I can tell he is very knowledgeable on speakers, I have never said otherwise, but so is Art for what matters.
The point is that it doesn't matter if 5th likes them or not, what matters and should clearly transpire from my posts is the fact that he has NEVER listened to those speakers and still, he refuses to listen to them and at the same time, he has a firm opinion on HOW they sound.
It doesn't matter how much you know about speaker design, you can't really fully judge something until you listen to.
I can personally tell you from experience that, from a listening standpoint, they are as accurate as a speaker needs to be and they have absolutely no detectable/audible peak that will bother your ears in any way.
They are indeed detailed, airy and very transparent speaker; they will detect if you have hooked up something that is not as good as they are. They will not sound good until you drive them with something good..and since many times stores or people assume as speaker does have to sound good regardless what you connect to it,they end up disappointing some people who are not ready to realize that the lower than expected performance is caused by either bad equipment or bad recording (Madonna, Michael Jacson, Metallica etc...)
It really takes skill to design something that will sound better than 800.
In my personal experience, I was unable to find actual problems that matched what 5th element claims of the true sound of these speakers are.
If I did, I would have had no problem confirming those statements and I am sure neither would have Art.
Regarding reviews, yes they are biased, but not to the degree you think they are. Michael Fremer to mention one, has no problem bashing something because it doesn't sound good (and I recall Stereophile reviewing really bad s silver signature speaker which was more for decore than sound).
The 800series has been a reference monitor for a long time and it's not just because of glowing reviews but for actual performance.
I challenge 5th design to bring it up to market and see what the critics says.
I can't really say how hos design will sound for sure I would be really interested to listen to those and to give him my most unbiased feedback (as long as he could take a potential no for an answer).
Do you mind explaining what you meant by "I have nothing against your R2D2's" ?
There's nothing really to explain, one has to interpret CSD diagrams and the mechanisms behind what's causing the issues to understand wha'ts going on.
In the case of the 15kHz problem you will first notice that there are twin peaks either side of the very narrow dip indicating that a resonant system, or two, are at work here. The raised output either side of the dip correlates extremely well with what you'd expect to find in the CSD with two pronounced ridges. The resonance however is centred on 15kHz and not the two ridges either side of it and the nature of the resonance itself is that whatever is resonating is doing it in antiphase with the system as a whole and this creates the sharp dip seen at 15kHz. In other words the dome of the tweeter will be producing output where the dip is centered on, but the anti phase resonance within its construction is causing the dip to occur as it interferes destructively with the tweeters nominal output.
Once the excitation of the system stops, ie the impulse response is over and the energy present within the system starts to die away, we see the dip reach a maximum as the resonance and the tweeters primary output presumably start to balance one another more equally, then as the tweeters primary output decays the resonance takes over and can now be seen as a ridge entirely of its own because it isn't being cancelled out by the tweeters nominal output.
There is nothing unusual about this and if you were to measure the tweeters dome output alone and then measure the output of the resonance alone, the CSDs would follow exactly what you'd expect them too, it's just in this case you've got both playing together and they do not add together constructively.
The notch at around 1kHz could be similar, but the trouble here is that the window applied to the impulse response is too short to really be accurate down that low so it could be a limitation of the system coming into play. Certainly the ridge that shows up at 1kHz doesn't look anything like what you'd expect that would account for the shape of the dip centred on 1kHz and neither does it decay in the way you'd expect either. In fact the frequency of the ridge appears to shift all over the place as time goes on, which isn't how resonances behave.
5TH, congrats you seem to have very good knowledge in matters of speaker design! However, can you please take a proper listen to the 800s and decide if what you hear matches your guesses about sound?
Stefanoo I have no intention of going out of my way to listen to a pair of speakers, I know from experience, will not suit my ears. From my point of view, regardless of the technical merits of any conversation that may transpire, all it seems you are interested in is having people say how awesome your speakers are.
It does not matter how nice you think the 800 series sounds, or if you think they are as accurate as a speaker needs to be. The truth of the matter is that they have their short comings and in rather large quantities. They are not accurate/neutral, their flawed frequency response tells us that and the off axis response issues will ensure that the B&Ws will sound significantly different depending on what environment you place them in. So Art says you can listen slightly off axis and this helps to reduce the impact of the 3.5kHz issue...this isn't actually a good thing. If such a small angle of difference in the vertical off axis (going from the tweeter axis down to the FST axis) can bring about such a dramatic change in the frequency response, then this means the speaker has a tiny primary listening lobe and this will only result in the speaker sounding quite different as you change your listening height by a tiny amount. These are not engineering choices in speaker design that one is to be proud of and they are the exact opposite of what you should be aiming for. A well designed speaker has as large a primary listening lobe and can be listened to at a wide variety of angles/listening heights/distances etc without sounding particularly different.
