Warning: In this page, I use a bunch of equations that might look very unfamiliar. These are called “s-domain expressions”, used in Laplace analysis. If these do not make any sense, don’t worry – it’s not really important that they do – so you can just skip over any math that looks ugly without guilt or concern. The reasons I used it here were twofold: firstly, these are taken from the equations in Bækgaard’s paper (mentioned below), and secondly, it’s the easiest way to show that something is missing (the “MissingPortion” in the intuitive equations below) in a two-way loudspeaker with one type of crossover.
When you build a two-way loudspeaker (one with a tweeter and a woofer), you have to divide the energy in the audio signal before sending it to the two drivers using a circuit called a crossover. This filters the signal sent to a tweeter using a high-pass filter (which only allows the high frequencies to pass through it) and the signal sent to the woofer using a low-pass filter (which allows the low frequencies to pass). The result is a signal that crosses over from the woofer to the tweeter as the frequency increases – hence the name. However, this is not necessarily the end of the solution, since high-pass and low-pass filters have some characteristics that we need to worry about.
One of those issues is that of the phase response of the filters. Although there are many different types of high-pass and low-pass filters, let’s take a simple example of the filters used in a second-order Butterworth two-way crossover – a very typical choice for passive loudspeaker designers.
In a typical (second-order Butterworth) two-way crossover, the two bands are 180° out of phase as can be seen in the phase responses calculated using Equations 1 and 2 and plotted in Figure 1.
Figure 1: The Low Pass (black) and High Pass (blue) components of a two-way loudspeaker with a 2nd-order Butterworth crossover network (assuming that both drivers had perfectly flat magnitude responses). Notice that the phase responses are 180º apart at all frequencies.
This phase difference isn’t a big problem at frequencies that are far away from the crossover frequency (where the two components have the same magnitude) because the quieter one isn’t loud enough to cancel the louder one very much. However, the closer you get to the crossover frequency, the more their magnitudes are alike, and so the more they cancel each other. In fact, at the crossover frequency itself, their magnitudes are identical, and, because they are 180° apart in phase, their sum is completely cancelled, resulting in no output at all. (Keep in mind here that, for the purposes of this posting, I’m living in a perfect world where loudspeaker drivers are perfectly linear, there are no time-of-arrival differences between the two drivers at the listening position, and things like diffraction and reflections do not exist…) The total result would therefore look like the responses shown in Figure 2.
Figure 2: The total result of adding the two components in Figure 1, if the polarity of one of the two is not inverted.
To avoid complete cancellation at the crossover frequency where the two signals have identical magnitude, the polarity of the upper frequency band is typically inverted. (This is expressed as the negative sign at the beginning of the right-hand side of Equation 3). However, this solution results in a total sum that has a “bump” in its magnitude response as well as an allpass characteristic (meaning that the phase response of the total output is not a straight line).
Figure 3: The total result of adding the two components in Figure 1, if the polarity of one of the two (in this case, the high pass component).
At the listening position, on-axis to the loudspeaker, in this perfect world, these two frequency responses for the low frequency and high frequency sections (expressed in Equations 1 and 2) are combined using Equation 3.
In May 1977, Erik Bækgaard (pronounced something like BECK-gore), a manager of electronic engineering at B&O, published a paper in the Journal of the Audio Engineering Society where he described a solution to this problem associated with second-order Butterworth crossovers. Take a look at that last equation… Since we want the output to equal the input, we want F_Total(s) = 1. Therefore we can calculate what is missing from the s-plane equation to make that happen.
In other words, what we want is:
Bækgaard’s solution to this problem was to insert that component missing in Equation 5.
The frequency response of that resulting “MissingPortion” is a first-order bandpass filter with a phase response that is exactly between the phase responses of the high pass and low pass components, as can be seen in Figure 4, below. By adding that missing link to the system, the phase response of the entire system is corrected, so Bækgaard called the additional loudspeaker driver a “phase link” driver. So now, if we add the high pass, low pass and phase link components, we get the following:
… which means that the output of the system equals the input – exactly what we wanted!
The Low Pass (black), High Pass (blue), and Phase Link (red) components of a Uni-Phase loudspeaker.
