horn mouth. K3 and K4 is low however we see a sharp rise starting at 12kHz.
I then optimized the crossover for integration with bass cabinet No.1798 using a simple first order crossover at 500Hz. I ended up using a very moderate L-pad since bass cabinet 1798 is already very sensitive at 101.5dB@1w.
Adding in the bass portion to complete a finished system is shown below. The response for the bass is shown ungated and so it includes nearby reflections in the response. So the graph provides a general representation of the system response.
I then measured the intermodulation distortion for the high frequency portion using the above crossover.
Side note: I attempted to cross lower than 500Hz but the overall sound quality was somewhat congested. This was also confirmed with IMD tests (not shown). A larger horn with a lower cutoff frequency would be required if wanting to go lower in frequency with this particular driver.
We have excellent results in the midrange between 500Hz-2kHz showing IMD -78dB. However we see a steady climb in IMD starting at 2kHz rising to only -60dB at 15kHz. For comparison’s sake, a recent tweeter tested here had IMD at -68 dB for the same test SPL. An earlier test found here had IMD at -78dB. The Axi2050 does not achieve my target for audiophile sound quality, at least not for the high frequencies.
As we can see from the above results, IMD is -50dB for the 95dB test, and only -40dB for the 105dB test when looking at the treble region. Distortion is only acceptably low in the 500Hz-1kHz region.
I then measured the intermodulation distortion of the complete system. (Includes bass cabinet No.1798)
The below tests are shown in sequence at 85,95, and 105dB SPL test signal levels.
We can see from the above results that the Axi2050 starts out favourable on the 85dB test, however by 95dB the distortion quickly rises well above that of the bass and midrange. By 105dB distortion is almost +15dB higher than the bass (100Hz vs 15kHz).
The step response for the system (LF + HF) is shown below. This isn’t relevant to the Axi’s performance but does confirm excellent physical time alignment between the LF driver and HF driver.
Burst Decay for the Axi is shown below. Generally I like to see any resonances well suppressed by 8 periods, however we see artifacts persisting until 18 periods in the 15kHz region.
The above CSD shows multiple small artifacts across the region from 5kHz upwards.
Testing on Large Format Midrange Horn No.1723
I then mounted the Axi2050 on horn No.1723 which is shown below. This horn uses the ES curvature out to the horn mouth and then abruptly ends where the flare geometry reaches 90 degrees to the horn axis. Without the wrap-around geometry beyond the horn mouth, we will see more artifacts in the frequency response as well as the off-axis polars. However the horn was designed to be a reasonable size.
The raw frequency response is shown below for 0,30,& 45 degrees off-axis. We see a reasonably linear response from 400Hz to 3kHz where we see a 12dB/octave falling response above.
For comparison, here is the response of the BMS 4599HE on the same horn. 
I then tested intermodulation distortion at 85dB and 95dB test SPL. Distortion is very low at -80dB for the 500Hz region rising to -70dB until 5kHz where things diminish slightly to -65dB.
Increasing the test SPL to 95dB we see distortion remain unchanged at -80dB for the 500Hz region increasing to -60dB by 5kHz, concluding that the driver meets our target for low distortion, at least for midrange applications.
Subjective Listening Impressions
Below is my ranking for my subjective listening impressions on the sound quality for the ES290 Biradial. I did not evaluate the subjective sound quality for the larger 200Hz horn No.1723.
I've divided up my listening impressions into two groups, 85dB listening level and 95dB listening level. I did this because the driver's sound quality changes significantly when output increases from 85dB to 95dB.
85dB Listening Level Results (About 80dB at the listening position 2m way)
Soundstage Depth --- 8/10
I found soundstage depth to be quite good. This might be a result of the single diaphragm configuration of the Axi.
Soundstage Width --- NA
I evaluated the sound quality with a single speaker in a large space with no side walls and so I could not evaluate soundstage width.
Smoothness --- 7.5/10
Even at the lower listening level I found the Axi to have a somewhat strident character on acoustic guitar, this being on a recording that should be smooth.
Coherence between midrange and treble --- 9/10
This was more a virtue of the physical time alignment of the bass driver and compression driver, along with the first order crossover. Coherence was on par with a single fullrange driver. The Axi maintains that coherence which is commendable.
Vocal Clarity Male --- 10/10
Again, this is more a virtue of the 1798 bass cabinet which uses the TAD TL-1601B bass driver.
Vocal Clarity Female --- 8/10
The Axi does very good on female vocals in terms of clarity, but I detected some strident character as mentioned earlier.
Accurate Musical Instrument timbre --- 8.5/10
Acoustic guitar and the ukulele came through sounding almost real, but it was let down by issues mentioned earlier.
Sense of Dynamic Range: --- 9/10
Dynamics seemed very good through the midrange, however suffered in the treble taking on a softer character. Leading edge transients in the treble left me wanting to add a super tweeter.
95dB Listening Level (about 90dB at the listening position 2m away)
Soundstage Depth --- 7/10
At the higher listening level there was a noticeable loss here.
Soundstage Width --- NA
Smoothness --- 6/10
The loss in smoothness left me with the urge to grab the volume knob and turn things down.
Essentially this would impress nobody.
Coherence between midrange and treble --- 9/10
Vocal Clarity Male --- 10/10
Vocal Clarity Female --- 7/10
Accurate Musical Instrument timbre --- 8/10
Sense of Dynamic Range: --- 8/10
Loss in dynamics seems to spread to the midrange as well.
Conclusion
With the driver mounted on the ES290 Biradial and contour circuit in place to flatten the response, the Celestion Axi2050's behavior changed drastically at elevated listening levels. It went from pleasant and musical to somewhat strident in nature at the higher listening level. This is in the context of using the driver for midrange and treble duty.
It’s clear that the Axi2050 performs better on a large format midrange horn as a midrange only solution, as we see with the 1723 horn. Distortion is within my target for audiophile applications, although the response linearity was not matched compared to the BMS 4599HE, although it was still acceptable.