Speaker Asylum

Posted by Dave_K (A) on January 18, 2015 at 14:42:05

This post is in response to beppe61’s request in a prior thread for some IM distortion measurements of my Wilson Benesch Vertex speakers.  I’m starting a new thread in case others are interested.  And it would be great if one of the speaker designers lurking here could help interpret these or point me to similar measurements of other speakers for comparison.

Some background: These speakers are small two-ways on stands with a high crossover frequency of 5 KHz.  The 170 mm mid-bass driver is connected directly to the amp and its frequency response is controlled only via mechanical/acoustical means.  Its pass-band in this design is close to 7 octaves.  We were wondering whether the demands placed on the mid-bass driver by this design would have a downside of higher distortion.  I had previously tried to measure THD using long sweeps at ‘normal’ listening levels (75-85 db) and it was buried in the background noise of my listening room.  beppe61 suggested measuring distortion in the midrange while a strong 100 Hz tone was being played, i.e. IM distortion.

I have a dedicated listening room but it’s not an anechoic chamber, so background noise is a challenge when trying to measure distortion.  Yesterday I found myself home alone for a while on a quiet day with no wind, no wildlife making noise, no neighbors running power equipment, etc. After shutting off a humidifier and furnace, the room was eerily quiet. I couldn't hear anything except some very faint transformer hum from my amp, and a reminder that I have some mild tinnitus lurking up around 10-15 KHz. The conditions were as perfect as I was going to get.

The tool I used for these measurements is Dayton Audio’s OmniMic.  I have several different measurement tools, and the one I use most often is REW.  However, the RTA in REW uses a relative scale, while OmniMic’s RTA shows SPL which is what I wanted.

Before getting on to the measurements of the WB Vertex, I need to show and discuss a spurious tone in the background noise of my listening room:



It’s a little bit hard to see, but there is a low level 19 dB SPL tone right at 1.0 KHz and a series of harmonics.  I was unable to find the source of this spurious tone.  It was not audible, but it kept showing up in the measurements of background noise all over my house.  I am fairly confident that it is an artifact of the measurement apparatus consisting of USB microphone + laptop.  I suspect it originates within the laptop and is carried into the microphone via USB ground.  Either that or something else :)  Whatever the source, it isn’t high enough in level to pose a problem measuring but it does show up in the plots below so I had to point it out.

One more important thing to mention is that these are IN ROOM measurements. As any of the resident speaker designers will know, I can't use pseudo-anechoic time gated measurement techniques to measure distortion because the S/N is insufficient. I have to use long duration sweeps or steady tones to achieve a high enough S/N to see the distortion products. And since I don't have an anechoic chamber, the measured response does include the contribution from the room. That raises an interesting question: how linear is the room? I don't have an answer, but I would like to know if anybody has looked into it before. We know very well how the room can affect the frequency response, but not so much the distortion. I don't know whether it is a significant contributor or not. And since I can't remove the room, I decided to make these measurements from the listening position with the speakers in their normal location.

All plots show absolute SPL from a single speaker measured at the listening position ~3m from the speaker.

The first plot is a 100 Hz test tone at 80 dB SPL at the listening position. I chose this level by ear because it seemed typical of what I would encounter during a moderately loud listening session. With average levels around 85 dB SPL, it would not be atypical to hear a 80 dB SPL note on the D or G string of a bass guitar. Here is the result:



There are 2nd, 3rd, 5th, and 7th harmonics present in the plot above, although fairly low in level. I actually expected distortion to be a bit higher at this frequency.

Next is a 1 KHz test tone at 80 dB SPL at the listening position (3m):



The third harmonic is dominant at -51 dB below the fundamental. I'm not sure whether the 2nd harmonic in the above plot is actually real since it's at the same level as the spurious tone mentioned above. Somewhat surprisingly, the distortion at 1 KHz is a little higher than at 100 Hz.

Now that you've seen those, here is a 100 Hz and 1 KHz tone mix, with the level set for 80 dB SPL @ 1 KHz:



Three intermodulation distortion products popped up at 900, 1100, and 1900 Hz. The highest is 54 dB down from the fundamental, which is a little below the level of the 3rd harmonic.

At this point, the results were looking a little too benign, so I repeated them with the levels raised by 10 dB. Unfortunately, the 100 Hz tone at 90 dB SPL started making my baseboard heating register resonate audibly, which screwed up the measurements. So I had to wedge some pieces of foam in the heating register to stop the metal from vibrating, and then chase down a couple of other rattles around the room. Then I got this result:



You can see the second harmonic is now only 48 dB down, so it is rising more rapidly than the fundamental. That is too be expected.

Here is 1 Khz @ 90 db SPL at the listening position:



The 3rd harmonic is still dominant, but now you can see the 5th starting to pop up as well.

And last but not least, here is the 100 Hz + 1 KHz tone mix again, this time with the level set for 90 dB SPL @ 1 KHz:



Finally, we see the spray of IM distortion products raise their ugly head. The highest IM product at 1100 Hz is 40 dB down, which is higher than any of the harmonic distortion products. -40 dB equates to 1% distortion.

I was going to raise the levels another 10 dB to see if the IM distortion was still higher in level than the harmonic distortion, but my room couldn't take 100 Hz @ 100 dB SPL without rattles and vibes which would have polluted the measurements, and my ears couldn't take 1 KHz @ 100 dB SPL. Even listening to a continuous 1 KHz tone at 90 dB SPL gets pretty annoying after a while and by the end my ears were starting to ring a little bit by the end.

Here are some of my takeaways:

1. At the lower of the two measurement levels, THD @ 1 KHz was higher than IMD @ 1 KHz. But at the higher of two measurement levels, IMD started to exceed THD, at least by a little. I expected both to increase, but did not expect IMD to increase faster than THD. Now I am really curious to know whether this is the case with other loudspeakers. It does make me wonder whether IMD is being neglected.

2. I was surprised by the low level of 100 Hz distortion. I've read a lot of Soundstage reviews of small two-way stand mounted loudspeakers and looked at their distortion measurements in the NRC chamber. They typically show higher levels of THD at 100 Hz measured at 90 dB @ 2m in the chamber than I measured at 90 db @ ~3m in room. It makes me question my results.

Unfortunately, I don't have much of a basis for comparing these results to other loudspeakers. Soundstage is the only place where I could find distortion measurements of high end loudspeakers, but they only measure THD and their distortion plots only have 45 dB of dynamic range. When THD+N is more than 45 dB down, it falls off their scale. Most of the DIY measurements I've seen are for subwoofers or drivers only, not complete loudspeakers of comparable size, and a lot of them are just showing the background noise floor. About all I can conclude is that there is no indication so far that Wilson Benesch has sacrificed distortion performance for bandwidth in this driver design.