Physically speaking, the solution to the problem was to add a third section in the crossover and an extra loudspeaker driver to fill in the missing phase component, linking the upper and lower frequency sections and avoiding the necessity for polarity inversion. This corrected the phase response of the entire system, eliminated the all pass characteristic and flattened the on-axis magnitude response, as can be seen in Figure 5.
The result was an entire range of loudspeakers, dubbed the “Uni-Phase” series, that was produced from 1976 to 1987. As is shown in Bækgaard’s paper, his system also improved the loudspeakers’ responses in the time domain, not only on-axis, but also off-axis in the vertical plane.
Figure 5: The total result if the three components shown in Figure 4 are added. Notice that the bump in the magnitude response from Figure 3 disappears, and the phase response becomes flat.
Figure 6, below, shows an example of one of the Uni-Phase loudspeakers. Without knowing what’s going on, it looks like a typical three-way loudspeaker with a woofer, midrange and tweeter. However, this is not the case. The woofer and the tweeter form a two-way loudspeaker and the middle driver is used as the Phase Link. So, instead of having two crossover frequencies, this loudspeaker has only one – and the peak in the bandpass response of the middle driver is at the same frequency as the crossover between the other two drivers.
Figure 6: BeoVox 120.2 (1983-1988) one of the many “Uni-Phase” loudspeakers with a woofer, tweeter and “phase link” driver.
Post-script: Of course, as I mentioned above, everything that I’ve said in this posting makes a lot of assumptions, not only about loudspeaker drivers, but cabinet effects and room acoustics. However, in order to keep things as simple as possible, it’s easier to isolate the issues described above as being the only problem with crossovers and loudspeakers. Sadly, this is not true…
“So how did the BeoLab 90 make us feel? When we closed our eyes in Bang & Olufsen’s special listening room, the pair of master reference speakers (#2 and #3 ever made)—along with the room—seemed to vanish the instant a song played. We weren’t listening to sound emanating from two specific points; instead, the Weeknd was singing his heart out right in front of us. Benny Goodman’s band performed an intimate set, and you could picture where each musician was sitting. The BeoLab 90’s ability to create such a lifelike three-dimensional sound stage is unparalleled when you’re sitting in the sweet spot. It certainly brings up the question of whether a speaker can be “too” good for the music—some now-classic albums weren’t necessarily well-recorded and mastered (think of when the Rolling Stones turned the basement of a rented French mansion into a makeshift studio slash drug den). But when all the variables align perfectly, the music engulfs listeners entirely and hits the guts. The result of such incredible technology and engineering happens to be a very visceral human experience.”
“Die Abbildung war phänomenal, jedes Instrument der gewählten Musik nahm ganz selbstverständlich den für sich bestimmten Platz im Raum ein, jedes Element war von Anfang bis Ende verfolg- und erlebbar. Aber nicht nur die Ortung verblüffte, auch die Detailgenauigkeit, mit der selbst kleinste Feinheiten bis zum erkälteten Backgroundsänger aufgedeckt wurde, sucht ihresgleichen.”
– modernhifi.de
“Vi kan bevidne, at effekten er besnærende. Højttalerne spiller sammen med lytterummet på en måde, vi ikke har oplevet før. Personligt har jeg aldrig hørt et mere holografisk realistisk lydbillede, hverken i eller uden for sweet spot. BeoLab 90 er også en fuldblods, fullrange-højttaler, der ikke overlader noget til tilfældighederne.”
– lydogbillede.dk
“I found that the size of the soundstage was consistently proportional to the size of the ensemble and the recording. I found that the bass was very well extended, taut, and satisfying. Most of all, I was impressed by the prototypes’ reproduction of detail throughout the audio band, and the uniformity of that quality across the soundstage.”
