Voicing

August 22, 2016
 by Paul McGowan

I am not a speaker designer, but I've spent much time with some of the best in the world.

I don't envy their task.

When I design electronics the initial work is relatively straightforward and the results predictable. Even first year engineers haven't too much trouble designing platforms with full frequency response and low distortion.

Once completed some of us take the extra steps of voicing our work, a process that takes years of experience to get it right.

None the less, even if you don't voice your work you're miles ahead of that which faces speaker designers, because the elements they have to work with are so flawed in the first place.

Here's an example:

ApexRibbonFreq

This is a response curve from a ribbon tweeter. To understand the difficulties faced by speaker designers, just look at the solid green line for starters. This represents what's known as the "on-axis" response. The measurement microphone points directly at the tweeter. Yikes!

Even straight on this thing's all over the map. It's about as far away from "flat" as a roller coaster.

But, now look at the other lines. They are what we hear off-axis, meaning the microphone (like the listener) sits to the side of the tweeter. These aren't different by a small amount. Each response differs by many dB.

And as Audiophiles we worry about tenths of dBs?

Subscribe to Paul's Posts

11 comments on “Voicing”

  1. My experience is that the biggest improvements of sound quality of a Highend stereo chain can be gained from the exchange of the speakers - or the hole chain of course! Exchanging cables or components only give minor or marginal differences in the sound from a given speaker - unless you run into a non matching amp for the speaker. Thus I be leave a well designed integrated (including DAC) from PS Audio could lift the level of the existing sound quality of a pure PS Audio system.

  2. The designer of this driver needs to go back to the drawing board. I had a career designing water measurement instruments, and in my core team we deprecated what we called "compensation engineers". These are designers who start with a concept, build a prototype and then add parts to correct deficiencies in performance, often the dreaded "trim pots". Our motto was "Anything that can be adjusted, will be adjusted WRONG."

    We practiced "reduction engineering", which means starting from a prototype and removing parts until it works. This, of course, was not linear. We would substitute a somewhat different circuit with fewer parts. The fewer parts, the easier it is to build and the less there is to fail. With less parts, assembly labor and packaging, you can afford better parts, which also means you are supporting better designers and manufacturers in the industry ecology.

    This driver requires too much "compensation engineering". The sound will be compromised by the parts you need to add to achieve flat frequency response. Since it is a ribbon, the vertical response is predictably worse.

    Not only do the off-axis responses at all horizontal and vertical angles add to the total effective frequency response in the room, the initial wavefronts of reflections are compared to the direct sound to construct a spatial map of the room, locate the source in the room, locate the listener in the room and identify the source by the spatial radiation pattern.

    The spectral components of a violin and oboe are quite similar, and normal room comb filtering in a concert hall makes it difficult to tell them apart by a spectrogram. The way the notes start and stop differentiate them assuming the speaker can start and stop quickly in response to the signal, which is why you need a waterfall plot to evaluate a driver or complete speaker.

    The spatial patterns - how the sound projects and bounces around the room - are radically different in a Violin and Oboe. String instruments are double sided, with the front soundboard perforated by "f holes" and a sound post under the bridge coupling the vibrations to the back sounding board. This makes them a quasi-dipole, but with the highest frequencies emanating from the small piece of spruce between the f-holes. If you want to make a sound like a violin, the speaker needs to have the same acoustic size and geometry.

    An oboe radiates from all the open holes and the bell, which means the shape of the sound has a lot of peaks and dips, and is different for every note because the pattern of open holes is different to change the pitch.

    Reproducing a whole ensemble is like listening through two windows, which theoretically are the size of the microphones, and re-radiating as a point or line. If you want to hear the room where the recording is made unadulterated, you need to suppress all the first reflection points with absorbtion or diffusion. If you want the "they are here" illusion, the speakers need to have flat frequency and phase response at all angles, so you have to start with drivers that have flat frequency and phase response in the pass band, at all angles horizontal and vertical.