Stefanoo you say you are an engineer, so how can you possibly fail to see what the objective measurements are saying about the loudspeaker? It's almost like you don't want to believe what the measurements are telling you because you only want to hear praise about the range of speakers you have. I suppose I should point out here that I am not saying that the 800 speaker is a terrible loudspeaker, clearly they do have a lot going for them, but they are a long way from being perfect. If you like them, great, but you would be wise to admit what the measurements say otherwise it ruins your credibility as an engineer.
It does not matter how nice you think the 800 series sounds, or if you think they are as accurate as a speaker needs to be. The truth of the matter is that they have their short comings and in rather large quantities. They are not accurate/neutral, their flawed frequency response tells us that and the off axis response issues will ensure that the B&Ws will sound significantly different depending on what environment you place them in. So Art says you can listen slightly off axis and this helps to reduce the impact of the 3.5kHz issue...this isn't actually a good thing. If such a small angle of difference in the vertical off axis (going from the tweeter axis down to the FST axis) can bring about such a dramatic change in the frequency response, then this means the speaker has a tiny primary listening lobe and this will only result in the speaker sounding quite different as you change your listening height by a tiny amount. These are not engineering choices in speaker design that one is to be proud of and they are the exact opposite of what you should be aiming for. A well designed speaker has as large a primary listening lobe and can be listened to at a wide variety of angles/listening heights/distances etc without sounding particularly different.
Stefanoo you say you are an engineer, so how can you possibly fail to see what the objective measurements are saying about the loudspeaker? It's almost like you don't want to believe what the measurements are telling you because you only want to hear praise about the range of speakers you have. I suppose I should point out here that I am not saying that the 800 speaker is a terrible loudspeaker, clearly they do have a lot going for them, but they are a long way from being perfect. If you like them, great, but you would be wise to admit what the measurements say otherwise it ruins your credibility as an engineer.
FWIW, some of the best audio experiences I have had involved B&W N802 speakers. What was common between the two setups (one of which I heard once only; the other several times) was that (1) they were not run full-range: one crossed over to sub-woofers at 80Hz, the other at 110Hz and (2) the rooms had considerable amounts of acoustic treatment, they were dedicated listening rooms, the latter being developed with said speakers in situ. What stood out for me in these systems was excellent dynamics, which exceeded my personal SPL comfort levels without distorting and smooth midrange. The system that I was most familiar with did have one idiosyncrasy: left-to-right imaging was good, but vertical seemed to be stretched to a wall of sound. For example, vocals could appear precisely between left and right, but did sound unnaturally tall, as if it came from a mouth that had a height and size 1/3rd of the room height.
I do agree with 5th Element's assessment of the speaker and its measurements and crossover design. They closely approximate my own thoughts on this speaker.
I do agree with 5th Element's assessment of the speaker and its measurements and crossover design. They closely approximate my own thoughts on this speaker.
Well at least you've altered your position. That's a good start. Fwiw I found an excellent example of a CSD plot where response peaks (or troughs) don't correlate with resonance behaviour, and vise versa.
An externally hosted image should be here but it was not working when we last tested it.
Clearly, you can have response irregularities (peaks and troughs) that are caused by constructive or destructive interference without a mechanical or electrical resonance involved.
This quote is also of value to the discussion:
Floyd Toole and his associate Sean Olive did considerable work on the audibility of resonances [39, 40]. It is generally held
that high-Q, high-but-narrow peak resonances are less objectionable than low-Q, low-but-broad peak resonances.
This too is of value:
The waterfall above is of the B&W 802D loudspeaker which uses exactly the same FST midrange unit and exactly the same crossover values as the 800 diamond. It's interesting that the 3.5kHz resonance is barely visible, and 100% decay is achieved within 1.26mS.
This suggests to me that the slightly inferior results obtained with the 800 diamond indicate a slight imperfection with its FST driver. IOW, it was a bad sample.
An externally hosted image should be here but it was not working when we last tested it.
The waterfall above is of the B&W 802D loudspeaker which uses exactly the same FST midrange unit and exactly the same crossover values as the 800 diamond. It's interesting that the 3.5kHz resonance is barely visible, and 100% decay is achieved within 1.26mS.
