Back in my more naive days I thought switchable +/- L-Pad's (instead of fixed) for midrange and treble (in a 3 way design) was a useful addition to make a speaker more "flexible" for different room placement, but I eventually learned the error of my ways. 😛
It's simply not a good idea to attempt to make a change to the frequency response of a speaker (for example for boundary compensation) by messing with the attenuation of each driver individually, even if the crossover frequency happens to fall at the frequency where you want the response to roll over.
Quite apart from the fact that it introduces an abrupt shelving type effect when what's usually required is a variable slope effect, it completely messes up the very delicate balance between the two adjacent drivers.
The phase response is screwed up, the crossover frequency is shifted, the relative frequency overlap changes, and so on. For example even a small drop in tweeter level can cause the midrange driver to become dominant in the low treble range, unless you're using a very steep filter, in which case you get a very abrupt change in response instead - neither are good.
Level mismatch at the crossover point always sounds "funny". A speaker should be designed with a fixed crossover with fixed L-Pads so that the response of each driver is level matched at the crossover point, and if any additional EQ is required for example to cope with boundary effects this compensation should be applied before the crossover (eg at line level) so that the phase and amplitude relationship between the drivers is not changed.
I pretty much agree - for preset bass equalization for dealing with room modes and gain a parametric EQ is the best, or narrow band graphic EQ if parametric is not available, but for general "balance" issues with variations in recordings a set of variable slope tone controls (and maybe the linear equalizer mentioned on the luxman) is all you need, and is preferable IMHO to either parametric or graphic equalizers, which offer "too much" control and complexity, but at the same time make it very difficult to apply accurate variable slope type corrections.
We're not talking about the general public though are we, most people who read this forum will have a greater appreciation of how to use tone controls or EQ in general, one would hope 😉
I really don't like Loudness controls, (although that's a whole other topic for me to go into why) and it's interesting to note that of all the array of tone controls the Luxman amp has, it does not have a tapped volume loudness control, when nearly every other amplifier of the era did. I think there's a good reason for that...
On the other hand even though my current amplifier does have a loudness control I never use it...
Last edited:
Hi Earl,
1. Someone asked when measuring a driver distortion in a production environement was useful, you said never, I pointed out one example. That's all.
2. "a 90 degeree phase shift does not decorrelate a signal and Greisinger does not say that either. He does say that they phase shift the LF by 90 degress and "it sounds better" but nowhere does he imply that this is because the two channels are decorrelated"
He most certainly does, it's clearly stated. The term is also used in a host of AES and IEEE papers denoting Butterworth hi and low pass xovers being"decorrelated" at the xover frequency. Its a common use of the term, perhaps one that you're unfamiliar with. Yes, in this context, its applied at a single frequency, not against a time varying complex multi-frequency envelope.
The portion of your feedback regarding room reflection randomizing the signal and leading to decorrelation I now agree with, well stated. I was wrong, randomness does decorrelate for a complex signal.
Dave
I know more than many what it means, both my parents families grew up under it and were displaced from their homeland because of it.
It was rhetorical to denote an ideological mentality forbidding and oppressing dissension. Apologies if it led to offense beyond that, none was intended. HumD extended it well beyond my intent.
In the context here we were talking about audio, not telephones, and in that context my statement was correct. I understand that when the device is in an adaptive loop that its nonlinearities are an issue, but not because they are audible, its because they screw up the algorithm.
There are adaptive algorithms for nonlinear systems - that is Wolfgang Klippels expertise.
Then its a misuse of the term because, from the definition, correlation is a time signal which also has a frequency domain representation - called the Cross Spectrum. If you apply this deffinition to two pure tone sine waves 90 degrees out of phase the answer will be a cosine at the same frequency. It is not zero except at a few locations. To call this "decorrelated" is taking a great deal of artistic license.
I taught a graduate coarse in signal processing at one time.
Last edited:
Favorite Topic. First where I am coming from. I listen to MUSIC I LOVE, not great recordings per se. I would rather listen to L Armstrong from the 30's on a recording that is poor by the SOTA standards than listen to an audiophile recording of the same music performed by marginally talented musicians. I listen to music from the 20's through the present. A lot of the artist I listen to are not with us, the recording we have are it, recorded well or not.
So ideally we start with a great pair of speakers that won't do drastic damage to the recording, Earl G has defined the characteristics of a properly designed speaker system (including bass requirements) for a small (home) listening space. Then we place them in a properly designed room (See RPC acoustics), and good electronics.
Having all that Flat is irrelevant . The recordings are all over the place, recorded with different technologies, in different rooms, with unknown mixing consoles (how many 741 op amps with their slewing problems
in a 70/80's era mixing console?). Speakers, how many poorly designed speakers with diffraction problems and so on. Mixing rooms with differing acoustics that have nothing to do with your listening room.
Given the differences of every recording in your music collection. Flat doesn't exist, never will.
What the world needs is a tone control to correct program material for YOUR listening space and YOUR music. I had a chance to hear a properly designed Tone control and what it could do for a recording. The Cello Audio Pallette, the operating points were: up to ±22dB at 20Hz 1db increments, ±12dB at 120Hz .5db increments, ±6dB at 500Hz .25db increments, ±6dB at 2kHz .25db increments, ±12dB at 5kHz .5db increments, and ±22dB at 20kHz 1db increments, the filters were low Q filters so the bands overlapped.
Levinson got the idea from Dick Burwin. The component cost ~10k when first introduced in 1984, it had >6,000 components IIRC lots of discrete amps (an audiophile no no). I was in NYC near Levinson's Cello showroom and stopped by, I got to listen to the complete system. The Pallette is what really surprised, I had been conditioned to think of tome controls as bad, but the Pallette transformed 'poor' recordings. There was a recording of The Goldberg Variations, bright unlistenable, using the Pallette transformed the recording to one that was great to listen to we went through a number of recordings same even 'good' recordings were made better using the Pallette. I played with 'correcting' recordings very intuitive and easy to use. Sadly I could afford the unit (still can't).
Changed my view of what was important. I wish I could have something like the Pallette. It would add value to all the recordings in my library.
D Burwin is selling something similar implemented in software. Levinson is (I think) selling it in conjunction with a D/A converter made by his new company. It is different in that it is intended to be used with loss less Digital files. He (Burwin) is also advocating adding some Hi Freq reverb to the digital recording. I would like to hear the software to see if it is usable.
So ideally we start with a great pair of speakers that won't do drastic damage to the recording, Earl G has defined the characteristics of a properly designed speaker system (including bass requirements) for a small (home) listening space. Then we place them in a properly designed room (See RPC acoustics), and good electronics.
Having all that Flat is irrelevant . The recordings are all over the place, recorded with different technologies, in different rooms, with unknown mixing consoles (how many 741 op amps with their slewing problems
in a 70/80's era mixing console?). Speakers, how many poorly designed speakers with diffraction problems and so on. Mixing rooms with differing acoustics that have nothing to do with your listening room. Given the differences of every recording in your music collection. Flat doesn't exist, never will.
What the world needs is a tone control to correct program material for YOUR listening space and YOUR music. I had a chance to hear a properly designed Tone control and what it could do for a recording. The Cello Audio Pallette, the operating points were: up to ±22dB at 20Hz 1db increments, ±12dB at 120Hz .5db increments, ±6dB at 500Hz .25db increments, ±6dB at 2kHz .25db increments, ±12dB at 5kHz .5db increments, and ±22dB at 20kHz 1db increments, the filters were low Q filters so the bands overlapped.
Levinson got the idea from Dick Burwin. The component cost ~10k when first introduced in 1984, it had >6,000 components IIRC lots of discrete amps (an audiophile no no). I was in NYC near Levinson's Cello showroom and stopped by, I got to listen to the complete system. The Pallette is what really surprised, I had been conditioned to think of tome controls as bad, but the Pallette transformed 'poor' recordings. There was a recording of The Goldberg Variations, bright unlistenable, using the Pallette transformed the recording to one that was great to listen to we went through a number of recordings same even 'good' recordings were made better using the Pallette. I played with 'correcting' recordings very intuitive and easy to use. Sadly I could afford the unit (still can't).
Changed my view of what was important. I wish I could have something like the Pallette. It would add value to all the recordings in my library.
D Burwin is selling something similar implemented in software. Levinson is (I think) selling it in conjunction with a D/A converter made by his new company. It is different in that it is intended to be used with loss less Digital files. He (Burwin) is also advocating adding some Hi Freq reverb to the digital recording. I would like to hear the software to see if it is usable.
I totally agree about the target!
I think we hashed this X-Curve out around page 20 of this thread(may have been another thread of similar origins). Steady state and anechoic are not equal measurements. Anyway, any meaningful standard will do--there needs to be a goal. A steady state FR above the modal region is not a meaningful standard. That's why the EBU/ITU/THX have meaningful standards spelled out--you could argue they go beyong meaningful in some ways.
Anyway, care to guess which of these meet THX standards?
audio blog: Review of Polar graphs
Notice that all but 1 are studio monitors--'flat' on axis sure looks like a goal to me. Can anyone else see that or am I alone? The off axis has some variation though, and that makes a big difference for sure.
Look at Dr. Toole's research, the room is not the same league of importance as the speaker. That's come up a few times as well.
Dan
It limits depth of convergence, but with correct design, doesn't need to screw up the algorithm. But we took two left turns at Albuquerque.
And now you can have +-0.5db adjustment capability anywhere across the entire audible spectrum along with much more advanced parametric EQs, all implemented in the digital signal chain for around $300. 😀
We've had "Flat" as a goal for quite a few years already and thus far that goal on it's own has not produced that many really great speakers at least by the standards of most readers that end up here.
It seems that both GedLee and SL have produced fantastic results by moving beyond "Flat" on axis because, by itself, it is simply not an adequate goal. In the end, both designers have included some form of further shaping both of the polar response AND by purposely implementing a design which DOES NOT provide a flat on axis response.
And thus many have simply concluded that there must be more than flat on-axis and for those that believe that there must be something better than the status quo of "Flat on axis" we are looking to improve the definition of a great sounding speaker and/or changes to the listening room.
absolutely I use these tools.
There are many options. I'm using the DEQ-2496 but that is just one option... depending on your design constraints there are many other pro-audio equipment or software on your computer that can do this for you.
If you are looking for something a little more boutique (i.e. inline with your kilo$ turntable and tube amps) you could look at http://www.holmacoustics.com/dspre1_introduction.php
Last edited:
I'm glad YOU have figured out that flat on axis isn't the only goal. It is IMO part of what works and that's backed by decades of credible research like it or not. Most are not that far along. Both SL and Geddes have very unique polar/power responses as far as the recording industry goes. I bet they'd both agree--at least I hope so. Dr. Geddes's design would probably work for THX however. I don't know that they have restriction on how narrow a monitor can be or of the axis dip would disqualify it? I have their requirements somewhere, but I'm too tired to look. Both are adjusting their listening axis's response. Maybe there's a pattern? These designs haven't been well tested compared to others in any meaningful way. A lot of sighted reviews.....😡 Making conclusions that apply to their designs and applying them to all designs would certainly be flawed reasoning. Surely you agree? I'll concede that saying flat is best for all speakers is equally flawed--I mentioned that 20 or so pages ago.
Dan
Last edited:
Can we perhaps all agree then (even just empirically) that "Roughly flat on axis, but with some design specific variation" is the correct goal then ?
In other words, straying vastly away from flat is likely to result in a worse sounding speaker, however specific relatively small deviations can result in a better sounding speaker than purely "anechoic flat on axis".
Can we also agree that the optimum magnitude and character of those small deviations is different for every different speaker design - in other words there is no "universal" correction curve away from "flat" that sounds best.
The question then becomes why is a slight correction necessary to go from a good sounding speaker to a great sounding one, and how best to determine the optimum correction for a given design - can it be done objectively (by numbers) or are we stuck with subjective ways (voicing) to do it.
As for the first part, although I don't think it's the whole story, it seems that polar response (power response, directivity etc) and the interaction with the room must be involved.
If you're well within the direct field of a speaker, (D>>R) even in a somewhat reverberant room - you hear primarily the listening axis response - usually near but not exactly the on axis response. On axis response seems to be a good correlation with perceived balance here, at least above 200Hz, and provided the response is relatively uniform over a small arc from on axis.
If you're far down the other end of a large reverberant reflective room and listening from well within the reverberant field (R>>D) then the perceived balance is modified greatly by the power response, although I'm not sure I would go so far as to say it is equal to the power response.
A typical home listening room will fall somewhere between these extremes, where listening axis response is dominant but there is some influence from the rooms power response.
Other factors that probably matter are baffle diffraction ripple signature, and crossover types. (Especially when crossovers can modify the power response)
What's needed is a carefully controlled study to try to find a correlation between speaker design topology, and perceived optimum deviation from flat on axis as found by skilled subjective "voicing" of the speakers.
Imagine a test where a variety of speakers of vastly different design and topology are used - all the way from small 2 way bookshelf's, through designs with 3 or or more drivers on large baffles, horns, panels, and so on - a good representation of both the common and more exotic designs.
Then use DSP technology to equalize all speakers flat on axis, (which can be done extremely accurately with modern FIR filters) and give them to some designers who are experienced in voicing speakers, and who are judged to have a good ear, but aren't necessarily familiar with the particular design topologies, and ask them to "voice" the speaker to sound as good as possible, as if it were their own design.
Part of that voicing would be to have more than one listener be the judge of whether changes are "good" or not, so it's not just one persons opinion on whether it sounds right, even if it is one person actually making the adjustments. Also have each speaker done independently by more than one engineer one after another, and also done with the listening tests in different room conditions.
At the end of this you should have a bunch of speakers which are slightly less flat but sound better. Then analyse the changes to try to find correlation with other characteristics of the speaker like directivity etc.
If some correlations are found, try to create a computer model that takes into account all the factors found in predicting the corrections required.
For example it might be possible to create a model where you feed in the RT30 of the desired listening environment, the rough direct to reflected ratio expected at the listening position, (eg are they close or far from the speakers) as well as the directivity/power/polar response of the speaker, and from that the computer can work out a "correction curve" that can be applied to that particular speaker which is in close agreement with that determined empirically.
If you truly could predict the change in this way it might make it possible for those who "design by numbers" to produce much better sounding speakers...and also reduce variations caused by subjective preferences of the design engineers.
Has any kind of extensive study like this been attempted with the goal of correlating subjective voicing corrections made by experienced designers with measured parameters of the speaker other than the on-axis response ?
Last edited:
My pet hate is nastiness in the upper midrange. Especially microphone resonances, vocal track levels, microphone overload, probably mixer distortion, lumpy speaker responses, speaker roll-offs, deaf or incompetent techs and misused BBC curves.
Have you noticed how such issues have changed over time? Listening to some 40's and 60's recordings last night it occured to me that these resonance issues seem to have increased in frequency by part of an octave per decade (ie: not quite doubled in frequency every ten years 😛) even up into the 90's on some recordings.
Remember many producers are reacting to what they think will sell as well as the equipment they are using in the studio. We have gone from large relatively efficient speakers in the 50's (home consoles) to small inefficient sealed boxes, am radio later fm then music videos playing through tv speakers, 8 tracks and cassettes to ipods boom boxes and ear buds. I lot of music is produced for the eq their customers are using. So a lot of pop recordings are really compressed and tinny.
So if you have a system that is accurate (good speakers in a well designed room), a lot of recordings will not sound very good. This is why a flexible easy to use tome control is needed.
Whilst reading your reply, I thought back and I seem to recall recordings sounding different when you bought them. Now I can't remember whether it was because I'd first listen to them in the car through a mono 5"x7" straight paper coned driver on AM, then when we bought the record I'd listen at home on the stereo with the 3 cubic foot 8" two-ways (how do I remember this stuff?). Or is it possible that there were two differently EQed cuts depending on the intended usage?
Radio used a lot of compression, especially AM in the states, Am was the bulk of music through the 60's with FM coming along afterwards. even then if people driving in cars were the audience there would be a lot of compression due to the road noise in a car.
Others more knowledgeable about radio could comment more intelligently. As to records the cartridges used back then would have influenced how the masters were cut and the limitations of 8 track and cassette again would have played a role.
IMO we need to recognize that the media we use and how it was created may or usually won't match the transfer function of our systems.
Others more knowledgeable about radio could comment more intelligently. As to records the cartridges used back then would have influenced how the masters were cut and the limitations of 8 track and cassette again would have played a role.
IMO we need to recognize that the media we use and how it was created may or usually won't match the transfer function of our systems.
Hmm, I recall AM as being quite a clean and natural sounding apart from the bandwidth limitation, and yes, the compression. Phono cartridges...well...
We didn't switch on FM until 1980, here. We tend to take the "if it ain't broke, don't fix it" attitude in Australia. I had an FM stereo in the seventies and it was as quiet as a church mouse.
We didn't switch on FM until 1980, here. We tend to take the "if it ain't broke, don't fix it" attitude in Australia. I had an FM stereo in the seventies and it was as quiet as a church mouse.
Have we stumbled upon a real use for a consumer grade graphic equaliser??
- Status
- Not open for further replies.