Midrange has lower spl than woofer and tweeter, because of wide overlap in-phase. It looks less steep than it is. Sometimes mid is called a filler.
https://musicanddesign.speakerdesign.net/Duelund_and_Beyond.html
Measured step response, please!
https://musicanddesign.speakerdesign.net/Duelund_and_Beyond.html
Measured step response, please!
Last edited:
The low-pass filter on the midrange rolls off at about 8dB/octave. The high-pass filter on the tweeter rolls off at about 5dB/octave. That's taken from the measurements. I think that slopes of 8dB/octave and 5dB/octave aren't very steep at all.
I'm also unsure of the sound quality benefits, if any, that are to be afforded by such a wide overlap between the outputs of the drivers. Was the intent to try and get a version of a 1st-order filter design with its notional ability to be transient perfect?
I'm also unsure of the sound quality benefits, if any, that are to be afforded by such a wide overlap between the outputs of the drivers. Was the intent to try and get a version of a 1st-order filter design with its notional ability to be transient perfect?
8dB/octave is a bit unusual since we typically work in multiples of 6dB to match 1st, 2nd, 3rd, 4th order filters—so 6, 12, 18, 24 dB/oct, and so on.
If you see something different, you'll need to extend the response.
It just means the frequency response hasn't fully settled yet.
If you see something different, you'll need to extend the response.
It just means the frequency response hasn't fully settled yet.
I don't agree with your assessment, although I can see that it would be challenging to assess the slope based on visually looking at my plots. My goal was 2nd order LR crossovers, and I think I got reasonably close to 12 dB/octave slopes.
Here is the responses of the three drivers, overlayed with an LR2 high pass filter at 2.3k . I think the tweeter response is a pretty nice approximation of 2nd order LR... yes?
Here is the overlay of the mid driver ideal low pass at 400 Hz + high pass at 2.6k. Here there is some deviation from the ideal to account for a smooth transition to the woofer through the BSC region, and to manage the power response and DI, and also to manage the 2k peak. Even with these compromises and the changes made for subjective voicing, the overall low pass and high pass slopes seem pretty close to 12 dB/octave... yes?
Perhaps I am not seeing something which you noticed. ?
j.
Slopes are not the same everywhere, I guess this is confusing witwald....symmetry and phase tracking from both sides of crossover point are much more important and nobody is trying to accomplish ideal 6, 12 or 18 db/oct slopes just for the sake of theory....in 3-way speaker, where tweeter is cutted higher lower order slopes are also OK and can protect tweeter from lower frequencies...
It seems that hifijim did good job with this one...
It seems that hifijim did good job with this one...
Ok, sure... I have never been clear on what I should learn from a step response
First is Impulse response
Now the Step response of 5.5 ms window
You have definitely achieved an excellent match with the target 2nd-order LR acoustic response functions that you aimed for. I recognise that the LR topology produces a much more rounded response shape than does a Butterworth topology. I think I was just expecting to see a response shape closer to a Butterworth than an LR from the point of view of.
Step response shows individual "ways" arrival time and polarity. From your plot we can see that the mid is reversed polarity and peak comes a tiny bit too late (0,1ms) to perfectly integrate to tweeter's peak's return cycle.
Regarding 42-60Hz fundamental and 1/8 note being picked, yeas we can't hear it coming late for the rhythm from the reflex, but as smearing of the transient attack, being muddy or rounded.
Regarding 42-60Hz fundamental and 1/8 note being picked, yeas we can't hear it coming late for the rhythm from the reflex, but as smearing of the transient attack, being muddy or rounded.
ps. Jim et al. - spl and step response with mid in right vs. wrong polarity will tell much about time integration!
Can you elaborate a little on what would be needed for the midrange unit to "perfectly integrate to the tweeter's return cycle"? Is that something that could even be attained without the use of first-order networks? How does the step response show us anything beyond what the overall frequency response curve (magnitude and phase) inherently portrays?
Timing mismatch comes from physical distance and/or phase mismatch. Typical even order symmetrical MT xo is critical because of short wavelength. This also causes vertical off-axis response to have a dip at MT xo.
