Chapter 5: Single Channel Measurements. Rational Acoustics Smaart v7

Chapter 5: Single Channel Measurements

Spectrum Measurement and Display Configuration

Single-channel spectrum measurements allow you to examine the spectral content of audio signals throughout a system. Spectrum measurements are extremely useful in many applications, including the identification of feedback frequencies in sound reinforcement systems, noise and sound exposure measurements, and general signal monitoring tasks. Data from spectrum measurements can be displayed as a conventional RTA (real-time spectrum analyzer) graph, or plotted over time in a threedimensional (level vs. frequency vs. time) Spectrograph chart.

Two basic groups of settings determine the appearance and behavior of RTA and Spectrograph displays in Smaart:

 Measurement settings affect how data is acquired. These, we have discussed in detail in Chapter

3. In this section, we will look more specifically at how some of those options affect the RTA and

Spectrograph displays.

 Display settings affect how spectrum measurement data is displayed after it is acquired but do not change the underlying measurement data. These options mainly reside on the Spectrum page of the options dialog window (Options menu > Spectrum), which we will be talking about later in this section.

In practice it isn’t really possible to draw a completely hard line between the two – for example, fractional octave banding is actually done at display time but we treat is as a measurement parameter for practical reasons – but that is the basic organizational intent, in terms of where the various options for spectrum measurements are located in Smaart.

RTA Measurements

The real-time spectrum analyzer, or RTA, is a familiar tool to most audio professionals and probably needs little introduction. It enables you to look at the frequency content of signals moment-by-moment in real time. Essentially the RTA is a graph of the energy in an incoming signal, broken down by frequency or frequency ranges, with frequency (in Hertz) on the x axis and magnitude (energy) on the y axis in decibels (dB). The graph is updated continuously when one or more live spectrum measurements are running, to produce a real-time display. By adjusting the scale and averaging of the display, we can refine measurement resolution and responsiveness to fit different tasks.

Octave banding 1/3 rd

Octave banding 1/12 th

Octave banding

Figure 42: RTA display with Octave, 1/3-octave and 1/12-octave banding.

60

Chapter 5: Single Channel Measurements

Time and frequency resolution are major trade-offs associated with spectrum measurements and realtime frequency-domain analysis in general. We always have to give up a little of one to get more of the other. On a basic level, the FFT size used to transform time-domain signals into frequency-domain spectral data limits the time and frequency response of RTA and Spectrograph. Larger FFT sizes give you tighter frequency spacing and more detail at low frequencies at the expense of integrating over a longer period, which can limit your ability to see fast changes in the signal. The other major factors affecting the degree of detail that you can see on the RTA graph and its responsiveness to changes in the input signal are averaging and banding.

Averaging and Banding Controls

For the RTA display, we typically average data from successive incoming FFT frames over some period of time to produce a display that is smoother and less jumpy. However averaging also limits how quickly the RTA can respond to rapid changes in the frequency content signals. Unlike banding, FFT size and averaging are baked into RTA data at the measurement level – when you capture an RTA trace you are capturing averaged data (if applicable). Note that spectrum averaging affects only RTA data. The

Spectrograph is always plotted from instantaneous (un-averaged) spectrum data.

Figure 43: Control Strip for Spectrum measurements.

Banding is actually a display parameter for Spectrum data. When you capture a snapshot of an RTA trace, you are capturing it at the original FFT resolution and can always change the banding after the fact. Using larger fractional octave bands can help reduce visual “noise” and make larger trends in the spectrum of a signal more obvious, but at the expense of limiting how much detail you can see.

When the active display is an RTA graph, the Banding and Averaging settings for the active spectrum measurement are adjustable from the control strip on the right side of the main Smaart window.

Banding is a global setting that applies to all spectrum measurements. Averaging applies only to the RTA display and can be set specifically for individual spectrum measurements. If the current active measurement uses the global setting for spectrum averaging then the Averaging selector on the control strip controls the global setting. Otherwise, it applies only to the active measurement.

61

Chapter 5: Single Channel Measurements

RTA Graph Types

Most commonly, the RTA is displayed as a bar chart with fractional-octave frequency scaling, however

Smaart can display banded RTA data as a line graph, or plot a combination of fractional octave data in bar chart form with the un-banded FFT data overlaid as a line graph. Un-banded (aka “narrowband”) FFT data is always plotted as a line graph. These three options for Banded Data (Bars, Lines or Both) are located in the RTA Display Settings section of the Spectrum page in the options dialog window (Options menu > Spectrum).

Fractional-Octave

Bar Graph (Bars)

Fractional-Octave

Line Graph (Lines)

Fractional-Octave Bar Graph and

Unbanded Line Graph (Both)

Figure 44: RTA graph types (Bars, Lines or “Both”).

