pedrojoaquimborges
New Member
- Joined
- May 8, 2026
- Messages
- 2
- Likes
- 4
Hi all! I've been a long time reader of ASR and, having been helped immensely by some of the info posted in the forum, decided to start a discussion that I feel may benefit from a different framing and could improve measurements as a representation of loudspeaker performance. I am very aware that touching the topics of transient response and time-domain behavior can open a can of worms in audio forum contexts. I am however, reintroducing the conversation due to suspecting that, with the right methodology, it can be both intuitively digestible and scientifically sound.
Onto the topic. In places where there's limited access to testing loudspeakers in the same room, Amir and Erin's measurements can be paramout in understanding a baseline of loudspeaker performance and, above all, brand design philosophy (Genelec/Neumann's pursuit of flat FR, Dynaudio's relative non-prioritization of directivity are some examples). The measurements themselves don't present enough information to express subjective taste, but if you know what you're looking for, I'd argue that they're a great start. One aspect of loudspeaker performance that I've consistently felt was underrepresented in the current measurement presentation is transient response and performance in the time-domain. I've recently upgraded a pair of Kali IN-5 to Neumann KH310 and their relative similarity in spinorama results made me ponder what measurement methodologies could better reflect the clear performance difference between them. This topic comes up a lot here and I've seen discussions get ugly. In fact I feel that this is a point of contention for the (recurring) Gearspace vs ASR debate. I have seen time-domain performance described as relatively unimportant for listening experience and would like to politely push back on that idea.
We're very sensitive to transient information and it can be considered a big part of listening experience and cognition. In music, as an example, transient sounds are vital. A very obvious and culturally relevant example are percussive or impact sounds. Time-domain linearity becomes really apparent when you judge the way a drum hits. For music producers and sound engineers, dialing in transients on drums is a very detailed process informed by listening - (colloquially) do you smooth them over, bring them out, do you leave more energy in the first 10ms, do you stretch it out over the first 80ms for a rounder transient, etc. Drums are just a blatant example, there's perception of element separation, depth, etc. all of it also depends on the way loudspeakers react to transient sounds. I might be stating the obvious, but bear with me.
The main recurring argument against documenting transient response beyond FR, IR/Step Response and group delay graphs is that frequency response (phase + magnitude) and impulse response are one and the same, since one can be derived from the other through a fourier transform. The crucial thing is that, while reproducing transient sounds speakers sometimes operate under a significant degree of non-linearity, which might not be represented in steady-state measurements. AFAIK you can only empirically measure non-linear behavior by actually operating a device under the conditions in which it operates non-linearly. You wouldn't try to measure a speaker's THD at 96 db SPL... without actually driving it to 96 dB SPL. So, if sine sweeps are our only test signal for measurement, there are parts of a speaker's transient response that might be underrepresented since the speakers are only tested in their linear behavior. Two speakers might measure same in FR, but produce a sound pressure level over different time intervals - 20ms or 10ms of 500hz can sound very different, despite integrating/averaging to the same total SPL.
I'm aware of a good thread by @pma suggesting methodology for measuring power handling and dynamic distortion. The tests described below do build on pma's own findings.
My somewhat speculative suggestions, before moving on to the case study.
A case study - Kali IN-5 vs Neumann KH 310 Woofer Transient Response
I recently upgraded one of my rooms from a pair of Kali IN-5 to Neumann KH 310s. ASR measurements were very helpful in understanding how the Kalis behave and projecting what other speaker could have a somewhat similar profile or design philosophy but offer higher performance. In short, I understood I liked to work with flat-ish FR speakers - seems obvious, but that fact is not a given in the pro audio speaker market.
It wasn't obvious, through the Klippel NFS measurements for both speakers, what differences between them would be significant. I got very curious on the specifics since they roughly both measure kind of flat FR-wise, have decent directivity, and neither has dramatic distortion figures. The KH310 does measure flatter, but 1 or 2 dB here or there isn't much when rooms can throw you 10dB up and down like it's nothing. But as far as measurements go, there's not much reason to believe they could have dramatically different performance for nearfield, low SPL applications.
