Just curious (based on Simon's comment about narrowing polar at the crossover frequency) if anyone can pick the crossover frequency from the attached plot. No cheating if you have seen my threadand a hint it Isn't between 2 and 2.1Khz
Tony.
*rolls dice*
3 khz!
*shrug*
I'm in this camp too.
I am too. I have been using 2" CDs for 6 months now and Im still enjoying them with the very narrow top end response (its EQed flat on axis with the MiniDSP). I tried the BMS 4590 (coax CD) for a bit but went back to the TAD TD-4001 because its mid range was definitely smoother, there wasn't much a sacrifice on the top end so its a minor compromise.
"Should HiFi loudspeakers be flat?"
Well, of course they should. They could hang on the wall like a poster. They could be posted like a letter.
I don't think you are going to do it though, not today at any rate.
Terry
Speaking of wall hanging speakers, I'd still love to have a pair of these, for the price it'd be hard to go wrong I think.
MMG - Magnepan, Inc.
Yeah, those are flat!
Speaking of cheap tweeters (not) I ran a pair of the good old Scan-Speak D3806 for years with a cheap 12" woofer. Hell of a combo. Giant room. Everyone who heard them, loved them.
But that's going with the expensive choice, not the cheap one. I'm mostly a woofer+horn kinda guy (It's a Classic!) but sometimes other stuff works too.
Speaking of cheap tweeters (not) I ran a pair of the good old Scan-Speak D3806 for years with a cheap 12" woofer. Hell of a combo. Giant room. Everyone who heard them, loved them.
But that's going with the expensive choice, not the cheap one. I'm mostly a woofer+horn kinda guy (It's a Classic!) but sometimes other stuff works too.
No thanks, it's all there on the recording. I just want to reproduce it, not recreate it.
What is cheap? I bought 2 boxes of 3/8" polydomes from Parts Express for $10 a box. That's 400 tweeters at 5 cents each, almost as little as manufacturers pay (and no shipping cost.) My previous record was about 175 of them for 28 cents each also from parts express. These are essentially the same as I paid $6 apiece for with the Dayton name stamped on it, $8 apiece for the Radio Shack name on it, and ??? for the Audax name stamped on it. BTW, among my collection of parts never used are 4 Tonegen Ribbons I bought as part of a lot for about $1 each. I thought they were JVC but Bill Legall at Millersound thought they were Tonegen. I have no idea what they're worth but they were probably not cheap at the time 25 years ago.
Thats simply not true. They are the only comments that I have.
Measurements do dominate my approach, you know that, but the comments are another data point for those who find them useful.
At this time in all my research there are zero controlled listening tests validating your speakers. Should we not hold your speakers to the same litmus test that ever other subjective claim gets?
There were several controlled listening tests early in the design cycle and these were reported. (Quite complimentary, so you would probably reject them as well.)
There are virtually never controlled listening tests done on any loudspeaker and even less so published. Its not because I am avoiding them, or fear them, quite the contrary, I would wholely endorse them, but they are so darn difficult to do - impossible for me actually.
I DO publish extensive objective data when most companies publish almost nothing.
Companies like JBL can do that kind of thing (very few others), but do you think that they are going to report data that is not complimentary to their designs? Do magazines report on "controlled" listening tests or even do fair and valid ones? I often gets requests from people who want to "review" my speakers, but the tests are never controlled so I decline.
I am a one man operation here - you have to consider reality.
I am not sure how you define "dynamic performance", so until I do I can't comment. But how do you account for the results that have been obtained on distortion showing it to be an almost irrellavent factor subjectively.
I used to think nonlinear distortion was a big deal and when I was working with B&C we studied this. We couldn't find any perceptual correlations. I now don't belive that it matters very much. There are things that we hear that people think are nonlinear distortion, but they are not. Lidia and I showed this as well.
Floyd Toole and Sean Olive don't believe that nonlinear distortion is a big deal either. I think that the data all points to this being the case.
Well, I don't think we can say "it's not a big deal" as a blanket statement, but we can say that the threshold that it becomes a big deal might be different, right?
So, how much non-linearity should we be willing to accept in a speaker?
When I said dynamic performance that was a lazy way of saying everything that's important about the large signal performance.