Peak Hold

When looking at dynamic signals on an RTA display the normal bar or line graph is typically averaged over some period. Without averaging, the display can be too jumpy to read, but averaging tends to smooth out some of the faster peaks in the signal. If you want to see both averaged power and a record of the highest level the peaks in the signal at each frequency or in each band, you can turn on Peak Hold in in the RTA Display Settings section of the Spectrum options (Options menu > Spectrum). Peak hold data is plotted as a second line trace on line graphs, or as a series of flattened bar segments on bar charts. When you capture an RTA trace with peak hold turned on, both the normal RTA trace and the peak hold data are preserved in the captured measurement.

Figure 45: RTA bar graph with peak hold.

62

Chapter 5: Single Channel Measurements

Options for Peak Hold include None, Timed or Infinite. The default setting is None, which turns the feature off. Infinite peak hold preserves the highest peak level recorded for each frequency until it is either replaced by a higher reading or you turn the feature off of press the [V] key to flush the averaging buffer. Timed peak hold allows the peak trace to decay after some period of time, as specified in the

Hold field.

By default, the peak hold function looks for the highest peaks in each incoming FFT, before the data goes into the average for the normal RTA display, so you may never see an averaged RTA trace come anywhere near the peak levels. If you want to look at the highest levels in the averaged signal instead, click the Averaged check box in the RTA Display Settings section in Spectrum options. When using this option, you may want to run the RTA for a few moments and give the live measurement(s) a chance to settle before turning on peak hold.

Plot Calibrated Levels

The RTA display in Smaart is calibrated to digital Full Scale by default, meaning that the largest magnitude value obtainable in a digital sinewave (given the current Bits per Sample selection in Audio

Device Options) is scaled to 0 dB and all lesser magnitudes end up being negative decibel values. This works very well unless you need to relate the RTA display to some external reference, such as sound pressure level (SPL).

If one or more of your spectrum measurement uses an input that is calibrated to SPL (which we will get to later in this section), and you want the RTA display to reference the calibrated levels, open Spectrum options (Options menu > Spectrum) and click the Plot Calibrated Levels check box in RTA Display Settings to enable it. This applies the specified input calibration offset for each input channel to spectrum measurements before plotting the RTA graph. It also sets the default RTA magnitude range to 20 dB to

120 dB (rather than 0 to −100 dB).

One potential problem you may encounter when using the Plot Calibrated Levels option is that it can result in a drastic difference in scaling between calibrated data inputs and uncalibrated measurements; for example, between calibrated microphone inputs and line level inputs. Uncalibrated spectrum measurements tend to “fall off” the graph when Plot Calibrated Levels is enabled. The best way to work around this issue is to add a dummy calibration offset to uncalibrated inputs channels used for Spectrum measurement, so that they rescale themselves along with calibrated measurements when the Plot

Calibrated Levels option is selected.

To assign a dummy calibration to an uncalibrated input, first bring up an RTA graph and run all spectrum measurements with Plot Calibrated Levels disabled, to make sure all measurements are visible on the graph and have signal present. With the RTA still running, turn on Plot Calibrated Levels in Spectrum options, then select Audio Device Options from the Options menu. Double-click the name of the device that is driving an uncalibrated measurement on the Input Devices tab of the Audio Device Options dialog to open its settings dialog. You will see that uncalibrated inputs have a Cal. Offset of 0.00 dB. You can click on any entry in the Cal. Offset column to edit this value, and then press the [Enter] key to set the change.

63

Chapter 5: Single Channel Measurements

If you are trying to match a line input to a microphone input that is calibrated to SPL, an offset of 120 dB is generally a good place to start. After changing the calibration offset, click the Apply button in the lower right corner of the dialog window and hopefully your measurement will appear on the RTA graph.

If you want to move it up or down, you can adjust the offset value and reapply the change.

Spectrograph Basics

Smaart’s Spectrograph straddles the time and frequency domains giving you a birds-eye view of the frequency content of a signal over time. The real-time spectrograph and the IR mode version are essentially the same display, oriented in two different directions.

If you are new to the spectrograph, then one way to think about how it works is to start with the idea of a spectrum analyzer. On a real-time spectrum analyzer (RTA) you typically have a bar graph or line chart with frequency on the x axis and magnitude in dB on the y axis, showing you the spectrum of some

chunk of signal at a given moment in time – perhaps something like the one shown in Figure 46a.

