Bit harsh there Mr. bentoronto! 
Dave Bullet was talking about a KEF C30 specified at +/- 3dB, and concluding that it doesn't sound good or accurate enough. Seems it might have a tweeter fault, so all conclusions are on hold. And, FWIW, the KEF was probably much better than that if we discount the lowest bass.
Mr. Juhazi is quite outrageous here in his prejudices:
I think good engineers have known about dispersion for a very long time, and in fact used measures like BW3 and diffusors to improve it, which is now a near lost art. It helped that a lot of them knew a bit about antennas and radar where similar issues arose, and I am thinking of Jim Thiel in particular.
The "British" or "BBC" sound back then was largely biased towards low efficiency and flat or uncoloured sound which is always compared to the original sound of concert or voice performance. Perhaps like the US "East Coast" sound.
The other school of thought was high efficiency or "West Coast" with its more ragged response, but arguably more detail. People always said it suited Rock music.
Well, I really don't know. +3dB is double the power, so +/- 3dB isn't saying much really. I can usually do better than that. But it doesn't mean the speaker wins a cigar. As people have wisely said, there are other factors.
An experienced toiler in the BBC vineyard is Alan Shaw of Harbeth. He is extremely interesting to listen to, IMO.
YouTube
Dave Bullet was talking about a KEF C30 specified at +/- 3dB, and concluding that it doesn't sound good or accurate enough. Seems it might have a tweeter fault, so all conclusions are on hold. And, FWIW, the KEF was probably much better than that if we discount the lowest bass.
Mr. Juhazi is quite outrageous here in his prejudices:
I think good engineers have known about dispersion for a very long time, and in fact used measures like BW3 and diffusors to improve it, which is now a near lost art. It helped that a lot of them knew a bit about antennas and radar where similar issues arose, and I am thinking of Jim Thiel in particular.
The "British" or "BBC" sound back then was largely biased towards low efficiency and flat or uncoloured sound which is always compared to the original sound of concert or voice performance. Perhaps like the US "East Coast" sound.
The other school of thought was high efficiency or "West Coast" with its more ragged response, but arguably more detail. People always said it suited Rock music.
Well, I really don't know. +3dB is double the power, so +/- 3dB isn't saying much really. I can usually do better than that. But it doesn't mean the speaker wins a cigar. As people have wisely said, there are other factors.
An experienced toiler in the BBC vineyard is Alan Shaw of Harbeth. He is extremely interesting to listen to, IMO.
YouTube
We all prefer speakers that sound great to us. I don't know if you have an ideal speaker. What frequency response you would target or does that depend on enclosure and driver materials?
Let me revise my bold *** claim..
It's ok if a designer deliberately puts a dip or tapered slope or no bsc etc into a design if they are making a conscious design decision. What I was referring to is a speaker response where the response is all over the place where breakup modes, baffle diffraction effects or overly smoothed curves used for simulation
It's ok if a designer deliberately puts a dip or tapered slope or no bsc etc into a design if they are making a conscious design decision. What I was referring to is a speaker response where the response is all over the place where breakup modes, baffle diffraction effects or overly smoothed curves used for simulation
Amongst nebulous waffling and opinions, I always think it is good to land on planet Earth and actually try real stuff.
I am playing around with this 8" plus 2" one at the moment:
It is not a million miles from the KEF C30 in fact. The £6.50 Monacor HT22/8 is generally despised here at diyaudio as being too low tech. I think it is a great and interesting tweeter. So does Joachim Gerhard, as it goes, though he found that only 20% of them were within any sort of accurate tolerance, which is too poor for his exacting standards.
The frequency response is fairly wobbly at +/- 2dB within a flattish overall envelope.
Test Lautsprecherchassis Hochtoner - Monacor HT22/8
It is hard to say how much of this response is the measuring microphone and how much is the driver. For sure, it is better than the Visaton TW70, which lacks resolution at the top. I did actually take a TW70 apart after an unfortunate incident of putting a screwdriver through the cone. It has a spider, which surprised me. No ferrofluid. It's actually constructed exactly like a bigger midbass in its own sealed enclosure.
