All of the Dunlavy models measured remarkably flat, anechoically.
43 responses Add your response
I think these are anechoic or simulated anechoic (time-gated spliced with close-miked) measurements, but don't know for sure. Click on "Measurements", on the right-hand side:
Hesson11, that is a very eductional set of curves. Thanks for posting the link.
In the 45 to 75 degree family of curves, you see that big bump between 2 kHz and 5 kHz? Unless you listen nearfield or under quasi-anechoic conditions, that bump is probably audible - and it's right smack in the region where the ear is most senstive, according to the Fletcher-Munsen curves. This excess off-axis energy in the lower treble region is caused by the tweeter having a very wide pattern just above the crossover point, and it's there on most speakers - but you cannot predict its presence from either the on-axis anechoic curve or the "listening window" curve. Kudos to SoundStage for measuring the off-axis response out to 75 degrees.
Thanks, Duke. I'm guilty of focusing too often on on-axis measurements, so it's good to get a bit of education from one as knowledgeable as yourself.
BTW, I assume most Audiogoners are aware of this, but if you go to the link below, you'll find a great many speaker reviews that feature the same kind of NRC measurements as shown for the Usher Be-718.
The K&H O 500 measurements are absolutely stunning (although the spectral decay plot may be hiding resonances as it cuts off at -25 db)
I have read good comments about these speakers (except they do not play very loud - probably a direct result of their tremendous linearity and even dispersion - as everyone knows you can EQ down to get flat but you lose efficiency).
Do you know what they sound like? I have not heard them.
This excess off-axis energy in the lower treble region is caused by the tweeter having a very wide pattern just above the crossover point, and it's there on most speakers
AHA ....now you have hit the nail on the head in respect of the deficiencies of most speaker designs and especially in contrast to the K&H O 500 you refer too, which has the most beautiful plots I think I have ever seen. WOW.
Shadorne, I think I heard that model (or possibly a predecessor) about five years ago at CES. My impression was they would indeed play very loud. They were very detailed, but to my ears they sounded forward and a bit bright. Now they may well have been just a few knob-twiddles away from nirvana, and the waveguides around the mid and tweet do appeal to me intellectually, but going by aural memory I'd prefer any of the big ATCs.
They were very detailed, but to my ears they sounded forward and a bit bright.
Thanks - although too forward is how many would describe ATC's. The plots indicate these may indeed be a bit brighter in the treble as there is no roll-off at all - beautiful plots on and off axis - apart from a hint of what might be resonance in the metal dome around 16 Khz - I doubt many could hear that but it might add some "glare" - who knows.
Of course as you said with such a perfect response one could simply tone down the treble to get any desired response "nirvana" at all...the key is the balance in on and off axis which remains whatever you do with tone control - beautiful plots!
Shadorne, I hadn't seen your post four up when I typed my reply three up. Congratulations on catching the implications and applicability of that off-axis bump. When working on a design, I place most of my attention on what's happening off-axis rather than on-axis.
Bob, no doubt much of the credit for those impressive measurements is due to a very well executed digital processing system. But signal processing cannot do much about radiation patterns, and the polar maps of those patterns are what I'd give the greatest consideration to (the green plots towards the bottom of the measurements page). You pretty much never see polar maps or any sort of off-axis data becaus it almost always sucks, and the K&H polar maps both look pretty darn good to me.
I recall a thread where there was an argument about the relevance or value of specifications. I can imagine some are rolling their eyes at this current discussion of beautiful looking plots.
For those rolling their eyes, and this is true: Dolly Parton once said that she was best known for just TWO things! Demonstrating so very well that people often see things all too differently. Her answer was, of course,.....her music and her lyrics ;-)
Bob, typically driver interaction has a large effect on radiation pattern in the vertical plane but not much in the horizontal plane. Even with the digital crossover, you can see the results of driver interaction in the vertical polar map.
I use a passive crossover, and my horizontal radiation patterns are as smooth if not smoother than the K&H. My vertical pattern probably isn't a good, but that's less critical (as long as the power response is smooth) because we listen within a fairly narrow vertical window anyway.
I found a website that said the big K&H speakers retail for thirty-six grand a pair, but you can mail-order them for a mere twenty grand a pair. That's actually one helluva discount.
Thanks Duke. You are right, I see that now in the plots at both crossover points. I was questioning based on their comment:
Due to the nearly ideal slope in the crossover region, the overlap in frequency spectrum between any two drivers is extremely narrow, actually only a few Hertz. As a result, vertical directivity remains optimal in this region, even off-axis.
That is a huge discount. But about 10 days ago a pair of JBL LSR4328 digital active monitors replaced my pair of M&K monitors driven by Bryston 7B SST amps (almost a $7,500 reduction in cost), so I'd have a difficult time justifying the price of the K&Hs.
It's amazing to see that some of my favorite speakers tend to have similar frequency response: Dynaudio and Thiel. The impedeance plot is very helpful in matching amp.
