The EQ Window is used to determine what EQ filters to apply to a response and to see the effect those filters would have on both the frequency and time domain behaviour. It always shows the response currently selected in the main REW window, which can be changed from the REW main window or by pressing ALT + a measurement number (e.g. Alt+3 selects the third measurement) or using ALT+UP/ALT+DOWN to move through the measurements.
The window has 3 main areas: a "Filter Adjust" graph of frequency responses, a second graph area showing the impulse response and waterfall, and a panel on the right with various settings related to the EQ functions and modal analysis. The right hand panel can be hidden/shown using the button at the top of the scroll bar.
The Filter Adjust plot shows the measured and predicted (equalised) response for the current measurement along with the target response and the response of the equaliser filters with and without the target. This plot, in common with all plots that have a frequency axis, also shows where any filters have been defined, displaying the filter's number along the top margin of the plot at the position corresponding to its centre frequency.
The frequency response of the measurement is labelled with the measurement name. The Predicted response shows the predicted effect of the measurement's filters. The Target trace shows the target frequency response for the measurement, including any desired House Curve response shape. If a House Curve has been loaded the symbol will be displayed by the trace value. The Target response includes the Bass Management curve appropriate to the speaker type selected for the measurement in the Target Settings. The Filters trace shows the combined frequency response of the filters for this measurement, along with the individual filter responses if this has been selected (see Filter Adjust Controls below). The Filters+Target trace shows the frequency response of the filters overlaid on the Target response. Selecting the filter responses to be drawn inverted and adjusting the filters so that this curve matches the measured response will result in the predicted response matching the target.
The control panel for the Filter Adjust graph has these controls:
The smoothing selector operates in the same way as those on the other graphs. When Invert filter responses is selected the responses of the filters are drawn inverted. This is useful for graphically matching the shape of a filter to the shape of the peak it is being used to correct, when the shapes match the overall response in that region will be flat. Fill filter responses fills the overall filter response. Show each filter draws the individual filter response shapes separately in different colours. Fill each filter fills the individual responses.
The EQ Filters panel is displayed by clicking
the button at the top of the EQ window.
The Waterfall plot shows a waterfall for the measurement and for the predicted result of applying the current filters to the measurement. The Predicted waterfall can be configured to update automatically as filters are adjusted (see Waterfall Controls below).
The Slice slider selects which slice is at the front of the plot - as the slider value is reduced the plot moves forward one slice at a time. The trace value shows the SPL figure for the front-most slice, the corresponding time for that slice is shown at the top right of the graph.
The X, Y and Z sliders alter the perspective of the plot, moving it left/right, up/down and forwards/backwards respectively. The check boxes next to the sliders allow the perspective to be disabled in that axis. Disabling the x axis can make it easier to see the frequencies of peaks or dips. Disabling the z axis turns off all the perspective effects.
The Predicted plot can be overlaid on the current measurement. The overlay is generated slice-by-slice, plotting a slice of the current measurement's waterfall, then a slice of the overlay, then the next slice of the current measurement and so on.
Transparency can be applied to the main plot, the Predicted overlay, or both. When transparency is set to 0% both plots are solid. The transparency mode can be switched between main/overlay/both to ease comparison between the plots.
If Predicted Waterfall Live Update is selected the waterfall will be regenerated as filters are adjusted - it may take a few seconds for the update to appear, depending on the speed of the computer and the frequency span of the measurement.
The Total Slices control determines how many slices are used to produce the waterfall. Fewer slices mean faster processing, but make it less easy to see how the response is varying over time.
The Time Range control determines how far the impulse response window is moved from its start position to generate the waterfall.
The width of the impulse response section that is used to generate the waterfall is set by the Window control (this control sets the Right Hand window width). The corresponding frequency resolution is shown to the right of the window setting. Longer window settings provide better frequency resolution.
The Rise Time control sets the width of the Left Hand window. Shorter settings give greater time resolution but make the frequency variation less easy to see. The default setting, 100 ms, is aimed at revealing room resonances. When examining drive unit or cabinet resonances with full range measurements a much shorter rise time would be used, 1.0 ms or lower, with time spans and window settings of around 10 ms. CSD mode is often more useful for such measurements as the later part of the impulse response can be noisy, obscuring the behaviour in the later slices.
The Smoothing applied to the waterfall slices can be increased from 1/48th octave (the minimum, and recommended) to as high as 1/3rd octave.
