Hi Geoff,
Have you considered connecting the crossovers to compare through something like AudioDiffmaker or FSAF where you can just measure and listen to the music through the crossover with music via headphones. Skip the the drivers, the room, the air?
Have you considered connecting the crossovers to compare through something like AudioDiffmaker or FSAF where you can just measure and listen to the music through the crossover with music via headphones. Skip the the drivers, the room, the air?
Perhaps trace-back some old discussions from here:
Due to impedance differences the respective inflexion/slopes around "nominal" crossover-point are not identical between series and parallel, despite oft-heard statements they must be. From your description, component values weren't tweaked to compensate, but FR remained within 0.5dB, vector-mathematically there must have been additional phase response offset/misalignment between drivers, in one XO or the other. This would likely show up in impulse response fidelity and (more controversial) imaging/soundstage.
I disagree, I created the parallel crossover 1st, then used the high and low pass responses ( with covers) as target frequency responses in LEAP 4, because I was able to use real life target values, LEAP was able to perfectly curve fit the frequency response of the series filters, the optimizer had little work do do. Not sure if that all made sense but I was able to make almost identical slope rates for both parallel and series and is was probably closure to .25dB error, LEAP is magical!. The finished slope rates and summation of the drivers will determine the the the phase response of the system , both series and parallel topologies should be the same. And no there was no difference in offset or alignment , I used the same drivers, just swapped crossovers. Before this test, I had long known about the Sonic differences between series and parallel .
But ....... There is a difference in impedance between series and parallel even when they are matched perfectly, I'm not 100% sure if this will cause any system phase differences, and if they did, I highly doubt that a few mS of delay is audible at high frequencies. I just got a LEAP 5 running, working on LEAP 4 both on a win 11 Machine.
My guess for the Sonic differences is that the impedance response is different and the amplifier sees a different reactive load with a series network. What is also interesting is that series Xos sound best with higher powered transistor amps where parallel sound best with tubes... Generally speaking. This may be due to a tube amp being more of a current source and a transistor amp being more of a voltage source, I really don't know , are there any EE's in the house?
Chat GPT says:
This is a great question that delves into the nuances of loudspeaker crossover design.
In theory, for ideal components and a resistive load (which a loudspeaker is not, due to its varying impedance with frequency), if series and parallel passive filter networks of the same order are designed to achieve the identical frequency response, then their impulse responses will also be identical.
Here's why:
* Relationship between Frequency Response and Impulse Response: The impulse response and frequency response of a linear time-invariant (LTI) system are a Fourier Transform pair. This means that if you know one, you can mathematically derive the other. Therefore, if two filter networks have the same frequency response (magnitude and phase), they must also have the same impulse response.
* Identical Transfer Functions (under ideal conditions): When designed correctly for the same target response (e.g., Butterworth, Linkwitz-Riley), both series and parallel crossovers of a given order can theoretically produce the same transfer function. A transfer function describes how a system modifies an input signal, encompassing both amplitude and phase characteristics.
However, the "ideal conditions" are key, and real-world loudspeakers introduce complexities:
* Loudspeaker Impedance Variation: This is the biggest factor. Loudspeakers do not present a constant resistive load across all frequencies. Their impedance varies significantly with frequency due to factors like voice coil inductance, driver resonance, and enclosure effects.
* Parallel Crossovers: In a parallel crossover, each filter section "sees" the impedance of its respective driver. Changes in driver impedance can affect the filter's actual cutoff frequency and Q (damping).
* Series Crossovers: In a series crossover, the filters are in series with the drivers and with each other. This means the impedance of one driver can influence the filtering applied to the other drivers. Some argue that this interaction can be beneficial in certain scenarios, as changes in one driver's impedance might partially compensate for changes in another, potentially leading to a flatter overall response under varying load conditions. However, it also makes the design more complex and sensitive to driver parameters.
* Back-EMF: When a speaker cone moves, it generates a "back-electromotive force" (back-EMF) that can interact with the crossover network. The way parallel and series networks handle this back-EMF can differ, potentially affecting damping and transient response.
* Component Sensitivities: While theoretically identical, practical component tolerances and parasitic elements can have slightly different impacts on series versus parallel designs, potentially leading to subtle differences in real-world performance, including impulse response.
In summary:
* Theoretically (ideal world): Yes, series and parallel crossovers of the same order, designed for the same frequency response, would have the same impulse response.
* Practically (real world): Due to the complex, non-resistive, and frequency-dependent impedance of loudspeakers, and how that impedance interacts with the different topologies, there can be differences in the actual frequency and impulse responses between series and parallel crossovers, even if they are nominally of the same order. These differences might manifest as variations in system damping, transient behavior, and overall "sound" (e.g., stereo imaging, detail).
Many loudspeaker designers have strong preferences for one topology over the other, often citing subjective sonic qualities that arise from these real-world interactions. Ultimately, the best crossover design (series or parallel) depends heavily on the specific drivers used and the desired acoustic outcome.
My guess for the Sonic differences is that the impedance response is different and the amplifier sees a different reactive load with a series network. What is also interesting is that series Xos sound best with higher powered transistor amps where parallel sound best with tubes... Generally speaking. This may be due to a tube amp being more of a current source and a transistor amp being more of a voltage source, I really don't know , are there any EE's in the house?
I just asked chat GPT about traniant response and with all things equal, will a series Crossover pass a square wave the same as a parallel crossover .
Chat GPT says:
This is a great question that delves into the nuances of loudspeaker crossover design.
In theory, for ideal components and a resistive load (which a loudspeaker is not, due to its varying impedance with frequency), if series and parallel passive filter networks of the same order are designed to achieve the identical frequency response, then their impulse responses will also be identical.
Here's why:
* Relationship between Frequency Response and Impulse Response: The impulse response and frequency response of a linear time-invariant (LTI) system are a Fourier Transform pair. This means that if you know one, you can mathematically derive the other. Therefore, if two filter networks have the same frequency response (magnitude and phase), they must also have the same impulse response.
* Identical Transfer Functions (under ideal conditions): When designed correctly for the same target response (e.g., Butterworth, Linkwitz-Riley), both series and parallel crossovers of a given order can theoretically produce the same transfer function. A transfer function describes how a system modifies an input signal, encompassing both amplitude and phase characteristics.
However, the "ideal conditions" are key, and real-world loudspeakers introduce complexities:
* Loudspeaker Impedance Variation: This is the biggest factor. Loudspeakers do not present a constant resistive load across all frequencies. Their impedance varies significantly with frequency due to factors like voice coil inductance, driver resonance, and enclosure effects.
* Parallel Crossovers: In a parallel crossover, each filter section "sees" the impedance of its respective driver. Changes in driver impedance can affect the filter's actual cutoff frequency and Q (damping).
* Series Crossovers: In a series crossover, the filters are in series with the drivers and with each other. This means the impedance of one driver can influence the filtering applied to the other drivers. Some argue that this interaction can be beneficial in certain scenarios, as changes in one driver's impedance might partially compensate for changes in another, potentially leading to a flatter overall response under varying load conditions. However, it also makes the design more complex and sensitive to driver parameters.
* Back-EMF: When a speaker cone moves, it generates a "back-electromotive force" (back-EMF) that can interact with the crossover network. The way parallel and series networks handle this back-EMF can differ, potentially affecting damping and transient response.
* Component Sensitivities: While theoretically identical, practical component tolerances and parasitic elements can have slightly different impacts on series versus parallel designs, potentially leading to subtle differences in real-world performance, including impulse response.
In summary:
* Theoretically (ideal world): Yes, series and parallel crossovers of the same order, designed for the same frequency response, would have the same impulse response.
* Practically (real world): Due to the complex, non-resistive, and frequency-dependent impedance of loudspeakers, and how that impedance interacts with the different topologies, there can be differences in the actual frequency and impulse responses between series and parallel crossovers, even if they are nominally of the same order. These differences might manifest as variations in system damping, transient behavior, and overall "sound" (e.g., stereo imaging, detail).
Many loudspeaker designers have strong preferences for one topology over the other, often citing subjective sonic qualities that arise from these real-world interactions. Ultimately, the best crossover design (series or parallel) depends heavily on the specific drivers used and the desired acoustic outcome.
Those are really cool programs , I have not heard of either will have to study them really interesting.
I'm not sure if it would be possible to mimic the series Crossover and if it was possible I don't think that headphones would reveal any differences, I believe that the amplifier is seeing a different reactive load and back EMF from the series network and this is why is sounds different, so it's really not about slope rates, delay or phase, it's about an electrical interaction with the amp and will therefore not show up in a headphone test.
Please note, I'm guessing here, I'm just going from my experiences and logical thinking ,your milage may vari
These programs allow one to measure the differences in audio interfaces, microphones, speakers, crossovers, even the room... literally anything or everything in your signal chain, and allow you to play back the "the Difference" (Audio Diffmaker) or "Residual" (REW's Fast Sub-band Adaptive Filtering)
If it's audible, then it might be worth caring about... so I suggested listening to the the Difference or Residual via headphones because it may make it easier to ascertain any differences
a) headphones allow you to block out background noise (particularly over-the-ear, or in-ear types)
b) $ for $, generally exhibit distortions less than speakers (sit close to the ear, and are not required to play very loudly)
VS the same speaker system with the series crossover?
If they are different then I agree with this belief.
If not, then someone with a deeper level of math understanding will have to explain why a series crossover might sound different to a parallel crossover.
***
We don't need to make a judgment on how it sounds different, only that we can hear the differenceIf it's audible, then it might be worth caring about... so I suggested listening to the the Difference or Residual via headphones because it may make it easier to ascertain any differences
a) headphones allow you to block out background noise (particularly over-the-ear, or in-ear types)
b) $ for $, generally exhibit distortions less than speakers (sit close to the ear, and are not required to play very loudly)
***
Do you have an impedance trace of the speaker system with the parallel crossover,VS the same speaker system with the series crossover?
If they are different then I agree with this belief.
If not, then someone with a deeper level of math understanding will have to explain why a series crossover might sound different to a parallel crossover.
Unfortunately those files are long gone but as mentioned, I have gotten LEAP 5 running and will try to run some designs of series Vrs parallel. I think I can run a reverse impulse response derived from syster FR, that would be interesting to compare .
I personally need no convincing that they can sound different but it bugs me that I done really know why.
My last amp was a 2A3 SET , I'm now going to a class D which will most likely sound good when mated to series Xos, we will see.
I may start another thread regarding cone material and why measurements don't really tell us how they sound .
I personally need no convincing that they can sound different but it bugs me that I done really know why.
My last amp was a 2A3 SET , I'm now going to a class D which will most likely sound good when mated to series Xos, we will see.
I may start another thread regarding cone material and why measurements don't really tell us how they sound .
This kind of tests are only valuable if you also publish the objective (measurment) data, otherwise they are an opinion (which is valid for you, but not for others). Then we can also easier see why it may or may not be better. Crossovers are always more precise if you work with measurements as base, as you know what you're doing then, otherwise it's guessing.
My own subjective experience with both types is that series is very good for low order (mainly 1st order) filters as it's easier to align phase, but from 2nd order the parallel wins, and most speakers need a higher order crossover than 1st order. That's what i also heared from more experienced designers.
My own subjective experience with both types is that series is very good for low order (mainly 1st order) filters as it's easier to align phase, but from 2nd order the parallel wins, and most speakers need a higher order crossover than 1st order. That's what i also heared from more experienced designers.