This suggests to me that the slightly inferior results obtained with the 800 diamond indicate a slight imperfection with its FST driver. IOW, it was a bad sample.
Last edited:
that's a good point actually.
If somebody helps me out I can make some measurements on my unit in my treated room. I have a dedicated microphone with calibration.
We can see if that peak shows up on my sig800 or not. 😱
If somebody helps me out I can make some measurements on my unit in my treated room. I have a dedicated microphone with calibration.
We can see if that peak shows up on my sig800 or not. 😱
I haven't exactly changed my position on the subject, all we're doing now is viewing multiple systems working together rather than a single one which is something quite different. You could essentially have two highly resonant systems working against one another, that when deactivated have an individual decay time of minutes, yet when measured appear to do nothing because they cancel each other out perfectly.
I never said that you could not, what I meant was that if you have a single system working on its own, that if a peak or a trough is present then the CSD will look exactly like you'd expect it to. If you've got interference occurring between two separate systems, such as a resonance somewhere cancelling out the output present in another part of the system, then of course you're going to get results that go against that statement.
Control theory however, as far as I am aware, dictates that all information present within the frequency response is present within the original impulse response. And if the frequency response was derived from an ungated impulse, then you can reverse FFT the frequency response to get the impulse response back. This means that the frequency response contains all of the information captured within the impulse response and you can then create the CSD out of it. So theoretically you should be able to predict exactly what's going to occur within the CSD from looking at the frequency response, although this would probably have to be completely unsmoothed.
Yes I know and theoretically speaking you could have a resonance of such high Q that it is never ever excited and is thus completely inaudible. The resonance in the FST spans more than half an octave however in the ears most sensitive region, this is not an inaudible resonance.
The FST drivers are not identical from one speaker to the next and do in fact come in different varieties. You can see evidence of some peak at around 4kHz in the 802D.
An externally hosted image should be here but it was not working when we last tested it.
Here's the frequency response of the 804D and this is appalling.
Obviously the presence of the peak at ~3.3kHz is dependent entirely on how the driver integrates with the tweeter. These are after all designs that are supposed to use 2nd order acoustic slopes centred on about 4kHz, so there's a lot of tweeter contribution within the region of the resonance. This can be seen in the off axis response as the presence of the peak is generally made easier or harder to see depending how the two are summing.
The real point here is that the frequency response is far from flat and even though you can arrive at decently flat with some combinations of off axis angles, this is a long way away from how a speaker should measure. On axis should realistically be as flat as possible (wave guide and horn effects not with standing) with the off axis mirroring the on axis as closely as possible for as long as possible, or with the off axis transitioning into a region of controlled directivity.
Zaph's own measurements of the FST show the presence of the 3.3kHz resonance. His measurements are done on a large flat baffle so don't obscure the measurement with any diffraction issues. For sure the peak doesn't look so terrible in his measurement because the scale is so large, nevertheless if one looks to the harmonic distortion you see its presence once again as an increase in 2nd, 3rd and 5th order harmonics.
You can see, if you like, the presence of this peak (and its associated decay) in the CSD plot.
Regardless of your opinion on these things it's clear that the peak should be avoided and most certainly filtered out if it is to be present within the transition band of the driver.
My own measurements of the same driver are basically identical to Zaph's so I don't think there's an issue of consistency here.
If you've got a measurement mic then that's as good a start as you need. This should help you get started with taking measurements though.
measurement guide
This is what you said:
Clearly it's an incorrect statement. You might be able to use active EQ to produce a flat measured response but you'll never be able to remove the resonance behaviour, no matter how hard you try. That's the whole point of CSD and why it was devised in the first place. It was discovered long ago that frequency response measurements alone could not sufficiently explain audible colourations.
But if control theory says that you can reverse FFT the frequency response to get the impulse response back and then use that impulse response to view the CSD, then technically speaking all the information present within the frequency response does show you all of the resonant behaviour.
The point I was trying to make in the case of a driver exhibiting a cone resonance or similar, such as the 3.3kHz resonance in the FST, is that the peak in the frequency response will always equal an elongated decay time in the CSD. If you then EQ that resonance flat the elongated decay in the CSD will follow suit. Also if another resonance, somewhere else in the construction of the driver is constructive with the nominal output of the driver then it will also cause a peak in the FR and an elongated decay in the CSD. If you EQ that flat, then the CSD will follow suit.
The only issue is that if a resonance is destructive with a drivers nominal output, then it will cause a dip in the frequency response instead of a peak. If this resonance has a long decay time then yes, you will see a peak appear in the CSD, but after the energy from the drivers nominal output has decayed.