– Kalman Rubinson, Stereophile magazine (print version, October, 2015)
“Wohl noch nie haben Lautsprecher die musikali schen Akteure so scharf ins Wohnzimmer projiziert. Ganz gleich, ob grosse Orchester oder kleine Jazz-Formationen – jedes ein zelne Instrument hat seinen exakten Platz im virtuellen Raum, der auch seine Tiefen dimension verblüffend genau zu erkennen gibt: Der Hörer kann zum Beispiel fast in Zentimetern abzählen, wie weit das Schlag zeug hinter dem Kontrabass placiert ist. Dass der Beolab 90 auch für schwärzeste BassTiefe, überbordende Dynamik und feinen, luftigen Obertonglanz steht, müssen TestHörer der Vollständigkeit halber natür lich ebenfalls zu Protokoll geben, aber das eigentlich Spektakuläre des Lautsprechers ist tatsächlich seine überragende räumliche Abbildung.”
– NZZ am Sonntag 18. Oktober 2015
This video is from the press event, recorded by the journalist from recordere.dk.
Jakob Dyreby designed/engineered the DSP that controls the beam width (or “directivity”, if you’re a geek) of the BeoLab 90. In this video, he discusses some of the processes involved in arriving at those filters.
One of the questions that has come up with regards to the specifications of the new BeoLab 90 is about the size of the woofers. The specifications state that it has one 13″ front woofer and three 10″ woofers for the sides and rear. However, if you look in the technical specifications in the Technical Sound Guide, you’ll see that the “effective diameter” of the front woofer is 260 mm (about 10″) and the remaining woofers is 212 mm (about 8″). Why is there a discrepancy?
The difference is in how a woofer – or any loudspeaker driver – is measured. When you say 13″ woofer, the measurement is the external diameter of the circular metal frame around the front of the driver. If you look on the first page of the datasheet shows that this diameter is 320 mm for the BeoLab 90’s front woofer – so it’s a 13″ driver. (I’ve copied the technical drawing from the datasheet below – see the two dimensions given on the left side of the drawing.) However, this diameter includes non-moving parts (at least they should not move – they’re screwed to the enclosure).
If you measure the moving parts of the woofer, then we have to decide on where to measure – what is the actual diameter of the diaphragm? Normally, the way to measure this is from the high points on the surround that encircles the diaphragm and connects it to the loudspeaker frame. As you can see in that same drawing, this diameter is 258.8 mm, which, in the “official” datasheet is rounded to 260 mm.
The technical drawing of the Scan-Speak 13″ Revelator front woofer from the official datasheet.
Most loudspeaker projects at Bang & Olufsen are conceived in the design or the Product Definition department. This means that someone decides something like “we’re going to make a loudspeaker, this size with this design” and then the project arrives at the Acoustics Department to find out whether or not the idea is feasible. If it is, then it continues through the development process until we reach the end where the is a product in a store. A better description of this process is in this posting.
BeoLab 90 was different. Instead of being a single project that began, evolved, and ended, it was more like a number of little streams coming together to form a river. Each stream was an idea that contributed to the final product.
One of the early “streams” was an idea that was hatched in the Acoustics Department itself around 2009. I went to the head of the department at the time, and offered to make a deal. If I were to pay for all the components personally, could I use my work hours and B&O resources (like the Cube) to build a pair of loudspeakers for home. These would be a “one chair – no friends” style of loudspeaker – so it would not really be a good candidate for a B&O loudspeaker (our customers typically have friends…). In return, I would keep the loudspeakers in the listening room at B&O for an extended time so that we could use them to demo what we are capable of creating, without our typical restraints imposed by design, development time, size, “normal” product requirements (like built-in amplifiers and DSP), and cost of components.
By early 2011, these loudspeakers were built (although not finished…) and ready for measurements and tuning. The photo below shows the “raw” loudspeaker on the crane in the Cube going out to be measured.
Figure 1. One of the original “seeds” that began the discussions about whether BeoLab 90 should be considered. The woofers are identical to the BeoLab 9 10″woofer. The mid-bass is a Scan-Speak 5 1/4″ Illuminator. The tweeter is a Fountek Pro5i ribbon.
Those loudspeakers lived in Listening Room 1 for about a year. We’re bring people in for a “special demonstration” of a loudspeaker behind the curtain. The general consensus was that the loudspeakers sounded great – but when the curtain was opened, many people started laughing due to the sheer size (and the ugliness of my design, apparently…) of the loudspeakers.
Figure 2: The finished version.
The second “stream” was an idea that was born from Gert Munch’s goal of building a loudspeaker with a smooth power response as well as a flat on-axis magnitude response. The experiment was based on a “normal” two-way loudspeaker that had an additional side-firing dipole on it. The basic idea was that the two-way loudspeaker could be equalised to deliver a flat on-axis response, and the dipole could be used to correct the power response without affecting the on-axis sound (since the on-axis direction is in the “null” of the dipole). For more details about this project, please read this post.
Figure 3: The “Shark Fin” loudspeaker – an experiment attempting to independently tune the power response without changing the on-axis frequency response. See this posting for more information.
Figure 4: The “shark-fin” loudspeaker – side view.
Figure 5: The “shark-fin” loudspeaker during the listening portion of the experiments in Listening Room 1.
After the “shark fin” experiment, we knew that we wanted to head towards building a loudspeaker with some kind of active directivity control to allow us to determine the amount of energy we sent to the nearby walls. Two members of the Acoustics Department, Gert Munch and Jakob Dyreby, had been collaborating with two graduate students (both of whom started working at B&O after they graduated), Martin Møller and Martin Olsen on exactly this idea. They (with Finn Agerkvist, a professor at DTU) published a scientific paper in 2010 called “Circular Loudspeaker Arrays with Controllable Directivity”. In this paper they showed how a barrel of 24 small loudspeaker drivers (each with its own amplifier and individualised DSP) could be used to steer a beam of sound in any direction in the horizontal plane, with a controlled beam width. (That paper can be purchased from the Audio Engineering Society from here.)
Figure 6: The results of a research project using 24 drivers placed radially symmetrically on a barrel. The top box is the DSP which feeds the bottom two boxes containing the amplifiers, one for each driver. The assembly is placed on the crane in the Cube.
The next step was to start combining these ideas (along with other, more developed technologies such as Thermal Compression Compensation and ABL) into a single loudspeaker. The first version of this was an attempt to reduce the barrel loudspeaker shown in Figure 6 to a reasonable number of loudspeaker drivers. The result is shown below in Figure 7.
Figure 7: Prototype 1. The idea for this version was a loudspeaker that could act a little like a lighthouse. The beam would be rotatable in any direction. There are 6 tweeters and 6 midranges arranged in a hexagon. The 4 woofers are arranged as a square. This version failed due to the lobing of the drivers – essentially, they are too far apart to result in an adequately-controlled beam.
This first prototype had a hexagonal arrangement of tweeters and midranges (6 of each) and a square arrangement of woofers. Each driver had its own DSP and amplification with customised filters to do the “usual” clean-up of magnitude response in addition to the beam steering much like what is described in the AES paper.
Unfortunately, this version was not a success. The basic problem when trying to do directivity control actively is that you need the loudspeaker drivers to be as close together as possible to have control of the beam width in their high-frequency band – but as far apart as possible to be able to control their low-frequency band. In the case of prototype 1, the drivers were simply too far apart to result in an acceptably constant directivity. (In other words, the beam width was different at different frequencies.) So, we had to try to get the drivers closer together.
For the second prototype (shown below in Figure 8, 9, and 10) we decided to try to forget about a steerable beam – and just focus (forgive the pun) on a narrow beam with constant directivity (the same beam width at all frequencies). In addition to this, we experimented with a prototype 8mm supertweeter that would take care of the band from about 15 kHz and up.
Figure 8: Prototype 2. This version did not have a rotatable beam, but it did have the capability of creating a very narrow beam due to the cluster of 5 midranges and 5 tweeters. Notice as well the cluster of prototype 8 mm super tweeters on the top. Also note that this version used 3 woofers instead of 4.
Figure 9: Prototype 2: side view
Figure 10: Prototype 2: Back view
Although Prototype #2 sounded great in the sweet spot, it lacked the versatility of the first prototype. In other words, it was an amazing loudspeaker for a person with one chair and no friends – but it was not really a good loudspeaker for sharing… So, we started working on a third prototype that merged the two concepts – now called “Beam Width Control” and “Beam Direction Control”. The result in shown below in Figures 11, 12, and 13.
As you can see there, the “cluster” of 3 tweeters and 3 midranges comes from prototype 2 – but we re-gained side-firing drivers and rear-firing drivers to be able to steer the sound beam in either of 4 directions. The Beam Width could only be controlled for the front-firing beam, since it is a product of the cluster. You’ll also notice that the super tweeter was still there in this prototype. However, we also changed to a different tweeter and were starting to question whether the extra 8mm driver (and its amplifier, DAC and DSP path) would be necessary.
Figure 11: Prototype 3. This version had almost as narrow a beam as the previous prototype, with the ability to steer the beam in 4 directions. The wired hanging from the front of the woofer cabinet connect to a thermal sensor on the magnet. At this point, we were still wondering whether to use an 8 mm supertweeter (on the top). It was eventually decided to not use this driver, since the Scan-Speak tweeters measure up to 40 kHz.
Figure 12: Prototype 3: Side view
Figure 13: Prototype 3: back view
One thing that any good acoustical engineer knows is that corners cause diffraction. This is well-known at B&O as you can read here. Looking at the enclosure for the midranges and tweeters in the three previous figures, you can see many flat surfaces and corners – which, we assumed, were bad. So, we set about on an informal experiment to find out what would happen if we smoothed out the corners in an effort to reduce diffraction – or at least to change it. This was initially done by applying putty to the MDF enclosures and measuring the off-axis response of the result. One example of this (in progress of bring coated with putty – this was not the final version – it’s just there to show the process) is shown below in Figure 14.
Figure 14. One of the tests for Prototype 3 was to find out whether there was a problem with diffraction due to the hard corners in the top cluster. Smoothing the shape with putty was done by hand.
Surprisingly, we found out in these tests that the “smoothing” of the structure around the midranges and tweeters either made no difference or made things worse. So, we continued on, knowing that the final result would be “smoother” anyway…
While that work was going on in the Cube, a third “stream” for the project was underway – the development of the Active Room Compensation algorithm. In the early versions, it was called “ASFC” or “Active Sound Field Control” – but as time went on and the algorithm evolved, we changed to a different system and gave it a different name. Photo 15, below, shows the listening room during one of the tests for the original algorithm. I count 9 microphones in there – but there may be more hiding somewhere.
Figure 15: One of the early Active Room Compensation experiments. Notice the many microphones distributed around the room. (I count at least 8 mic’s there…)
All of the prototypes shown above are just loudspeaker drivers in MDF enclosures. All of the DSP and amplification (in a worst-case, 17 channels in total per loudspeaker) were outside the loudspeakers. In addition, the amplifiers needed active cooling (a fancy way to say “fans”) so they had to be in a different room due to noise. The photo below shows the rack of DSP boxes (on top) and amplifiers (4 channels per box) used for the prototyping.
Figure 16. The electronics needed to drive the prototypes. The left stack is for the left loudspeaker, the right stack is identical. The top box is the DSP prototype board with 1 analogue input channel and 20 line outputs. The remaining 5 boxes contain ICEpower amplifiers – 4 channels per enclosure.
While this equipment was being used to evaluate and tune the complete prototypes, a parallel project was underway to find out whether we could customise the amplifiers to optimise their behaviour for the use. For example, if you know that an amplifier will only be used for a midrange driver, then it doesn’t need to behave the same as if it were being used for a full-range loudspeaker. I’ll describe that development procedure in a future blog posting, since it’s interesting enough to deserve its own story.
Finally, we were at a point where we built a first prototype of the “real thing”. This was hand-built using 3D-printed parts and a lot of time and effort by a lot of people. The first example of this stage is shown below on the crane in the Cube, sitting next to Prototype #4 for comparison. Notice that, by now, we had decided that the supertweeter was unnecessary, since the Scan-Speak tweeter we’re using was reliable up to at least 40 kHz. The only significant difference between the 4th prototype and the mechanical sample is that the wooden version has only one tweeter and one midrange pointing directly backwards. The “real thing” has two, aimed slightly towards the Left Back and Right Back. (See the Technical Sound Guide for more detailed information about this.)
Figure 17 Prototype 4 (on the floor) was basically the same as Prototype 3, but with a different woofer arrangement. Instead of 4 matched 13″ drivers, this version used three 10″ drivers and one 13″ in the front. On the crane is the first sample with the original mechanical design.
Once the measurement of the first mechanical samples were done and the correct filters programmed into it, it was time to move a pair into the listening room to see (or, more importantly, to hear) if they performed the same as the wooden prototypes. The first setup of device numbers 2 and 3 (the first one stayed with the electronics team for testing) in Listening Room 1 in Struer is shown below in Figure 18. For reference, the room is 6 m deep x 5 m wide – and that’s a 55″ BeoVision 11 on the wall.
Figure 18: Samples #2 and 3 in the listening room (they’re still there today, since this is the “master reference” pair). These have internal electronics and did not require the external DSP or amplification.
When we did the measurements on the samples shown in Figure 18 – both in the Cube and in the listening room, we could see that there was an unusual (and unexpected) dip in the on-axis magnitude response of about 1 dB at around 8o0 Hz. Unfortunately, it did’t seem to be easily correctable using filtering in the DSP, which meant that it was probably the result of a reflection somewhere off the loudspeaker, cancelling the direct sound at the listening position. After a day or two of playing with putty placed in various locations around the loudspeaker, we found that the problem was caused by a reflection off the “shelf” just below the face of the top unit. That can be seen in Figure 19, below.
Figure 19: The “shelf” below the midranges caused a reflection at the listening position from the top midrange. The result was an approximately 1 dB dip at about 800 Hz. This was redesigned, resulting in a new shape, shown below.
The way to correct this problem was to bring the height of the shelf up, which also meant that it was closer to the face of the top cluster. (Note that the front panel is missing in Figures 19 and 20 – the actual face is the pink panel seen in Figure 22.) This fixed the problem, but it meant changing the mould for the aluminium enclosure. In the meantime, while that change was happening, we were able to 3D-print an insert of the same shape that could be used for the listening reference pair of loudspeakers. This meant that we didn’t have to wait for the new aluminium versions to start tuning.
Figure 20: The repaired “shelf”. The pink 3D-printed portion raised the shelf closer to the midranges and is almost flush with the faceplate that is mounted in front of the loudspeaker cluster. The bad paint job on the top loudspeaker enclosure is a sealant. The enclosure is 3D-printed plastic which (we found out) is not airtight, to it had to be sealed to eliminate leaks.
Of course, the electronics team developed their components on a test bench, piece by piece. Eventually, all of those pieces came together into a single unit (minus the loudspeaker drivers and enclosures) which could be used for testing and software development. An example of one of those test boards (actually, the first one of its kind to be made – and one of the few with all of the amplifiers attached…) is shown below in Figure 21.
I’ll probably show some better photos of the DSP board in a later posting.
Figure 21: The entire electronics assembly for the BeoLab 90. The top portion in the brown MDF box is the power supply. Moving down the photo, the 4 PCB’s are the woofer amplifiers. Next are the input and DSP boards. On the bottom are the 14 ICEpower amplifiers. The circle not he right is the light ring comprised of 72 LED’s.
Finally, everything came together into a product that, acoustically and electrically, was identical to the production model. This is the version that we use for sound design. It’s shown (about to go out into the Cube for yet another round of measurements) in Figures 22 and 23, below.
Figure 22: The early sample, on the crane in the cube, undergoing measurements to ensure that it is ready to start the sound design process.
B&O received a question on one of its social media sites this week, and I was asked to write up an answer. The question was:
Hi Bang & Olufsen
I just wanna be sure of a myth that’s been going around my audio community recently. The myth is that condenser microphones are more prone to produce feedback than dynamic microphones as a result of higher sensitivity in (and reproduction of) the treble.
Is this true of false? Thanks.
The short answer
This is false.
The long answer
Feedback happens when you have a system where the input to a microphone is amplified and sent to the output of a loudspeaker, AND the output of the loudspeaker is received at the microphone at a level loud enough to cause the signal to get louder (instead of quieter) each time it circulates through the system. The result is a “howling” or “squealing” sound from the loudspeaker. This effect will happen first at whatever frequency has the highest gain (amplification) in the system.
That frequency could be due to a peak in the magnitude response of the microphone or the loudspeaker, or some acoustical effect of the room (such as a room mode), or something else. (For example, if you put your hand over the microphone diaphragm, making a resonant cavity, you could result in a peak in the total system’s magnitude response that would not be there if you moved your hand away.)
So, the basic problem is one of signal gain. The higher the gain (or “amplification”) of the signal, the more likely you are to have feedback. The question is: what determines this total loop gain in a typical sound reinforcement system?
the sensitivity of the microphone
This is frequency-dependent, since the magnitude response of the microphone is likely not perfectly flat.
It is also spatially dependent. If you have an omnidirectional microphone and a cardioid microphone that have the same sensitivity on-axis (in front of the microphone), they will be very different behind the microphone. Also note that this directional pattern is also frequency-dependent. A cardioid is not a cardioid at all frequencies…
the gain of the microphone pre-amplifier
the additional gain applied after the microphone preamplifier
this may be frequency-dependent, like an EQ applied to the microphone signal, or an EQ applied to the entire mix sent to the loudspeaker amplifiers
the gain of the loudspeaker amplifier(s)
the sensitivity of the loudspeaker(s)
the distance between the loudspeaker(s) and the microphone
the radiation pattern of the loudspeaker(s)
many loudspeakers are directional, so they’re louder in front than behind
this is also frequency-dependent. Bass is usually omnidirectional, high frequencies are usually directional
the orientation of the microphone relative to the loudspeakers (i.e. is the loudspeaker in front of, or to the rear of the microphone), especially if the microphone is directional (like a cardioid or a hypercardioid pattern)
the coupling to room modes due to
the strength of the modes themselves (a function of the room construction and its materials)
the location of the loudspeaker(s)
the location of the microphone
There may be some other things – but that’s certainly enough to worry about.
IF you have two microphones, one is a dynamic microphone and the other is a condenser microphone, and they both have the same polar patterns, the same magnitude responses, the same sensitivities, they’re both in the same location in the room with the same orientation to the loudspeakers, and all other components in the system are identical, THEN the risk of getting feedback with the two mic’s is identical.
IF you have two microphones with different polar patterns, different magnitude responses, different different sensitivities, etc. etc. THEN the risk of getting feedback with the two mic’s is different. Whether the basic electromechanical construction is based on a condenser or a dynamic design is not the cause of the difference.
That said, it is true that microphones (both condenser and dynamic) are built with particular uses in mind. For example, (dynamic) Shure SM58 is designed to be tolerant of noises caused by it being hand-held (the diaphragm assembly is vibration-isolated from the housing) this is not true of a (condenser) AKG 451 which is designed to be mounted on a stand and not touched while you’re using it. However, this difference is not caused by the fact that one is dynamic and the other is a condenser – it’s a result of the mechanical designs of the microphones housing the “business end” of the devices. (Note, however, that this example has nothing to do with feedback – it’s just an example of microphones being designed for different purposes.)
It is also true that many condenser microphones have a magnitude response that extends to the high frequency bands with less roll-off than many dynamic microphones (there are exceptions to this statement – but I used the word “many” twice…). And, a higher sensitivity in any frequency band will result in a greater risk of feedback. However, this increased risk is a result of the magnitude response of the microphone – not its electromechanical construction. If you have a condenser microphone with a roll-off in the high end (say, an older, large-diaphragm mic, especially off-axis) and a dynamic microphone with an extended high-frequency range (i.e. a ribbon microphone, which typically has a flatter high-frequency response than a moving-coil microphone), then the dynamic will be at higher risk of feedback.
So, like I said at the start – the myth is false. If you get feedback in your system, it’s because
the person running the system was not paying attention to the gain
the person with the microphone moved too close to a loudspeaker while the person running the system was not paying attention to the gain
Either way, it’s the fault of the person controlling the system – not the construction of the microphone. As the old saying goes: “It’s a poor craftsman that blames his tools.” Or, as a friend of mine once told a class he was teaching: “If it’s too quiet, you turn it up. If it’s too loud, you turn it down. That’s the way I remember it.”
