Phase Coherence - A Desirable Attribute In A Loudspeaker?

[colhd]

Hi Alan,

I recently spoke to an audio retailer in England about an Accuphase amplifier for my soon to arrive SHL5s. Once he found out that I would be using Harbeths he immediately started talking about phase coherent loudspeakers, which he implied would be far superior to a design that was not phase coherent. He seemed to think that Harbeths, along with other BBC inspired designs, were not. I know that you take painstaking care in designing the crossovers and that there are always compromises so can you tell me if phase coherence important and is it a Harbeth design goal? It won't change my mind about the SHL5s, I chose them because to me they sound better than the competition, but it would be nice to know when a retailer is talking rubbish, perhaps to try and gain a sale.

For what it is worth, said retailer stated that if Harbeth speakers were phase coherent he would be interested in selling them.

Thanks,

Col

[A.S.]

Could you explain please the meaning of 'phase coherent'. Conversely, what, I wonder, is the opposite?

Would this phase coherence only be detectable with specialist measuring equipment or could the casual listener (i.e. me) 100% detect such a speaker from mere listening? Or only 80% of the time? Or 50/50? Assuming that this phase thing was mandated as the most important aspect of the design, would any compromises have to be accepted elsewhere in the design? What would they be? Coloration? Repeatability? Distortion? Off axis response? Naturalness? Bass? Could this be (another) fixation on just one corner of the design jigsaw forgetting that what matters is not how the painting was done but what it looks like from a distance - or in loudspeaker terms, how it sounds overall?

A look across our customer base reminds me that we are a band of unconventional mavericks who eschew dogma and admire pragmatism and free thinking The extreme irony is that our products look conventional and sound good because they ignore current convention. Long may it continue!

P.S. As Dudley Harwood (Harbeth's founder, ex-BBC research engineer) is reported to have replied to someone who asked him about the technical core of his speaker design philosophy .... "You just fiddle about until they sound right". The truest words spoken on the subject of loudspeaker design - by someone who really did know the subject inside out!

[colhd]

Alan,

I am afraid I have told you everything I know - that's one of the reasons I asked. My assumption (with all the dangers making asumptions entails) is that the retailer meant that the drive units are arranged, physically or electrically, so that the energy from each arrives at the listener at the same time (I know Stereophile test for this in their reviews, for what that is worth). He used the term "phase coherent" as though it was a label, in much the same way that a speaker could be described as "first order" as opposed to "second order", "bass reflex" as opposed to "sealed", "active" as opposed to "passive". The fact that you, an experienced designer, don't know instantly what he was talking about suggests to me that there is at least some marketing BS involved. He did use Wilson Audio and Focal/JM Lab as examples of "phase coherent" speakers.

Col

[A.S.]

Actually, I guessed that you might mean staggered drivers (physically). That seems like a 'good idea' in concept but for one thing. Well, a few things actually! The first one is that if the designer is adamant (for marketing reasons?) to make a kangaroo-like speaker with the voice coils in vertical alignment (sounds like a good idea - right?) he has worked himself into an intellectual corner because he can then only use the most primitive crossover design. But the crossover is as complex as it needs to be to minimise overlap between the drive units so that the tweeter isn't trying to reproduce bass and the bass/mid isn't attempting to be a tweeter.

A simple, first-order network of the type often associated with 'phase coherent' speakers has the chronically serious problem (in my humble opinion) of letting bass/mid bleed through into the tweeter's band and tweet into the bass mid. So the drive units fight each other. A (first order) crossover of just 2 components can not behave as if it is 4 or 6 or 8 components so this cross-pollution has to be accepted as inescapable for simple networks. With more components the designer can 'sharpen-up' the filters and prevent this out-of-band coloration.

Furthermore, every drive unit has it's optimal operating range - as a car engine has an optimal rev range. For the cleanest sound (smoothest acceleration) you need to know when to change gear, and that's why a proper crossover, with many components is really essential. Think of the crossover as an acoustic gearbox with many cogs and gears that click-in at just the right times to get the very best overall smoothness. How can you compare a 2 gear car with a 6 gear? No comparison in smoothness.

Add to that the fact that with more components you can equalise and match the sensitivity of the drivers in fairly narrow bands (rather than globally with a simple network). Finally - and this may be a surprise - but the complex network actually slows down the electrical signal as it works its way through the components, to the tweeter and out to your ear as sound. That delay is the equivalent to the typically 5-10 cms at crossover frequency, because sound travels at 340mS.

Here's the clever bit - remember the tweeter's voice coil is physically in front of the bass unit. If the (complex) crossover circuit is well designed in the transition band and incorporates just the right amount of time compensation, the sound waves that reach your ear from the bass/mid and tweeter are correctly in acoustic vertical alignment! How about that then! All the claimed advantages of vertical alignment - none of the disadvantages - and by using a normal rectangular cabinet, minimised cost!

We've looked at the function of the crossover here.

[colhd]

Alan,

That sounds like your speakers are "phase coherent" as I understand the retailer was using the term, achieved electrically as opposed to physically, plus the other advantages you refer to. Can you tell me a good resource for reading up on loudspeaker design so I can figure out what exactly those other advantages are? Maybe that way I won't pester you with questions all day as I wait for my Harbeths to arrive!