    AND, the right way to do it is violin speakers for the violins and oboe speakers for the oboes. In this case, drivers that have the same off-axis response as the target instrument are the best.

    1. And still the best possible speaker 'shows' a 6' tall singer only 3' tall - as the tweeter on ear level determines! Most frustrating. Could it be that stereo was invented for sitting singers only? 🙂

      1. Which is why my stage speakers all consist of a woofer base and a treble speaker on a pole, so I can adjust the height to match the sound source.

        This is one of the big problems with PA speakers, they are not on the stage at the same height as the instruments and mouths. Even if you are assuming everything in "the mix" is coming from one pair of speakers (a poor choice locked in by a preliminary experiment in 1933!), the tweeter and midrange should be at ear height, the optimum position for the pair of virtual windows into the performance space.

    2. "We practiced “reduction engineering”, which means starting from a prototype and removing parts until it works."

      When ALL of the parts are gone....so are all of the problems they created. 😀

    3. I don't think speaker building is all so hard. They just do not need to be perfect. Even back in the 50's AR ran some very convincing live vs recorded demos of simple quarter music. The key to the success, I believe, was that the space that they played back in was a large room. The real problem with home Hifi is the need to make the speakers work in a typical home listening room. Now the off axis performance becomes critical. Bass integration with room nodes even more so. It's only as you start to push the limits of dynamics, something AR could not do so well, competent speakers do become extremely hard to build.

      1. Speaker design for people who were content to listen to speakers in the past is not so hard. OTOH, comparing live music to speakers is like comparing oranges to artificially sweetened applesauce sherbet.

        Even in Edison's famous test proving his cylinder recorder was indistinguishable from the real thing, he used a soprano (which fit the bandwidth of the cylinder machines) and people who had never heard an operatic soprano nor set foot in Carnegie Hall, then a unique acoustic environment in New York. If he had let the soprano open up to full dynamics, she could have overpowered your WATT/Puppies and then the real live music would have been obvious.

        I have determined that 1" dome tweeters can't reproduce the dynamics of a string quartet, so your speakers are probably more than 10dB below realistic levels for acoustic music, let alone electric levels.

        I have speakers that are 10db louder than a Steinway Model D, and they sit next to one that is played regularly by concert pianists. The tweeters have 40 times the surface of a 1" dome, which makes them 32dB louder for the same linear displacement. I have recordings of the Steinway and the speakers amplifying the Steinway for comparison, and the sound is close. I challenge ANY audiophile speaker to an A-B-Steinway comparison, and then tell me how easy it is to design speakers.

        1. I'm not sure who your replying too. Was the Wilson WATT/Puppie comment just a generic audiophile example?

          It appears you are in general agreement with me. " It’s only as you start to push the limits of dynamics, something AR could not do so well, competent speakers do become extremely hard to build."

          I do have to disagree with you regarding the Steinway test. I have a 5'4" "Boston" baby Grand in room, and I generally have to reduce the volume of my various speakers, including ones with ~ one inch tweeters to match the loudness level. A couple of caveats. One: this is not a concert grand and two: Of course this is in the room and not me trying to fill an auditorium with a pair of speakers. I've never done an exact recording but I can play concert piano music over the speakers and try to play parts of it at home and the sound is close, forgiving my skills.

          I might add that the speakers with domes cross over above 3500 hz and not at the usual 2k range that is so popular these days.

  3. "I don’t envy their task."

    Why not? If it gets good reviews in the magazines it will sell and they'll make a lot of money no matter what it sounds like. Even if they all sound like "canned music"......which they do.

Leave a Reply

Stop by for a tour:
Mon-Fri, 8:30am-5pm MST

4865 Sterling Dr.
Boulder, CO 80301
1-800-PSAUDIO

Join the hi-fi family

Stop by for a tour:
4865 Sterling Dr.
Boulder, CO 80301

Join the hi-fi family

linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram