Hello, I posted this topic as part of an active thread but I feel that it deserves its own.
I built crossovers for a system using both series and parallel topology. These tests were done in a semi- anachoic room so off axis / power response was not a thing, overall frequency response between crossovers was .5dB or less. There was virtually no difference in the frequency response but the stage presentation of the two was very different, the changes were enough where untrained listeners could easily hear differences.
Parallel topology:
This gave a very strong center lock on vocals, very topical of what you hear in a good system
Series topology:
The center lock was much less noticeable but the depth and width of the stage were much more defined. Low level detail was also more pronounced. There was more space between instruments.
Both topologies sounded good and, I can't say that one was better in absolute terms and it comes down to system synergy and personal preference. Some people love the strong center lock , some love the more 3D ambient sound , speakers and amplifier choice play a large role in what subjectively sounds better .
Why do they sound different.... I don't know but we suspect that if may have do do with back EFM that the amp sees and reactive loading on the amplifier is different between topologies. I Would be interested to hear people's experience with series Xos and a possible explanation for the Sonic differences.
G.
I built crossovers for a system using both series and parallel topology. These tests were done in a semi- anachoic room so off axis / power response was not a thing, overall frequency response between crossovers was .5dB or less. There was virtually no difference in the frequency response but the stage presentation of the two was very different, the changes were enough where untrained listeners could easily hear differences.
Parallel topology:
This gave a very strong center lock on vocals, very topical of what you hear in a good system
Series topology:
The center lock was much less noticeable but the depth and width of the stage were much more defined. Low level detail was also more pronounced. There was more space between instruments.
Both topologies sounded good and, I can't say that one was better in absolute terms and it comes down to system synergy and personal preference. Some people love the strong center lock , some love the more 3D ambient sound , speakers and amplifier choice play a large role in what subjectively sounds better .
Why do they sound different.... I don't know but we suspect that if may have do do with back EFM that the amp sees and reactive loading on the amplifier is different between topologies. I Would be interested to hear people's experience with series Xos and a possible explanation for the Sonic differences.
G.
Last edited:
Very interesting, thanks for sharing .
I have been interested in trying a series xo for my two way system.
I have been interested in trying a series xo for my two way system.
I was able to model series Xos in LEAP 4 but was not able to optimize them, the program was not designed to do that . I'm not sure if you can in LEAP 5 or not . I was however able to see the final frequency response and with a lot of trial and error come up with an XO that summed flat at my desired crossover point. This system was a 2 way and what is also interested is that the values became almost symmetrical, for both the woofers and tweeter, I did a 2nd order LP and HP and got the same values for the caps and inductors, example, both the tweeter and woofer used an 8uf cap and 1mH inductors. I'm not sure why it worked out this way but found in interesting, will post some very old schematics.
You say the frequency response was the same (within 0.5dB).
I assume that was the magnitude response.
Do you know how the two phase responses looked? That could be a major factor in the presentation.
Jan
I assume that was the magnitude response.
Do you know how the two phase responses looked? That could be a major factor in the presentation.
Jan
I admire your patience
High order filters series seems over whelming to me.
Often high order is essential for accurate response.
So would go straight to parallel
Having done very basic first order series filters.
It only worked because it was 2 way.
Wideband with a Woofer.
Low crossover point made it work.
The bass driver was side mounted or 90 degrees.
Additional accuracy that 1st order would not achieve.
Made possible since the woofer is off axis, provided a additional filter.
Stereo image is a stereo image.
To many variables to state series produce this " image" and parallel that " image"
Since most of that depends on the recording.
We get this" image" thing with amp descriptions too.
Amps are pretty neutral regardless of the magic talk.
The distortion is either high or low, and they either drive difficult loads well or dont.
So the "image" depends more on the recording, any extra detail that contributes to a "image"
Is usually just basic driver distortion or improved bandwidth by not forcing drivers to cover wide bandwidth.
Any instrument or vocal set wide in a mix, will change this so called holographic blah blah blah image thing.
Typically sounds like non engineer types, parroting typical audio reviews with romantic blah blah blah.
Since mids or high presence will vary widely with recording. And any phase issue cancelation/peaking
will vary from one speaker design to another. Series or parallel crossover would not define this
High order filters series seems over whelming to me.
Often high order is essential for accurate response.
So would go straight to parallel
Having done very basic first order series filters.
It only worked because it was 2 way.
Wideband with a Woofer.
Low crossover point made it work.
The bass driver was side mounted or 90 degrees.
Additional accuracy that 1st order would not achieve.
Made possible since the woofer is off axis, provided a additional filter.
Stereo image is a stereo image.
To many variables to state series produce this " image" and parallel that " image"
Since most of that depends on the recording.
We get this" image" thing with amp descriptions too.
Amps are pretty neutral regardless of the magic talk.
The distortion is either high or low, and they either drive difficult loads well or dont.
So the "image" depends more on the recording, any extra detail that contributes to a "image"
Is usually just basic driver distortion or improved bandwidth by not forcing drivers to cover wide bandwidth.
Any instrument or vocal set wide in a mix, will change this so called holographic blah blah blah image thing.
Typically sounds like non engineer types, parroting typical audio reviews with romantic blah blah blah.
Since mids or high presence will vary widely with recording. And any phase issue cancelation/peaking
will vary from one speaker design to another. Series or parallel crossover would not define this
Last edited:
This may be of interest. Note the tone burst responses.
https://www.tubecad.com/2025/06/blog0622.htm
https://www.tubecad.com/2025/06/blog0622.htm
You said semi-anechoic, but you probably need to provide more details about the room, since soundstage depth/width are often affected by off-axis response, how the speaker is positioned, etc.
Also, you said the frequency response was the same, but were the individual driver responses? Which feeds back into the point above - was off-axis behavior the same? I know you said it doesn't matter, but so far it doesn't feel like you've given enough information to prove that's the case.
Also, you said the frequency response was the same, but were the individual driver responses? Which feeds back into the point above - was off-axis behavior the same? I know you said it doesn't matter, but so far it doesn't feel like you've given enough information to prove that's the case.
Last edited:
I'm sure I wouldn't notice it, but 0.5 dB is more than enough to get audible differences in an AB test with someone with good hearing. In double-blind tests meant to test something else than frequency response differences, the frequency responses and gains are usually equalized and matched within 0.1 dB.
Sure it's possible to hear small differences but as mentioned, we were not listening in the frequency domain. We were specifically listening to the sound stage presentation which is not really determined by frequency response changes especially small ones.
This is not a one off event, I have designed many series Xos with the same effect, this was the the only time I " cloned" a parallel system and brought in outside listeners.
ABX testing is great for picking up differences in the frequency domain and that is about it, I find it useless in evaluating high end systems.
Here is an analog to ABX testing:
Let's say that I have been drinking purified water for a month and I have grown accustomed to it. I now taste tap water ( I live in Los Angeles the tap water is not so good) I will easily be able to discern that the tap water tastes bad and even articulate the subjective reasons of why . Now let's reverse the situation and I have been drinking tap water for a month and taste the purified water . I will not know that this water tastes different or better, I will say it tastes like water . It's easy to taste " poluttents, it's not easy to taste the lack of them
Let's now repeatedly go back back and forth from the pure water to the tap water , you will not be able to tell the difference unless you clear your pallette. This is the same as ABX testing.
The exact reasons for taste are not the same mechanisms as out auditory system but I feel like this is a good analogy.
This is why people who do subjective testing, live with different components in their system for a few days, Weeks or months, then switch back to an evaluation.
Also, these differences in Xos are only noticeable if you have a good sound system. A great way to tell if your system is " coherent" is to play a well recorded piece with stage ambience . When you close your eyes, your speakers should disappear , you should not be able to localize them, if you can hear that the music is coming from drivers, something is not right. A good system will immerse you in the stage and you will have zero localization of your speakers / drivers
This is not a one off event, I have designed many series Xos with the same effect, this was the the only time I " cloned" a parallel system and brought in outside listeners.
ABX testing is great for picking up differences in the frequency domain and that is about it, I find it useless in evaluating high end systems.
