I had a set of Dynaudio Audience 42's once, worst speakers i've ever owned and i've had a few over the years.
The Rega floor standers i'm currently using have quite simple cross overs and sound very good indeed.
The Rega floor standers i'm currently using have quite simple cross overs and sound very good indeed.
Dynaudio's are capable of the best ... and the worst sound - depending upon what they're driven by, 😉.
At the audio show, one room with Dynaudio had the best sound I've ever heard from a "raw" system. In another, it was the pits - midfi crud ...
At the audio show, one room with Dynaudio had the best sound I've ever heard from a "raw" system. In another, it was the pits - midfi crud ...
I'm going to be so bold as to quote a small bit of SY's Linear Audio paper here - the Testing One,Two, ... one - to make a point.
In an experiment they noted this:
If you can't hear anything that's like a click then you're in good shape - but once you have found a sound aberration, and mentally zoomed in on it, that's what you use to distinguish things ...
In an experiment they noted this:
When one listens for distortion in playback, what you need to do is listen for the "clicks" in the sound. That is, sounds that are unrelated to the music as it would have sounded at the time of the recording, they are anomalies in every sense of the word, they just plain don't belong. And these "clicks" will sound different, very different when the distortion spectrum changes by only a tiny amount, just like the change of placement in this experiment.
If you can't hear anything that's like a click then you're in good shape - but once you have found a sound aberration, and mentally zoomed in on it, that's what you use to distinguish things ...
Frank,
You really don't need to work so hard to evaluate a set of speakers. A simple test that will tell you much in the critical mid-range would be a simple test with female vocals. If you can't get that right nothing will ever sound as it should. Bass extension these days is fairly easy to do, clean upper octaves are more difficult as always but it seems that many don't make much of a point about that as so many seem to have lost that upper octave hearing so just ignore those top frequency bands. Otherwise I would think many more would complain about some of the very fatiguing high frequency sound out of so many devices.
That test protocol you reference from Sy is an obviously flawed setup if it was that easy to identify the relay sound. That should have been addressed and the tests rerun.
You really don't need to work so hard to evaluate a set of speakers. A simple test that will tell you much in the critical mid-range would be a simple test with female vocals. If you can't get that right nothing will ever sound as it should. Bass extension these days is fairly easy to do, clean upper octaves are more difficult as always but it seems that many don't make much of a point about that as so many seem to have lost that upper octave hearing so just ignore those top frequency bands. Otherwise I would think many more would complain about some of the very fatiguing high frequency sound out of so many devices.
That test protocol you reference from Sy is an obviously flawed setup if it was that easy to identify the relay sound. That should have been addressed and the tests rerun.
Yes, I agree about this character of a neutral sound. I'm just trying to point out that there are MANY attributes of a good sound and they are difficult to achieve all at once. You can achieve attribute X but then you lost attribute Y and so on. So when you try to achieve a FLAT response you might lose the other attribute, especially IF you/we don't realize that there is this "trade-off" issues to solve.
IMO, flatness is not really critical. People mention that 1dB can change the sound considerably, but I think it is not the decibel that is really in action.
In this case you might be examining the seamless blend, instead of the tweeter individual operation.
With the woofer on, it will mask the problem with the tweeter. Seamless blending between drivers is important (and is another problem to solve), but it is also important that each driver operates flawlessly. IME, tweeter is the major source of distortion and fatigue.
It is not uncommon that female voice sounds like a cat, a doll, a kid, or simply not natural coming out of the tweeter. I think it is a good idea to listen to a voice that have accent and that you know very well. It is strange how a small capacitance can change the human voice from recognizable to not at all. Something like this cannot be predicted by software or measurement tools, that's why I prefer to rely on ears, and the tools will automatically become complementary.
At the beginning you still need the measurement tools to make sure that your ears can be trusted. With positive results, you don't even have the need to confirm anything.