Use CSD Mode should be selected if the later slices of the waterfall are contaminated by noise in the measurement. It would commonly be used when examining drive unit or cabinet resonances. CSD mode anchors the right hand end of the window at a fixed point and only moves the left side. This does mean, however, that the frequency resolution reduces (and the lowest frequency that can be generated increases) as the slices progress, as each has a slightly shorter total window width than the previous slice.
The control settings are remembered for the next time REW runs. The Apply Default Settings button restores the controls to their default values.
The Impulse plot shows the impulse response of the measurement and of the predicted result of applying the current filters to the measurement.
The area to the right of the graphs contains a group of collapsible panels containing settings that affect the EQ functions.
The Equaliser panel is used to select the type of equaliser that will be applied to the current measurement. Changing the equaliser type updates the filter panel, applying the settings appropriate to the selected equaliser. Filters already defined are retained where possible, but parameter values will be adjusted if necessary to comply with the ranges and resolutions of the chosen equaliser. The currently selected equaliser is shown in the panel title and in the EQ Filters panel. Details of the various equaliser types can be found here.
The Target Settings panel is used to tell REW what you expect or want the response to look like, so it knows what to aim for when applying EQ. The first selection (Speaker Type) is what kind of speaker the measurement is from. If it's a Full Range ("Large") speaker the basic target is flat. If it's from a Bass Limited ("Small") speaker the target has low frequencies rolled off to include the effect of the bass management filter, then the Crossover setting lets you tell REW how steep the bass management filter is and Cutoff is what frequency it is set to, typically 80Hz in Home Theatre systems. Same for Subwoofer, except the high frequencies are rolled off. The crossover slope would typically be 24dB/octave for a subwoofer and 12dB/octave for a bass limited speaker, however the 12dB/octave figure for a speaker is used because the speaker itself is expected to have around a 12dB/octave acoustic roll-off, hence the overall effect of the filter and the speaker's roll-off is around 24dB/octave - the 24dB setting may be a better match to the measured response in those cases.
For Subwoofers and Full Range speakers the LF Cutoff and LF Slope settings also come into play. They are to allow for the fact that although we might like to have speakers or subs that are flat right down to 0 Hz, in practice they have a lower limit so it is better for the target to reflect that. For a typical 'Full Range' speaker that might be 40 Hz, for a typical sub it might be 20 - 30 Hz. You can usually see from the measurement where it is rolling off and adjust these settings for the target until it matches. Setting the LF Cutoff to zero results in a target response that remains flat to 0Hz.
The LF Rise and HF Fall settings are there to allow a house curve to be superimposed on the response, the LF rise is used to have bass rising as frequency gets lower, which is often subjectively preferred, while the HF fall is used to reflect the downward tilt in HF which is normal for most listening position speaker measurements. The target curve will rise below the LF Rise start frequency at the slope selected until reaching the LF Rise end frequency. Similarly, the target curve will fall above the HF Fall start frequency at the slope selected.
Finally the Target Level control lets you move the whole target response up or down until it sits in the right place relative to your measurement. When the target is right the bits that go above it are the peaks you want to tame and it usually runs more or less through the middle of the measurement. Set Target Level automatically adjusts the level of the target response to provide a good match to the measurement over the range selected for EQ, but don't be afraid to manually adjust the level to suit.
The default speaker type, crossover slope, cutoff, LF rise and HF fall to use for new measurements are specified in the Equaliser Preferences.
The Filter Tasks panel is used to control REW's automatic filter adjustment feature. REW can automatically assign and adjust filter settings to match the Predicted response to the target response.
The Match Range defines the frequency span over which REW attempts to match the target response, and within which filters will be assigned. REW can apply filters anywhere across the band, but it is usually best to limit filters to low frequencies (less than 200 Hz or so) unless you are compensating for some general characteristic in the speakers (an example might be a dip in the mid range or a bit too much HF) - that is using EQ as a fancy tone control.
Individual Max Boost sets the maximum boost that REW will allow for any individual filter. This can be set to zero to prevent REW assigning any boost filters.
Overall Max Boost sets the maximum boost that REW will allow for the combined effect of all the filters. This can be set to zero to prevent REW allowing any overall boost, but individual filters may still have boost.
In addition to the gain limits, boost filters are subject to Q limits to avoid inadvertently creating artificial resonances. The Q of boost filters is not allowed to exceed a value which would cause the filter's 60dB decay time to exceed approximately 500 ms (the actual Q limit value depends on the filter's gain).