An RTA is very useful tool, but if you want to get a better understanding how the spectrum of a signal changes over time, you either need a really good memory or a different kind of graph. One solution might be to just keep sliding the old data to the back instead of erasing it as new data comes in, to form

a 3-D graph with time on the z axis, as in Figure 46b. If you did this with a 3-D area chart instead of a bar

graph you would have what is commonly called a waterfall chart, but let’s continue with the bar graph example, as both have the same limitation. The problem with this approach is that as new data comes in, higher-level values in front will cover up some of the data in back, so that you only get a partial

picture of the history, as in Figure 46b.

Figure 46: Turning a spectrum analyzer into a spectrograph.

64

Chapter 5: Single Channel Measurements

You could rotate the graph in space until you can see all of the bar tops, but when you do that it becomes harder to discern how tall they all are. Assuming that you have a color display (waterfall charts were popular before anyone had color monitors) you might try to alleviate that problem by painting the

tops of the bars different colors based on their relative magnitude as we did in Figure 46c. But at that

point, the chart would much more readable if you just dispensed with the bar graph idea altogether and plotted it as a 2-D chart instead, with frequency on one axis, time on the other, and magnitude indicated

by color (Figure 46d). That’s a spectrograph.

Generally, the “domain” of a graph is the independent variable, e.g., time or frequency, which is normally assigned to the horizontal (x) axis, but the spectrograph display has two independent variables.

You can orient it whichever way is most convenient. In real-time mode in Smaart we want to relate the spectrograph to other frequency-domain graphs, so we plot it with frequency on the x axis and time on the y axis. In IR mode, we most often want to look at it in the context of other time-domain graphs, so we do it the other way around – in that case, time goes on the x axis and frequency on the y axis.

The Real-Time Spectrograph

Now that we have a general idea of how a spectrograph works and what it can tell u s, let’s look more specifically at the real-time spectrograph. Smaart’s real-time spectrograph display is a plot of a signal’s spectrum over time, with frequency (in Hertz) in the x axis, time on the y axis, and magnitude in decibels represented in color. The time axis of the graph is unreferenced, because the update frequency can vary somewhat, depending on how busy your computer is at any given time.

Figure 47: The real-time mode Spectrograph display.

65

Chapter 5: Single Channel Measurements

Spectrograph Dynamic Range

The dynamic range of the spectrograph is controlled by two arrowhead shaped widgets that appear on the left edge of the Spectrograph chart.

These controls are echoed on the RTA graph. The upper of the two widgets sets the maximum (Max) threshold; the lower one sets the minimum (Min). The spectrograph scales its color spectrum between these two extremes. Any FFT bin whose magnitude exceeds the specified maximum is mapped to the color white. Values falling below the minimum are mapped to black. Note that you can also specify threshold values for spectrograph dynamic range in the Spectrograph Settings section of Spectrum options (Options menu > Spectrum).

Figure 48: Spectrograph dynamic range and color mapping.

The real key to creating a useful spectrograph is getting the dynamic range right. If you set the range too wide, the display loses definition and important features may get lost. Set it too narrow or set the lower threshold too high and you might miss some important features altogether. One of the major advantages of Smaart 7’s real-time spectrograph is the ability to adjust spectrograph thresholds dynamically, so that you can see the affect that adjusting the Min/Max thresholds has on data already on the screen and make an “apples-to-apples” comparison.

Buffer Size and Slice Height

Another advantage of the Smaart 7 spectrograph is its ability to maintain a lot more history than ever before. In older versions of Smaart, the number of “slices” of spectrograph history that you could maintain was limited by how many pixels on the screen that could be allocated to the graph. In Smaart

7, the limit is more a matter of how much memory you want to allocate.

The amount of spectrograph history that you can display is still limited by screen resolution and graph size and the spectrograph slice height, but you can keep as many as 2000 slices of history in memory.

That works out to at least 80 seconds worth of data at a maximum update speed of 24 frames per second. When the history size exceeds the graph size, you can scroll the graph backward and forward using the up and down arrow keys on your keyboard (assuming that it is the active graph).

The Slice Height setting in the Spectrograph Settings section of the Spectrum options dialog determines the vertical height of each horizontal “slice” of spectrograph. The smaller the slice height, the more data you can fit on the screen. Larger slices may make small features easier to see and may also consume fewer graphics processing resources.

Grayscale

One additional option for the real-time spectrograph is to render the graph in grayscale, rather than in color. Users who are colorblind or otherwise have trouble distinguishing between some colors may find the grayscale spectrograph easier to read than the color version. The grayscale option may also produce better results when making screen shots for printed documents, if you don’t plan on printing the document in color. To change the spectrograph to grayscale, open Spectrum options (Options menu >

Spectrum) and click the check-box labeled Grayscale in the Spectrograph Settings section.

66

Chapter 5: Single Channel Measurements

Spectrum Options

We have already covered the use of most settings in Spectrum options but not quite all. To pick up the ones we did not talk about and recap the ones that we have discussed in this section, here a full list of all the settings on the Spectrum options page of the Options dialog.

