In this blog post I test the Scanspeak H2606 25mm dome tweeter in horn No.2007 which I featured here and here. Originally I tested the 2007 horn with the Peerless OC25SC65-04 with retails for $22 USD each, while the H2606 is still affordable at $44.60 USD. The FS is lower on the H2606 compared to the OC25SC65-04 at 1030Hz vs 1382Hz due to the hollow pole piece. The plastic waveguide is removable on the Scanspeak by removing three small metal clips.
I begin by measuring the raw frequency response. The tweeter exhibits a very linear response until 8kHz where we see some breakup. The Peerless tweeter was more linear through this region but we are only talking (±)1dB variation.
We can see that by extending the response graph out to 50kHz that the tweeter extends nicely out to 20kHz where it falls sharply above with no breakup peaks.
Installing a 6.8uF capacitor flattens the response nicely. Moving forward the test data will include the capacitor so that we get a more realistic result in a typical application.
Off-axis is well behaved at 0,15,30, and 45 degrees.
Starting with burst decay we can see a relatively clean result with the exception of a small resonance at 9.2kHz corresponding to the small anomaly in the frequency response. Also note that since the FS is below the passband of the tweeter, we do not see it represented in the burst decay.
However looking at the CSD plot we do see a resonance centered at the FS dying down after 1.97ms which is only 2 periods at 1kHz. This is why we can barely see it in the burst decay. If we look at the rest of the bandwidth we can see that the tweeter is completely died off at less than 0.49ms. The resonance at 9kHz is about the same as what we see with the Peerless tweeter.
Measuring harmonic distortion at 85dB 1m we see predominately H2 at -61dB with H3 and H4 below -80dB.
Increasing the test SPL to 95dB 1m we see a corresponding loss of dynamic range at -51dB for H2 with H3 and H4 below -68dB.
Looking at intermodulation distortion using a 12 band/octave multitone test signal at 75dB test SPL at 1m is shown below. We see distortion averaging -85dB across the spectrum. This is the noise floor for the Focusrite Mic Preamp.
Increasing to an 85dB test signal we see -65IMD performance through it's bandwidth.
Increasing to 95dB we see a corresponding loss of dynamic range to -55dB IMD.
In a previous blog post introduced a new distortion metric which is a variation of the harmonic distortion test where the final distortion figure is weighted against human perception in terms of what we find audible and offensive. Some types of distortion are audible, but not necessarily offensive, while some are not audible, and others are audibly offensive! Moving forward I will be including this metric in my testing using Virtins measurement software.
Test Signal Parameters
Test Tone Type: Single Test Tone
Test Tone Frequency: 2kHz
Test Tone SPL level: 75dB
Gedlee Metric (Gm) Score: 0.0029
THD-A from same result: 0.0304%
For reference I've shown the Virtins software screen which uses a real time spectrum analyzer to display to the harmonic side band products as well as two windows with the corresponding Gm and THD-A values.
I should note as well that I attempted to use a two-tone multitone test signal in order to obtain the Gm value but I was getting distortion figures above "10" which did not seem correct. I can only conclude that the Virtins software does not support multitone for Gm measurement. Below is a screen shot where I attempt a 2kHz F1 tone along with a 10kHz F2 tone. The software must think that F2 is harmonic distortion and includes it in the total calculation which is unfortunate.
Increasing the test SPL to 95dB we see Gm increase to 0.0165 with THD-A at 0.3314%.
I decided to input these results in excel to see the relationship between Gm and THD-A as I increase the test SPL from 75dB, 85dB and then 95dB.
We can see that the rise in Gm is much more exponential as we increase the test SPL, rising only 65% and then 243% while THD-A increased in a linear fashion at 223% and 237% between the three test SPL levels. To me, the Gm metric more closely reflects how a driver quickly falls into harshness as we raise test SPL beyond it's comfort zone. So one point for Gm.
At this point the actual Gm value is meaningless to me in terms of what should be a target value for good sound quality. My hope is that as I test more drivers I can develop a reasonable sense for a good Gm threshold amount for retaining sound quality. Technically speaking, the Gm value can range from 0 to 100, with values below 10 having a strong subjective objective correlation based on the study results documented in Convention Paper 5891. For reference also, the ranges for Gm in the Convention Paper were between 0.34 and 23.4. The lowest Gm distortion value in the study is x20 higher than the distortion figures I was able to measure even for the highest test SPL. This means that my measurement figures are not even in the same ballpark as the distortion figures used in the study, meaning I can't use the study to correlate audibility.
The Scanspeak H2606 offers essentially the same performance as the Peerless OC25SC65-04 in terms of flat response and low distortion. The Scanpeak has a little flatter response through the lower treble and does a little better keeping the FS below the bandwidth of the driver. I plan on testing more tweeters in the weeks to come.