The Flatness Target controls how tightly REW tries to match the Predicted response to the Target Response. The lower the Flatness Target, the more filters will be required.
Allow narrow filters below 200 Hz determines whether target match uses filters narrow enough to counter modal resonances at low frequencies. This should be selected when applying EQ to a room measurement, but is best not selected when applying EQ to a device response (headphone EQ, for example). When this option is not selected the highest Q used will be 5.0.
Match Response to Target starts REW's automated filter assignment and adjustment process. REW assigns filters to match the Predicted response to the Target Response, beginning with the area within the Match Range where the measurement is furthest from the target. After assigning filters, REW adjusts the settings of the filters to get the closest match. It is best to apply the 'variable' smoothing to the response before running the target match.
For best results it is essential to first ensure the shape of the target response is correctly selected to suit the type of speaker whose response is to be equalised and set the Target Level so that REW does not end up applying filters to try and correct a level difference - equalisers are not volume controls!
Note that REW will not apply filters below the frequency at which the measurement first exceeds the target or above the frequency at which the measurement last drops below the target to prevent trying to boost a response beyond its natural roll-offs, if you wish to lift the low or high end response this can be done with manually applied filters but beware of exceeding the excursion limits or headroom of the woofer or power handling limits of the tweeter.
The Filter Tasks panel also includes a set of controls to optimise the settings of the current filters. Note that only filters that lie within the Match Range will be adjusted. Optimise gains will adjust the gains of all 'Automatic' PK and modal filters to best match the target response. Optimise gains and Qs will adjust the gains and Qs of all 'Automatic' PK filters and the gains of all 'Automatic' modal filters. Optimise gains, Qs and frequencies will adjust the gains, Qs and centre frequencies of all 'Automatic' PK filters and the gains of all 'Automatic' modal filters - it is equivalent to Match response to target without the automatic assignment of filters. Centre frequencies will be adjusted to within 10% of their initial setting and will remain within the match range.
If REW can read from the equaliser Retrieve filter settings from equaliser will be enabled, selecting it will read the settings either directly from the equaliser or from a file exported by the equaliser, depending on the equaliser type. Send filter settings to equaliser will transfer the current filter settings to the equaliser if the equaliser offers that capability. Save filter coefficients to file writes the biquad coefficients for the current filters at the selected sample rate in a form the equaliser can import. Export filter settings as text generates a text file with the filter types and settings. Reset filters for current measurement will clear all the filters.
REW can analyse the low frequency part of the measured response to search for modal resonances. The search is controlled by the settings in the Modal analysis panel. To determine the modal characteristics a parametric analysis of a segment of the impulse response is carried out to identify the frequencies, amplitudes and rates of decay of the resonant features that make it up. Such an analysis is not constrained by the frequency resolution limits of an FFT, allowing precise values for each mode's parameters to be determined. However, the accuracy of the results depends on the signal-to-noise ratio of the measurement. The better the measurement, the better the results. To get the highest measurement quality for modal analysis set the sweep end frequency to match the highest frequency of interest, use the longest sweep and adjust levels so that the peaks of the captured signal are around -6 to -12 dB.
The modal analysis panel includes selections for the room dimensions. These are used to determine the room's theoretical modal frequencies up to 200 Hz, which may be plotted on the SPL & Phase, Group Delay, Spectral Decay, Waterfall and Spectrogram graphs. If any dimension is set to zero the corresponding modal frequencies will not be plotted, if all the dimensions are zero no modal frequencies will be plotted. The colours used for the modal frequencies are the same as those used in the Room Simulator.
The panel controls select the range to search for resonances (which will be restricted to the range of the measurement if smaller), the duration of the impulse response to analyse and a threshold for filtering out spurious resonances due to noise in the measurement. Best results are obtained by keeping the frequency span to around 100 - 200 Hz. The Analysis Length, 500 ms by default, may be reduced if the measurement is noisy or increased if the measurement has particularly low noise (noise floor of the impulse more than 60 dB below the peak).
Small alterations of the analysis length, 10-20 ms or so, can help establish whether the modal resonances identified are accurate - modes with consistent frequency, amplitude and decay time at differing analysis lengths indicate reliable data. When Find Resonances is clicked the analysis begins, it usually completes after a few seconds. The results are shown in the Resonances panel.