In other words:
1) sufficient SPL headroom (at all frequencies) beyond not just the average SPL levels but also beyond anticipated peak SPL levels
2) Distortion not rising excessively when reaching those high peak SPL's (staying below the distortion vs SPL curve's knee)
3) Minimal amounts of power compression as well as frequency dependant compression. (Eg minimal frequency response deviation from the small signal response)
To put it simply, ensuring that speakers are sufficiently over-rated for the intended use with a good safety margin.
I know that you are a big believer of "distortion doesn't really matter", (you've expressed this view many times before) yet your designs will inherently have very low distortion at higher SPL compared to the typical audiophile speaker due to the use of large high performance drivers such as wave-guides and large woofers. In other words your designs are vastly over spec'ed for the probable playback SPL levels, (that's not a criticism, I think its a good thing) and achieve low distortion at normal SPL as a by-product of that.
Despite saying that distortion doesn't matter and doesn't correlate with listener preference, you used it as a criticism of dome tweeters earlier in this thread:
So which is it ? Does distortion matter, or doesn't it ?
Clearly when a driver is overloaded two things happen, one is dynamic range compression (often frequency dependent) and the other is a large increase in distortion. Drivers don't sound good when they're overloaded, and its hard not to conclude that distortion is the main reason.
If that's the case there must be some threshold of distortion for a given frequency range above which the sound is unsatisfactory. Perhaps the threshold is higher than most people think, perhaps once you're significantly below that threshold further decreases aren't audible. Perhaps distortion below this threshold can't be used to rank speaker preference for that reason, eg once distortion is "low enough" at all SPL's then its out of the equation, and trying to reduce it to vanishingly small levels is wasted effort. Maybe. But what is the threshold then ?
I'm not as convinced as you though, and I still think that lower levels of distortion which are below the "gross distortion" threshold (whatever that might be) may still be audible in subtle ways which aren't perceived as distortion per se, but for example as "congestion" in a spectrally rich piece of music like massed strings where a lot of intermodulation products will be present.
I'm open to being wrong about this though.
Last edited:
Now we are getiing into the details.
Serious levels of clipping type of distortion are a problem, but tracking THD etc. at anything less than a clipping type situation are meaningless. I will agree that many systems all appart at higher levels because the drivers are under spec'd. What I mean by my distortion position is comparing the THD levels of two drivers operating within their design range and saying that the one with the lower THD will sound better. That is simply ridiculous.
Yes, I am a big believer in "head room" but not because the THD is lower. It has to do with thermal performance. I am a big believer in thermal modulation (which is distinctly different than thermal compression). I am preparing to do a controlled experioment in this regard to prove my point. Large efficient speakers are going to be orders of magnitude lower in thermal modulation than smaller inefficient ones. This is why I say that a small tweeter is not acceptable. The difference in thermal modulation between a 1" compression driver on a waveguide and a 1" dome tweeter is going to be 50 to 100 times.
So yes, as you say, my designs are inherently low distortion at high SPLs, but thats more an added benefit than a stated design criteria.
How much linearity should we accept in a speaker? That depends entirely on what the shape of the nonlinearity is. I prefer some third order nonlinearity over the whole stroke to a sharp abrupt change in the BL(x) curve at some displacement. If you look at the Klippel data for the B&C drivers that I use you will see that they do precisely this. The idea that a flat BL(x) curve is a good thing is false. Pure second and third order nonlinearities are virtually in audible - so why not use them to advantage.
And once again, I will say this for the 10^x time - I do NOT claim that all distortion in a any loudspeaker is inaudible only that it is eminately reasonable to design a speaker such that this is true. If it is not true, then either 1) the speaker is broken 2) poorly designed, 3) used beyond its limits. I just don't use poorly designed or broken transducers and I design for the kinds of limits that one often sees in Home Theater, so nonlinear distortion is simply not a factor in my design criteria.
It is perhaps semantics, but I hear all too many claims that such is such is "better" because it has "lower distortion". As a blanket state that is just as untrue as the opposite is. In reality the situation is far more complex.
Serious levels of clipping type of distortion are a problem, but tracking THD etc. at anything less than a clipping type situation are meaningless. I will agree that many systems all appart at higher levels because the drivers are under spec'd. What I mean by my distortion position is comparing the THD levels of two drivers operating within their design range and saying that the one with the lower THD will sound better. That is simply ridiculous.
Yes, I am a big believer in "head room" but not because the THD is lower. It has to do with thermal performance. I am a big believer in thermal modulation (which is distinctly different than thermal compression). I am preparing to do a controlled experioment in this regard to prove my point. Large efficient speakers are going to be orders of magnitude lower in thermal modulation than smaller inefficient ones. This is why I say that a small tweeter is not acceptable. The difference in thermal modulation between a 1" compression driver on a waveguide and a 1" dome tweeter is going to be 50 to 100 times.
So yes, as you say, my designs are inherently low distortion at high SPLs, but thats more an added benefit than a stated design criteria.
How much linearity should we accept in a speaker? That depends entirely on what the shape of the nonlinearity is. I prefer some third order nonlinearity over the whole stroke to a sharp abrupt change in the BL(x) curve at some displacement. If you look at the Klippel data for the B&C drivers that I use you will see that they do precisely this. The idea that a flat BL(x) curve is a good thing is false. Pure second and third order nonlinearities are virtually in audible - so why not use them to advantage.
And once again, I will say this for the 10^x time - I do NOT claim that all distortion in a any loudspeaker is inaudible only that it is eminately reasonable to design a speaker such that this is true. If it is not true, then either 1) the speaker is broken 2) poorly designed, 3) used beyond its limits. I just don't use poorly designed or broken transducers and I design for the kinds of limits that one often sees in Home Theater, so nonlinear distortion is simply not a factor in my design criteria.
It is perhaps semantics, but I hear all too many claims that such is such is "better" because it has "lower distortion". As a blanket state that is just as untrue as the opposite is. In reality the situation is far more complex.
All of my extensive testing say that this is not the case in loudspeakers. This is mostly because loudspeakers cannot have high order nonlinearities at lower levels because they are mechanical devices. There are exception to this rule however. An "oil can" type of nonlinearity in a suspension is an example, but very rare. A rub that occurs at rest position crossing is another example, but we call this "broken".
High order nonlinearities are not only possible in electronics but almost guaranteed without careful design. This is why you can hear a .1% distortion in an amp played through a speaker with 5% distortion.
Possibly related to the matter of identifying nonlinear distortion by ear -- sort of anecdotal (but something that can be heard for yourself). I made a scheme a few years back that could separate nonlinear (harmonic or IM) distortion from linear (frequency response) distortion in speaker playback, and could provide separated recordings of each.
A funny thing came up when checking out a part of a recording that sounded like what I would have thought would have clearly been non-linear distortion -- it turned out to be linear distortion.
There was a Speaker Builder article back then about it, text and WAV files linked here: Distortion Isolation in the Time Domain.
The WAV file referenced is the LSLB DI Low level.wav if anyone wants to hear (linear portion is in the left channel, nonlinear is in the right).
edit: oops. Looks like the wav files are no longer up on the site. If anyone wants to hear them, I can try to track down some backup copies....
A funny thing came up when checking out a part of a recording that sounded like what I would have thought would have clearly been non-linear distortion -- it turned out to be linear distortion.
There was a Speaker Builder article back then about it, text and WAV files linked here: Distortion Isolation in the Time Domain.
The WAV file referenced is the LSLB DI Low level.wav if anyone wants to hear (linear portion is in the left channel, nonlinear is in the right).
edit: oops. Looks like the wav files are no longer up on the site. If anyone wants to hear them, I can try to track down some backup copies....
Last edited:
Waveguide
The referenced article shows that a compression driver will produce far better results from a waveguide than a dome tweeter will. It also shows that the best performance will be obtained in the graphs shown with a crossover frequency not lower than 3 to 5 Khz. The combination of compression driver and suitable waveguide produces remarkably flat response over a very wide angle. However, the problem is that while the relative response is flat, the absolute response isn't. As you move laterally, output increases or decreases with angle although uniformly across the spectrum. This would be unimportant were it not for the fact that other drivers in the system do not behave the same way. Therefore the overall output on axis will change as a shelf for the tweeter but not for lower frequency drivers. This will alter the ovrall perceived spectral balance especially in an acoustically dead room.