General Settings

There really is not much to talk about here.

The only setting in this section is the Data

Window, used to precondition timedomain signals before performing an FFT to find their spectrum. The default setting is

Hann and unless you have some good reason to want to change it, that is a good place to leave it. A discussion of why one might want to use a different window function for real-time measurements, or not use a window function at all, is a little

Figure 49: The Spectrum options dialog tab.

beyond the scope of the document. Note that this setting applies only to un-banded (narrowband) spectrum measurements. For fractional octave banded measurements, Smaart always uses a Hann window.

Graph Settings

Frequency Scale – The frequency scale selector sets the frequency scale and grid ruling options for the

RTA display. Actually, there are only two scaling options: linear (Lin) and logarithmic. All of the others are grid-ruling options for logarithmically scaled frequency.

 Decade plots the RTA graph with logarithmic frequency scaling and decade (base 10) vertical grid ruling.

 Octave plots the RTA graph with logarithmic frequency scaling and vertical grid lines spaced at one-octave intervals.

 1/3 Octave plots the RTA graph with logarithmic frequency scaling and vertical grid lines spaced at 1/3-octave intervals.

 Lin plots the RTA graph with linear frequency scaling and linearly-spaced vertical grid ruling.

Magnitude Range (dB) sets the decibel range for y axis of the RTA graph. Note that the default y ranges

(calibrated and un-calibrated) for RTA graphs are hard coded. These settings control only the current display range and are overwritten if you zoom the plot using mouse or keyboard shortcuts or reset to the default range.

67

Chapter 5: Single Channel Measurements

Y-Zoom Increment (dB) sets the increment used for keyboard zoom in on the y-axis of the RTA graph.

When an RTA display is selected in the plot area pressing the [+/=] or [−] keys will increase or decrease the vertical scale of the graph by the number of decibels specified here.

Y-Scroll Increment (dB) sets the increment for keyboard scrolling on the RTA or Spectrograph displays.

When an RTA or Spectrograph display is selected in the plot area, each press of the up arrow [↑] or

Down arrow [↓] keys will scroll the plot up or down by the number of decibels specified.

Y-Grid Interval (dB) sets the y-axis grid-ruling interval for the RTA graph in decibels.

RTA Display Settings

The Banded Data setting in the RTA Display Settings section of Spectrum options determines the type of chart used for octave and fractional octave band displays.

 Selecting Bars plots banded RTA graphs as a bar chart.

 Selecting Lines changes the banded RTA display to a line graph.

 The “Both” option is a hybrid display that superimposes a narrowband spectrum trace over a fractional-octave banded bar graph.

Plot Calibrated Level applies input calibration offset (e.g., for SPL calibration) to RTA measurements (if applicable) and makes the default RTA display range 20 dB to 120 dB. Note that this will result in a drastic difference in scale between data from inputs calibrated to SPL and inputs calibrated to digital Full

Scale. When this option is not selected, Smaart ignores input sensitivity calibration and references all spectrum measurements to digital Full Scale. The default magnitude range for RTA displays in that case is 0 to −100 dB.

When Show THD is enabled and the RTA graph is set to a fractional-octave resolution of 1/12th-octave or higher, the notation THD: n%, will appear in the cursor tracking readout above the RTA graph, where n is the THD percentage value calculated for the current cursor frequency. THD values in Smaart are the ratio of the power in the fractional octave band at the cursor frequency, to the sum of the power in the next three harmonic frequencies. If (and only if) the cursor is positioned at the frequency of a single sine wave being used to stimulate the system under test, this value should be indicative of the total harmonic distortion present in the system at that frequency. Otherwise, it is generally meaningless.

Track Peak causes Smaart to track and display magnitude and frequency of the data point with the highest magnitude in the front trace on the RTA plot when enabled.

Peak Hold displays a line or bar segment at the peak magnitude measured for each frequency point on the RTA graph. There are three possible settings for peak hold:

 None turns peak hold off.

 Timed peak hold allows the peak trace to decay after some period of time, as specified in the

Hold field.

 Infinite peak hold preserves the highest peak level recorded for each frequency until it is replaced by a higher reading or you turn the feature off of press the [V] key to flush the averaging buffer.

68

Chapter 5: Single Channel Measurements

Hold – this field sets the decay time for Timed peak hold in seconds.

Averaged peak hold bases the peak hold trace on averaged RTA data when selected (if applicable).

Otherwise, the peak hold function looks for the highest peaks in each incoming FFT, before the data goes into the average for the normal RTA display.

Spectrograph Settings

Slice Height sets the height in pixels for each row in the Spectrograph display.