Coincident speaker definitely is very friendly toward SET amp while Thiel and Dynaudio will rquire lots of low impedance driving capability.
Looking at the measurements for several speakers I know well, it surprises me how similar many of them are. For example, Dynaudio C1 and Thiel 2.4 look quite similar (and both very flat) to my uneducated eye, although they don't sound the same. Interesting to look at something like the Reference 3a MM decapo, which seems to measure pretty horribly. Or the Silverline Sonatina, which has a big midrange dip. Even the Watt/Puppy is a bit ragged. How important is this, I wonder.
Would love to see measurements for some models of Harbeth.
The ATC has a very flat response. Do you find that flat responses sometimes limit the 3rd dimension to the soundstage? In other words, do you find it is more like a two dimensional picture than a three dimensional picture depth-wise?
No not at all. A flat "power response" without the usual treble and bass boost can sound extremely natural and is hard to distinguish from live or real instruments. The vast majority of speakers have a strong off axis response in the bass and treble with a dip in the midrange. Sometimes called the "BBC dip", this has become so popular in consumer speakers since the 80's that a flat power response is rare and is usually interpreted as "forward" or "harsh".
There is a strong correlation to soundstage (although most depends on the recording)...there is a tendency for instruments to sound closer as if they are in the room with you - so you could argue that there is less depth with this pattern - conversely you could argue that other speakers will never reach as forward into the room or close to the listener forever relagated to sounding as if you are at the first balcony....so I think the soundstage moves forwards somewhat but is no less or more compressed than speakers with a BBC dip.
According to Dr. Toole's group that has been researching subjective listening for the past 2 decades, it seems to be very important. Not necessarily flat, but smooth.
The not necessarily flat point is important. Supposedly people like a deviation from flatness in the mid-bass and on top, according to Paul Barton (who also worked with Toole).
The thing that doesn't impress me with their methods is that they cherry pick their testers and instruct them in what qualities they should listen for in the speakers before using them in the tests. They only keep the people who demonstrate an ability to pick out the qualities they are instructed to pick out for the tests.
It is true that the listeners go through an automated "training" program to improve their listening skills. Seems reasonable to me. But, I agree that it seems like it could introduce biases.
They found that audio reviewers and audio enthusiasts made poor subjects because they were so inconsistent. The training has proved necessary.
Not necessarily flat, but smooth.
Agreed - exactly! Very few people are aware of this. When I talk about tone control adjustments audiophiles cringe ....but they should not. It is no crime to tweak in this way. In fact a ruler flat on axis response is NOT all that important to our listening pleasure...it is actually the off axis that matters most.....in fact the MOST important thing is that the on axis and off axis behave in at least a uniform way with consistent gentle roll off as you get to wider angles...this is what sounds natural!
What sounds unnatural to our ears is a speaker transducer that has either strong directivity or wobbles and bumps off axis such that the on axis and off axis curves do not all behave together in a smooth way - in many poor designs they often crisscross as the cone breaks up or there is a sharp discontinuity at the crossover region (suckout or sudden abrupt change in dispersion pattern)
See this discussion SM75-150S. You will see the author correctly emphasizes that it is NOT the smooth on axis frequency response that is so important in a driver but the smooth consistent behavior between on and off axis and a wide dispersion.
SMOOTH is the key ingredient. This is why many designers pay so much attention to smooth phase response which is a warning sign of unwanted things going on with the design.
Does anyone have any information and a link that quantifies "smooth?" I've been wondering exactly what the term means in respect to frequency response, not as much in phase response.
I'd also be interested in what Paul Barton meant by additional mid-bass and extra sizzle on top, if anyone has a link to numbers.
Shadorne's point is well taken. That is another aspect of Dr. Toole's work. He has called stereo listening an antisocial activity, because typically only a single person is allowed to enjoy good sound. His work has been to expand good sound to a larger area, so off axis response is critical. That is the basis of JBL's Linear Spatial Reference design philosophy:
LSR Linear Spatial Reference Design
Does anyone have any information and a link that quantifies "smooth?" I've been wondering exactly what the term means in respect to frequency response...
I don't, but Toole points out that high resolution measurements are necessary to get a clear picture. A 1/3 octave FR make look quite smooth, but when expanded (zoomed in on) by say 1/12 octave resolution the truth reveals itself. He also talks about Q and audible thresholds, but I don't recall the details.
Jkalman, in my opinion Robert E. Greene of the Absolute Sound (also a high-ranking professor of mathmatics at UCLA) has a genuine understanding of what matters when it comes to a loudspeaker's in-room performance. Here is a link to his website:
See in particular the measurement supplements to his TAS reviews of the Gradient Revolution and Jamo 909, and the measurements of his Harbeth monitor 40 (that page is currently "under construction" but he has some measurements posted).