As soon as I had both pairs in the studio, I started side-by-side comparisons with coherent room positioning. At first I was taken aback by how subtle the difference was in regards to overal tonal balance and presentation, while listening to music. The KH do have a more extended low end response, but I didn't get them just for a few extra Hz down there. The Neumanns' superiority started to show upon detailed listening. When I focused on sense of depth (testing in mono), element separation, transient response, and masking, the difference became quite obvious. I suspected that, among other things, these perceptual differences may correlate with aspects of time-domain behavior - a speaker's ability to not mask adjacent frequency bands and lower amplitude sounds and its ability to reproduce fast transient events at a given frequency band. So, having both pairs in the room, I decided to quickly and informally test their transient response to the best of my knowledge and abilities.
The test signal - sine bursts with a variable onset/attack window. A series of sine wave bursts with a increasingly shorter onset times - such that the amplitude onset becomes progressively shorter in time, which should make speakers behave increasingly non-linearly. This test should ideally be performed on a large series of frequenciesa along the spectrum, but the difference in transient response for these was only clearly relevant below 1kHz - so I ran the test at 20 different frequencies between 100hz-1khz, with consistent results throughout. The same test could be run for the burst's decay - meaning we can independently test a speaker's ability to both accelerate and decelerate. When I tested it, the results for the decay/offset were proportional to the onset/attack.
In the listening portion of the test is noting transient quality and non-linear behavior as transients get progressively shorter.
To not make the thread too long, I'll just present the results for a single frequency - 220Hz - as, in this case, I wanted to compare transient performance in this frequency region. the results were consistent for all the discrete frequencies testes (20 frequencies between 100hz-1khz).
Time-domain (waveform) representation of the test signal
Frequency Response representation of the test signals
Left to right: 7 cycles onset (attack/fade-in); 3 cycles onset; 1 cycle onset; 0 cycles onset.
EDIT: I have come to realize that the Pure Data patch I quickly assembled for the test wasn't outputting the cleanest sine waves - that's why the FR looks the way it does. I've since fixed it, but don't have the Kalis on hand anymore to repeat the rest. Although not ideal, this shouldn't be a problem since the results will be presented in direct comparison to the input signal.
NOTE: All FR graphs made with a Hanning window, 4096 block size and normalized to 0 dBFS
I tested both speakers of each pair, to rule out unit anomalies. Both woofers were positioned the same in-room and were captured by a Behringer ECM8000 25cm away at aprox 75 dB SPL. Since I am testing in a room, which is necessarily going to be excited by the transient information, the presented data has limitations. It does, however show the effectiveness of the test method, especially in anechoic or half-space outdoor conditions.
Results - the transient response is audibly VERY different and quite distinct in FR measurements. Regarding the audible difference, I don't mean it in a nuanced way - anyone would be able to tell the results apart.
Listen to the results (with headphones, ideally) in the link below. Each speaker plays 4 sine bursts for every onset/attack length, starting with the Kali. Meaning you'll hear 4 bursts from the Kali with a 7 cycle onset, 4 from the Neumann at the same onset, then 4 bursts from the Kali at a 3 cycle onset, and so on.
https://www.dropbox.com/scl/fi/8ht8...ey=m5pxav7tczz15j7esn9pwbape&st=f322fgfy&dl=0
Of course, listening to this with headphones is not the intended metric for the test. It just serves to demonstrate that the difference is so stark that it can be easily measured even without anechoic conditions.
Time and frequency domain results
I believe the results of this rest should considered with a grain of salt. The data has influence from room excitation and a bit too much background noise to be scientifically rigorous. It does, however, illustrate that reproducing the measurements in close to anechoic or even half-space conditions could represent transient non-linearity very well.
However, both speakers were tested in the same position and SPL, meaning that they'll both excite the room modes. Therefore, the main point in analyzing these results is not as much individual speaker performance, but rather comparing transient FR between the two.
Frequency Response
Time-Domain Response (Waveform) - Kali in Red, Neumann in Green.
I don't personally think there is much useful info to extract from these graphs.
Final Thoughts
First, it's clear to me that, even without changing measurement methodology, we could have slightly better representation of tested loudspeakers' transient response. The stark audible difference in transient response between the Kalis and Neumanns highlights the importance of having access to this info. Comparing their current measurements in the forum does not appear to fully capture the magnitude of the subjective differences I observed. To be clear, I don't believe transient response is the end-all response to the gap between measurements and audible perception - at all. It might, however, be a small step in the right direction.
Also, I don't really believe this to be the case of a highly unusual test signal engineered to bring out speakers' non-linearity. Transient signals like these the bread and butter of recorded music. In fact, this exact test signal is not too different from an 808s tom sound.
My current hypothesis is that, in this case, some of the observed transient limitations came down to distortion. However, it's not clear to me whether: a) the existent klippel distortion figures simply don't represent transient response; b) transient-related distortion may be perceptually more salient than steady-state distortion.
My central point in this post is mostly highlighting TR's perceptibility and its importance as a metric. This test is easily reproducible and can be made much more exact, scientifically speaking, so I hope being able to perceive this difference in person sheds some light on how relevant it might be for this forum's purposes.
About me
For a very short intro, for the sake of context, i work full time as a mixing and mastering engineer, having had the luck of worked on a few well-regarded records in Portugal, studied audio in college and have worked in acoustics engineering years ago. I've recently become more intensely invested in browsing ASR and DIYAudio while studying loudspeaker design.
Onto the topic. In places where there's limited access to testing loudspeakers in the same room, Amir and Erin's measurements can be paramout in understanding a baseline of loudspeaker performance and, above all, brand design philosophy (Genelec/Neumann's pursuit of flat FR, Dynaudio's relative non-prioritization of directivity are some examples). The measurements themselves don't present enough information to express subjective taste, but if you know what you're looking for, I'd argue that they're a great start. One aspect of loudspeaker performance that I've consistently felt was underrepresented in the current measurement presentation is transient response and performance in the time-domain. I've recently upgraded a pair of Kali IN-5 to Neumann KH310 and their relative similarity in spinorama results made me ponder what measurement methodologies could better reflect the clear performance difference between them. This topic comes up a lot here and I've seen discussions get ugly. In fact I feel that this is a point of contention for the (recurring) Gearspace vs ASR debate. I have seen time-domain performance described as relatively unimportant for listening experience and would like to politely push back on that idea.
We're very sensitive to transient information and it can be considered a big part of listening experience and cognition. In music, as an example, transient sounds are vital. A very obvious and culturally relevant example are percussive or impact sounds. Time-domain linearity becomes really apparent when you judge the way a drum hits. For music producers and sound engineers, dialing in transients on drums is a very detailed process informed by listening - (colloquially) do you smooth them over, bring them out, do you leave more energy in the first 10ms, do you stretch it out over the first 80ms for a rounder transient, etc. Drums are just a blatant example, there's perception of element separation, depth, etc. all of it also depends on the way loudspeakers react to transient sounds. I might be stating the obvious, but bear with me.
The main recurring argument against documenting transient response beyond FR, IR/Step Response and group delay graphs is that frequency response (phase + magnitude) and impulse response are one and the same, since one can be derived from the other through a fourier transform. The crucial thing is that, while reproducing transient sounds speakers sometimes operate under a significant degree of non-linearity, which might not be represented in steady-state measurements. AFAIK you can only empirically measure non-linear behavior by actually operating a device under the conditions in which it operates non-linearly. You wouldn't try to measure a speaker's THD at 96 db SPL... without actually driving it to 96 dB SPL. So, if sine sweeps are our only test signal for measurement, there are parts of a speaker's transient response that might be underrepresented since the speakers are only tested in their linear behavior. Two speakers might measure same in FR, but produce a sound pressure level over different time intervals - 20ms or 10ms of 500hz can sound very different, despite integrating/averaging to the same total SPL.
I'm aware of a good thread by @pma suggesting methodology for measuring power handling and dynamic distortion. The tests described below do build on pma's own findings.
My somewhat speculative suggestions, before moving on to the case study.
- It would be useful to have a solid representation of a speaker's transient response in ASR measurements, as we already have for FR, directivity, group delay and distortion. Specific suggestions follow.
- I actually think the current Klippel measurements might have enough information to describe linear time-domain performance if presented in different ways. The current set of graphs and parameters aren't particularly good at representing transient response. Transient response is highly frequency dependent. It's not uncommon to have tweeters and/or mids with great transient reproduction paired with woofers that sacrifice transient response to reach lower. So looking at an IR or step response by itself is fairly uninformative - most speakers are likely to be able to decently reproduce mid and high transients, which dominate the full-range IR graph and give the impression of a transparent TR through the whole frequency range. A suggestion: Presenting filtered IR magnitude graphs for a few frequency bands, in anechoic or klippel measurements, can probably be somewhat representative of transient reponse. Example: showing a bandpassed IR magnitude for 100Hz, 300Hz, 500Hz, etc (using filters with appropriate time behavior).
- I suspect that a good representation of transient reponse can be reached by measuring with sine bursts of variable 'window' lengths of onset/attack and offset/decay, such as in the informal test i conducted below (basically fading in/out slower/faster on a series of sine bursts). I've experimented with this and think that it might be a solid method, if properly fleshed out to the frequency spectrum and multiple SPL. With a fast enough onset/offset, speakers might start to distort and behave non-linearly, which allows us to have FR and IR graphs that both show distortion in time and frequency domain. I'll expand on it below. NOTE: I clearly didn't come up with this, sine bursts are an established measurement test signal.
- I don't intend for this post's subtext to express that a "fast" transient response is superior. I personally take a lot of pleasure from listening and working on speakers (and rooms) that provide a very fast and controlled transient response - like a sealed speaker in a dry room. However, I imagine a lot of people would much rather listen to speakers with a "rounder" bass response - like some ported speakers and subwoofers - and know, for a fact, most musicians producers work under those conditions. However, faithfully representing transient information in input signals probably does often require speakers that accelerate and decelerate faster, since that is a physical limitation for drivers' motors, suspension and diaphragms.
- Aside from a speaker's ability to reproduce a transient sound by itself, an important consequence of transient response seems to be how much transients are masked by adjacent frequency bands. A frequently reported feature of speakers with great transient response is the ability to still faithfully reproduce, say, transients at 300 Hz when there's a lot of sub information in the input signal - transient response under higher excursion distances.
A case study - Kali IN-5 vs Neumann KH 310 Woofer Transient Response
I recently upgraded one of my rooms from a pair of Kali IN-5 to Neumann KH 310s. ASR measurements were very helpful in understanding how the Kalis behave and projecting what other speaker could have a somewhat similar profile or design philosophy but offer higher performance. In short, I understood I liked to work with flat-ish FR speakers - seems obvious, but that fact is not a given in the pro audio speaker market.
It wasn't obvious, through the Klippel NFS measurements for both speakers, what differences between them would be significant. I got very curious on the specifics since they roughly both measure kind of flat FR-wise, have decent directivity, and neither has dramatic distortion figures. The KH310 does measure flatter, but 1 or 2 dB here or there isn't much when rooms can throw you 10dB up and down like it's nothing. But as far as measurements go, there's not much reason to believe they could have dramatically different performance for nearfield, low SPL applications.
As soon as I had both pairs in the studio, I started side-by-side comparisons with coherent room positioning. At first I was taken aback by how subtle the difference was in regards to overal tonal balance and presentation, while listening to music. The KH do have a more extended low end response, but I didn't get them just for a few extra Hz down there. The Neumanns' superiority started to show upon detailed listening. When I focused on sense of depth (testing in mono), element separation, transient response, and masking, the difference became quite obvious. I suspected that, among other things, these perceptual differences may correlate with aspects of time-domain behavior - a speaker's ability to not mask adjacent frequency bands and lower amplitude sounds and its ability to reproduce fast transient events at a given frequency band. So, having both pairs in the room, I decided to quickly and informally test their transient response to the best of my knowledge and abilities.
The test signal - sine bursts with a variable onset/attack window. A series of sine wave bursts with a increasingly shorter onset times - such that the amplitude onset becomes progressively shorter in time, which should make speakers behave increasingly non-linearly. This test should ideally be performed on a large series of frequenciesa along the spectrum, but the difference in transient response for these was only clearly relevant below 1kHz - so I ran the test at 20 different frequencies between 100hz-1khz, with consistent results throughout. The same test could be run for the burst's decay - meaning we can independently test a speaker's ability to both accelerate and decelerate. When I tested it, the results for the decay/offset were proportional to the onset/attack.
In the listening portion of the test is noting transient quality and non-linear behavior as transients get progressively shorter.
To not make the thread too long, I'll just present the results for a single frequency - 220Hz - as, in this case, I wanted to compare transient performance in this frequency region. the results were consistent for all the discrete frequencies testes (20 frequencies between 100hz-1khz).
Time-domain (waveform) representation of the test signal
Frequency Response representation of the test signals
Left to right: 7 cycles onset (attack/fade-in); 3 cycles onset; 1 cycle onset; 0 cycles onset.
EDIT: I have come to realize that the Pure Data patch I quickly assembled for the test wasn't outputting the cleanest sine waves - that's why the FR looks the way it does. I've since fixed it, but don't have the Kalis on hand anymore to repeat the rest. Although not ideal, this shouldn't be a problem since the results will be presented in direct comparison to the input signal.
NOTE: All FR graphs made with a Hanning window, 4096 block size and normalized to 0 dBFS
I tested both speakers of each pair, to rule out unit anomalies. Both woofers were positioned the same in-room and were captured by a Behringer ECM8000 25cm away at aprox 75 dB SPL. Since I am testing in a room, which is necessarily going to be excited by the transient information, the presented data has limitations. It does, however show the effectiveness of the test method, especially in anechoic or half-space outdoor conditions.
Results - the transient response is audibly VERY different and quite distinct in FR measurements. Regarding the audible difference, I don't mean it in a nuanced way - anyone would be able to tell the results apart.
- Both in-room and in the recording the Neumann's transients sound, for lack of better wording, more transparent and uncoloured.
- The most immediate difference is that the Kalis seem to introduce more distortion and to already do it at longer onset times.
- The Neumanns seemed to perceptually reproduce a louder and sharper transient, despite the same SPL, throughout the tested spectrum. NOTE: above 1khz the Kali actually performed audibly quite well.
Listen to the results (with headphones, ideally) in the link below. Each speaker plays 4 sine bursts for every onset/attack length, starting with the Kali. Meaning you'll hear 4 bursts from the Kali with a 7 cycle onset, 4 from the Neumann at the same onset, then 4 bursts from the Kali at a 3 cycle onset, and so on.
https://www.dropbox.com/scl/fi/8ht8...ey=m5pxav7tczz15j7esn9pwbape&st=f322fgfy&dl=0
Of course, listening to this with headphones is not the intended metric for the test. It just serves to demonstrate that the difference is so stark that it can be easily measured even without anechoic conditions.
Time and frequency domain results
I believe the results of this rest should considered with a grain of salt. The data has influence from room excitation and a bit too much background noise to be scientifically rigorous. It does, however, illustrate that reproducing the measurements in close to anechoic or even half-space conditions could represent transient non-linearity very well.
However, both speakers were tested in the same position and SPL, meaning that they'll both excite the room modes. Therefore, the main point in analyzing these results is not as much individual speaker performance, but rather comparing transient FR between the two.
Frequency Response
Time-Domain Response (Waveform) - Kali in Red, Neumann in Green.
I don't personally think there is much useful info to extract from these graphs.
Final Thoughts
First, it's clear to me that, even without changing measurement methodology, we could have slightly better representation of tested loudspeakers' transient response. The stark audible difference in transient response between the Kalis and Neumanns highlights the importance of having access to this info. Comparing their current measurements in the forum does not appear to fully capture the magnitude of the subjective differences I observed. To be clear, I don't believe transient response is the end-all response to the gap between measurements and audible perception - at all. It might, however, be a small step in the right direction.
Also, I don't really believe this to be the case of a highly unusual test signal engineered to bring out speakers' non-linearity. Transient signals like these the bread and butter of recorded music. In fact, this exact test signal is not too different from an 808s tom sound.
My current hypothesis is that, in this case, some of the observed transient limitations came down to distortion. However, it's not clear to me whether: a) the existent klippel distortion figures simply don't represent transient response; b) transient-related distortion may be perceptually more salient than steady-state distortion.
My central point in this post is mostly highlighting TR's perceptibility and its importance as a metric. This test is easily reproducible and can be made much more exact, scientifically speaking, so I hope being able to perceive this difference in person sheds some light on how relevant it might be for this forum's purposes.
About me
For a very short intro, for the sake of context, i work full time as a mixing and mastering engineer, having had the luck of worked on a few well-regarded records in Portugal, studied audio in college and have worked in acoustics engineering years ago. I've recently become more intensely invested in browsing ASR and DIYAudio while studying loudspeaker design.
Attachments
Last edited:
