Hello everyone,
we have run some preliminary hearing tests with "informed" college students on whether different crossover slopes can be heard out. We used a Behringer DCX 2496 to "model" different slopes for a "typical" two-way (2k XO) and a three-way (300Hz & 3KHz XO) system. We tested first order Butterworth, second order Bessel (tweeter phase reversed), third order BW, and fourth order L-R filters. Also, we modeled some asymmetrical designs, such as a first order BW, second order Bessel combination (with tweeter 180 phase reversal). The outputs of the six channels of DCX were summed through a high-quality analog mixer and re-recorded in high-resolution (96kHz/24bit). The result, after normalization (careful matching of levels), was compared to the re-recorded direct sound through the analog mixer of two outputs of the DCX with NO filters active. The results can a bit as a surprise to me: there is VERY little difference in the sonic signatures in the actual crossover phase changes. Most of the students had trouble distinguishing ANY change. I am tempted to deduce from this that the sonic signature that people attribute to various crossover slopes MUST come from the individual driver's frequency and time response. You might think that this awareness is rather obvious, but I thought I would bring it up for discussion since people seem to debate so hotly the sound quality that a change of filter slopes may introduce. It is also possible that Behringer's digital model of the analog filters do not accurately model the actual phase changes that occur with the analog counterpart - I admit that should be next on my list of inquiry, but I thought I would open up the discussion here. I would be very curious about what people have to say on this topic. Forgive for not (yet) providing you with a rigorous scientific experiment. I would be happy to make the sound files available, however.
Thank you for your input,
Cheers
RB
we have run some preliminary hearing tests with "informed" college students on whether different crossover slopes can be heard out. We used a Behringer DCX 2496 to "model" different slopes for a "typical" two-way (2k XO) and a three-way (300Hz & 3KHz XO) system. We tested first order Butterworth, second order Bessel (tweeter phase reversed), third order BW, and fourth order L-R filters. Also, we modeled some asymmetrical designs, such as a first order BW, second order Bessel combination (with tweeter 180 phase reversal). The outputs of the six channels of DCX were summed through a high-quality analog mixer and re-recorded in high-resolution (96kHz/24bit). The result, after normalization (careful matching of levels), was compared to the re-recorded direct sound through the analog mixer of two outputs of the DCX with NO filters active. The results can a bit as a surprise to me: there is VERY little difference in the sonic signatures in the actual crossover phase changes. Most of the students had trouble distinguishing ANY change. I am tempted to deduce from this that the sonic signature that people attribute to various crossover slopes MUST come from the individual driver's frequency and time response. You might think that this awareness is rather obvious, but I thought I would bring it up for discussion since people seem to debate so hotly the sound quality that a change of filter slopes may introduce. It is also possible that Behringer's digital model of the analog filters do not accurately model the actual phase changes that occur with the analog counterpart - I admit that should be next on my list of inquiry, but I thought I would open up the discussion here. I would be very curious about what people have to say on this topic. Forgive for not (yet) providing you with a rigorous scientific experiment. I would be happy to make the sound files available, however.
Thank you for your input,
Cheers
RB
Interesting, did you make sure that the acoustic phase was sorted out??
For a given set of drivers in the same cabinet you cannot swap and change (even if the resultant individual acoustic slopes are correct) filters and assume each is going to sum correctly.
Did all the setups bar the 1st order provide a reverse null when the polarity of one driver was reversed? and if so were they all good notches of the expected shape?
If I apply a 4th order LWR acoustic slope to a woofer and tweeter they might sum reasonably well in phase, but when reversed not a good notch, time delay on the tweeter needs to be added for the phase to work well. Similarly a 2nd order LWR acoustic could sum perfectly and provide a good notch without any time delay.
For a given set of drivers in the same cabinet you cannot swap and change (even if the resultant individual acoustic slopes are correct) filters and assume each is going to sum correctly.
Did all the setups bar the 1st order provide a reverse null when the polarity of one driver was reversed? and if so were they all good notches of the expected shape?
If I apply a 4th order LWR acoustic slope to a woofer and tweeter they might sum reasonably well in phase, but when reversed not a good notch, time delay on the tweeter needs to be added for the phase to work well. Similarly a 2nd order LWR acoustic could sum perfectly and provide a good notch without any time delay.
I should emphasis that this hearing "experiment" was done WITHOUT any drivers. These are results only from the summing of the outputs of the actual CROSSOVER NETWORK. We wanted to see how much the actual crossover contributes to the sonic signature of the overall loudspeaker system.
5th Element: yes, we did reverse the polarity for an imaginary tweeter involving second order Bessels. EVERYONE indeed could hear when the polarity was NOT reversed in this setup.
Cheers
5th Element: yes, we did reverse the polarity for an imaginary tweeter involving second order Bessels. EVERYONE indeed could hear when the polarity was NOT reversed in this setup.