Jay,
I am now having to solve a problem as you are pointing out, nothing to do with audio but a similar problem. My client/ partners became so fixated on a single issue with a design that I worked through that they forgot about what the basic function of the design was supposed to do. Now I am having to come up with a fix to solve a problem that didn't exist until this other issue became a dominant driving factor. Design always is a balance and what to compromise should be ranked in order of importance, otherwise as you have stated you end up just chasing your tail in circles.
I am now having to solve a problem as you are pointing out, nothing to do with audio but a similar problem. My client/ partners became so fixated on a single issue with a design that I worked through that they forgot about what the basic function of the design was supposed to do. Now I am having to come up with a fix to solve a problem that didn't exist until this other issue became a dominant driving factor. Design always is a balance and what to compromise should be ranked in order of importance, otherwise as you have stated you end up just chasing your tail in circles.
In effect I'm doing both ... listening for seamless blending and the tweeter operation - by very specifically positioning my ear precisely in line with the tweeter diaphragm. If even in this extreme position I can't locate the treble speaker then the system is working well
I agree about the female voice - I've mentioned Odetta before, and these recordings very, very easily can convey her voice as a stereotype of the black female voice - it sounds like a caricature, rather than a real person ... good material for evaluation.
Higher order crossovers have higher power loss. More resistance there. They do not "sound" the same with simpler crossover (of the same drivers of course). But deep discussion is necessary about this topic.
The only reason for me to choose high order crossovers is to avoid breakups (which I'm very sensitive with). If it is doable with first order, I will, but IME it is impossible.
My latest speaker, I tried 2nd order electrical with both tweeter and woofer. This should be my preferred crossover if doable. But most of the time it is not doable because of phase issue, so people (and me in this project) have to usually go with 3rd order electrical with the tweeter.
With the woofer, 2nd order was only justified with dual notch filters (very expensive as I use MKP and foil inductors).
We're not even talking about enclosure yet (which is an integral part of the crossover!)
Yessir, Ed Vilchur and Henry Kloss had something there. I am told by friends into speakers that acustic suspension, when done right, will start to roll of the bass earlier than bass reflex, but will do so in a much more phase coherent manner. Also, their efficiency is usually lower than BR, and for good bass extension one needs drivers designed specifically for acoustic suspension, or would need large enclosures to do it right.
In its time, AR3A was one of the best around, although not cheap, it was as neutrally voiced as the best of them. And because it had mid and treble level controls, it could be married to many a room. Although a good friend had them for years, it was a long time ago and I can't say I remember them well. If I was to judge from my own AR94, all of it was true, however, they do best with high current capability amps. With its advertised 90A of peak current (without a clearer spec, like for how long), my H/K PA 2400 doesn't mind, and even in the midrange class, it's H/K which does best with them. Bass lines are clean, clear and well controlled, which is why I keep them as sort of stand-by speakers.
Excellent point. Modern bean counters will not use one resistor more than absolutely necessary, and if they do an amp with say 30 transistors, that's only because it couldn't be done with 28 transistors.
Richard, there are more levels than blak and white bad vs. good sound. In fact, many devices sit somewhere in between, not bad but not as good as could (should?) be. In all honesty, I've come across a situation that a say speaker I know is capable of more with better amps actually sounds outright bad only a few times, and mostly with badly and obviously mismatched players, like driving a $5k speakr with a $200 integrated amp. Confirmtion: change the amp for a better one, and the sound improves. Some people have crazy theories which typically mess up the sound.
That's an important part of the game, to sense when something is falling short and doing something about it.
Also, we need to always bear in mind that eah and every one of us hears differently from someone else. No-one has a truly linear hearing, and this cannot but influence our conclusions. I do not claim my speakers are linear only because I hear them as such, but also because the measurements tell me so. Without them, I wouldn't rely on my own hearing only to claim it's so.
At first hearing in my room, the sound was let's say different from what I was used to until then by my AR94 speakers. They are relativel good speakers, but are no match for my 1041 monitors. As I half expected, the main differences were all in the mid to high range, which cleaned up, became even more detailed and overall clarity was much improved. More space around the perfirmers, and the high range, now portrayed by a 1" titanium dome in lieu of a 2" cone, really stood out for its perceived extention and clarity. No fizz, thank you.