The Resonances panel includes controls to filter the results list according to the T60 decay times of the resonances and their amplitude. The list of resonances may be sorted by frequency, SPL ("Peak dB") or T60 decay time by clicking on the column headers in the table. Clicking on a resonance in the table will show a plot of its shape on the Filter Adjust graph, multiple resonances can be selected by clicking and dragging or using Ctrl+click or Shift+click. Clear Selection clears any selections made.
Filters which accurately counter specific resonances can be generated by selecting the "Modal" filter type and setting the Target T60 value to the T60 time determined by REW. Modal filters are normal parametric EQ filters whose Q or bandwidth is adjusted by REW as their gain is changed to ensure they target the specified T60 value as closely as the equaliser settings resolution permits.
REW provides a Pole-Zero plot as an alternative way of viewing the results of the modal analysis. This may be an entirely unfamiliar way of viewing a response to many, but it does have some virtues when looking at resonances and filters. However, little would be lost by ignoring this section.
The pole-zero plot is a graph of complex numbers with the real part along the horizontal axis and the imaginary part along the vertical axis. There is a circle on the plot with a radius of one unit (referred to as the "unit circle") which corresponds in a way to the frequency axis of a frequency response. The plot shows results up to a frequency a little above the end of the modal analysis search, the upper frequency span of the plot is shown just to the left of the unit circle, near the point (-1, 0). As we move around the upper half of the unit circle the frequency increases from zero at the right side to the upper limit of the plot at the left. The lower half of the circle corresponds to negative frequencies, but for the signals we are looking at the bottom half is always a mirror image of the top half and can be ignored.
The plot shows poles, represented by crosses, and zeroes, represented by circles. Poles are places where the response becomes infinite, zeroes places where it becomes zero. The closer a pole gets to the unit circle, the more it pulls the frequency response upwards. Conversely, zeroes pull the response towards zero. Poles and zeroes at the same location cancel each other out completely, poles and zeroes close to one another partially counter each other's effects. If the plot has many pole/zero pairs that overlap they can be reduced by increasing the Noise Threshold setting. Poles outside the unit circle would correspond to an unstable system, none should appear there. Zeroes outside the circle would mean the response is not minimum phase, but the analysis may not start at the zero time of the impulse so this plot is not necessarily a good indicator of whether a response is minimum phase, for the correct method of determining that (using the excess group delay plot) refer to the Minimum Phase help topic.
Each modal resonance has a corresponding pole (actually a pair, the second is a mirror image below the axis). The frequency of the pole can be seen by drawing a line from the (0,0) point out through the pole to the point it reaches the unit circle, in the plot above the pole is at approximately 92.7Hz. REW shows the frequency value, and the SPL level at that frequency (81.3dB above). The closer a pole gets to the unit circle, the longer its T60 decay time. REW shows the T60 time corresponding to the cursor position, in the example above it is 440ms. If a resonance is selected in the Resonances panel its pole will be highlighted on the plot. The plot can be zoomed in on to get a closer view, clicking the button just above the x axis zoom buttons will reset the axis ranges to show the upper half of the unit circle.
Filters also have poles and zeroes, a parametric EQ filter has a pair of poles and a pair of zeroes (one pole and one zero above the axis, the other below). The locations of the filter's poles and zeroes vary as the filter's settings (frequency, Q/bandwidth and gain) are adjusted. If the settings of a filter are adjusted so that its zero is directly over the pole of a resonance, it completely counters the effect of that resonance in the time and frequency domains. Seeing how filter zero locations compare to response pole locations is where the pole-zero plot can be useful. In the case of the "Modal" filter type REW makes the adjustments that keep the filter's zero at a distance from the unit circle that matches the filter's target T60 time.
Filter poles and zeroes are shown in colour on the plot, corresponding to the colour used for that filter on the filters panel and the Filter Adjust plot. An example of a set of filters is shown below. Filters that cut (negative gain) have their zeroes closer to the unit circle than their poles, filters that boost have their poles closer to the unit circle than their zeroes.
Show cursor annotations controls whether REW draws a line from the origin through the cursor position to the unit circle and labels the frequency, response SPL and T60 values. If Show 500ms T60 boundary or Show 1000ms T60 boundary are selected REW will draw circles on the plot corresponding to those T60 times, any pole outside those circles has a T60 time greater than the circle value. If Show Resonance Poles Only is selected the pole-zero plot will only show the poles for the resonances displayed in the Resonances panel, otherwise it shows all poles found during the analysis.