Musings on Loudspeakers, Part 1 (of 4): Motivation

[kaiser_soze]

Musings on Loudspeakers, Part 1: Motivation

I’ve listened to many speakers over the years, including some that sounded significantly better to me than my pair of Advent 5012 speakers from the early ‘80s, and including some that sounded every bit as pleasing as that sound you get when you scrape your fingernails on a chalk board. In recent months, I started thinking again about buying a new pair of speakers. That was motivated in large part by the fact that my Sennheiser HD-580 headphones sounded so much better to me. About three years ago, I bought a Cambridge Soundworks Newton MC500 to replace the center channel that I had been using. The MC500 was not a good timbre match for the Advents, but it was closer than the other center channel that I took home to listen to, so I kept the MC500. Almost immediately, I noticed that the MC500 sounded uncannily similar to the HD580 headphones, except for not having deep bass of course. I never attributed the similar sound of the MC500 and the HD580 to more accurate timbre though, and being inquisitive by nature, I wanted to have an objective, quantitative understanding of why the MC500 and the HD-580 sounded similar, and the Advents sounded dissimilar. I was reluctant to buy a new pair of speakers until I could find a way to be confident that I would not make the mistake of replacing them with speakers less accurate. I also wanted to improve my understanding the theoretical limitations of 2-way designs that use low-order crossovers.

 

[kaiser_soze]

Musings on Loudspeakers, Part 2 (of 4): Theoretical Considerations

Back in the ‘70s and ‘80s, you would frequently see speakers and headphones tested using square waves, and there was a lot of talk about phase-coherency and time-alignment, which are related concepts. Phase coherency and time alignment aren’t emphasized nearly as much these days, and I don’t even recall the last time that I saw an image of the output of a speaker driven by a square wave. The explanation seems to be that, given a complex waveform consisting of two or more pure sine waves, scientific investigations into how we perceive sound cast doubt on whether anyone can hear the difference when the relative phase of the individual tones is varied. That doesn’t surprise me, because that capability is fundamentally separate and distinct from the mechanism that we use to localize sound, which relies on detecting the phase difference between the arrival of a wave at the two ears. It is difficult to imagine why humans would have evolved the capability to detect phase differences between two different frequencies arriving at the same ear.

Yet, phase coherency remains important when it influences frequency response, and that is why acoustic phase coherency between two drivers is important in their overlapping frequency domain. If each of two drivers is expected to contribute one-half of the total sound energy in the vicinity of the crossover frequency, but the acoustic waves are out of phase at a given listening location, acoustic cancellation will occur at that location for specific frequencies in the octaves above and below the crossover frequency.

If two drivers move in and out in unison, the acoustic output will be in phase at any location that is equidistant from the two drivers. But when the listener is not equidistant from both drivers, cancellation will occur at specific frequencies within the overlap region. Moreover, unless all the drivers have flat faces that are aligned, perfect acoustic phase coherency in the crossover region will be elusive even if the electromotive force experienced by the voice coils is phase coherent. Aligning the voice coils won’t eliminate the effect of different parts of the woofer cone being at different distances to the listener. That effect will be insignificant as long as the wavelength is substantially longer than the depth of the cone, but if a woofer having a 2” cone depth is filtered at 3 dB per octave, in all likelihood its output will remain significant even at wavelengths approximating the woofer’s depth. The theoretical benefits of using multiple woofers of smaller size, has to do with more than directivity.

The counter-argument is that in a 2-way design using a first-order crossover, the outputs of the two drivers will make up for the fact that neither driver will exhibit a flat frequency response over its operating range. The criticism of the problem with drivers not being uniform in frequency response over their operating range is certainly valid, but the allegation that this can be solved by allowing two dissimilar drivers to overlap throughout the region between the centers of their respective operating ranges, is dubious. The low-order crossover is accompanied by phase shifts that vary according to the frequency, so it is not possible to keep the electromotive force experienced by the two voice coils in a consistent phase relationship over the overlapping frequency range covering multiple octaves to either side of the crossover frequency. The drivers will move in and out in unison at specific frequencies, but not over a usefully broad range of frequencies.