Here is an analog to ABX testing:
Let's say that I have been drinking purified water for a month and I have grown accustomed to it. I now taste tap water ( I live in Los Angeles the tap water is not so good) I will easily be able to discern that the tap water tastes bad and even articulate the subjective reasons of why . Now let's reverse the situation and I have been drinking tap water for a month and taste the purified water . I will not know that this water tastes different or better, I will say it tastes like water . It's easy to taste " poluttents, it's not easy to taste the lack of them
Let's now repeatedly go back back and forth from the pure water to the tap water , you will not be able to tell the difference unless you clear your pallette. This is the same as ABX testing.
The exact reasons for taste are not the same mechanisms as out auditory system but I feel like this is a good analogy.
This is why people who do subjective testing, live with different components in their system for a few days, Weeks or months, then switch back to an evaluation.
Also, these differences in Xos are only noticeable if you have a good sound system. A great way to tell if your system is " coherent" is to play a well recorded piece with stage ambience . When you close your eyes, your speakers should disappear , you should not be able to localize them, if you can hear that the music is coming from drivers, something is not right. A good system will immerse you in the stage and you will have zero localization of your speakers / drivers
I was the senior loudspeaker system designer engineer at Aurasound in the late 90s to early 2000s. We had an acoustic chamber around 40 x 40 feet with a 15 foot ceiling, About 10" of quality damping material on the walls, not a true anechoic chamber but good down to around 100Hz. Please note that im mentioning this event because if was the only actual test I performed but I have built dozens for series systems with similar results. The speakers were about 12 feet from the wall so there really wasn't any room interaction and power response was not a real consideration, they were 6.5" 2 ways crossed over around 2k. They were nice speakers that could create a beautiful sound stage. I actual brought my boss in who was the VP, a well known guy in the industry with a background in consumer electronics and absolutely not high end audio , he would give me the stink eye if I brought up something that was beyond the realm of measurements. I believe that he said something like " the series system sounded more "ethereal" and 3D and less focused.
I was also the designer for a small high-end company in the early 90s called speaker Art, I learned about series xos. from the owner who had zero measurement equipment but designed really nice sounding systems by ear, he turned me on to them as well as AMT drivers and Gorlich woofers used in ensemble speakers, (another story)
Very interesting article but in my opinion, I'm going to disagree that the phase difference / group delay is audible at these frequencies. I would be really interesting if someone did an active filter that could be switched from a standard filters to FIR filters that have zero delay
If an amplifier measured like most speakers do, it would be trashed as completely defective.
And rightly so.
And rightly so.
I think it stems from the process of both.
Parallel networks treat the drivers separately to bring them together.
Series networks start with drivers together and proceed to carefully separate them.
Both are means to an end, but work differently to try and achieve the same results.
Parallel networks treat the drivers separately to bring them together.
Series networks start with drivers together and proceed to carefully separate them.
Both are means to an end, but work differently to try and achieve the same results.
Agreed but from an acoustical standpoint, there should be no difference, the impedance however is very different for the 2 networks, probably something their to explain sonic differences
I have no professional audio experience but I faced this series-vs-parallel dilemma in my very first diy a few years ago. Since then I have done many more (search Fullrange Photo Gallery by: wchang) and prefer series 1st-order. In fact I've asked diyaudio the same question as you. For your higher-order "symmetric" XOs, I would guess that the parallel depended on worse phase alignment (vector math) to realize the same amplitude frequency response, hence presented worse imaging (especially depth) and coherence (small transcient details). Alternatively, the series had more very-high frequency at the ears somehow.
My comp techniques:
My comp techniques:
Interesting! I don't think it's phase issue with series Xos but I will have to double check. The phase response / group delay is dependant on the slope rate of the Xos not the topology, I also do not believe that phase issues play a role in soundstage presentation, the delay is super low in the 2kHz and above range probably not audible. I would love to be proven wrong, as mentioned, this can be tested with DSP
Let me correct myself, phase response ( not delay) will affect soundstage when it comes to inter driver spacing and crossover point you will get vertical lobing, , not good for stage