Thanks for taking the time to answer.

Col

[A.S.]

'The Loudspeaker Cookbook' would probably be a good start as a contemporary design guide book. It covers the major basic systems (sealed, vented, TL) crossovers design etc.. I own a couple of older editions. It's on Amazon here. What is really interesting is the feedback comments on Amazon. Bearing in mind that I said that this was a good start - it is - the reader comments reflect my own in large part ....

Reader comment 1:
===============
"This was a great starter book for me. I have no electrical knowledge or education, so the prospect of designing my own speakers left me at a loss. However, the reviews of this book led me to purchase it. After reading it 3 times, I know have a grasp of the basics that all the expert amateurs are talking about. Terminology, box design, and some of the principles of crossover design are all much clearer to me.However, the book left me wanting much more. Each chapter only imparts basic information, where as I wanted much more detailed information about each subject. So, buy it, but don't expect to be an expert after reading just this one book.".

Reader Comment 2:
===============
"Warning: This book is not for beginners. Hard core hobbyists or speaker design professionals may find some useful models, empirical studies and references in this book. For the beginner, the minimum vital data needed to design and build a speaker is interwoven in to a jumble of esoteric empirical tests and computer simulations. For the more experienced speaker builder, the depth of this book is great and the footnote references are very complete. While the breadth is great, the depth of technical explanation for fundamental speaker enclosure modeling physics was a lacking. I guess I need to buy another book or research the footnotes myself.

This book could be a lot better with some solid editing. The writing - for the most part - is clear, but there are sections that are hard to read and there are some errors. The fonts are too small (many of the subscripts are illegible). The graphics are horrible. The book has over 100 frequency domain plots (developed with speaker CAD packages), which are illegible and poorly labeled. Most of the graphics are nowhere near the text.. I give this book two stars because of the breadth and the references, but it's no cookbook. You won't be able to look up a recipe and cook up a speaker".
.................................................. .................................................. .................................................. .................

Now, taken together that's a fair summary. What disappoints is that there is not one hint anywhere in the entire book that making changes to the drivers, the cabinets, the crossover almost certainly have an effect on how the system sounds. The inexperienced reader/speaker designer would assume that he has a vast canvas on which to paint his masterpiece and that it mattered not a jot which brush or colours he used - it would all come out equally attractive. So there is no mention or discussion at all of the sonic signatures of the cones themselves and the reader would assume, in the absence of knowledge or experience, that the cone material was merely a cosmetic issue unconnected to measurement or perceived quality. The preference of say, a shiny aluminium cone over a dull black plastic one. We here know that the cone material is all critical to the perceived sound quality. It couldn't be more critical but it receives not one word in the book. Not one.

But as I mentioned in my previous post (above) every aspect of the engineering has or is likely to have a measurable consequence and may/may not have an audible one. The major section on crossover design completely sidesteps key audibility issues and presents circuits which may well give a superb theoretical response but I know from experience will sound terrible. Terrible to a Harbeth user - probably perfectly normal and unremarkable to other manufacturer's customers.

I'd say that a better approach (the one I've always adopted) is a curiosity driven from long listening experience as to what sounds right and then wherever possible building a mental experience-base of how that listening observation correlates with physical parameters of the design. Taking that approach, high school maths would be more than enough providing you have a curiosity about the hows and whys, and avoid dogma. You must build you own mind-map. In my case, I discovered right at the start of my career that my personal listening experience correlated perfectly with the pragmatic loudspeaker engineering underway at the BBC by people paid to be objective at the taxpayers expense. Everything I've experienced in the successive 35 years has convinced me that I nailed my colours to the right mast. Not perfect for every conceivable situation (such as heavy duty rock music) I freely admit, but that's an irrelevance for our market, our users.

I'll send you a link to a masterpiece of a paper later - from 1958, and presumes absolutely nothing of the reader's knowledge. You'll get far more out of that. Read it, read it again and read it again and do that for the rest of your life - you'll learn something new every time.

P.S. Just as well I can commit thoughts to keyboard quickly without editing otherwise I'd be typing all day and I assure you that I'm not!

[colhd]

Alan,

Thanks - I look forward to receiving the link. It will help me ignore the terrible weather here in East Kilbride as I wait for all my new toys to get here: trial of firewire dac, a Weiss DAC2, which arrives tomorrow, SHL5 and various amps Friday. It will be fun and hopefully will result in musical bliss. Should be enough to take my mind off the poor performances of a certain Govan-based football team.

Cheers,

Col

[A.S.]

Ah, Govan District (General) Hospital. My birthplace.

OK here is the link to the BBC's D.E.L. Shorter seminal paper "A Survey of performance criteria and design considerations for high-quality monitoring loudspeakers" from 1958. Shorter was drafting this paper the year that I was born. As I've commented before, this is a masterpiece of lucidity and technical writing for the layman. It riveted me when I first read it and it is so exceedingly well crafted that whether you are mildly curious about audio or a speaker designer, every time you read it some new gem pops out. It is so well written by a true master that you can not help but feel uplifted - even euphoric - that fifty + years on, you are walking the same path as Shorter. If you read nothing other than this, it's all here. Print it, read it, rediscover it in a few months and read it again. Repeat forever! Proof that the fundamental issues concerning quality loudspeaker design (and listening) are completely unchanged from the perspective of a 'serious' listener.

