Thomas & Stereo posted this interview with Harley Lovegrove and much of the discussion is about speaker measurements. Harley provides numerous insights on speaker design and discusses the pros and cons of a implementing a flat frequency response. I enjoyed the discussion and wanted to share.
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Interview about Measurements with Pearl Sibelius Speaker Designer
- Thread starter brandonhall
- Start date
Well, I found this quite fascinating. But I'm eager to hear the reactions from folks here. I imagine there will be some disagreement or rebuttals to some of the points Harley makes. (For those unaware of who Harley is, he is the founder/designer of the Pearl Sibelius speakers.)
Quote from video: "Lets say the speaker measure with a big dip problem... by textbook case it is bad... if you measure it close enough there is a big dip, but it sounds great... so what do I do? Do I correct it?"
"... but I'm bring a speaker to market too and this is a question, a dilemma I'm facing, do I ask them to correct it...it sounds fantastic" ()
This is a good example of how, despite all the educational videos by Amir, Erin and countless posts on the ASR forum and elsewhere regarding loudspeaker measurement and evaluation of measurement results, not much in the way of information has reached the "measurement skeptics".
The way the questions are formulated alone shows that he apparently only looked at one frequency response, presumably the on-axis FR. Nobody can judge a loudspeaker only on the basis of one frequency response measurement - apart from completely failed speaker tunings.
Sure, in the ideal case the frequency response of a loudspeaker should be relatively flat, but that only applies if the horizontal and vertical frequency responses are also relatively ideal.
So the answer to the question is it depends on how the horizontal and vertical frequency responses behave.
To do this, you look at the individual frequency responses, but also at sound power (and PIR). SPDI and sonograms, normalized and non-normalized, also help with the assessment,....
This is exactly what the measurements are for.
And during the development of a loudspeaker, as Thomas is probably doing right now, you can explore many points and peculiarities of a loudspeaker in advance by simulation, discuss them and change them by design adjustments (before even the first wooden panel was touched).
If the speaker sounds "fantastic", then of course the dip should be left in place. The only important thing is by evaluating the crossover simulation tool (e.g. VCAD), to make sure that it does not show any indicators of a "colored sound".
To do this, one considers the above-mentioned parameters provided by the crossover simulation tool. If there is a contradiction to the heard, further listening sessions are recommended to exclude a coloration.
Despite the fact that the "measurement skeptics" always make it seem otherwise, even if everything has been measured down to the last detail, when it comes to fine-tuning, it is always the ear that decides.
"... but I'm bring a speaker to market too and this is a question, a dilemma I'm facing, do I ask them to correct it...it sounds fantastic" ()
This is a good example of how, despite all the educational videos by Amir, Erin and countless posts on the ASR forum and elsewhere regarding loudspeaker measurement and evaluation of measurement results, not much in the way of information has reached the "measurement skeptics".
The way the questions are formulated alone shows that he apparently only looked at one frequency response, presumably the on-axis FR. Nobody can judge a loudspeaker only on the basis of one frequency response measurement - apart from completely failed speaker tunings.
Sure, in the ideal case the frequency response of a loudspeaker should be relatively flat, but that only applies if the horizontal and vertical frequency responses are also relatively ideal.
So the answer to the question is it depends on how the horizontal and vertical frequency responses behave.
To do this, you look at the individual frequency responses, but also at sound power (and PIR). SPDI and sonograms, normalized and non-normalized, also help with the assessment,....
This is exactly what the measurements are for.
And during the development of a loudspeaker, as Thomas is probably doing right now, you can explore many points and peculiarities of a loudspeaker in advance by simulation, discuss them and change them by design adjustments (before even the first wooden panel was touched).
If the speaker sounds "fantastic", then of course the dip should be left in place. The only important thing is by evaluating the crossover simulation tool (e.g. VCAD), to make sure that it does not show any indicators of a "colored sound".
To do this, one considers the above-mentioned parameters provided by the crossover simulation tool. If there is a contradiction to the heard, further listening sessions are recommended to exclude a coloration.
Despite the fact that the "measurement skeptics" always make it seem otherwise, even if everything has been measured down to the last detail, when it comes to fine-tuning, it is always the ear that decides.
Here is an example to the above ("Should a dip in the FR go or stay"), which is a key point in the video.
A few years ago I designed a speaker (hobby DIY) that sounded really good with flat tuning and low sound pressure level (SPL) when listening (with some pieces of music you could hear small conspicuities, but since SPL was low it wasn't annoying yet).
But at high sound pressure level the speaker very quickly became unbearable.
The best crossover tuning that made it possible to listen to music relaxed even at high sound pressure levels looked like this - on-axis frequency response (FR), FR of woofer, midwoofer, tweeter:
At low SPL the speaker sounded "warm" on-axis, at high SPL the speaker sounded "right" - how can that be?
As an explanation for the "warm" sound at low SPL, the on-axis FR is sufficient in this case. At low SPL the direct sound dominates and the 3dB dip around the ear canal resonance (around 2.7kHz) leads to the sound perceived as "warm".
But why does the speaker sound just right with 3dB dip at high SPL and is unbearable with flat FR at high SPL?
To answer that you have to look at the horizontal and vertical FR. To keep the example as simple as possible, we restrict ourselves here to the horizontal FR.
To see what happens we examine the horizontal FR normalized to the on-axis (this would correspond to an ideally flat equalized speaker):
We can see that in this case theangular off-axis frequency responses in the range 1-4kHz show a hump and partly have a higher SPL than the on-axis FR.
With increasing volume, the perception of diffuse sound increases. The sound power (the radially emitted sound of the loudspeaker) has a hump exactly in the area of the ear canal resonance (where many people are particularly sensitive to sound power peaks).
It's quite different when we look at the final crossover tuning:
The on-axis dip compensates for the "errors" (caused by the loudspeaker cabinet, driver arrangement and edge diffraction) of the off-axis FR.
This is also confirmed by the sound power FR - blue curve. The black curve shows the on-axis FR and the red curve the (horizontal) directivity index, which is also a good indicator of the speaker's design flaws.
The sound power curve is so to speak an indicator or shows a tendency for the sound at high SPL in a normal listening room - of course not in the sense of a measuring device, where every tenth deviation shows a linear dependence.
Thus, measurements and listening impressions complement each other perfectly and quickly lead to a good result without the risk of overlooking important characteristics of the loudspeaker.
Update:
The analysis is greatly simplified by omitting the vertical FR, so please do not apply the statements to loudspeakers in general.
For example, unevenness in the DI caused by the vertical radiation is to be judged differently than by the horizontal radiation, etc....
A few years ago I designed a speaker (hobby DIY) that sounded really good with flat tuning and low sound pressure level (SPL) when listening (with some pieces of music you could hear small conspicuities, but since SPL was low it wasn't annoying yet).
But at high sound pressure level the speaker very quickly became unbearable.
The best crossover tuning that made it possible to listen to music relaxed even at high sound pressure levels looked like this - on-axis frequency response (FR), FR of woofer, midwoofer, tweeter:
At low SPL the speaker sounded "warm" on-axis, at high SPL the speaker sounded "right" - how can that be?
As an explanation for the "warm" sound at low SPL, the on-axis FR is sufficient in this case. At low SPL the direct sound dominates and the 3dB dip around the ear canal resonance (around 2.7kHz) leads to the sound perceived as "warm".
But why does the speaker sound just right with 3dB dip at high SPL and is unbearable with flat FR at high SPL?
To answer that you have to look at the horizontal and vertical FR. To keep the example as simple as possible, we restrict ourselves here to the horizontal FR.
To see what happens we examine the horizontal FR normalized to the on-axis (this would correspond to an ideally flat equalized speaker):
We can see that in this case the
With increasing volume, the perception of diffuse sound increases. The sound power (the radially emitted sound of the loudspeaker) has a hump exactly in the area of the ear canal resonance (where many people are particularly sensitive to sound power peaks).
It's quite different when we look at the final crossover tuning:
The on-axis dip compensates for the "errors" (caused by the loudspeaker cabinet, driver arrangement and edge diffraction) of the off-axis FR.
This is also confirmed by the sound power FR - blue curve. The black curve shows the on-axis FR and the red curve the (horizontal) directivity index, which is also a good indicator of the speaker's design flaws.
The sound power curve is so to speak an indicator or shows a tendency for the sound at high SPL in a normal listening room - of course not in the sense of a measuring device, where every tenth deviation shows a linear dependence.
Thus, measurements and listening impressions complement each other perfectly and quickly lead to a good result without the risk of overlooking important characteristics of the loudspeaker.
Update:
The analysis is greatly simplified by omitting the vertical FR, so please do not apply the statements to loudspeakers in general.
For example, unevenness in the DI caused by the vertical radiation is to be judged differently than by the horizontal radiation, etc....
Last edited:
I didn't like much of what was in there. His logic that he uses colored mics so speakers can be colored made no sense to me. Why would I want to perform double coloration that way? Wouldn't I want a neutral speaker to hear what coloration he added specifically by the use of the specific mic? Why would I want every piece of music colored all over again by the speaker?
And as @ctrl is rightly stating, it doesn't seem like they know about any measurement than on-axis response.
And as @ctrl is rightly stating, it doesn't seem like they know about any measurement than on-axis response.
OP
Well to me ASR was quite an eye-opener when it comes to amp and DAC purchases. Speakers, I don't know yet. I plan on purchasing one of the Revel bookshelf speakers just to check if they do better in my ears.
Over the years Ive heard and owned maybe over 100 different speakers from all decades and price points and one of the speakers I have my fondest memories of was a pair of Magnepan MMG, which I foolishly sold for greener pastures. Now we all know how the Magnepan LRS measurements turned out. Another speaker I very much liked was the Goodmans Dimension 8 (which by all means can not measure well) and I still own a pair of Bose 6.2, which are just fun and great for background listening, but come on Bose does not build great speakers, they just sell them well. So yeah for speakers, I really do not know...
Over the years Ive heard and owned maybe over 100 different speakers from all decades and price points and one of the speakers I have my fondest memories of was a pair of Magnepan MMG, which I foolishly sold for greener pastures. Now we all know how the Magnepan LRS measurements turned out. Another speaker I very much liked was the Goodmans Dimension 8 (which by all means can not measure well) and I still own a pair of Bose 6.2, which are just fun and great for background listening, but come on Bose does not build great speakers, they just sell them well. So yeah for speakers, I really do not know...
I don't agree with much of what he says here. He's obviously looking at speakers with the idea that they are musical instruments which I think is the wrong starting point. On the other hand there is a market for such speakers and he is in business to sell speakers.
The problem with his approach is that you don't end up with a loudspeaker that is a jack of all trades, and that to me is what a 'good' speaker should be. I am tired of people saying 'It measures poorly but it sounds fantastic.' Really? It sounds fantastic with every recording ever made? And you listened to every recording ever made to come to this conclusion? No you didn't. What you mean is you listened to a few recordings and they sounded fantastic. What about the other ten million recordings you didn't test with?
Here's a review of the Sibelius https://hifiplus.com/articles/pearl-acoustics-sibelius-sg-loudspeaker/
''If the term ‘SG’ sends you racing to the Gibson SG guitar and it being the preferred weapon of choice for AC/DC axe-grinder of note Angus Young, you are reading the wrong review. While these loudspeakers are perfectly comfortable playing almost all genres of music, the Sibelius’ power handling and maximum power output figures suggest a loudspeaker unsuited to playing power chords at high volumes, and this is confirmed in the listening
hat effortlessness also applies to its portrayal of dynamic range, although this is more about subtle micro-dynamic cues that draw you into the music than the rolling thunder of timpani and Led Zep drummers.''
So it doesn't sound fantastic if you want to listen to some rock music (god forbid that you would). And we can extrapolate from that that it will not sound fantastic with a million other recordings too.
I am biased, I admit, this is the sort of speaker I would run a mile from. Aside from the mediocre efficiency and the low power handling I know from experience that these speakers with their dinky little single driver require you to keep your head dead still in the sweet spot, move even an inch and the soundstage will collapse into the left or right speaker. I can't live with that even if there were no other failings.
The problem with his approach is that you don't end up with a loudspeaker that is a jack of all trades, and that to me is what a 'good' speaker should be. I am tired of people saying 'It measures poorly but it sounds fantastic.' Really? It sounds fantastic with every recording ever made? And you listened to every recording ever made to come to this conclusion? No you didn't. What you mean is you listened to a few recordings and they sounded fantastic. What about the other ten million recordings you didn't test with?
Here's a review of the Sibelius https://hifiplus.com/articles/pearl-acoustics-sibelius-sg-loudspeaker/
''If the term ‘SG’ sends you racing to the Gibson SG guitar and it being the preferred weapon of choice for AC/DC axe-grinder of note Angus Young, you are reading the wrong review. While these loudspeakers are perfectly comfortable playing almost all genres of music, the Sibelius’ power handling and maximum power output figures suggest a loudspeaker unsuited to playing power chords at high volumes, and this is confirmed in the listening
hat effortlessness also applies to its portrayal of dynamic range, although this is more about subtle micro-dynamic cues that draw you into the music than the rolling thunder of timpani and Led Zep drummers.''
So it doesn't sound fantastic if you want to listen to some rock music (god forbid that you would). And we can extrapolate from that that it will not sound fantastic with a million other recordings too.
I am biased, I admit, this is the sort of speaker I would run a mile from. Aside from the mediocre efficiency and the low power handling I know from experience that these speakers with their dinky little single driver require you to keep your head dead still in the sweet spot, move even an inch and the soundstage will collapse into the left or right speaker. I can't live with that even if there were no other failings.
OP
- Thread Starter
- #12
100%. That stuck out to me as well.
The great thing about the SPIN data is you can learn the character of the speaker in a single chart. Some folks love Maggies and others like Genelecs. Yet, even more, love Revels or Elacs. Completely different sound signature for all of those products. Do you want something exotic? Maggies. Do you want studio monitor detail which reveals flaws? Genelecs. Do you want something forgiving of bad recordings with a friendly sound signature? Revel or Elac
Isn't this the Harbeth philosophy as well..?
Even though I do not agree to all of Harley's statements, I like to watch his YouTube channel though. His video about Jacqueline Du Pré was very good...
Vacceo
Major Contributor
- Joined
- Mar 9, 2022
- Messages
- 2,669
- Likes
- 2,822
I will never understand why a "good speaker" is only good unless Napalm Death starts blasting through it.
In theory, bands such as Napalm Death or Cannibal Corpse should be easier to play due to how compressed their recordings are and how small their dynamic range is, yet here we are...
It's only good if it can play everything competently.
I understand the niche market for speakers like the Sibelius, I know people who only listen to classical trios and quartets, they could live with a speaker like this as they are never going to listen to any Napalm Death. That important point is not made in the discussion although it is in the review I linked to.
If I was designing a 2 way bookshelf speaker for the mass market (as the interviewer appears to be doing) I would not follow the advice Harley gives him.
Vacceo
Major Contributor
- Joined
- Mar 9, 2022
- Messages
- 2,669
- Likes
- 2,822
If it can play a Mozart´s 19th String Quartet well, a minute after Plague Rages, and finish your morning with Saint Saëns´ Danse Macabre; it is an objectively a better speaker.
Point is, there´s plenty of speakers that can do all those three equally well. And probably, cheaper.
I own a pair of Sibelius speakers. They are excellent, with a very flat, neutral "uncolored" response and an incredibly detailed soundstage. The sound is natural and musical without any hint of brightness or distortion at all frequencies (down to 38hz). I think it's the absence of a crossover, a great full range driver, and the finely tuned solid 1 1/2" French Oak cabinet that allows you to hear all the details, even at low volume. I listen to a lot of rock and jazz, as well a some classical, and can say that they sound fantastic on all types of music. As long as you aren't looking for front-row rock concert volume, they wont disappoint.
His logic is to use a more intuitive approach to designing a speaker. Rather than obsessing of measurements and striving for perfection, he tunes by ear to find a sweet spot. I'm fairly certain that Harley directly compares his speakers with real world instruments.
Thankfully, Harley has a trained ear. Because tuning by ear can have pretty divisive results as shown by Grado headphones which sound very colorful.
Now the reverse could be true with budget speakers which measure really well but don't sound great. I'm looking at the JBL Stage A130. You have to look at more than just the frequency response however. You'll find flaws.
Sorry to disagree with you in my first post here... been lurking for a while and respect and enjoy much of the hard work that is showcased here.
I felt compelled to respond because I feel you are missing a pair to fundamental points here. I don't think anyone is advocating a colored loudspeaker, though they are all colored. More neutral or less neutral exists, but no loudspeaker playing back in a real environment is laboratory perfect. I think we can all agree on that. And I think Harley was simply saying that once he found what he felt was the best balance between neutral and frequency extension he considered his speaker ready for the market.
The reason that a recording engineer would pick a "colored" mic is that the recording process is a creative process not a scientific process. The musicians and the engineers are attempting to produce a compelling musical recording that we will want to play back in our homes, cars, or wherever... they are not trying to create a scientific example of the acoustical output of a particular Stradivarius violin or Steinway piano. The final product that we listen to is one where the sound that a mic has picked up typically has gone through various equalizers, compressors, limiters, noise gates, and all manner of other processes. A good engineer like a good vocalist will have favorite microphones for different tasks. Many old ribbon microphones with limited bandwidth are cherished for the type of coloration they provide.
Of course if you are making a simple XY mic'd recording of a live performance you will want very flat responding microphones, but even then you may want to add a bit of low end lift so that the recording doesn't sound thin.
I have never met Harley or heard his speakers, but from my experience building (and measuring) speakers for approaching 40 years, I don't find what he is saying to be outlandish or snake oil like so much of what is out there.
Last edited:
Similar threads
- Poll