I am tempted to say though that if you EQ that dip flat then the extra energy within the CSD will vanish too. Why? Because the extra energy injected to remove the dip will now work against the energy present within the resonance and dampen it out and I am guessing that the scale at which the energy injected into the resonance scales in a different way vs dBs otherwise you'd never be able to EQ it out.
Obviously, if a resonance causes increased distortion then no amount of EQ is going to remove the distortion, which is why you generally want to avoid resonances, but some are benign enough so that you can get away with only needing to EQ them.
I yet have to see/listen to a practical example of even top of the quality EQ that sounds better than no EQ and treated room. This once more to prove that perfect frequency response is not necessarily the key to good sound.
One more example, I have mentioned it already, but let's try again: I can provide yo I with two amplifiers, one that provides medium distortion and frequency response that you would define not perfect but good enough, and another with perfect distortion extremely low and very high bandwidth; one being said uses a enormous amount of feedback to produce near perfect specs, like the input of a precision instrument let's say (but with power), while the other one will use very moderate amount of feedback current only and a less traditional topology and therefore will offer you apparently less bench mark performance.
Which one is going to likely sound better? This is obviously a rhetorical question, but now let me ask you why the same principle doesn't apply to speakers as well?
This to tell you that sound is not only about measurements, which are important but should never take the precedence over good sound.
Regarding the link for measurements, thank, I will check it out tonight....that way when I have some time I can take measurements and post it here so we can comment them.
One more example, I have mentioned it already, but let's try again: I can provide yo I with two amplifiers, one that provides medium distortion and frequency response that you would define not perfect but good enough, and another with perfect distortion extremely low and very high bandwidth; one being said uses a enormous amount of feedback to produce near perfect specs, like the input of a precision instrument let's say (but with power), while the other one will use very moderate amount of feedback current only and a less traditional topology and therefore will offer you apparently less bench mark performance.
Which one is going to likely sound better? This is obviously a rhetorical question, but now let me ask you why the same principle doesn't apply to speakers as well?
This to tell you that sound is not only about measurements, which are important but should never take the precedence over good sound.
Regarding the link for measurements, thank, I will check it out tonight....that way when I have some time I can take measurements and post it here so we can comment them.
Yes, all of that is correct. The impulse response contains all the information, within certain measurement limits of course, and theoretically it's possible to correct the impulse response, but to do so requires an accurate impulse response measurement and DSP (processing), and assumes that the loudspeaker behaviour is the same at all power levels. The only tool that I'm aware of that can do this is DEQX.
Response irregularities are caused by factors other than just driver resonances, so there's no correlation between magnitude of a peak in the FR and the CSD time. A peak that's caused by multipath interference has a very different CSD signature to that of a driver resonance.
Unless you're referring to DSP equalisation of the impulse response of course, because that's the only way that you can correct a driver resonance.
When I talk about EQ I do not mean a generic box of tricks that one adds into their system, I am talking about the general application of crossover filters, these can be active, passive, DSP based or analogue. EQ simply stands for frequency response equalisation and it can be applied wherever you want it to. In the case of loudspeakers generic boxes of tricks should only be used for certain things and the removal of midrange resonances is not really one of them. If you tried to notch out the on axis 3.3kHz peak from the FST you'd affect the output from the tweeter too, which is not necessary and only detrimental to the overall sound quality.
Frequency response linearity is the no.1 criterion for good sound quality and that has been proven time and time again. Why people continue to try and refute this I do not know.
I am not talking about correcting for frequency response issues on the fly, I am simply talking about measuring the loudspeakers and processing the information within those measurements.
The point is that via the correct processing and mathematics one can go from the impulse response to the frequency response and back again without losing any data.
But if from above, the frequency response contains all of the information present within the original impulse response, then the frequency response surely does contain all of the information presented within the CSD as the CSD comes from processing the impulse response.
Where can I find recordings with no eq at all?
As far as I know, every acoustic recording has eq build in. The mics used, the room, the placement of the mics, and of cause the mixing and mastering stages usually involve eq.
"Perfect" frequency response is an other thing that doesn't exist with speakers, simply because the room causes frequency deviations, often quite large! And "perfect" means something different for different people.
If you want "transparency" though, you need a frequency curve of your room speaker combination that is flat within a few dB's.
Agree may i quote this post member ra7 did over full ranger forum (path to full post http://www.diyaudio.com/forums/full-range/270634-first-fullrange-suggestions-14.html#post4248570). And thanks the measurement guide #110 very helpful guide.