Troels Gravesen tells more http://www.troelsgravesen.dk/Time-Alignment.htm
Passive speakers can use stepped or tilted baffle to better align mid and tweter (on design-axis)
Active/dsp speakers can use delay for the tweeter.
Troels Gravesen tells more http://www.troelsgravesen.dk/Time-Alignment.htm
Passive speakers can use stepped or tilted baffle to better align mid and tweter (on design-axis)
Active/dsp speakers can use delay for the tweeter.
I think there might be a bit of misunderstanding about the step response.
A step response is actually defined differently and typically involves working with a DC signal.
(Hence the name step)
ARTA displays only impuls responses
(Also see manual)
There is an option that can translate the data into a step response if needed.
For more detailed explanations, I’d recommend checking resources like Google, Wikipedia, or some standard literature on the topic.
Unfortunately, I can’t go into the very basics here, but those sources should help clarify things.
A step response is actually defined differently and typically involves working with a DC signal.
(Hence the name step)
ARTA displays only impuls responses
(Also see manual)
There is an option that can translate the data into a step response if needed.
For more detailed explanations, I’d recommend checking resources like Google, Wikipedia, or some standard literature on the topic.
Unfortunately, I can’t go into the very basics here, but those sources should help clarify things.
Out of curiosity, I thought I'd use XSim4 to simulate the 3-way system that was developed by @hifijim. I used pure LR2 highpass and lowpass filter response functions as appropriate. The midrange was connected with inverted polarity. The woofer low-frequency response was modelled as a 4th-order Butterworth high-pass filter with a −3dB point of 35Hz. The tweeter's high-frequency roll-off was modelled using a 2nd-order Butterworth lowpass filter with a −3dB point of 25kHz. The various individual and the summed frequency response functions are shown below. These results were obtained after a little bit of tweaking of the midrange sensitivity. Also note that the acoustic centres of the drivers are all coincident, so there are flight time differences to the listening point.
Unsurprisingly, this looks very much like the set of responses obtained in the real loudspeaker system.
And, using XSim4's step response modelling capability, we obtain the following step response function. This is very similar to the step response that was shown earlier, but which included the effects of the various peaks and dips in the frequency response function.
If we add 0.5inch of rearward offset to the midrange unit, we get the following step response. Here we can see that the midrange's output does arrive a little bit later.
The associate frequncy response function is shown below. The change in the frequency response is much more clearly obvious, and amounts to some phase-related sound cancellation at high frequencies, leading to a droop in the high-frequency response above about 4kHz.
Unsurprisingly, this looks very much like the set of responses obtained in the real loudspeaker system.
And, using XSim4's step response modelling capability, we obtain the following step response function. This is very similar to the step response that was shown earlier, but which included the effects of the various peaks and dips in the frequency response function.
If we add 0.5inch of rearward offset to the midrange unit, we get the following step response. Here we can see that the midrange's output does arrive a little bit later.
The associate frequncy response function is shown below. The change in the frequency response is much more clearly obvious, and amounts to some phase-related sound cancellation at high frequencies, leading to a droop in the high-frequency response above about 4kHz.
The input step response is a constant DC signal, that much is quite clear. However, when a step input is applied to a bandpass system (e.g., a loudspeaker), there can be no acoustic response at DC. The woofer's voice coil will of course reach a constant displacement due to the DC part of the step's input signal, assuming that the driving amplifier is DC coupled, but this of course happens after the cone has finished responding to the initial part of the step.
So, after translation using the appropriate mathematical transformation, is not the step response shown? Those responses do look very much like the step response one would expect from a dynamic system with a high bandwidth, which is what has been built here.
No it's not and it's not how a step and/or impulse responses are defined.
Again, check a decent resource to find out the difference between a step response and impulse response, since you don't seem to be very familiair with the subject.
Just out of curiosity, I've used XSim4 to create a model of your loudspeaker system using BW3 (acoustic 3rd-order Butterworth filter response functions). With a little bit of tweaking of the midrange sensitivity, the following set of frequency response functions was obtained. As you mentioned, the knees are all a lot sharper. In this instance, coincident drivers were assumed to simplify the model.
The step response of the idealized loudspeaker based on the BW3 filters is shown below. This step response is very similar to the one that was obtained earlier using the LR2 filter functions.