Sphere® Technology Whitepaper

Townsend Labs Sphere L22 with 3D Sphere - 90 Degrees Off-AxisOver the years there have been a handful of products designed to model microphones. Typically these products took a conventional studio microphone, such as a Shure SM57, as a source and applied processing to make it sound like some other microphone, such as a Neumann U47. Some microphone modeling products have used a custom designed microphone as the source, but in essence, these are still just conventional microphones with a known response.

These products apply various forms of filtering to attempt to “morph” one microphone into another. As any audio engineers can attest, no amount of EQ or filtering can make one microphone sound like another. This is largely because EQ doesn’t take into account the three-dimensional polar response of the microphone. The best you can do with EQ, or any other type of single-channel processing, is model the on-axis response of the microphone, but the sound of a microphone is of course also greatly dependent on the off-axis response.

Even if these other products can model the on-axis frequency response, in general, it can only be done at a single distance, because proximity effect varies with different microphones. Proximity effect causes the well-known bass boost in directional microphones that increases the closer you get to the microphone.

 

Now For Something Completely Different

In contrast, Sphere® technology models the full three-dimensional response of a wide range of microphones, including proximity effect. It is able to do this by independently sampling the pressure and velocity of the soundfield at a point in space, which allows the modeling algorithms to reconstruct the transient and frequency dependent polar response of a microphone.

Essentially all microphones have polar patterns that are frequency dependent. This has a lot to do with what gives a microphone its character. When a microphone’s stated pattern is cardioid or omni, for instance, that is only its nominal response. The actual polar response is much more “complicated” and varies with frequency.

For example, typical cardioid large dual-diaphragm condenser mics are cardioid in the 1 to 2 kHz region and tend towards omnidirectional at low frequencies. At high frequencies, these mics may move towards figure-8 or other more complicated patterns.

U87ai-CardPolarPattern365x353

Figure 1: U87ai Cardioid Polar Diagram

That large diaphragm condensers move towards omni at low frequencies helps explain why they make for such good vocal mics. Omni microphones tend to be much less sensitive to plosives and don’t have proximity effect. This property means that these microphones are reasonably directional over most frequencies, but have less plosive sensitivity at low frequencies where it really matters. On the other hand, small single diaphragm condensers tend towards figure-8 at low frequencies and could explain part of the reason why they are not often used on vocals.

Figure 1 shows the published polar pattern of a Neumann U87ai in the cardioid setting at 125 Hz (blue), 1 kHz (red) and 4 kHz (green). Notice how only at 1 kHz is the pattern actually cardioid. At 125 Hz the pattern approaches omnidirectional. At 4 kHz the pattern is roughly hypercardioid.

By using a dual-capsule microphone it becomes possible to more completely capture the soundfield—including the directional and distance information otherwise lost with a conventional single-channel microphone. This additional information makes it possible to reconstruct how different microphones respond to the soundfield.

 

Microphone

To capture the directional and distance information the Sphere microphone has two back-to-back capsules with independent analog XLR outputs. This is why we named the microphone “Sphere L22” (two diaphragms and two outputs). With a conventional microphone that has a single output, it is impossible to capture any of this directional or distance information.

In some respects, the Sphere microphone is related to an Ambisonics soundfield microphone. An Ambisonics microphone has four capsules/outputs and can sample the pressure response as well as the velocity response in three axes (front/back, left/right, up/down). This allows the microphone to encode, height, depth, and width.

The Sphere microphone with two capsules is able to sample the pressure response and velocity response in the front/back dimension (when looking at the microphone on-axis). The two output design is most closely related to the handful of dual back-to-back capsule mics, such as the Sennheiser MKH 800 Twin, Pearl TL-44, and the Microtech-Gefell UM 930 twin. The Schoeps Polarflex system is similar but uses discrete omni and figure-8 capsules to derive different frequency-dependent patterns.

The cool thing about these microphones is the ability to change the polar pattern in post-production by recording both outputs. Summing the two outputs creates an omni pattern. Subtracting the rear output from the front creates a figure-8 pattern. Just using the front output gives a cardioid pattern. With the correct mixing levels, any other first order pattern is possible, including hypercardioid, supercardioid, and subcardioid.

Sphere also allows for changing the polar pattern after recording, but it can also change the mic type, filter, proximity, axis, and various other settings by using the Sphere plug-in to process the two-channel audio. In other words, you can Re-Mic your tracks long after they were recorded. More on that in a moment.

Sphere microphones use capsules and electronics with very tight tolerances so that the sound is consistent from mic to mic. This also means that any two Sphere microphones can be used as a stereo matched pair with excellent results.