Here is Shorter's masterpiece; I've added Raymond Cooke's printed obituary. I have in my file Raymond Cooke's original hand written obituary. This is a big PDF of 2.7MB.

{Moderator's comment: link was repaired but damaged again in server restore. Now finally fixed. Apologies to all.}

[A.S.]

Finally the link problem was resolved. Sorry for the confusion.

I just thumbed through D.E.L. Shorter's seminal 1958 paper. Every time you read it, as your knowledge grows, some new feature reveals itself. I note for the first time the gently rolling-off high frequency response in a number of on-axis responses. And I wonder what story lies under that. Example: Fig 3, fig. 15.

[witwald]

I recently spoke to an audio retailer in England about an Accuphase amplifier for my soon to arrive SHL5s. Once he found out that I would be using Harbeths he immediately started talking about phase coherent loudspeakers, which he implied would be far superior to a design that was not phase coherent.

I think that what is meant by a "phase coherent" loudspeaker is one that has a largely linear phase response over much of its passband. It's not possible to achieve complete phase linearity because the rolled off bottom end and top end of the loudspeaker's response adds phase shift that is non-linear in nature (i.e. not linear phase).

If a speaker is "phase coherent" or "linear phase", then it should probably be reasonably good at reproducing square wave signals. This is because there is a particular phase relationship between the harmonics that go to make up the square wave, which needs to be carefully maintained in order to accurately reproduce the square wave's shape. However, all the harmonics need to be reproduced accurately, and eventually these higher harmonics are attenuated by the loudspeakers rolloff at high frequencies. Such a loudspeaker might also perform well at accurately reproducing the waveform of a toneburst. All of this, of course, assumes a nominally flat frequency response (in the loudspeaker's passband).

For a well designed loudspeaker system, using multiple well designed drivers, I cannot say how phase coherency affects the perceived sound quality. I have read that our hearing mechanism is relatively insensitive to moderate amounts of phase distortion (i.e. lack of phase linearity), such as that which is often present in multiway loudspeaker systems.

[witwald]

Attached is a brochure for the B&W DM6, which was marketed as a phase coherent loudspeaker back in its day. It is a 3-way loudspeaker system, and was developed around 1977. It consists of a 250 mm bass unit, a 150 mm midrange unit, and a 19 mm high-frequency unit. The design is interesting in that it utilises a stepped front baffle to allow for the different arrival times of the various drivers (see page 2). Its crossover frequencies were specified to be 500 Hz and 5 kHz, and the crossover network appears to use first-order filters. A contour control for adjusting mid-frequency response is included, working in the 500 Hz to 5 kHz band, as well as a high-frequency energy control. These controls allow the listener to tailor the sound reproduction to some degree.

The brochure shows a number of different response traces to various stimuli (see page 3). These include a sin² pulse, a mid-frequency squarewave, and a 50 ms-wide pulse. Some of these results are compared to a full-range electrostatic loudspeaker and a normal 3-way dynamic system. Many tone burst oscillograms are also shown (see page 4), taken at one-octave steps in frequency starting at 125 Hz and ending at 31.5 kHz.

Apart from the usual magnitude response plots, of which there are quite a few illustrating various aspects of the DM6's design, there is also a phase response plot of measurements taken on-axis at 3 metres. This shows that the DM6 has a quite linear phase response over the frequency range 150 Hz to 20 kHz.

[A.S.]

Assuming that 'phase coherence' is a 'good thing', why, despite the strong technical arguments you advance did the concept not catch on? Why aren't all quality loudspeakers built with a stepped baffle, kangaroo style? The reason must be all or some of these:

1. The consumer couldn't hear a difference between the phase coherent speaker and the best examples of standard, non-phase coherent speakers of the time (KEF 104, Harbeth Mk1/2, BC1 etc.)
2. The consumer did not like the sound of the phase coherent speaker and actually preferred the sound of the conventional speaker, for whatever reason
3. The cost of implementation priced the speaker out of the pocket of the consumer.

What do you think?

As I have noted before, the problem with loudspeaker design is of one step forward, 1¼ steps backwards. There is always a hidden snag to using new technology, sometimes serious, sometimes just a minor inconvenience.

[witwald]

I think that all three of the reasons proposed above resulted in this type of design approach being dropped. Compared to a speaker such as the Spendor BC1, which I'm at least somewhat familiar with, the DM6 is certainly significantly more complex and I expect much more expensive to implement.

It also appears that phase coherence is not a major requirement for high sound quality. Still, in purely engineering terms, the DM6 was quite an accomplished design. How the DM6 sounded compared to a Spendor BC1, which I did audition once many, many years ago, I do not know.

[A.S.]

Perhaps another of my archive papers gives a good summary - see attached. It's nice to have >10,000 accumulated PDFs to draw on!

>