The "target curves" that you see in the Gradient and Jamo supplements are similar to a target curve published by Bruel & Kjaer, which I can't find online but it's basically flat below 100 Hz and then slopes down by a little less than 1 dB per octave from there on up. I presume this is a non-time-gated measurement from the listening position.
I think one of the misconceptions about the Toole research was that he trained listners to pick a certain loudspeaker over another. Actually, what they did was to evaluate and train listners to hear various types of distortions added to program material. As well, this was for research purposes, not in any way a marketing ploy to sell a particular brand of loudspeaker.
Also, to add a couple of items to the off-axis dispersion discussion: This is generally accepted to be simply good loudspeaker engineering. Most of the Canadian companies that use the NRC or NRC-based research see this as critical to proper loudspeaker design. But in addition you'll find that Doug Schneider, who assists in the loudspeaker measurements at SS!, and John Atkinson at Stereophile (has measured many speakers), have come to agree that smooth, controlled off-axis dispersion equate with good sound in real rooms.
Lastly, I actually discussed the Athena Project with Paul Barton at CES this year. What they determined was not so much that a midbass bump was desirable, but instead a smooth rise in the bass starting at 200Hz and going down to 20Hz -- I think it was about a 6dB rise over that range. The key being a _smooth_ rise. Anthem is actually basing their new room correction software on this target curve. As for the top end, simply google the EBU target curve and you'll find a good graph.
Hope this helps.
The SoundStage! Network
Actually, what they did was to evaluate and train listners to hear various types of distortions added to program material. As well, this was for research purposes, not in any way a marketing ploy to sell a particular brand of loudspeaker.
The testing also taught listeners how to identify changes to different ranges of the frequency response. Though that could be considered added distortion. I read an interview where Voeks discussed the process. If people couldn't learn to identify changes to frequency response, and didn't identify which one was incorrect, or not flat, they weren't used for the blind testing. People with suboptimal lower and higher frequency hearing were also excluded. I don't think any of this was a ploy, but I do think it is cherry picking, and that will likely affect the final outcome.
Is it a good thing or a bad thing? I don't know. It is good that speakers have some type of technical boundary within which they should measure and that people have tried to shed light on that boundary. It does make me wary though as to exactly what those boundaries are and how much they can differ between different individuals. Especially if those individuals had no training.
I'm planning on experimenting with my own setup to see what I like best. Part of that will include speakers that measure extremely flat. I would consider Revel Ultima2 Salons to be one of the speakers worth experimenting with, but no one around me is carrying them to allow me to demo them at a store... ATC is another one I've been considering (the Pro versions). The Magico Mini II as well.
As for the top end, simply google the EBU target curve and you'll find a good graph.
If you could give a link to something I would appreciate it. I tried Google-ing "EBU target curve" and I didn't get anything useful on the first bunch of pages.
I have been trying to find the quote someone else I know got their information from concerning the "top-end sizzle," but so far no luck (he might have been wrong about the top-end, which is exactly why I want to find some quotes on it - I'm not being cantankerous or anything, just looking for a source).
Search Yahoo for "Listening conditions for the assessment of sound programme material" and the EBU article will come up.
I understand your suspicion re: the Toole research, but one thing that the EBU study shows is that there is much consensus about many of these issues. Take the off-axis response issue. EBU doc states "Frequency response curves measured at directional angles +10° and +30° should not differ from the frequency response measured on the main axis (directional angle = 0°) by more than the following permissible deviations:" ...and then it gives the data.
Also in that doc is the operational room response curve. What you find many designers doing is tapering the HF response of their loudspeakers about a dB or two starting at 16kHz or so (varies with designer of course) and with the absorptive nature of room furnishings that can easily translate into the desired curve.
I do think it is cherry picking, and that will likely affect the final outcome.
It'll definitely affect the outcome. You'll be able to place more confidence in the results. Common practice of controlling the number of variables in experiment design. You wouldn't include people that are color blind in an experiment that measures color perception. The number of subjects needs to be large to be of value, but there are practical matters of managing the subjects and running the tests, so you wouldn't want to waste a "slot" on someone that has a known physical disability which impacts the quantity you're trying to measure.
It makes sense that a flat frequency response is a good starting point. I don't know if people will inherently like it better without training. I can't even do that kind of experiment on myself, because I've done the Moulton Labs training courses...
It is very easy to take a flat frequency response and alter it after the fact by dropping dBs wherever it suits you if you find you would like to accentuate parts of the spectrum. My understanding is that it is much harder to fix a signal that is inherently flawed.
I liked it. It is cool to be able to understand when you are hearing things like ranges of the frequency response being out of proportion. Of course, with my 40-90Hz bass issues, I already knew there was a problem because of the kick drum from hell.
It is also nice to be able to identify compression, distortion artifacts, phase issues, reverb effects.
Some of the stuff, like effects, is fairly obvious if you play electric guitar (or other instruments where effects are used often), but others, like being able to single out areas of the frequency response when they are boosted, are pretty neat.