For Sphere, it’s also important that the sensitivity level of the front and rear outputs are closely matched. The gain tolerance of the Sphere L22 circuit is matched to within at least 0.05 dB between the front and rear channels. To achieve this level of accuracy the electronics were custom designed for high consistency and linearity.

Sphere L22 PCB Rev-A

Typical condenser microphone circuits are not especially consistent, partly because they use discrete components, such as FETs, which have a relatively high degree of variation. Even with careful matching and selection of components to minimize variation, this can still be an issue.

Instead, the Sphere microphone uses a precision op-amp based circuit to achieve high linearity and consistency, as well as low noise. Sphere uses relatively expensive 0.1% resistors to ensure the gain matching between channels is within a small fraction of a dB.

 

Preamp

For this to all work properly the microphone outputs need to be connected to a mic preamp with equal gain settings on both channels. For this reason, we recommeApollo_Twin_Topnd using preamps with precision digital gain control or analog stepped gain adjustment so the channel levels can be easily matched.

Preamps with digitally controlled gain include those found in Universal Audio’s Apollo series interfaces (except Apollo 16), the Presonus Studio 192, and other products from companies including Apogee, RME, MOTU, Avid, Roland, and Focusrite. Apollo is a particularly good choice because the Sphere plug-in can run on the internal DSP with ultra-low latency processing.

There are also pure analog designs with stepped attenuator switches, such as a Neve 1073 or a Millennia HV-3D, which work nicely with Sphere. Not all preamps with stepped attenuators are perfectly matched, but most are plenty accurate.

Using calibration mode on the microphone you can easily verify the gain matching of your preamp. When calibration mode is switched on the front capsule output is fed to both XLR outputs, so there are identical signals on both channels. This makes it easy to use the metering in your DAW or in the plug-in to verify and set the gains of preamp channels if using a non-calibrated mic preamp.

To help select which preamp to use, we compiled an extensive list of over 250 mic preamps that have precision stepped gain adjustment. http://www.townsendlabs.com/prelist

 

Plug-in

The plug-in is where much of the interesting work happens. The DSP algorithms in the plug-in process the two-channel audio to create the microphone modeling. The algorithms apply complex filtering to each output of the microphone to derive the desired three-dimensional response. In the telecommunications industry, this is often referred to as “beam-forming” or “spatial-filtering”.

The front microphone channel is routed to the left input of the plug-in the rear channel is routed to the right. It is important to keep the gain from the outputs of the microphone to the inputs of the plug-in matched between channels.

The modeling algorithms are designed to be as robust as possible to a mismatch in gain between channels. Although we recommend a 0.1 dB gain matching for optimal performance, 1 dB or so of mismatch usually works well enough. A difference, if audible, tends to be most noticeable in the off-axis response. For example, with a cardioid pattern, the rejection of sound coming from directly behind the microphone might not be quite as good if the gain is mismatched slightly.

The plug-in also provides a number of other unique and useful features, such as adjustable proximity effect, axis, capsule alignment and Off-Axis Correction™, which take the Sphere microphone way beyond what is currently possible with any conventional microphone. And as with the modeling all of these adjustments can be made before or after tracking.

 

Proximity Effect

The bass response of all directional microphones increases the closer the sound source is to the mic. This is commonly referred to as "proximity effect". Proximity effect occurs because the figure-8 component of the microphone response increases at low frequencies as the distance to the source is decreased.

The Proximity control works by adjusting the amount of figure-8 response at low frequencies and thereby directly changing the amount of proximity effect produced by the mic. This actually results in the polar pattern of microphone changing at low frequencies, in a way that is similar to physically changing a microphone capsule to have more or less proximity effect. Turning Proximity to the left makes the pattern more omnidirectional at low frequencies and turning it to the right makes it more figure-8.

Because Proximity only changes the figure-8 component of the response it has no effect on omni microphone models. Partly for this reason, a Proximity EQ control that allows the user to cut or boost low bass content in a frequency range that is similar to where proximity effect occurs. Unlike Proximity, this control always has an effect, even for omni pattern types.

For a complete explanation of proximity effect Shure has an excellent article titled Why does Proximity Effect Occur?

 

Axis Shift

It’s common for recording engineers to rotate a microphone slightly off-axis to achieve a different tonal quality. Sometimes this is done to tame high frequencies and reduce sibilance of a vocal.

This effect primarily occurs at high frequencies (typically above 5 kHz) where the wavelength of sound is comparable to the size of the microphone’s diaphragm. For directional, microphones the boost in bass frequencies due to proximity effect is also reduced when used off-axis. In fact, at 90 degrees off-axis there is essentially no proximity effect.

The Axis control allows the user to virtually shift the microphone axis so that the off-axis response of the microphone becomes the on-axis response. For example, if the user selects an Axis of 45 degrees then the on-axis frequency response of the microphone will match the frequency response of the microphone being modeled as if it was turned 45 degrees.

This method does not actually shift the polar response as occurs when physically rotating a microphone, but often it is more useful for the polar pattern to stay constant and only shift the tonality of the mic.

For more information on this topic check out our blog Dark Side of the Mic.

 

Stereo

Another powerful feature of Sphere is that you can record in stereo with a single microphone. Sphere achieves this by creating two virtual mics with one pointing forward and the other pointing backward.

For stereo recording, the mic is rotated 90°, so the front output becomes the left channel, and the rear output becomes the right channel. This provides a 180° (back-to-back) capsule arrangement. To use the microphone in stereo mode, you must instantiate the Sphere 180 plug-in.

 

SingleMicStereoDrumKitBackground1-Fade

Because the left and right diaphragms are coincident, perfect phase coherency is always maintained, even when the channels are summed to mono. For a coincident mic technique, this produces the widest stereo spread possible, but typically leaves a hole in the center of the stereo field.

This occurs because the center of the stereo field corresponds to 90° off-axis from each capsule, which is where the high-frequency response is typically lowest. Nonetheless, if the goal is to produce a wide stereo spread with room in the middle for center-panned material, then this is a benefit, not a detriment. The best results are usually achieved by setting the polar pattern to somewhere between cardioid and supercardioid, with supercardioid providing a wider stereo field.

For more information see the Stereo Recording chapter in the Sphere User Guide.

 

Simultaneous Mic Models

Often recording engineers use two or more conventional microphones on a source, such as a guitar amp, to allow blending of different responses, but because the microphones can never be located at the same point in space phase problems can be in issue. Typically careful alignment of microphone diaphragms is necessary to get good results.

The Sphere plug-in has two virtual microphone models that can be blended together to produce a mono output or with the Sphere 180 plug-in are panned hard right and hard left to produce a stereo output. Because the virtual models are derived from a single physical Sphere microphone this has the added advantage that phase coherency is maintained.

If the Mix control is set to 50% then the response will be an equal combination of both Mics 1 and 2. At 0% the response is fully Mic 1 and at 100% the response is fully Mic 2. Blending the virtual models not only changes the combined frequency response but also changes the polar pattern as occurs when blending two conventional microphones.

It’s interesting to look at cases where combining distinctly different patterns yields a totally new pattern. For example, combining an omni mic with a figure-8 mic with a 50% Mix will produce a cardioid pattern. This occurs because the back side of the figure-8 has opposite phase as the front which cancels with the omni pattern and greatly reduces the rear pickup.

In effect, all cardioid mics can be thought of as equal portion of omni and figure-8 patterns. By combining a cardioid mic with a figure-8 mic in equal proportion a hypercardioid pattern is produced. And combining a cardioid mic with an omni mic gives a sub-cardioid pattern.

 

Alignment

Although exact phase alignment is generally desirable, sometimes slight misalignment of phase can produce great results. This is often the case when miking a guitar amp because slight misalignment can cut some of the harsh high-frequencies in a way that is not possible with a typical EQ.

The Align control in the Sphere plug-in allows for up to 2 cm of adjustment between the two virtual microphone models. If, for example, the distance is set to 1 cm then the relative phase is adjusted as if Mic 2 is 1 cm farther from the source (as shown in Figure 2). If the distance is -1 cm then Mic 2 is moved closer to the source, relative to Mic 1.

SM57-AlignControl-Diagram

Figure 2: Physical representation of the Sphere plug-in's Align control

It's important to keep in mind that this control only adjusts the relative phase between each mic model. If the Mix control is set such that only one model is being used then the control has no effect.

 

More Polar Patterns

It is possible to select polar patterns not available with the original microphone because the modeling can synthesize intermediate patterns. For example, you can create an 87 model with a super or hypercardioid pattern.

Most multi-pattern microphones use a dual-diaphragm capsule (invented by Braunmuhl and Weber in the 1930s) and internally within the microphone electrically combine the two capsules in different proportions to produce the various patterns. When synthesizing patterns with Sphere the results are very close to what occurs if the microphone was electrically modified to support those patterns.

We have tested this by comparing synthesized patterns with measured patterns for various microphones. For example, with a 414 we compared a synthesized hypercardioid model produced from an equal combination of cardioid and figure-8 with a hypercardioid model made directly from measurements of the hypercardioid setting and got nearly identical results.

The plug-in has nine available patterns for most microphone types, and patterns available in the original mic are highlighted in blue. But keep in mind that it is not possible to generate other patterns from just omni or from just figure-8.

For example, there is no meaningful way to take a figure-8 ribbon mic and make it omni or cardioid, at least not in a way that has any physical basis in reality. In these cases where it is not possible to generate intermediate polar patterns, the pattern control in the Sphere plug-in is disabled, and only one pattern is allowed.

For more information on this topic check out our blog Plentiful Polar Patterns.

 

Off-Axis Correction

Once we have this additional information is captured from the soundfield, it also becomes possible to correct anomalies in the polar response of the microphone using our proprietary DSP algorithms. Difficulties can arise because of the frequency dependence of polar patterns. As already mentioned, a typical microphone with a cardioid pattern will only be cardioid in a narrow range of frequencies. These inaccuracies in the polar pattern can lead to excessive bleed or colored room pickup.

In Figure 3 the frequency response from directly behind the mic (180 degrees off-axis) of a U87ai is compared with Off-Axis Correction. The green line in this plot shows the frequency response of a Neumann U87ai at 180 degrees off-axis and 1 meter from the source with ⅙ octave smoothing. The blue line shows the response of the Sphere microphone with Off-Axis Correction™ enabled.

U87ai Cardioid 180 Degrees Off-Axis vs Sphere Off-Axis Correction

Figure 3: U87ai Cardioid vs Sphere Off-Axis Correction at 180 Degrees Off-Axis (directly behind the mic)

As you can see the Sphere response is much flatter. The U87 response varies by as much 16dB, whereas the Sphere mic only varies by about 6dB and overall the fluctuations are smoother. Not surprisingly audio coming from 180 degrees off-axis will sound very colored with the Neumann. With the Sphere mic the result is much smoother and mostly what you hear is the natural room ambience.

A key thing to understand about Off-Axis Correction is that when enabled it overrides the pattern selection of Mic 1 and Mic 2, but it still retains the on-axis tonal qualities of these mic models, which can change dramatically depending on what patterns are set for Mic 1 and Mic 2.

See the Off-Axis Correction chapter in the Sphere User Guide for more information on this subject.

 

Flat 20 to 20

If the user so desires, it is possible to create a microphone with an essentially flat response. To do this the user simply selects the “Flat On-Axis” mic model so the actual on-axis frequency response is flat and linear phase from 20 Hz to 20 kHz. This “Custom” model is omnidirectional, but it can be made directional by enabling Off-Axis Correction.

Because of proximity effect, it is impossible to make a directional microphone with a completely flat response at all distances, but Off-Axis Correction allows the user to set which distance the mic will have a flat response, as specified by the On-Axis Distance control. For many practical applications, this actually results in a flat response, as long as all sources are at approximately the same distance from the mic.

For more information on this topic check out our blog Why Directional Mics are Inherently Colored.

 

Accuracy

The Sphere L22 microphone uses a large diaphragm condenser capsule and is optimized for modeling large diaphragm mics most accurately. For vintage large diaphragm condenser microphones, in particular, our target is to meet or exceed the accuracy of currently available reissues, clones or other modern versions of the corresponding mic. In most cases, we achieve or surpass that standard of accuracy.

For other types of microphones, such as dynamics, ribbons, and small diaphragm condensers, we refer to these models as “hybrids” because they combine the response of those mics with a little bit of the quality of a large diaphragm condenser. These models are generally accurate when used on-axis. Beyond about 45 degrees off-axis the models may differ substantially, but the overall polar pattern is still maintained. Generally, this means that close miking will be more accurate than distant miking.

Keep in mind for sources that are large relative to the mic capsule, such as a guitar amp speaker, a fair amount of the sound pickup will be off-axis. Also, models of ribbon mics with long ribbons can sometimes be less accurate at very close distances because different sections of the ribbon pick up different sounds from the source.

The Sphere microphone is designed to have a very low amount of noise. For the Sphere L22 microphone the equivalent noise spec is 7 dB-A. This approach provides the user with the lowest noise possible, even if that is substantially lower than the original modeled mic.

While Sphere does model the subtle harmonic characteristics of microphones, including transformers, tubes and discrete electronics, it does not model the full overload characteristics. The Sphere microphone is designed to have a large amount of headroom and dynamic range, so it can handle a very soft and very high sound pressure levels. Our approach provides maximum headroom even if the original microphone being modeled didn’t have as much headroom. We believe this is a major benefit of the technology since an overloaded microphone is rarely a desirable thing.

Handling noise and plosive sensitivity are not directly modeled. The plosive sensitivity of the mic is comparable to typical high quality side-address large diaphragm condenser microphones but does change in proportion to the selected polar pattern and mic type in the Sphere plug-in. Omni generally has the least plosive sensitivity and figure-8 has the most. Cardioid is somewhere in between. At 60mm in diameter the Sphere L22 headbasket is relatively large (the same size as a U47), and all other things being equal will have less plosive sensitivity than a mic with a smaller headbasket.

Microphones, in general, have more variation from unit to unit than just about any other piece of equipment in the recording signal chain, so when comparing any two mics even of the same type there will likely be some difference. The Sphere algorithms model the specific microphones that we used, which might not exactly match similar mics in your collection.

For comparison, current production Neumann microphones with this specification list a tolerance of +/-2 dB from the nominal response (e.g., U47 FET, M 147, TLM 103, TLM 49, KM series), and some AKG microphones are spec’d with a tolerance as high as +/-2.5 dB (e.g., C12 VR). This means that any two of the same type of microphones could have up to a 4 or 5 dB difference at some frequencies, and this is when the mic is brand new. If we’re talking about a vintage mic that’s 50+ years old then the variation can be much greater!

The microphone measurements assume a free field response, which basically means that the microphone is floating in infinite space without a mount. In practice, this is done by using a low profile mount and stand. For some microphones, such as an SM7 or MD421, the mount is built-in to the microphone in which case the mount is included in the measurement. For most end address dynamic mics, such as the SM57, the included microphone clip was used, since that is the most typical use case.

 

Polar Meter

Another unique feature in the Sphere plug-in is the Polar Meter, which displays the approximate amplitude and direction that sound is arriving from. The Polar Meter can be particularly useful for determining the direction that bleed from other instruments is arriving from and setting the polar pattern to best reduce the bleed. The moving yellow line in the Polar Meter shows the approximate amplitude and direction that sound is arriving from. The blue line shows the currently selected microphone polar pattern.

The image below shows the Polar Meter with a sound source that is 135 degrees off-axis. In most cases, sound from 135 degrees off-axis would be considered bleed, in which case the user would set the polar pattern to reject as much of this sound as possible.

The two “shadowed pie wedges” show where the polar pattern has the most rejection. Assuming, the desired sound source is muted and the off-axis leakage is occurring, then you want to line up the area of most rejection to cover as much of the yellow shape as possible.

Because of how the Sphere microphone is constructed the Polar Meter cannot differentiate between sound coming from the left, right, up or down. In this example, sound coming from 135 degrees to the right will be displayed on the Polar Meter as coming from both 135 degrees to the left and 135 degrees to the right. But because most microphones, including the Sphere microphone, are bilaterally symmetric the polar pattern is the same whether or not you are off-axis to the left or off-axis to the right, so there is no way to adjust the pattern to reject sound just on the left or just on the right.

In general, it’s best to use the Polar Meter as a guideline, and then dial in the exact pattern by ear, based on what audibly produces the most rejection. There are also some inherent inaccuracies in the Polar Meter due to how the microphone responds to proximity effect. This means that sometimes low-frequency content can distort the Polar Meter response. Some measures are taken to minimize this effect, but if the distance to the source is known then this can be compensated. When Off-Axis Correction™ is enabled the distance controls are used to help produce a more accurate Polar Meter response.

In a highly reverberant environment the Polar Meter will be more circular showing that sound is coming equally from all directions, as is shown here:

It’s also important to keep in mind that the polar pattern plot (blue curve) shows the polar pattern at 1.0 kHz, and the pattern for most microphones are likely to be significantly different at some frequencies. But most conventional microphones nominal patterns are usually measured in this frequency range, so it will usually match up with user expectations.

An original U47 is an interesting example because it has two patterns one of which is labeled cardioid, but the actual pattern is actually somewhere between supercardioid and hypercardioid. In this case, what is displayed in the polar meter is a truer representation of the actual pattern of the microphone.

 

Further Reading

Sphere technology is patented and only available from Townsend Audio Inc. If you really want to understand the gory details of how it works, you can read the patent at the following link. But be warned, it’s really long and dense!

https://www.google.com/patents/US8666090

The Sphere User Guide is also full of additional detailed information if you want to dive in further.