I am playing around with this 8" plus 2" one at the moment:
It is not a million miles from the KEF C30 in fact. The £6.50 Monacor HT22/8 is generally despised here at diyaudio as being too low tech. I think it is a great and interesting tweeter. So does Joachim Gerhard, as it goes, though he found that only 20% of them were within any sort of accurate tolerance, which is too poor for his exacting standards.
The frequency response is fairly wobbly at +/- 2dB within a flattish overall envelope.
Test Lautsprecherchassis Hochtoner - Monacor HT22/8
It is hard to say how much of this response is the measuring microphone and how much is the driver. For sure, it is better than the Visaton TW70, which lacks resolution at the top. I did actually take a TW70 apart after an unfortunate incident of putting a screwdriver through the cone. It has a spider, which surprised me. No ferrofluid. It's actually constructed exactly like a bigger midbass in its own sealed enclosure.
Actually, it is hard to say how much of this is smoothing.
We smooth FR curves to satisfy our eyeballs and because there just isn't much we can do (or need to do) about drastic but narrow-band variations. The true FR is unsmoothed and it looks like the Rocky Mountains, with far bigger variations between notes. And even that unsmoothed FR is subject to the testing context.
For sure, the ⅓ octave smoothing* that manufacturers supply (with "truthful hyperbole" as a certain well-known person says) provide a reasonable sense of the over-all FR envelope and a basis for parametric EQ to follow. They do not represent what our ears hear when we do a slow sweep.
B.
*manufacturers generally don't tell us what smoothing was applied (or anything else about FR tests... and very rarely information on GD or distortion which they easily could provide), thus moving from "truthful hyperbole" more towards "disinformation"
Last edited:
Of course it's the overall envelope we are interested in. And +/-2dB is doable.
This is interesting, if you put your faith entirely in flat frequency response.
Did anyone actually read Joachim Gerhard of Sonics? NO.
YouTube
Or listen to Alan Shaw of Harbeth. NO.
Now come on. Tell me something I don't know, instead of everything I do.
Joachim Gerhard said:
This is interesting, if you put your faith entirely in flat frequency response.
SpeakerBuilding.com - Interview with Joachim Gerhard of Audio Physic, Page 1Joachim Gerhard said:
Did anyone actually read Joachim Gerhard of Sonics? NO.
YouTube
Or listen to Alan Shaw of Harbeth. NO.
Now come on. Tell me something I don't know, instead of everything I do.
The simulation response can be very different from the real deal:
For this speaker project Usher 8945A +9950-20tweeter soft dome (note: usher speakers are back on sale on partsexpress!!!)
Xo point is 2250hz, in simulator more 1600!
Note the : real response is maybe the last design, but in any case it is very close to the actual XO simulated.
Here is the attachments of the simulated response in the baffle, tweeter is copied from the Scan-Speak and should match the Usher tweeter (1/5 price)
Predicted in simulator:
Real response:
For this speaker project Usher 8945A +9950-20tweeter soft dome (note: usher speakers are back on sale on partsexpress!!!)
Xo point is 2250hz, in simulator more 1600!
Note the : real response is maybe the last design, but in any case it is very close to the actual XO simulated.
Here is the attachments of the simulated response in the baffle, tweeter is copied from the Scan-Speak and should match the Usher tweeter (1/5 price)
Predicted in simulator:
Real response:
Attachments
Last edited:
Floyd Toole says at least 20dB or more because fine details are perceivable - but who listens that guy anyway. There are so many internet gurus that know so much more.
I gather from this thread from local gurus that loudspeaker frequency measurements should be as non linear and as smoothed as they can be. Certainly not +/-1.5dB because then they don't sound natural. Oh yes, and anechoic measurements are waste of time.
All in all - cool.