When toed inward, as you move from left to right, you should be able to notice the difference. The best angle to aim it at is the one at which the rise in output as you move away from the speaker laterally and closer to its axis, decreases with increasing distance at the same rate as it increases with approaching on axis. That is they should compensate each other. This will be different for different distances from the speakers so the best results should be obtained in rooms that are not relatively deep or where the speakers are placed along the long wall of the room. The best compromise is to find this angle halfway between the speakers and the wall behind you.
Like any speaker this design will produce many reflections which will not have the same spectral relationships as the incident wave. That part of the field, the reverberant part will remain relatively constant with angle mitigating to a degree the change in absolute output from the directly ariving sound as you move around the room.
The referenced article shows that a compression driver will produce far better results from a waveguide than a dome tweeter will. It also shows that the best performance will be obtained in the graphs shown with a crossover frequency not lower than 3 to 5 Khz. The combination of compression driver and suitable waveguide produces remarkably flat response over a very wide angle. However, the problem is that while the relative response is flat, the absolute response isn't. As you move laterally, output increases or decreases with angle although uniformly across the spectrum. This would be unimportant were it not for the fact that other drivers in the system do not behave the same way. Therefore the overall output on axis will change as a shelf for the tweeter but not for lower frequency drivers. This will alter the ovrall perceived spectral balance especially in an acoustically dead room.
When toed inward, as you move from left to right, you should be able to notice the difference. The best angle to aim it at is the one at which the rise in output as you move away from the speaker laterally and closer to its axis, decreases with increasing distance at the same rate as it increases with approaching on axis. That is they should compensate each other. This will be different for different distances from the speakers so the best results should be obtained in rooms that are not relatively deep or where the speakers are placed along the long wall of the room. The best compromise is to find this angle halfway between the speakers and the wall behind you.
Like any speaker this design will produce many reflections which will not have the same spectral relationships as the incident wave. That part of the field, the reverberant part will remain relatively constant with angle mitigating to a degree the change in absolute output from the directly ariving sound as you move around the room.
Here is a link to two of the WAV files mentioned above -
LSLB DI1.wav
The original (not played through speakers, not separated) is here:
LSLB.wav
LSLB DI1.wav
The original (not played through speakers, not separated) is here:
LSLB.wav
Ah, good question. In order to achieve the desired result consistently the engineer must take every variable that affects the ultimate performance into account and design systems that will compensate for them. This begins with a profound understanding of how sound works and of the recording process. Short of this the results will be inconsistent and never satisfactory for those who expect that a high fidelity sound system will reproduce the sound of live musical performances convincingly to critical listeners. This may have been nothing more than a pipe dream in the 1970s but science and engineering have come a long way in 40 years. I think if manufacturers of so called high end audio equipment cannot achive this simple goal compared say to the achievement of creating a laptop computer then the asking price for their products is way out of line with what you get for your money. Am I going to tell you how this problem is correctly and completely analyzed and solved? No. Do I know how to?.......
So, nothing specific to add to the discussion then ? Just more vague and sketchy doublespeak. Ok then...
So, nothing specific to add to the discussion then ? Just more vague and sketchy doublespeak. Ok then...
Asking questions and pointing out major discrpancies no one can answer to doesn't mean I have to supply the answers....if I have them. A lot of people got paid a lot of money for the wrong answers. How much would be paid for the right answers. Considering the near impossiblity of recovering revenues lost from stolen intellectual property these days even when you have a patent, it doesn't pay to give away anything worth knowing that isn't already in the public domain.
Interesting paper. Thanks for attaching it.
I wouldn't agree that it shows compression drivers outperforming domes with waveguides. The only sub par performers were the 20mm SEAS before the waveguide was flush mounted and the 25 mm dome before the phasing disc. Afterwards both are beyond reproach.
I am frequently doing these calculations for theaters but flipped to the vertical plain so the level increase that comes from moving onto the beam of the horn compensates for the varying distance from the front of the theater to the back. You can equalize level across a plane in your listening room, but that will not be adequate to achieve any time/intensity trading. Movement from one side to the other will still see the image quickly collapse to the near speaker.
David S.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.