Cheers
I'm surprised you could hear any effect at all! (except for cases where there is a frequency response notch)
Your surmise about why effects can be heard with real drivers is right on the mark. Another aspect of this is lobing and polar pattern, which is different with different crossovers.
Also, distortion from running drivers too close to their LF limits is a very characteristic "sound" of first-order crossovers.
Your surmise about why effects can be heard with real drivers is right on the mark. Another aspect of this is lobing and polar pattern, which is different with different crossovers.
Also, distortion from running drivers too close to their LF limits is a very characteristic "sound" of first-order crossovers.
For me, when I state that I can very much hear the overall 'slope' design of a given loudspeaker system, when out of the room, well..that trick is not all that difficult, if you put in the time to learn to recognise the effect, of which I am about to speak of.
It is the direct sound (sitting in front of) a given speaker, vs the total acoustical power of a loudspeaker. With the first order slope and the 24db crossovers, they are different enough, that a reliable difference can be heard, when listening to the different slope types, when the observer is OUT of the room the speakers are located in. Each has characteristic 'sound' of direct vs. total power.
You'd have to try it with a group of speakers, and after a while you'd learn to recognize the sound of each. This is, of course, when the crossovers I speak of, are created via LCR passive design work.
We all hear differently,and we all have different levels of personal skill involving the art of observation, as tied to any particular method we may each have of gathering data. In my opinion, it is not the ear alone, in terms of given sentitivites to sound that is the arbiter of that kind of skill, but the obsevationalist tied to that given ear. Ie, how 'aware' are you?
(the sneaky bits are recognizing, with the ear...in that whole acoustical mess of sound...the phase noise of the crossover types, that's the -real- trick here. Most folks would say I'm full of **** with such a statement. Spend enough time with speakers and you won't feel I'm full of it any more. So..no..I'm not full of ****. I've just put in the time required to reach that point of aural understanding.)
It is the direct sound (sitting in front of) a given speaker, vs the total acoustical power of a loudspeaker. With the first order slope and the 24db crossovers, they are different enough, that a reliable difference can be heard, when listening to the different slope types, when the observer is OUT of the room the speakers are located in. Each has characteristic 'sound' of direct vs. total power.
You'd have to try it with a group of speakers, and after a while you'd learn to recognize the sound of each. This is, of course, when the crossovers I speak of, are created via LCR passive design work.
We all hear differently,and we all have different levels of personal skill involving the art of observation, as tied to any particular method we may each have of gathering data. In my opinion, it is not the ear alone, in terms of given sentitivites to sound that is the arbiter of that kind of skill, but the obsevationalist tied to that given ear. Ie, how 'aware' are you?
(the sneaky bits are recognizing, with the ear...in that whole acoustical mess of sound...the phase noise of the crossover types, that's the -real- trick here. Most folks would say I'm full of **** with such a statement. Spend enough time with speakers and you won't feel I'm full of it any more. So..no..I'm not full of ****. I've just put in the time required to reach that point of aural understanding.)
If you are running a summed signal I doubt anyone would tell a difference. I can clearly hear a difference when using different actual drivers with this setup, and my ears are nothing spectacular. I would check that the summed signal is any different than the input signal by digital comparison, but I believe you will just be doing a decomposition and summation, with a net change of zero (or very slight amplitude differences near crossover frequencies). The digital filters are likely not phase-accurate to an analog filter, so an analog crossover could sound different (plus it is nearly impossible with off-the-shelf components and tolerances to match a digital filter's performance for the low-pass and high pass regions, especially with respect to frequency accuracy).
The differences I hear are mostly due to the directionality of the different sized drivers (lobing, etc), as well as the blending across drivers (combing, etc) or brick wall transition across different drivers.
I would suggest Audio DiffMaker (Liberty Instruments) to see if things are any different, other than a slight change in response at the crossover frequency. To be phase accurate, it would need the RLC equivalent model of each driver as well as the passive components, or a model phase transfer function, and that's why I don't think Behringer has programmed the imaginary portion of the FFT apodization function for each of these crossover models.
The differences I hear are mostly due to the directionality of the different sized drivers (lobing, etc), as well as the blending across drivers (combing, etc) or brick wall transition across different drivers.
I would suggest Audio DiffMaker (Liberty Instruments) to see if things are any different, other than a slight change in response at the crossover frequency. To be phase accurate, it would need the RLC equivalent model of each driver as well as the passive components, or a model phase transfer function, and that's why I don't think Behringer has programmed the imaginary portion of the FFT apodization function for each of these crossover models.
Interesting, and good to know at least that the DSP doesn't have much coloration coming from the filter processing. If there were distortions from the processing this test should tend to reveal that, because you must have used more processing power for the steeper slopes. I would expect that 48 dB slopes might prove less than transparent with this type of test.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.