Taking and Interpreting Measurements with REW (FREE eBook)

[Keith_W]

MODS: I wanted to create this thread in "Audio Reference Library" but I have no permission to do so. Would you be able to move this thread there if you think it is appropriate and delete this message. @amirm @RickS

A few years ago, Amir wrote Room Measurement Tutorial for Dummies Part 1 and Part 2. The third part has never materialized despite many requests. In the meantime, there is a clear need for clear guide on how to take room measurements because we are seeing more and more people take and post their measurements on ASR and making a lot of mistakes in the process.

LINK TO DOWNLOAD EBOOK

I have written two documents on taking and interpreting room measurements and posted it in my Google Drive. You can find both documents in the link above. You may find other free eBooks or resources in that link that I have written and decided to share.

The first document, which I will reproduce below, is an extremely basic guide to how to place your microphone and interpret the frequency response. I have tried to keep it simple and approachable. Note that as time goes on, the eBook will be updated, whereas I can not edit old posts on ASR.

The second document, which I am still writing, is a far more extensive guide that covers some speaker measurements and how to interpret other measurements in REW. At the moment you will be able to see it in Google Docs format, but when it is finished it will be replaced with a PDF.

 

[Keith_W]

Part 1: Initial Setup​

Read this section, make sure you tick every requirement, and move on to Part 2.

You will need:

1. A PC running Windows, MacOS, or Linux capable of running REW. Download the latest version of REW here: https://www.roomeqwizard.com/
2. A calibrated USB or XLR microphone. This needs to be an omnidirectional condenser microphone.
3. If you are using an XLR microphone, you will also need a microphone preamp with 48V Phantom Power, ideally an audio interface.
4. If you are using a USB microphone, you will need a DAC with suitable drivers for your operating system.

You will also need:

1. Appropriate cabling
2. A microphone tripod
3. (Optional) SPL meter

It is important to:

• Point your microphone at the ceiling if you have the appropriate calibration file, otherwise point your microphone at a point between the two speakers.
• Always use a microphone tripod
• Make sure the calibration file is loaded
• Check for correct channel assignments before you take measurements.
• Read “Getting started with REW” in the REW manual: https://www.roomeqwizard.com/help/help_en-GB/html/gettingstarted.html

All the above points are explained more extensively in the larger document.

Part 2: Room Measurements​

Please read the appropriate section in the REW manual: https://www.roomeqwizard.com/help/help_en-GB/html/makingmeasurements.html#top

The goal of a room measurement is to measure the loudspeaker response together with the room. This is different to a loudspeaker measurement, where we want to measure the response of the loudspeaker alone and independent of the room. You might need to do this if you want to DSP your loudspeaker. That is a much more difficult undertaking, and it will not be discussed here.

Measurements also need to be evaluated for quality. It needs to have an adequate signal to noise ratio, taken at a representative volume, and demonstrate what you wish to examine. If it is a measurement of a room, the microphone needs to be positioned in a representative listening position so that the specific pattern of reflections at that location is captured. Let us examine each of these in turn.

Signal to Noise Ratio (SNR)​

As the term implies, there are two terms - signal and noise. There are three variables we can manipulate to improve the SNR:

1. Choose to measure at a time when ambient noise is low, e.g. at night or weekends;
2. Increase the height (SPL) of the signal,
3. Increase the width (time) of the signal.

Noise can be a real problem and may be beyond our control. The noise in our measurement can be easily seen if we look at the waterfall graph and extend the time scale to 1000ms. I have coloured the noise floor in red:

Note that the noise floor in most typical listening rooms has a rising bass response. This is because rooms act as low-pass filters. High frequencies are easily attenuated by normal furnishings but leave long wavelengths untouched. This example shows a particularly quiet listening room, with noise levels around 30dB for higher frequencies, yet the noise rises up to 50dB for low frequencies.

Noise isn’t usually a problem with our measurements depending on what we are looking at. If there is so much noise that it is almost equal to our signal, it makes our signal unrepresentative - this typically occurs at low frequencies where most speakers have a bass drop-off at the same time the noise floor is rising.

Unless the noise floor is very high, (as a rule of thumb, less than 20dB to our signal), the measurement is valid. This is true except for all measurements which examine decay - the RT60/T30/T20, spectrograms, and waterfalls. Some measurements which have components below the noise floor, e.g. distortion, may also be contaminated. In these cases we need to increase the SNR as high as possible.

Increasing signal height / SPL. If we play our test signal louder, we can improve the SNR. However, this has its own problems as will shortly be discussed.

Increasing signal width / time. Taking a longer sweep will improve the SNR. It is said that a 45 second sweep from 10Hz to 24kHz provides 90dB of noise rejection. REW allows you to specify the length of the sweep through the above drop-down menu in the Measurements dialog box. Bear in mind that taking a very long sweep (2M, 4M) sends a lot of energy to the driver and may overheat and damage the driver if multiple sweeps are taken. The REW manual states that these longer sweeps are meant for testing electronic components and not drivers because there is a risk of mechanical damage!

