Derek Hughes has worked in the audio industry for over half a century. He is one of the United Kingdom’s premier loudspeaker designers in both the professional and consumer sectors. When Spendor, the company his father started, was sold, he became a freelance consultant whose services were employed by many audio companies throughout England and Europe, including award-winning designs for Harbeth, Stirling Broadcast, and Graham Audio. More recently with Graham he developed a high-power, high-quality music and effects loudspeaker for the Royal Opera House in London. The following interview was conducted via Zoom in November 2021.
PS: How do you go about designing a loudspeaker?
DH: Let me describe my process. The first thing you start with is size. Do you want a big loudspeaker or a small one? Then the drive unit. Very simple things like that. You know what frequency range you want the drive units to cover, so you select or design or, in our case, a bit of both to suit that spread of frequencies. But immediately you are faced with the fact that there is no such thing as a perfect drive unit. Real Life is that drive units are extremely imperfect devices which have all sorts of resonances, frequency anomalies, phase anomalies throughout their band. So compromise is the name of the game. You’re always working with something with known restrictions and problems—you cannot eliminate them, but you can minimize them at different stages in the design process. The same is true of crossovers.
PS: Since you started with the box, in another interview you described the thin-wall construction as an “engineering solution”. Can you elaborate?
DH: A lot of the loudspeakers at the BBC [British Broadcasting Corporation] were primarily directed at accurate speech reproduction. But the areas in which thick-wall cabinets and cabinets in general resonated were often up in the sort of five, six, seven hundred Hertz areas, which interfered with accurate speech reproduction considerably. The end result of the thin-wall damped design is it that it reduces the Q of the cabinet, that is, how long the resonance or resonances last. It also pushes the major resonances down into the hundred-fifty to two-hundred Hertz area where they are out of major speech regions. Voices cover broadly two hundred to three hundred Hertz up to eight kilohertz, and if you don’t get that right, then you haven’t got anything right, basically. The voice has to sound right is the key. Then you extend out to other regions of the spectrum.
PS: The LS5/5 represents a significant departure from the designs you’re known for, being, first, considerably larger; second, a three-way; and third, slot-loaded, which, if I’m not mistaken, has been absent from speaker design for a very long time.
DH: A departure, maybe, but you can also look at it the other way around. The midrange driver of the LS5/5 is essentially the same driver we use in the LS8/1. From a strictly engineering point of view, the 5/5 is basically an 8/1 with extended, more powerful bass.
P.S. But even apart from the bass, they don’t quite sound identical.
DH: The 8/1 does as good a job as I was expecting from it—even I was surprised! But if you put it and the 5/5 in a big room with a big, decent sized amplifier playing material with a significant amount of really clean bass, then you will notice the difference, more dynamic range, and generally more authority to the sound. You can’t escape from the multi-layered three-way design if you want to extend the bass end any with a significant amount of power.
PS: And the slots, which widen the dispersion?
DH: I’m not keen on super-wide dispersion. The BBC approach was to get more consistent off-axis response over a certain area, which in the case of the 5/5 is around sixty degrees. You have to keep in mind that most of our speakers are BBC licensed and designed as professional studio monitors. So if you imagine a large mixing desk, the engineers want a more consistent off axis response as they move from side to side. By the way, that’s also the reason the tweeter is mounted between the midrange and the woofer, as opposed to the conventional way on top—so that it’s optimally placed for the mixer’s ears at the console on the stands in use at the BBC. There’s about fifteen to twenty millimeters between the top of the tweeter and the bottom of the midrange slot, which is more or less the optimal vertical axis in terms of the integration of the two drivers. This also allows for an optimal height in most domestic environments.
PS: What about the argument that wide dispersion adversely affects precision of imaging?
DH: I believe I can tie those two things together. Put simplistically, stereo imaging has basically two parts. One, obviously, is an accurately matched response between the two loudspeakers, preferably with a neutral tonal balance. But the whole off-axis dispersion consistency feeds into it because if you have a very uneven off-axis response, a large part of what you’re hearing is the room, which is to say fairly uncorrelated reflections, and that is going to affect stereo imaging, mostly negatively. The BBC did a lot of work on the accuracy of a stereo representation and investigated these things in great detail and discovered that stereo imaging is either preserved or damaged by the listening environment more than by almost anything else. But a loudspeaker with a pretty flat response, low coloration, and a fairly consistent off-axis response up to a point is a good starting place. That’s what we aimed for in the 5/5.
PS: You specify a response of forty Hertz to twenty kilohertz plus or minus two dB in the 5/5.
DH: Yes, that’s basically an anechoic response, but it’s only the first step. One of the advantages of having access to an anechoic chamber is that you can measure a speaker and then put it back into a room and reverse engineer your measurements—we can adapt our measurements in a room so that we can measure it in more or less the same manner. The BBC approach to loudspeakers is monitoring in the broadest sense—like Peter Walker said, trying to approach the original sound. But sometimes it’s judicious to allow changes in the frequency response in order to ameliorate resonances and colorations of the loudspeaker which you cannot otherwise fully control. Changes like that are made to allow the speaker to perform better in real rooms outside the anechoic chamber. For example, in real rooms you probably want a loudspeaker with a slight rise at the bass because listening in an ordinary room and listening in a concert hall are two totally different environments and will give you different impressions. This is one reason why you don’t tend to end up with a dead flat response.
PS: I read the BBC paper on the development of the 5/5 written by your father and Dudley Harwood. The decision to use slot-loading required an extraordinary amount of mathematical calculation and cut and try experimentation.
DH: There’s some horrendous math in that thing, which I certainly don’t understand! But even they acknowledged that their initial analysis of the slot and the way it behaved didn’t quite work out in practice because it’s a complicated interface with the cone and the air and the slot. There’s potential for two sorts of resonance. There’s the resonance of the actual wood itself because you’ve got an unconstrained edge around the slot. The other resonance is the acoustic resonance of the air mass between the slot and the cone. You can narrow the width of the slot down to a certain point, and then the frequency with which the air resonates behind the slot comes into the pass band. So the trick is we’ve allowed the baffle to be thicker at the edges, so it partly fills in the air volume behind the slot. And this actually pushes the resonant frequency up and reduces the variation of it. That’s one example of the kind of tweaks we did in the newer version.
PS: So you’re constantly going back and forth between theory and practice?
DH: Yes. We come at it as needed empirically. You do something and you think, well, that’s not quite right and you change this and that until you get the result you want. One of the engineers in the research department was a mathematician, very clever guy—he estimated that in a loudspeaker drive unit, there were something like two to three hundred variables, only probably forty or fifty of which at that time were readily measurable. His point was that the empirical practical approach has to be largely the one you take because there are so many unknowns and so many things that are difficult to measure that you have to balance the measuring and listening to get some sort of reasonable result.
PS: Do you use program material or, as Peter Walker used to say, a combination or burps, beeps, buzzes, and other non-musical electronic tones?
DH: Again, a mixture. When you initially assess drive units, obviously they’re not in the box, so there’s no possibility of using music. So your starting point is a moderately level frequency response in the anechoic sense. But as I said previously about adjusting anechoic response once rooms are brought into the equation, the same goes when you start listening to material that you are very familiar with and trust to some degree, which obviously brings you to the other big problem with loudspeaker assessment, namely, do you have material you trust and to what degree do you trust it? Because at the end of the day, unless you do it yourself, you don’t know how these things have been recorded.
PS: What do you use—vinyl, CDs, proprietary material?
DH: I have a tape of various people I know, and they were recorded in an anechoic chamber. As for CD or vinyl, I use CD. Vinyl is out of the question because you can’t get the consistency. There are too many variables in it. I mean, you’ve already got too many variables in the recording process to put in more variables about tracking weights and cartridges and capacitor loads.
PS: I have a question about the LS8/1, the revision of the fabled BC-1. My understanding is that the reason for the supertweeter in the original design owed to the limitations in the tweeters of the day, so a second was needed so that some very high frequency FM carrier signal could be monitored. But that isn’t a problem with modern tweeters, so why is the extra tweeter retained? Nostalgia?
DH: Well, yes, mostly. But another reason for the extra tweeter was that in those days any loudspeaker with three drivers was considered professional and did not require a purchase tax! So there’s a pragmatic aspect to it as well. But really, we felt that the 8/1 is a legacy product, which we wanted to honor as well as improve.
PS: Last question. If you were given an unlimited budget, what kind of speaker would you design?
DH: I don’t think it would be very different from the kinds of speakers I’ve already designed. But if really had all the money I ever wanted, I wouldn’t spend it on the loudspeaker, at least not right away. I would spend it on a great deal more investigation into the weaknesses and limitations of loudspeakers so that we could improve them all. I guess that doesn’t quite answer your question, but it’s where I would put the money.
Tags: INTERVIEW LOUDSPEAKER