All told, I'd say the AR94 would classify as good but not outstanding in any way, whereas the 1041 I would class as outstanding in practically every way except for some really wild downward extension, with its -3 dB point being at around 36 Hz vs. the AR's advertised 44 Hz point. I haven't gained so much in extension as in clarity, easily demonstrated with some high power organ music (which I am very fond of).
That's an important part of the game, to sense when something is falling short and doing something about it.
Also, we need to always bear in mind that eah and every one of us hears differently from someone else. No-one has a truly linear hearing, and this cannot but influence our conclusions. I do not claim my speakers are linear only because I hear them as such, but also because the measurements tell me so. Without them, I wouldn't rely on my own hearing only to claim it's so.
At first hearing in my room, the sound was let's say different from what I was used to until then by my AR94 speakers. They are relativel good speakers, but are no match for my 1041 monitors. As I half expected, the main differences were all in the mid to high range, which cleaned up, became even more detailed and overall clarity was much improved. More space around the perfirmers, and the high range, now portrayed by a 1" titanium dome in lieu of a 2" cone, really stood out for its perceived extention and clarity. No fizz, thank you.
All told, I'd say the AR94 would classify as good but not outstanding in any way, whereas the 1041 I would class as outstanding in practically every way except for some really wild downward extension, with its -3 dB point being at around 36 Hz vs. the AR's advertised 44 Hz point. I haven't gained so much in extension as in clarity, easily demonstrated with some high power organ music (which I am very fond of).
I feel reasonably certain that an important prt of Frank's process is twofold. First, he has to find the oustanding bottleneck and do something about it, and then proceed to the enxt bottleneck. Second, I imagine he at least sometimes goes out of his way to try to push a system to its rational limits just to hear what it can do, where "rational" stops short of rebuilding it from beginning to end.
Where the rational point is is hard to define. As an example, right out of the box my Yamaha CDX 993 player sounded very "warm" and very "analog", somebody took great pains to make it sound "analog", i.e. "non digital" (the phrases we hear a lot). It took me about a week to ask myself: where was the price for that located? I started studying the schematics and concluded that the I/V stage op amp, a ubiquious NJR cheaper than dirt op amp. Took it out, installed a gold plated Amphenol socket, took a wild guess and installed an OP FET input amp which never failed to improve Marantz and Philips CD players until then (OP 275). Not so here, so the search went on and finally settled on an AD826. That one didn't take anything away, but instead provided a feeling of ease of operation and defintely a lot more detail. It's inside to this day. Took me about 2 weeks of trying this or that op amp, but it was well worth it.
What pleased me most though was that I also tried running the player hotter by adding small increments of bias to its discrete buffer and output stages. This produced no audible changes in the first two or three steps, and then a degradation set in, which made me think that somebody at Yamaha must have tried the same and settled on what really appears to be the optimum for these circuits.
Where the rational point is is hard to define. As an example, right out of the box my Yamaha CDX 993 player sounded very "warm" and very "analog", somebody took great pains to make it sound "analog", i.e. "non digital" (the phrases we hear a lot). It took me about a week to ask myself: where was the price for that located? I started studying the schematics and concluded that the I/V stage op amp, a ubiquious NJR cheaper than dirt op amp. Took it out, installed a gold plated Amphenol socket, took a wild guess and installed an OP FET input amp which never failed to improve Marantz and Philips CD players until then (OP 275). Not so here, so the search went on and finally settled on an AD826. That one didn't take anything away, but instead provided a feeling of ease of operation and defintely a lot more detail. It's inside to this day. Took me about 2 weeks of trying this or that op amp, but it was well worth it.
What pleased me most though was that I also tried running the player hotter by adding small increments of bias to its discrete buffer and output stages. This produced no audible changes in the first two or three steps, and then a degradation set in, which made me think that somebody at Yamaha must have tried the same and settled on what really appears to be the optimum for these circuits.