 

[kaiser_soze]

Musings on Loudspeakers, Part 3 (of 4): Measuring

Anyone who has tried to measure the frequency response of a speaker knows that this is no easy task. I used the test CD from Rives Audio, which has tones spaced at 1/3 octave, and even includes a set of tones that are supposed to compensate for the anomalies of the Radio Shack SPL meter. Reproducibility is the key to knowing if you are really measuring anything useful. If you move the speaker to different locations having different relationships to the walls, then if you get the same measurements when you use the same position for the meter, you can be reasonably confident that you have actually measured the response of the speaker.

I started by measuring the frequency response of several of my headphones. That was easy enough except that I couldn’t get a useful measurement of the bass response as you hear it, because I didn’t have a way to reproduce the effect of the space between the headphone and the head being closed off. But I was able to get some useful measurements of the HD-580, and my suspicions were confirmed. The response follows an essentially straight line from 100 Hz to about 2.5 kHz, with a very slight steady rise with increasing frequency. At 2.5 kHz, it starts to climb abruptly and there is a +6 dB plateau between 3.5 kHz and 5 kHz, then a dip to +2 dB at 6.5 kHz, then another 6 dB peak at 8 kHz, followed by a smooth decline and not descending below –2 dB even at 20 kHz. The frequencies covered by that region of excitation are associated mostly with the higher harmonics that various musical instruments exhibit as the fundamental frequency decays. By exaggerating those frequencies, the HD-580s enhance the harmonic qualities of various musical instruments, as do many other headphones, including the Sennheiser products in general. This specific deviation from the ideal behavior of a transducer is a major reason why many people like them. The example that will be familiar to many people is the squeaking sound made as the guitar player’s fingers slide along the strings over the frets, which produces a distinct sound partly because of the excitation of longitudinal waves, lengthwise along the string.

The measured response of the MC500 was remarkably similar to that of the HD-580, particularly from 100 Hz to 2 kHz. The MC500 remained flat out to about 5 kHz, and then it presented a single, smooth peak of about 6 dB between 5 kHz and 8 kHz, centered at 6.5 kHz. It then returned to level between 8 kHz and 14 kHz. Except for that smooth peak between 5 kHz and 8 kHz, and ignoring the very gradual straight tilt that yields an increase of about 2 dB by the time you reach 5 kHz, the response easily stays within a +/- 1 dB window from 100 Hz to 5 kHz. This makes that peak especially unfortunate, because were it not for that, the frequency response of the MC500 would be nothing short of amazing. The reason for that peak is that it is the natural resonance of the tweeter in its enclosure. The silk dome is coated with some sort of substance that is no doubt there to preclude the set up of any standing waves on the surface of the little dome. While I applaud the avoidance of those nasty metallic sounds, I wish that this had been accomplished through a means that did not cause the tweeter’s natural resonance to occur smack in the middle of its operating range. Much as I dislike this elevated tweeter response, given a choice between it and the shrill sound that the greater majority of metal domes make due to having dimensional resonances within the audible range, I’ll take this any day. Cambridge Soundworks uses this same tweeter on all their full-range speakers, along with either the same midrange or a similar midrange of slightly different size. In all likelihood, other Newton speakers will have a similarly uniform response with the same exaggerated treble at the tweeter’s natural resonance. Many listeners will not find that objectionable, and may even appreciate it, especially those who appreciate the sound of Sennheiser HD-580 headphones.

The on-axis response of the Advent 5012 exhibited some unevenness, which is likely due to the drivers not maintaining a consistent phase relationship throughout the overlapping region. Yet, the response still managed to stay contained within +/- 3 dB from 50 Hz to 10 kHz. Moreover, the response follows a smoothly rounded plateau over that region, and if you apply a broad correcting curve to flatten out that very broad rounding of the wide plateau, then the response containment would be within +/- 2 dB over that wide range. As you move off-axis, particularly vertically off-axis, the irregularities became more pronounced, but in the room response, those zigs and zags tend to be relatively inaudible as long as the average response within each octave is consistent from one octave to the next. All in all, notwithstanding the theoretical concerns, the Advent 5012 still holds its own against a whole lot of speakers that cost a whole lot more, just like it did back in the day. (It’s okay for me to say that, because I don’t say it about every speaker I come across.)