I gather from this thread from local gurus that loudspeaker frequency measurements should be as non linear and as smoothed as they can be. Certainly not +/-1.5dB because then they don't sound natural. Oh yes, and anechoic measurements are waste of time.
All in all - cool.
Last edited:
1/12 octave smoothing seems to be industry standard.
As an aside, does anybody share my feeling that SPL trace methods give wildly inaccurate and oversmoothed phase in simulators? Actual measurements for FRD and ZMA as Dayton nobly do seems to work better.
I was looking at gabdx's speaker with typical drivers to get a feel for it. I think the sim got it right.
Don't take this personally, gabdx, but I found an impedance that dropped to 2 ohms due to a very bad tweeter filter and some severe level problems. I don't think the actual driver responses will save you here. This needs more work.
As an aside, does anybody share my feeling that SPL trace methods give wildly inaccurate and oversmoothed phase in simulators? Actual measurements for FRD and ZMA as Dayton nobly do seems to work better.
I was looking at gabdx's speaker with typical drivers to get a feel for it. I think the sim got it right.
Don't take this personally, gabdx, but I found an impedance that dropped to 2 ohms due to a very bad tweeter filter and some severe level problems. I don't think the actual driver responses will save you here. This needs more work.
Attachments
No offence, but this is more a demonstration of flawed modelling and measurement technique rather than room effects. Using the Scanspeak model was misguided - the Usher tweeter while 'visually inspired' by SS tweeters is not the same as any SS tweeter.
With correct technique there is no reason that simulation and measurement can't agree within 1dB or so, e.g: Zaph|Audio - ZRT - Revelator Tower
Room effects should certainly not substantially change the effective XO frequency.
Indeed, much good stuff there, jameshillj. I have heard one of Joachim's very expensive Manger/Scan speakers. Faultless execution. Joachim and Alan Shaw are very generous with their views, based on huge experience, knowledge and extremely accurate measurement and tolerances.
Joachim continues to tell us interesting things:
Sonics by Joachim Gerhard cabinets and kits.
Now I had never thought of notching a metal dome at 27kHz!
At gabdx and TMM, that Usher does look a bit Scan 18W/8531G00ish. A horrible driver, IMO! The Usher has a useful natural rolloff at 3kHz which might help things along.
Attachments
Last edited:
Yes, The lateral thinkers still lead the way in speaker development, IMO
I'm fascinated by some people spending a huge amount on amps and speakers that supposedly will give them a flat freq response and then they set them up in their lounge room and get a very 'non-flat' result and much 'agro' - we lag so far behind 'pro-audio' it's not funny.
There's a company here that uses these same drivers in a number of their speakers - from the graph, it looks okay up to maybe 700Hz - maybe they want them to sound 'off' for a market 'difference' or something
I received the Transcendental Audio newsletter recently that mentioned the "Fixer', a tone control preamp - much easier to use than a dsp (I'll bet that gets a few comments!)
I'm fascinated by some people spending a huge amount on amps and speakers that supposedly will give them a flat freq response and then they set them up in their lounge room and get a very 'non-flat' result and much 'agro' - we lag so far behind 'pro-audio' it's not funny.
There's a company here that uses these same drivers in a number of their speakers - from the graph, it looks okay up to maybe 700Hz - maybe they want them to sound 'off' for a market 'difference' or something
I received the Transcendental Audio newsletter recently that mentioned the "Fixer', a tone control preamp - much easier to use than a dsp (I'll bet that gets a few comments!)
Bass and treble one controls became unfashionable for a while, putting something in the way an' all that, but I find them invaluable.
Terribly dependent on listening level, of course, but when you get the balance right, it seems taking the tone controls up and down from flat has maximum effect.
It's very hard to say whether closed box or reflex does the right things. I prefer closed box for studio recordings like Mark Knopfler's stuff. But for a BBC prom of Mahler's 4th with the big orchestra with the big bass foundation of the Albert Hall, reflex packs the wallop that makes it sound real. We know about room gain, of course.