Slices in History sets the maximum number of rows for the spectrograph history. This can exceed the number of rows you are able to display on your screen at a given time, allowing you to scroll back to see transient events or other features that have scrolled off the screen. The caveat is that the more rows in the history, the more memory is required. For large FFT sizes and very long histories the memory requirements can be quite large

Max Memory Required calculates the memory requirements for the specified spectrograph history size based on the current FFT size selected for spectrum measurements in Measurement Config.

Grayscale changes the spectrograph to shades of gray, rather than colors.

Dynamic Range (dB FS) sets the upper and lower (Min/Max) boundaries for the spectrograph display in decibels. Frequency data points whose magnitude values fall below the specified minimum (Min) value are displayed in black. On the high end, out-of-range values that exceed the specified Max value are set to white.

Application Examples

What follows are some examples of spectrum measurements used in “real world” applications. These are intended as walk-though exercises that you can use to get a little hands-on practice, as well as examples of useful things that you can do with Smaart.

Distortion and Overload

For this exercise, we will start with a very simple measurement setup and provide detailed instructions for every step in the process. If you already know your way around Smaart then just skip over the parts that you already know.

If the headphone output on your computer is capable of overloading its

Figure 50: Loopback setup for distortion and overload example.

line inputs, the measurement setup for this example may be as simple as a 3.5 mm (1/8") TRS patch cable. Otherwise, you may need to route the output of your audio I-O device through a gain stage of some kind, such as a mixer or preamp, and then back into an input. The idea is that we want to be able to drive a sinewave hard enough to clip the input on your audio I-O device.

69

Chapter 5: Single Channel Measurements

Click the Spectrum button in the display control section of the control strip to set the graph area to a single graph pane and the graph type to RTA. If you don’t already have a spectrum measurement set up for the input channel that you are using on your audio I-O device, click the little hammer and wrench icon on the control strip or select Measurement Config from the

Options menu to open the Measurement Config dialog,

On the Group tab in Measurement Config, click the New Spectrum

Measurement button. In the pop-up window, type a Name for the measurement, select the Device and input channel (Ch) that you are using, and then click OK to create the measurement. In the Global Spectrum

Settings section of Measurement Config, make sure the FFT size is set to at least 8K or larger. Set

Banding to 1/48 and Averaging to 8, then click OK to exit Measurement Config.

Figure 51: Measurement configuration for distortion and overload example application.

Back in the main Smaart window, you should now see a control block for your new measurement in the lower portion of the control strip. Click the run (

►

) button for the measurement to make sure it works.

All we care about at this point is that the run button turns green and Smaart doesn’t throw any error messages. If there’s any problem, double-click in the background of the measurement control block to open its tab in Measurement Config, and make sure you selected an input device that is currently connected to your computer. You can change it right there if necessary.

Now, click in the center portion of the signal generator controls (the output level readout) on the control strip or press [Alt/Option]+[N] on your keyboard to open the Signal Generator control panel.

70

Chapter 5: Single Channel Measurements

Select Sine as your Signal type, then set the master Level value to −12 dB and the Level 1 gain to 0 dB.

Set the frequency (Freq) to 1000 Hz and then click OK to exit the dialog window. The signal generator on/off button on the control strip should now say Sine (instead of Pink Noise). Go ahead and click it with your mouse or press [G] on your keyboard to start the signal generator.

Figure 52: Signal Generator setup for distortion and overload example application.

If the measurement that we set up is not still running, click its run button to turn it on and then adjust the signal level until it’s running in the yellow portion of the measurement’s input level meter. You can use the −/+ buttons on the signal generator or your external gain stage if applicable. If all goes well, you

should now see a nice clean spike on the RTA display at 1000 Hz like the one in Figure 53a. If you don’t

see anything, click anywhere in the left margins of the RTA graph to reset it to the default y range.

If you can’t see the noise floor of your I-O device on the RTA graph, use the up/down arrow keys on your keyboard to slide the range of the graph up and down and the plus/minus keys to scale the y range until you can see everything from the noise floor up to 0 dB.

Now we want to increase the signal level a little bit at a time until input starts to clip. As the signal level meter for the measurement starts to max out in the red zone, you should start to see additional spikes

rising up on the RTA graph at integer multiples of 1 kHz as in Figure 53b. Those are harmonic distortion

products forming as overloaded input clips the signal and our nice clean sinewave starts to resemble something more like a square wave.

71

Chapter 5: Single Channel Measurements

a) Undistorted sinewave on a log frequency scale b) Clipped sinewave and distortion products on a log frequency scale c) Clipped sinewave and distortion products on a linear frequency scale

Figure 53: Measurement results for distortion and overload application example.

The relationship between the harmonic frequencies becomes even more apparent if you switch the frequency scale of the graph to linear, so click on the word Spectrum at the top of the control strip or select Spectrum from the Options menu to open spectrum options. In the Graph Settings section of

Spectrum options, change the Frequency Scale setting to Lin and click the check box labeled Show THD in the RTA Display Settings section. Click the OK button in the lower right corner of the dialog window to apply changes and exit the dialog. You should now see the distortion products from the overloaded input spaced at even intervals along the frequency axis of the plot. Also, if you put your mouse cursor on the peak a 1 kHz, Smaart will calculate THD and display the value in the cursor readout.

Be sure to go back in and change the frequency scaling for the RTA back to one of the logarithmic options (Decade, Octave or 1/3 Octave) before moving on to the next exercise. Linear frequency scaling is great for looking at harmonics and comb filters, but generally not that great for most things that we do with Smaart.

72

Chapter 5: Single Channel Measurements

Feedback Frequency Identification

For this next example, we need an actual sound system. It doesn’t have to be a very elaborate sound system, just a vocal microphone and a mixer driving an amplifier and loudspeaker or a powered speaker will do the trick. The vocal microphone is routed through the mixer to the amp and loudspeaker and we acquire the output signal of the mixer as our measurement signal.

In Smaart, set up a spectrum measurement for the input on your I-O device that you’re connecting to the mixer. See the previous example application (Distortion and Overload)

Figure 54: Measurement system setup for feedback study.

for instructions on how to do this if you need them. Click the Spectrum button in the control strip, and then click the button with the split rectangle

graphic in the next row down. Your graph area should then look like Figure 55, with an RTA graph on top

and a Spectrograph below. Set the Banding selector in the control strip to the right or the graph to 1/24octave, so we can see a little more detail on the RTA and Spectrograph.

Feedback Frequency

Constant tone shows up as a straight line on the

Spectrograph display

Figure 55: Feedback study.

73

Chapter 5: Single Channel Measurements

Click the run button on your live measurement to start acquiring data and then slowly bring the microphone gain up until the system begins to feed back. You should see a vertical streak forming on the

Spectrograph at the feedback frequency, like the one at 1.24 kHz in Figure 55. If you position your

mouse cursor over the streak on the graph, you can read off the precise frequency in the cursor readout at the top of the graph area. For extra credit, plug a music player into the mixer and repeat the experiment with music playing. You should still find it pretty easy to identify the feedback frequency on the Spectrograph, whereas it becomes much harder to find on the RTA graph in the presence of a complex dynamic signal like music.

The spectrograph is also a powerful tool for helping to track down and identify audible problems other than feedback during performances. It is a common practice for mix engineers to route the solo bus output of a mixing console to an input for Smaart. Monitor engineers find this particularly useful for viewing the spectral content of their mixes, or in general viewing the spectrum of any input signal before it is amplified by the sound system. FOH engineers will often have a microphone set up as well to monitor the acoustical output of the system in real time. Even the most trained ears can benefit from this. For example, if you are hearing a low mid build up between 160Hz and 220Hz, the spectrum analyzer set up you can help you see exactly what frequency is the main offender, and how much attenuation is needed to get back in line.

Examining Interaction Patterns with Spectrograph

The following is a simple technique that uses the Spectrograph for examining coverage and interaction patterns in loudspeaker systems. Simply put, you excite the system with pink noise – which should produce a relatively constant level/color at all frequencies on a spectrograph plot – and then move the measurement mic through the listening environment. Level variations from interactions, like the audible comb filtering caused by reflections, can be seen as interaction patterns on the spectrograph plot of the mic signal. Adjusting the dynamic range helps to better highlight the interaction patterns.

Figure 56: Spectrograph plot of comb filter interaction patterns.

74

Chapter 5: Single Channel Measurements

Sound Level Measurement: SPL and LEQ

Sound pressure level (SPL) and Equivalent Sound Level (LEQ) are two ways of measuring and characterizing the loudness of sounds. Actually they are very much the same thing, the main difference being how the measurements are averaged and for how long.

Exponentially averaging sound level meters (SLMs) that measure SPL have been around for several decades. These rely on first-order lowpass filters for time integration, which are easily implemented in analog hardware – actually, just a resistor and a capacitor will do the trick. Most meters nowadays are digital but the specifications for exponential time integration have not changed. The time constant for the Fast integration setting is 125 ms. The Slow setting has a time constant of one second.

Integrating sound level meters used for LEQ measurement average linearly, over longer periods; often several minutes or even hours. LEQ is regarded as a better way of characterizing such things as background noise, environmental noise and dosages for sound exposure in very noisy environments such as manufacturing facilities, construction sites and rock concerts. Traditional SPL measurements use shorter integration times and are arguably still a better tool for measuring intermittent sounds or fluctuating sound fields.

Smaart can perform both SPL and LEQ measurements and given a good microphone and proper calibration, it can perform them very accurately. The DSP processes used for sound level measurements in Smaart were actually designed to the old Class 0 (laboratory instrument) specifications and exceed the requirements of current standards for Class 1 precision instruments. Unfortunately, there are no provisions in the current standards for certifying DSP processes independently of data acquisition hardware. The standards for sound level measurement equipment are written with hardware devices in mind and their test and certification procedures all pertain to the “complete instrument” – which in our case would include your computer, microphone(s), audio I-O interface and all cabling, in addition to the

Smaart software. All this is to say that if you require standards traceability for your sound level measurements, you pretty much need a standards-compliant sound level meter.

Both SPL and LEQ are typically measured using standardized A or C weighting curves. The A weighting curve mimics the frequency sensitivity of the human ear at low levels and is also a good predictor of hearing damage from exposure to loud sounds. The A curve rolls off more drastically on high and low ends than the C curve, which simulates the sensitivity of human hearing at higher sound levels. There was also a B curve that split the difference between the A and C curves, but no one ever really used it for anything. The B curve has been dropped from the most recent standards but Smaart 7 still supports it.

Signal Level / Sound Level (Broadband) Meter

The large numeric readout in the upper right corner of the Smaart v.7 main window can be configured to function as a standard Sound

Pressure Level (SPL) meter, an Equivalent Sound Level (LEQ) meter, or a peak signal level meter calibrated to digital full scale. The text labels above and below the large numeric value in the center of the display function as drop-down menus. The upper menu selects which audio

75

Chapter 5: Single Channel Measurements

device and input channel drives the display. The one in the lower right sets the measurement type.

Pressing the [K] key on your keyboard replaces this display with a clock, and pressing [K] again changes it back to a meter.

The selections for SPL and LEQ in the measurement type menu reflect the current averaging and weighting settings for these measurements selected in the Sound Level Options dialog (see below). For example, “dBC SPL Slow” means the meter is configured to display sound pressure level with C weighting and Slow exponential time integration. Similarly, “dBA LEQ 10” would mean equivalent sound level (LEQ) with A weighting and a 10-minute integration period.

The Max value keeps a record of the highest signal or sound level encountered since the measurement began or since the last reset. The little red circle ( ● ) to the right of it functions as a reset button for the

Max value in SPL and peak level modes. In LEQ mode, the reset button changes from red to green ( ● ) once the buffer for the rolling LEQ average is fully populated, and clicking it will clear the LEQ buffer and restart the LEQ measurement.

Sound Level Options

Clicking anywhere in center portion of the meter opens the

Sound Level Options dialog window. Here you can select weighting curves and integration times independently for both

SPL and LEQ measurements, enable logging and specify a log file. The Calibrate button in the lower left corner opens the

Amplitude Calibration dialog, where you can calibrate your input device(s) and microphone(s) for accurate SPL and LEQ measurement.

SPL and LEQ Settings

The Type selector in the SPL Settings section of the Sound Level

Options dialog sets the integration period for exponentially averaged sound pressure level (SPL). The two options are Fast and Slow. Fast SPL averaging uses a first-order exponential time average with a 125 ms time constant. The Slow option has a one-second time constant.

Figure 57: Sound Level Options dialog.

Weight gives you a choice of A, B or C frequency weighting or no weighting (None), which could be considered equivalent to Z weighting, provided that your microphone and input device are flat within standardized tolerances from 10 Hz to 20 kHz. There are separate settings for SPL and LEQ weighting.

The Minutes field in the LEQ Settings section of Sound Level Options specifies the length of the LEQ averaging period in minutes. The value in this section is appended to the “LEQ” notation in the lower right corner of the signal/sound level meter when measuring LEQ. Typical integration periods for LEQ vary by application. The default in Smaart is 10 minutes, which is a fairly common choice for environmental noise measurements. Integration periods as short as 1 minute may suffice in situations where sound levels are not expected to vary by much over time. Often, community noise regulations or

76

Chapter 5: Single Channel Measurements

architectural specifications will specify the integration period to be used for LEQ measurements, so be sure to check any that may apply. In any case, the integration period for LEQ should always be noted in statements of measurement results.

Logging

The logging function for SPL, LEQ or signal level data simply writes the current meter reading to a text file at a specified interval. When enabled, it will keep writing to file until you tell it to stop or close the program. Logging is turned off by default each time you run Smaart to prevent the log file from inadvertently eating up all the storage space on your computer, so in the event of a program crash or power loss you would need to go back in and turn it on again to restart logging if desired.

The log files are formatted in tab-delimited ASCII text, suitable for import into a spreadsheet, with one log entry per line. The format of the data depends on the measurement type. SPL and dB FS logs have columns for time and date of each entry, the meter reading in decibels and the measurement type. LEQ logs have additional columns.

Sound Level Calibration

To Calibrate one or more input channels for SPL measurement in Smaart , press [Alt] + [L] on your keyboard or select SPL/LEQ from the options menu, then click the

Calibrate button in the Sound Level options dialog. This opens the Amplitude Calibration dialog window shown in

Figure 58. At the top of the calibration dialog are three drop-

list selectors for Input Device, Input Channel and Microphone.

Select the device and channel for the microphone that you want to calibrate, and if the input device happens to be a

Smaart I-O, then the Microphone selector becomes available as well.

Calibrating with a Sound Level Calibrator

In the realm of DSP, the only real reference we have for amplitude values is how big they are relative to the biggest

Figure 58: Amplitude Calibration dialog for

SPL/LEQ calibration.

number that you can get from a sample word of a given number of bits. For example, a 24-bit sample word gets you max PCM amplitudes of +/− 8388607 (2^23

− 1). We typically normalize those maximum full-scale numbers to +/− 1, so that full scale works out to be 0 dB and all lesser amplitudes are negative numbers on a dB scale. In order to relate that internal amplitude value to some number of volts or Pascals of pressure in the external world, we need to calibrate it to signal of known amplitude – and to do that with a microphone we need an acoustical signal of known amplitude. This is where a sound level calibrator comes in.

77

Chapter 5: Single Channel Measurements

Calibrating an input channel and microphone for SPL measurement using a sound level calibrator is a two-step process consisting of: a) Measuring the digital signal level of an input channel that is being driven by a signal of known amplitude, such as a microphone with a Sound Level Calibrator, and b) assigning the reference amplitude value to the measured Full Scale amplitude.

To measure the reference input signal level, you connect a microphone to the input channel that you want to calibrate, affix your sound level calibrator to the microphone, turn on the calibrator and adjust input gain to a desired level, then click the Calibrate button in the Amplitude Calibration dialog to run Smaart’s calibration routine. Once calibrated you will need to leave the input gain set exactly where it is in order to maintain calibration (unless you are using a Smaart I-O), so a little forethought here can save you needing to repeat this procedure again later. Some things to think about are the loudest sounds you need to measure and the max SPL rating of your microphone.

If your microphone is rated for 120 dB max SPL, then you would properly need a different microphone, to measure

Figure 59: Calibration Progress dialog.

sounds louder than that. On the other hand, if your microphone is rated for 140 dB then, hopefully you won’t need to get anywhere near that and might want to pick a lower number. Whatever you decide on as your max SPL figure, subtract it from the reference level for your calibrator (e.g., 94, 104 or 114 dB) and if the result is a negative number that’s your maximum full-scale amplitude for calibration. If your target full scale amplitude works out to >= 0, then 0 dB full scale would be your absolute max, but you might want to call it more like −1 or −2 dB, just to make sure you don’t clip the input during calibration.

On the lower right in the Amplitude Calibration dialog is a meter that shows you the peak full-scale signal level for the selected channel. With your microphone calibrator running, adjust the gain for the input channel to your target full scale amplitude level and then click the Calibrate button. Smaart measures the input signal over a period of a few seconds and reports the full scale level. If you are happy with the result, make sure the value in the Set this value to field in the pop-up dialog matches the reference level of your calibrator, and then click the OK button.

Back in the Amplitude Calibration dialog, you will see that Smaart has calculated the necessary Offset value to calibrate the selected input to SPL. If the input device is a Smaart I-O, the Sensitivity field is populated as well. If you are using a Smaart I-O, click the Save Mic button and give it a name – Smaart will select the new microphone automatically when you have done so. Otherwise we are done. You can click OK to exit the dialog or click Apply and then select another input to calibrate.

Calibrating Based on Microphone Sensitivity

(Smaart I-O Users)

The Smaart I-O is a special case for calibration because Smaart knows the electrical sensitivity of its inputs and can read its preamp gains settings. This makes it becomes possible to calculate the combined sensitivity of the preamp and microphone, provided that the microphone sensitivity is known.

78

Chapter 5: Single Channel Measurements

Measurement microphones often come with individually measured sensitivity and frequency data, and so if your input device is a Smaart I-O and you know the sensitivity of your microphone in millivolts per

Pascal, you can enter the number in the Sensitivity field. Smaart will calculate the required offset for SPL calibration when you press the [Enter] key to set the change. You can then save this as a named microphone by clicking the Save Mic button and giving it a name.

79