Which speakers are the Classical Music Pros using?

[Floyd Toole]

The Harbeth speaker you show looks more like the old fashioned "smiley face" than anything - curious as it comes from someone who had been involved with at least some of the BBC monitor developments. One wonders what kinds of listening tests they did to end up with this . . . Starting with a neutral speaker and turning up the bass and treble controls could do it too - with the benefit that neutrality can be restored.

Playing with spectral balance to create impressions of distance, proximity, "detail", etc. and to improve speech intelligibility are well known tools of the trade. However, one should start with neutral speakers to reveal the truths of what the mics are picking up and work on modifications electronically.

Recording engineers are human, and they respond to manufacturer's claims, industry gossip, etc. How else can one explain the widespread use of Yamaha NS-10s - in essence an Auratone 5C with more bass, and the UREI 811B, an Auratone 5C with even more bass and serious horsepower. None of these speakers is remotely neutral, and all have had imitators since then. Habits die hard, even bad ones. The original notion that such a strongly colored sound resembles the "average" consumer loudspeaker was never true. Having tested hundreds of speakers over the decades I can state that the true "average" of them all is essentially flattish. Individually they cover all possibilities of being wrong. Now, things are much improved, and even little bluetooth wireless speakers can be quite pleasant within their obvious limitations. That happens because electronics and transducers are married; always a good idea.

 

[tuga]

The Harbeth speaker you show looks more like the old fashioned "smiley face" than anything - curious as it comes from someone who had been involved with at least some of the BBC monitor developments. One wonders what kinds of listening tests they did to end up with this . . . Starting with a neutral speaker and turning up the bass and treble controls could do it too - with the benefit that neutrality can be restored.

Playing with spectral balance to create impressions of distance, proximity, "detail", etc. and to improve speech intelligibility are well known tools of the trade. However, one should start with neutral speakers to reveal the truths of what the mics are picking up and work on modifications electronically.

Recording engineers are human, and they respond to manufacturer's claims, industry gossip, etc. How else can one explain the widespread use of Yamaha NS-10s - in essence an Auratone 5C with more bass, and the UREI 811B, an Auratone 5C with even more bass and serious horsepower. None of these speakers is remotely neutral, and all have had imitators since then. Habits die hard, even bad ones. The original notion that such a strongly colored sound resembles the "average" consumer loudspeaker was never true. Having tested hundreds of speakers over the decades I can state that the true "average" of them all is essentially flattish. Individually they cover all possibilities of being wrong. Now, things are much improved, and even little bluetooth wireless speakers can be quite pleasant within their obvious limitations. That happens because electronics and transducers are married; always a good idea.

Since you've mentioned the Yamaha NS-10 I wonder if you've read this piece extolling its unique set of skills and fitness for purpose:

THE YAMAHA NS10M: TWENTY YEARS A REFERENCE MONITOR. WHY?
Philip R Newell, Keith R Holland, Julius P Newell

 

[eliash]

Wonder if this discussion about flat frequency response on axis filled up with standardised side reflections is really hitting the point, whether classical (orchestral, chamber) music sounds good or bad in a certain architectual/acoustical situation. +/-3dB on axis response is anyway overwhelmed by room reflections, and the beforehand displayed studios exhibit all kinds of room architecture...

Floyd Toole has mentioned other things like crossover frequency speaker beam pattern change (...and there may be more like enclosure movement/vibrations modulating the sound...), which tend to be responsible for personal judgement as well.
Any responsible studio person will certainly work on the set-up and maybe also use response correction (acoustic or electronic) to optimise cognition and measurements...

Btw.: Listening to some medium size B&W speakers recently in a well-equipped HiFi-studio for quite a long time I could not make friends. Not because of the speaker/room frequency response, but because the set-up sounded distorted* with various musical material to me...wonder if Revels would impress me better...

*a late addition: what I mean with distortion is a harshness and/or roughness perception in their natural meaning, not necessarily according to their scientific definition

 

[tuga]

Wonder if this discussion about flat frequency response on axis filled up with standardised side reflections is really hitting the point, whether classical (orchestral, chamber) music sounds good or bad in a certain architectual/acoustical situation. +/-3dB on axis response is anyway overwhelmed by room reflections, and the beforehand displayed studios exhibit all kinds of room architecture...

Floyd Toole has mentioned other things like crossover frequency speaker beam pattern change (...and there may be more like enclosure movement/vibrations modulating the sound...), which tend to be responsible for personal judgement as well.
Any responsible studio person will certainly work on the set-up and maybe also use response correction (acoustic or electronic) to optimise cognition and measurements...

Btw.: Listening to some medium size B&W speakers recently in a well-equipped HiFi-studio for quite a long time I could not make friends. Not because of the speaker/room frequency response, but because the set-up sounded distorted with various musical material to me...wonder if Revels would impress me better...

Also important is the fact that in a control room or studio the early reflections are treated, either with absorption, deflection, diffusion, or a combination so the engineer will mostly be listeing to direct sound, perhaps not always on axis depending on the width of the desk.

The domestic playback room is a different beast, hardly ever treated and with massive diversity in terms of shape and size. And on top of that there's listener preference.

 

[Frgirard]

Also important is the fact that in a control room or studio the early reflections are treated, either with absorption, deflection, diffusion, or a combination so the engineer will mostly be listeing to direct sound, perhaps not always on axis depending on the width of the desk.

The domestic playback room is a different beast, hardly ever treated and with massive diversity in terms of shape and size. And on top of that there's listener preference.

The Moulton design does not trait early reflections.

In studio when you need to create a stereo sound stage, you don't want your introduce a stereo picture.

 

[tuga]

The Moulton design does not trait early reflections.

In studio when you need to create a stereo sound stage, you don't want your introduce a stereo picture.

I'd never hear of the Moulton design before, thanks.

http://www.moultonlabs.com/more/brave_new_world_control_room_design/

My preferred approach is just the opposite of this, in what I think of as an Early-Reflection Fast-Decay design (and call, egotistically, a “Moulton Room”). We know, from a lot of research and accumulated experience, that humans integrate the early reflections (for up to 50 ms.) with the direct sound. We also know that the sound decaying for the period from 70 to 150 ms. really interferes with the clarity and intelligibility of the direct sound. The Moulton Room supports early reflections for up to 50 ms. for that integration, and then begins absorption, as completely as possible.

It turns out that this is comparatively cheap to do (in essence - make the wall behind the loudspeakers anechoic, leave the other walls alone). Further, the sound resembles the sound in end-users rooms quite effectively for the first 50 ms., really improving the Listening Ahead performance, and not hurting the Listening Back performance at all. The downside of this is that it requires speakers with good dispersion, and room-to-room compatibility requires reasonably similar dimensions.

 

[eliash]

I'd never hear of the Moulton design before, thanks.

http://www.moultonlabs.com/more/brave_new_world_control_room_design/

My preferred approach is just the opposite of this, in what I think of as an Early-Reflection Fast-Decay design (and call, egotistically, a “Moulton Room”). We know, from a lot of research and accumulated experience, that humans integrate the early reflections (for up to 50 ms.) with the direct sound. We also know that the sound decaying for the period from 70 to 150 ms. really interferes with the clarity and intelligibility of the direct sound. The Moulton Room supports early reflections for up to 50 ms. for that integration, and then begins absorption, as completely as possible.

It turns out that this is comparatively cheap to do (in essence - make the wall behind the loudspeakers anechoic, leave the other walls alone). Further, the sound resembles the sound in end-users rooms quite effectively for the first 50 ms., really improving the Listening Ahead performance, and not hurting the Listening Back performance at all. The downside of this is that it requires speakers with good dispersion, and room-to-room compatibility requires reasonably similar dimensions.

So to speak rather a DELE room...anyway I was also thinking about much shorter delay times, not in the ms range, but in the µs range caused by enclosure vibrations, maybe call it jitter. E .g. a bass driver modulating a tower enclosure with the tweeter by a realistic 100µm front-to-back movement causing some 300ns time jitter. Any DAC showing such a figure here at ASR would go immediately into the bin (the pink panther crushed in thousand pieces). The question how this is perceived by the general audience and what physical or audiological effects are relevant or not?

 

[tuga]

So to speak rather a DELE room...anyway I was also thinking about much shorter delay times, not in the ms range, but in the µs range caused by enclosure vibrations, maybe call it jitter. E .g. a bass driver modulating a tower enclosure with the tweeter by a realistic 100µm front-to-back movement causing some 300ns time jitter. Any DAC showing such a figure here at ASR would go immediately into the bin (the pink panther crushed in thousand pieces). The question how this is perceived by the general audience and what physical or audiological effects are relevant or not?

I think that we can find a clue at the begining of the text I quoted, where Moulton says "My preferred approach".
The most accurate reproduction of the signal would happen in an anechoic chamber: the speakers transduce the audio signal and it reaches the listener's ears without being adulterated by room interference. Yet most people would agree that such accuracy would sound bad, they'd all prefer to add some room-generated distortion.
How much distortion, and the type or types depends on who you ask; in other words it's a matter of personal preference. Likewise, some people prefer narrow directivity speakers, other wide, others still omni or dipole, and some people like "dead" rooms whilst others prefer reflective ones. There's no one-size-fits-all solution.

 

[Frgirard]

I think that we can find a clue at the begining of the text I quoted, where Moulton says "My preferred approach".
The most accurate reproduction of the signal would happen in an anechoic chamber.

The European Broadcast Union has define what is an accurate reproduction.
https://tech.ebu.ch/docs/tech/tech3276.pdf


Northward Acoustic designs studio following the Non Environmental prerequesite except Northward introduce "noise" at the listening seat to give a normal ambiance at the engineer...

Hear music in anechoic chamber is a suffering. An accurate reproduction is made for a human.

 

[Floyd Toole]

Since you've mentioned the Yamaha NS-10 I wonder if you've read this piece extolling its unique set of skills and fitness for purpose:

THE YAMAHA NS10M: TWENTY YEARS A REFERENCE MONITOR. WHY?
Philip R Newell, Keith R Holland, Julius P Newell

Hi, I know the document and I know the authors. There is much to say about it, including noting the comment early on that the Yamaha NS-10M "appeared to have a sound character that mixing personnel had been looking for" - in other words the speaker was selected for being a pleasing equalizer for pop/rock music, not a neutral revealer of recorded "truths". There is the admission later on that the obvious mid frequency excess will be reflected in recordings, saying "the resultant balance of low- vs mid-frequencies will probably be correctable using equalisation should that be deemed necessary during the mastering process". This is a classic "kick-the-can-down-the-road" attitude. They assume that mastering engineers will be using neutral monitors and therefore might hear the coloration and can fix it.

This is precisely what professional recording engineers should not be doing - in my opinion . . .

It is important to understand that loudspeaker performance at low frequencies is minimum-phase behavior. That is, its time domain performance is predictable from its amplitude response. Apply ANY EQ and the time domain behavior changes - are there any control room systems that do not have some EQ?. Woofer performance is predictable from contemporary mathematical modeling of transducer/enclosure interaction. Of course, superimposed on speaker behavior are the medium-to-high-Q room resonances. Listening at about 1 m over the top of a console work surface is one thing. Listening a 2 to 3 m in a normally-reflective domestic space is another. All their measurements were on axis only; no off axis data that might have added insights. Anechoic chambers are not anechoic at low frequencies, and even attempts at corrective EQ have limitations (I know because I have "calibrated" chambers for use below cutoff frequency).

Section 12.5.1 "Old-School Monitoring" in the 3rd edition of my book shows measurements on several relevant loudspeakers, showing clearly that the Pro version of the NS-10 was modified to do an even better job of imitating "ye olde original crappy speaker" the Auratone 5C, an inexpensive 5-inch full range paper-cone speaker that was widely used in CRT television sets of the period.

BTW, the designer of the NS-10M and the NS-1000M visited me at my NRC lab in Canada to experience the measurement process and double-blind listening tests. They left with many physical measurements and photographs intending to duplicate some of the facility and processes. The original speakers were designed to exhibit flat sound power (believed, incorrectly, to be what listeners heard in the far field), which my measurements showed they did extremely well. The problem was that the two-way NS-10 ended up with a very non-flat on-axis response, but the three-way NS-1000M, with more uniform directivity with frequency, was an exemplary loudspeaker at the time (1974) - see Figure 18.3 (e).

I could go on, but fortunately I don't need to because the recording industry has done it for me; it has moved on. The fad has substantially passed, and the current norms for monitor loudspeakers (including Yamahas) are not different from the current objectives for neutral sounding domestic entertainment loudspeakers, which is as it should be.

 

[Floyd Toole]

The European Broadcast Union has define what is an accurate reproduction.
https://tech.ebu.ch/docs/tech/tech3276.pdf


Northward Acoustic designs studio following the Non Environmental prerequesite except Northward introduce "noise" at the listening seat to give a normal ambiance at the engineer...

Hear music in anechoic chamber is a suffering. An accurate reproduction is made for a human.

The EBU Tech 3276 and ITU-R BS.1116-3 standards are essentially identical, and both are wrong. They require that the loudspeakers be flat on axis (good!, but a 4 dB tolerance for 1/3-octave measurements is quite generous) and also flat in the listening room, using EQ if necessary. The topic is fully discussed, with measurements, in Section 13.2.2 in the 3rd edition of my book. A flat ORR (operational room response, a.k.a. steady state room curve) will be perceived as being to bright in all normally reflective listening rooms using forward-firing (not omnidirectional) loudspeakers.

 

[Frgirard]

The EBU Tech 3276 and ITU-R BS.1116-3 standards are essentially identical, and both are wrong. They require that the loudspeakers be flat on axis (good!, but a 4 dB tolerance for 1/3-octave measurements is quite generous) and also flat in the listening room, using EQ if necessary. The topic is fully discussed, with measurements, in Section 13.2.2 in the 3rd edition of my book. A flat ORR (operational room response, a.k.a. steady state room curve) will be perceived as being to bright in all normally reflective listening rooms using forward-firing (not omnidirectional) loudspeakers.

Page 7 this is not the case. after 2KHz they preconise a slope of 1db by octave.

No one preconise a flat response at the listening position.

 

[Duke]

A flat ORR (operational room response, a.k.a. steady state room curve) will be perceived as being too bright in all normally reflective listening rooms using forward-firing (not omnidirectional) loudspeakers.

What are your thoughts about steady-state room curves for omnidirectional or polydirectional loudspeakers? Would there be any advantage to reducing the spectral discrepancy between the direct and reflected sound, something your old bioplar Mirage M-1's may have done?

 

[Floyd Toole]

Page 7 this is not the case. after 2KHz they preconise a slope of 1db by octave.

No one preconise a flat response at the listening position.

I guess we are not seeing the same thing. The upper tolerance limit for the ORR is flat. The lower limit is flat to 2 kHz where it slopes downward expanding the allowable performance - there is a lot of room for spectral balance variations - including broadband flat or downward sloping above 2 kHz.. But the standard states: "Lm is the mean value of the levels of the 1/3–octave bands with centre frequencies from 200 Hz to 4 kHz." It is shown in the figure as a horizontal straight line running the full bandwidth to 16 kHz.

In my observation all of this ignores the physics of loudspeakers which become progressively more omnidirectional at low frequencies - with a flat on-axis response the steady-state room response bass must rise (or, depending on your perspective, the higher frequencies must fall). With forward firing loudspeakers the subjectively preferred products yield downward sloping steady-state room curves. Flat is too bright as it tilts the on-axis, direct sound, curve upwards. Figure 13.2 in my book shows this. Note that the axial curves have 1/20-octave resolution, not 1/3-octave.

Also, see this open access paper: Toole, F. E. (2015). “The Measurement and Calibration of Sound Reproducing Systems”, J. Audio Eng. Soc., vol. 63, pp.512-541. This is an open-access paper available to non-members at www.aes.org http://www.aes.org/e-lib/browse.cfm?elib=17839

 

[Floyd Toole]

What are your thoughts about steady-state room curves for omnidirectional or polydirectional loudspeakers? Would there be any advantage to reducing the spectral discrepancy between the direct and reflected sound, something your old bioplar Mirage M-1's may have done?

My personal experience is too limited to answer with any certainty, and nobody I know of has pursued this as a research question. The well-behaved multi-directional speakers I have heard sound just fine, and are a viable alternative to forward firing designs. They appear to add a bit more "space" to a stereo presentation, and for those mixes that have hard-panned L and R images they might present a 'softer' image. But real investigation is needed to give a certain answer.

 

[DSJR]

Harwood may have been expressing his opinion. His own design, the Harbeth HL did sport the dip in the presence region (see below).

But more importantly EQ'ing the presence region does affect how we perceive phantom image proximity - audio engineers know and use this (psychoacoustic) effect - and in small rooms with orchestral music it is not an unpleasant effect, quite the contrary.
I wonder if this isn't one of the reasons why so many classical music production monitors are B&Ws...or perhaps because it's what everyone else is using.

That plot looks to be the very first HL speaker model. The HiFi Choice test shows similar and there are good documented reasons why Dudley apparently didn't use voice in the design (unlike current issue models). The HL was 'flattened' quite quickly I remember and we sold a lot of them. the HL III matured this particular design and neutral or not (I suspect a gentler smiley-face response as the tweeter used had a rising response).

To bring this brand up to date, an Australian review with measurements of the C7ES£-XD - HF held much tighter but not 'up' in level.

https://www.audiomagic.com.au/files/products/951/documents/Harbeth_Compact_7ES-3XD.pdf

 

[Duke]

My personal experience is too limited to answer with any certainty, and nobody I know of has pursued this as a research question. The well-behaved multi-directional speakers I have heard sound just fine, and are a viable alternative to forward firing designs. They appear to add a bit more "space" to a stereo presentation, and for those mixes that have hard-panned L and R images they might present a 'softer' image. But real investigation is needed to give a certain answer.

Thank you VERY MUCH.

"Well-behaved multi-directional speaker" is the niche I'm exploring, and it's reassuring to hear from you that such can be a "viable alternative to forward firing designs".

You see, there is really NO point in a little bitty company like mine trying to beat Harman at its own game, namely well-behaved forward-firing designs, be they cone-n-dome or bighorn-n-bigwoofer. We might as well do something DIFFERENT, especially if such resulting in a "viable alternative" is a possibility.

I was also unable to find where anyone had researched this area to an extent that would answer the questions which arose, so we conducted some simple tests of our own, trying to figure out "where the goal posts are" for at least one multidirectional configuration.

One such test involved a fixed-volume controlled-pattern forward-firing array being paired with a variable-volume controlled-pattern rear-firing array. Listeners used a remote control to blindly adjust the volume level of the rear-firing array to their preference (this was done in stereo, so both rear-firing arrays were adjusted simultaneously). Listeners were instructed to set the volume of the rear-firing array by ear to the level where it was subjectively a net benefit AND did not degrade clarity. The results were educational to me and my colleague, though of course their applicability beyond the specific test conditions is marginal at best.

Despite my hopes of a well-behaved multi-directional speaker being a "viable alternative" for home audio, I DO NOT see any scenario in which a well-behaved multi-directional speaker would be useful for a recording studio, as I do not think it would provide the recording engineers as clear a picture of what's on the recording as a good forward-firing design would. By way of analogy, it would likewise be out of place to "employ tasteful upmixing to embellish the sense of space” during the the production of a two-channel recording. [Quote lifted from the 3rd edition of your book, but I can't locate my copy at the moment to give the page number.]

Once again, thank you very much for taking the time to reply.

 

[q3cpma]

Thank you VERY MUCH.

"Well-behaved multi-directional speaker" is the niche I'm exploring, and it's reassuring to hear from you that such can be a "viable alternative to forward firing designs."

Are you not afraid of such speakers sounding good only situationally, as almost all music is produced with monopole speakers and headphones?

 

[Duke]

Are you not afraid of such speakers sounding good only situationally, as almost all music is produced with monopole speakers and headphones?

Excellent question!

To the best of my knowledge, that hasn't happened yet with our current version when set up as recommended. I can see how it might have happened with some of our early versions in some rooms, as they had less setup adjustability.

Here is a simplified description of what we're doing: Imagine starting with a well-behaved wide-pattern speaker (a big Revel perhaps) and then chop the radiation pattern in two. Using aggressive toe-in, aim half of the pattern at the listening area, and aim the other half off in a different direction, such that it arrives after some time delay. So basically what we have done is, pushed back in time the "center of gravity" of WHEN the reflections arrive. This simulates what might happen in a larger room, wherein the longer reflection path lengths likewise push back the temporal "center of gravity" of the reflections. One of the results seems to be reduced "small room signature", such that with a good recording, we are more aware of the spatial cues on the recording. And thus far it seems that even with a poor recording the net result is not detrimental.

Putting it another way, we are manipulating (including aiming) the radiation patterns to approximate having a "reflection free zone", at least in the horizontal plane, which is a room treatment technique that seems to give good results, and which likewise pushes the "center of gravity" of the reflections back in time somewhat.

 

[q3cpma]

Excellent question!

To the best of my knowledge, that hasn't happened yet with our current version when set up as recommended. I can see how it might have happened with some of our early versions in some rooms, as they had less setup adjustability.

Here is a simplified description of what we're doing: Imagine starting with a well-behaved wide-pattern speaker (a big Revel perhaps) and then chop the radiation pattern in two. Using aggressive toe-in, aim half of the pattern at the listening area, and aim the other half off in a different direction, such that it arrives after some time delay. So basically what we have done is, pushed back in time the "center of gravity" of WHEN the reflections arrive. This simulates what might happen in a larger room, wherein the longer reflection path lengths likewise push back the temporal "center of gravity" of the reflections. One of the results seems to be reduced "small room signature", such that with a good recording, we are more aware of the spatial cues on the recording. And thus far it seems that even with a poor recording the net result is not detrimental.

Putting it another way, we are manipulating (including aiming) the radiation patterns to approximate having a "reflection free zone", at least in the horizontal plane, which is a room treatment technique that seems to give good results, and which likewise pushes the "center of gravity" of the reflections back in time somewhat.

So, if I understand well, the goal is to have a combination of narrow dispersion with quite delayed reflections following it?