I also did some listening tests using pink noise. The change in timbre, as you move vertically off-axis was inescapable. At first, I attributed that to the most obvious of all possible explanations, i.e., the effect of energy declining with distance coming into play as I varied the relative distances from the two drivers. But when I moved steadily away from the speaker while maintaining the same off-axis angle, the sound quality associated with that off-axis angle didn’t change. That tells me that the explanation for the observed differences in sound quality at different vertical off-axis angles likely has more to do with phase cancellation than it does with either the fall off in energy with distance or with the woofer’s directivity. I now have the Advents sitting upside down on a pair of small tables that I use for speaker stands, which places the woofer’s center closer to ear level, and which makes the distances to the tweeter and the woofer’s voice coil more equal.

One theoretical consideration that I haven’t said much about, is the way that different speakers vary in how well they permit the listener to discern the location of individual instruments and voices along the one-dimensional line between the two speakers. If the frequency response of the speaker is not consistent as you move off-axis, the frequency response of the two speakers will differ enough to make it difficult for the ear to localize the individual instruments. The harmonic output of a given instrument will be ever so slightly smeared across a horizontal space between the two speakers. While this effect is real, I feel that its importance is greatly exaggerated except possibly for speakers that sound so bad that no one would want to listen to them anyway. Not that the consistency of the frequency response off-axis isn’t important. It’s just that the reason it is important is that you want a flat frequency response no matter where you sit, not because of the barely perceptible effect that it may have on localization of individual instruments.

The Advent 5012 manages to defy the theoretical concerns for the most part, but I still feel that those theoretical concerns are entirely legitimate, and as I continue my search for new speakers, my understanding of those concerns will help me to make an informed, educated decision. The argument for those tall skinny towers that use a vertical array of drivers, especially when they use high-order crossovers with multiple midranges surrounding the tweeter, is compelling.

 

[kaiser_soze]

Musings on Loudspeakers, Part 4 (of 4): Tweaking the MC500’s Crossover with Pencils

If you aren’t laughing, shame on you. You do not yet know how I’m going to use a couple of pencils to improve the accuracy of my MC500 center channel, so you should definitely be laughing.

Even though the similarity of the sound of the MC500 and the HD-580 headphones was inescapable, the timbre of the MC500 did not match the Advents, and this was distracting when listening to anything with center channel content. That included the handful of Dolby Surround CDs that I bought before DVD and before Dolby Digital 5.1 appeared on Laser Discs. Those are still some of my favorite CDs, and while I liked listening to them using the MC500 for the center, it was apparent that the timbre of that speaker was excessively “bright”. After I performed the measurements, I had a precise understanding of what was going on. Being in the hunt for new speakers to replace the Advents, I would need to decide whether to replace the MC500 along with them, which I was considering even though I hadn’t had it very long. I decided to try tweaking the crossover before deciding whether to replace it.

To reduce the output of the tweeter, you merely have to place a resistor of appropriate value in series with the tweeter, but that increases the impedance of the tweeter circuit relative to the woofer circuit, which will shift the crossover point to a higher frequency. To compensate for that and maintain the crossover point at the same frequency, you have to place a second resistor in parallel with the tweeter and the first resistor, so that the resulting parallel resistance or impedance is the same as that of the tweeter alone. Ideally, you would measure the complex impedance of the tweeter and compute the resistance at the original crossover point, but not having an impedance bridge, I had to hope that calculations based on the tweeter’s DC resistance would suffice. Looking at the frequency plot, I determined that if I reduced the tweeter output by 3 dB, the peak would then be at +3 dB relative to the midrange response, and it would be preceded by a dip that would reach –3 dB, with the roll-off above 10 kHz occurring a little sooner.

To reduce the power by 3 dB, you have to reduce the square of the voltage across the tweeter by half. I ended up subtracting 1 from the square root of 2 and multiplying that by the resistance of the tweeter, which told me that the resistor that I would put in series with the tweeter would need to be 1.65 Ohms. To find the resistance for the parallel resistor, I added 2 to the square root of 2 and multiplied that by the resistance of the tweeter, and got 13.6 Ohms. I had long since forgotten the resistor color codes and the standard logarithmic sequencing of resistor values. I had taken the plate that has the tweeter and midrange off the MC500 in order to measure the resistance of the tweeter, and once I start something like this, I don’t like having to wait for a couple of small resistors to be delivered. You can try calling the local electronic supply houses, but these days they tell you that they only sell on-line and no longer even have counter service. Times have changed.

I wondered whether I might be able to make resistors from some wire that I had lying around, but even using 28-gauge copper wire, I would need 25’, and that was for the 1.65-Ohm resistor. I started looking around, and when I pulled open one of the drawers where I keep an assortment of tape and glue and small parts and such, I spotted a few pencils. Pencil lead is made of graphite of course. I cut off both ends to expose the graphite and measured the resistance for the entire length. I determined that while the length that would be needed for the small resistor would be too short to be able to achieve the desired accuracy, if I scrapped away about half of the “lead” over most of the length of that pencil, I could use it to make the 13.6 Ohm resistor. Then I remembered that I had a couple of those fat pencils for little kids, and after measuring its resistance, I determined that I could make the 1.65 Ohm resistor with a few inches of it, without scraping the lead. I’ve never previously heard of anyone using pencils to correct the exaggerated output of a tweeter, so this may well be a first.

Truth be told, it took me a couple of hours to make the resistors. I had to cut the wood very carefully to expose enough of the lead to insure reliable contact, but not so much that I would end up breaking the pencil. I had to be especially careful with the regular sized pencil, where I needed to scrape off a little more than half of the lead over most of the length. Wait, you are wondering how I made contact with the graphite, right? Using small alligator clips, of course. I pressed their little teeth into the soft wood on the opposite side from the exposed graphite, and then used some very fine elastic cord to secure each alligator clip and taped it all over carefully. After some soldering, and some heat shrinking and some more taping, I used paper package tape to secure both of them to the inner reaches of the sub-enclosure used for the midrange/tweeter assembly, then dropped the mounting plate in place and put the screws back in.

I thought that maybe I should test it before wiring it back up normally, but I went ahead and set it on top of the rear projection television and quickly found the Telarc CD “Switched on Bach 2000”, which is encoded in Dolby Surround. It definitely sounded much better, but I was surprised at how much of a difference it made. I switched on the white noise test tones for Dolby Surround, and whereas previously the difference in timbre between the MC500 and the Advents was like night and day, they were now much better matched in timbre, albeit hardly perfect. I repeated that with the 5.1 test tones generated by the DVD player just to check the level matching, and again was pleased at the timbre match evident in the white noise test tones.

In order to measure the change in frequency response, I had to take it back off the television and set in on the stool, connected to one of the main channels. By this time I was very proficient at running through the 31 test tones quickly using the fast-forward control and jotting down the numbers. I plotted the results, and it was exactly what I had anticipated. Whereas previously the response was essentially flat nearly to 5 kHz, now there was a visible drop starting at about 2 kHz but very gradual to about 3.5 kHz, where it became a little steeper but not reaching –3dB until 5 kHz. Then it turned sharply up and reached a +3dB peak at 8 kHz. Just as I had planned, the area under the negative deviation is nearly an exact match for the area under the positive deviation. The level at 16 kHz was reduced by about 2 dB.

I had to increase the level slightly for both the Dolby Surround, which is set at the receiver, and for Dolby Digital, which I set at the DVD player due to my using the decoder in the player. With the treble reduced and the level increased, the smooth middle bass and midrange of this speaker comes forth like it never did before. I had no idea that it could sound as good as it now sounds. I was always puzzled why a center channel speaker this large didn’t make any more bass that it did. I wouldn’t have thought that such a simple change could make such a big difference, but the difference is enormous. Whereas previously I did not like listening to regular stereo with the Dolby Surround decoder engaged, I now prefer that, especially for television programming. (I have HD cable but only use a stereo audio connection, which will be changing as soon as the Denon AVR-2807 finds its way to the store shelves.) If there were a way to effect a similar change to the sound of the HD-580 headphones, I would do it in a heartbeat.

I started out thinking that I was going to replace my Advent 5012 speakers with new speakers that would match the timbre of the MC500 center channel speaker and the HD580 headphones. I ended up turning the Advent speakers upside down and installing a couple of pencils inside the MC500. It just goes to show you: you never know where you’re going ‘till you get there.

- Kaiser Soze

 

[Sheep]

OMG wallhacks! its WmAx's clone!

SheepStar

 

[corey]

I skimed it, but couldn't find a reason to read the whole thing.

 

[kaiser_soze]

Ah, but Corey, the question is, were you *looking* for a reason to read it, or were you looking for a reason to let others know your opinion on something?