I mentioned my current project:
I am very much at the listening and fine tuning stage for whatever it does well. It is a design that is hardly revolutionary, quite retro in fact. The BBC used Zobel impedance correction a lot. Makes for flattish impedance at the input and the drivers, which must be win-win.
But, TBH, it works very well without them with a bit of redesign into simpler circuits.
The treble quality is interesting. With the accurate Zobel it is very polite but rather rolled off, lacking air. Without it, it has the all-too-familiar splashy, extended and bright top end.
I am thinking I need something inbetween. What is a half-Zobel? Twice the resistance and half the capacitance, it turns out. 15R and 0.68uF. Put two in parallel and they make the full 1 unit.
Why does that seem an interesting idea on theoretical grounds? Because according to our latest theories about quantum theory and information, and Claude Shannon I am looking at you, a half is the minimum unit of information and Planck's constant of action. Hope that isn't too nebulous.
I must try this. Ideas are everything.
Terribly dependent on listening level, of course, but when you get the balance right, it seems taking the tone controls up and down from flat has maximum effect.
It's very hard to say whether closed box or reflex does the right things. I prefer closed box for studio recordings like Mark Knopfler's stuff. But for a BBC prom of Mahler's 4th with the big orchestra with the big bass foundation of the Albert Hall, reflex packs the wallop that makes it sound real. We know about room gain, of course.
I mentioned my current project:
I am very much at the listening and fine tuning stage for whatever it does well. It is a design that is hardly revolutionary, quite retro in fact. The BBC used Zobel impedance correction a lot. Makes for flattish impedance at the input and the drivers, which must be win-win.
But, TBH, it works very well without them with a bit of redesign into simpler circuits.
The treble quality is interesting. With the accurate Zobel it is very polite but rather rolled off, lacking air. Without it, it has the all-too-familiar splashy, extended and bright top end.
I am thinking I need something inbetween. What is a half-Zobel? Twice the resistance and half the capacitance, it turns out. 15R and 0.68uF. Put two in parallel and they make the full 1 unit.
Why does that seem an interesting idea on theoretical grounds? Because according to our latest theories about quantum theory and information, and Claude Shannon I am looking at you, a half is the minimum unit of information and Planck's constant of action. Hope that isn't too nebulous.
I must try this. Ideas are everything.
Attachments
The simulator got it right that the woofer response is within 1 db for most, actual room gives more like +/- 3db.
but no 4 db bump at 50 hz, not measured or observed.
Simulator would suggest a 8ohm instead and a 2.1 Input inductor.
However listening test proved 10ohm 2.4 a lot better for high frequency.
For the tweeter:
The filter with 7.5 ohm actually corrected the dip in the tweeter response, with no filter the impedance is not 2 ohm it is 0 ohm in the simulator.
More filter with more impedance would make the tweeter sound not equal.
The response that you see is close to the final design which I can't pin-point in the 50 simulations I did.
It is from 2 meters off-axis and gives a great sound, sweep tones are quite controlled.
I just say that the actual XO is surprising if you look what I would come with in the simulator to correct the bumps which are not measured in real life.
I admit the tweeter is very badly simulated... lack of data
but no 4 db bump at 50 hz, not measured or observed.
Simulator would suggest a 8ohm instead and a 2.1 Input inductor.
However listening test proved 10ohm 2.4 a lot better for high frequency.
For the tweeter:
The filter with 7.5 ohm actually corrected the dip in the tweeter response, with no filter the impedance is not 2 ohm it is 0 ohm in the simulator.
More filter with more impedance would make the tweeter sound not equal.
The response that you see is close to the final design which I can't pin-point in the 50 simulations I did
.It is from 2 meters off-axis and gives a great sound, sweep tones are quite controlled.
I just say that the actual XO is surprising if you look what I would come with in the simulator to correct the bumps which are not measured in real life.
I admit the tweeter is very badly simulated... lack of data
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.