Last edited:
Dejan, that's about it - I'm extremely lazy, I only do what absolutely has to be done - I'll never do something just for the heck of it, I'm slowing down more and more all the time, 🙂. What gets me excited is a new angle, a concept of dealing with sound issues that's fresh - I get a buzz out of investigating whether there's a benefit or not - is there a better shortcut to achieving worthwhile sound?
If the view of many parts in a passive x-over is what makes preconceived notions grow and interfere with the evaluation, x-over implementation with analogous (IIR) DSP filters would kill the problem.🙂
I have taken the route of DSP and although I am very satisfied with the ease of implementation, flexibility in testing / measurements and the end results, I have to say I find it very hard to settle on a certain kind of x-over IIR filter.
It may be the syndrome of the too many available choises they are offered to me as I am going between 6,12,18,24,48 dB/octave (plus BW, LR, Bessel with some of them) for two years without a robust conclusion.
Well, while playing with high volume setting, there are really fewer available options, as the 6dB and maybe the 12dB/oct are wiped off but at moderate and low volumes I doubt my choices frequently.
There is a strong factor of preconception in my case I have to point to:
Apart from brief listening tests and very long listening evaluations, I take measurements too and the phase diagrams I take –combined with the established view of “rapid phase change is a bad thing”- influence my decision.
I can’t go to another DSP platform for now, so I can not test constant phase FIR filters.
People who can implement both IIR-FIR filters may find it less hard to acoustically evaluate different x-over topologies and hopefully can inform the rest of us ( I know, every final opinion is valid only for the specific driver units/box/room).
Thanks for that post Jay (and your next one).
George
Last edited:
Hi Tom,
Noted your contribtion to the latest edition of Handbook for Sound Engineers
It's been on my list for sometime - why do these books need to be so expensive?? (Rhetorical question of course).
Jan
Well, mine registered +/- 1.5 dB 40-18.000 Hz at 100 dB SPL at 1 m. If we allow the industry standard +/-3 dB tolerance, that becomes 36-22.000 Hz. Not perfect, but as good as I have ever seen from anybody at more than 1W.
33 Hz is, BTW, system resonance.
Are-you sure ? And what kind of flatness are you referring to ? Response curve in the axis ? Response curve at 30° ? Phase response ? Group delay curve ? Waterfall ?
You can modify the overall response curve with an equalizer.
Did- this mod change really the *character* of the instruments you are listening to ?
Anyway, a flat response curve is not the one i prefer. I tend to prefer a slightly descending curve (3dB on all the bandwidth), as strait as possible, but it is, for me, not the most important part.
Dynamic, and specially micro dynamic is what i'm most aware of. Power handling and distortion is an other important point.
When i have to chose a speaker, the first thing i'm looking at, after the power handling and the efficiency, is the impedance curve. With a little habit, you can see immediately at witch frequency the cone will begin to fractionate, and the resonance frequency. it will tell you immediately if you can use-it for your purpose. It eliminates 4/5 or more of the speakers at first sight.
Then, you will, of course, have a look at the response curve, to see if any high Q accident between or close to those two points. Correlated with the impedance curve, it will give-you indications about the cause of the defect (cone, suspension, chassis etc.). It will not stay a lot of usable speakers in your shopping list at this point. Then the waterfall: i don't like my speakers to continue to play music when my amp is off ;-)
Last edited:
I would have to say that the last point that Christophe just made is perhaps the real killer of most speakers when it comes down to it, the waterfall response. You can have a great looking frequency response and a terrible waterfall plot. any high Q resonance hanging on after input has ceased will kill the sound or cause fatigue after not to long a time. Besides all the other common factors such a bandwidth and FR flatness if you see a waterfall response with some real aberrations it is time to look elsewhere. Take a look at most titanium dome tweeters and the real reason they are so fatiguing jumps out in the waterfall plots in the top octave, the resonant frequency of that material generally falls right at the top of our hearing range. I would conjecture even those who test as not having good high frequency hearing detection will still find this types of high frequency noise will still find that sound fatiguing after some time unless you have truly lost all ability to hear high frequencies which is often not really the case, we just hear at much lower levels with age those high frequencies.
- Status
- Not open for further replies.