Another method to increase signal width is to take multiple measurements and average them. This is my preferred approach, I set REW to take 5 sweeps and then walk away and have a coffee.

Measurement Volume​

The first thing to realize is that loudspeakers are nonlinear devices. Measuring at different volumes will probably give you a different frequency response!

This is a measurement of the Fosi SP601 taken by Erin’s Audio Corner. Link: https://www.erinsaudiocorner.com/loudspeakers/fosi_sp601/

The loudspeaker was measured at 76dB, 86dB, 96dB, and 102dB. All measurements were then processed to show deviation at different volumes and frequencies from the 76dB measurement, which is depicted as a flat black line.

As can be seen, when this speaker plays louder, it becomes progressively more nonlinear with more than 3dB bass drop-off and a downwards treble shelf. This phenomenon is especially true for small speakers with limited excursion, but nearly all loudspeakers will have a similar problem to a greater or lesser degree.

My recommendation is to measure at normal listening volume. Play some music, adjust the volume until it is what you typically listen at, and then leave the volume control untouched. If the resulting measurement looks horrendous, then it’s time for new speakers!

You will note that REW nags you to use an SPL meter, and I have avoided recommending that you buy one. An SPL meter is essential if you are taking a formal measurement, e.g. for publication. It is also required if you wish to do certain types of measurements, e.g. the dynamic compression measurement shown here, to set the noise floor, and so on. You do not need an SPL meter. As a hobbyist, you can take perfectly adequate measurements without one.

How to calibrate microphone SPL without an SPL meter​

Most smoke alarms are calibrated to produce a minimum of 75dB at 1m. Set up your microphone at 1m from the alarm, then set off the alarm by pushing the button (or by smoking a cigarette) and use REW’s controls in the Preferences menu to set the SPL.

Measurement Quality​

Every time you set up for a new measurement, you need to quickly examine the first measurement for quality and take appropriate remedial steps. If the first measurement is OK, then all subsequent measurements will likely be fine. In no particular order, here are some aspects of quality you may wish to examine:

1. Appropriate volume. Check the Distortion tab and look at THD. If this exceeds 1-2% (higher thresholds for low frequencies, up to 5%), the loudspeakers were likely played too loud.
2. Adequate signal to noise ratio. Examine the waterfall plot as previously described.
3. Correct timing reference. REW sometimes fails to “hear” the tweeter reference. Look at the delay and make sure that it makes sense.
4. Stupid mistakes. For example, forgetting to turn the subwoofer on, sweeping the wrong speaker, etc. Apply smoothing and look at the frequency response curve.

 

[Keith_W]

2.1. Room Measurements​

The intention of a room measurement is to capture loudspeaker response over a listening position, or a listening area together with reflections introduced by the room and furniture.

With all room measurements, it is highly recommended that you follow best practice:

1. All speakers and subwoofers need to be measured individually.
2. All measurements must have a loopback or acoustic timing measurement.
3. Correct microphone orientation, appropriate calibration file, and microphone tripod to be used as detailed in Part 1.

Some might argue that (1) and (2) are not strict requirements. Indeed, they are not. But if you are a beginner, it is better that you follow these recommendations with every measurement. Because, if you don’t, you might discover that you need to repeat the measurement later.

Some people recommend moving furniture out of the way when taking room measurements. My opinion is that it is not necessary, unless early reflections are interfering with your timing measurements. The typical symptoms are inconsistent deltas between one measurement and another, or wildly unrealistic deltas.

Having furniture in the way will contaminate your measurement with early reflections. My position is that you want to see these reflections in your measurement, along with the influence these have on your frequency response. Large pieces of furniture will affect measured bass response, and you want to correct for it.

The typical measurements taken are:

• Single point swept sine wave (SPS)
• Multi-point averaged swept sine wave (MPA)
• Moving Microphone Method (MMM)

2.1.1. Single Point Swept Sine Wave (SPS)​

The purpose of a SPS is to obtain timing information for all sorts of applications, especially time alignment. The frequency response taken with this method is at a very specific microphone position which is not representative of real life listening, so it should not be used exclusively for DSP correction, and it should be interpreted with caution.

The procedure is simple:

1. Mount your microphone on a tripod,
2. Place your microphone between your left and right speakers at the main listening position, pointing the mic at a central point between your speakers,
3. Read the REW manual instructions on how to take a sweep https://www.roomeqwizard.com/help/help_en-GB/html/makingmeasurements.html#top then take the sweep with a timing reference. You may optionally use an offline measurement if it is more convenient.
4. Examine for quality and decide to accept or reject the measurement.

A little note on microphone centering is in order. REW lacks a microphone centering tool. So how do you know if your microphone is centered between the speakers or not? And is it even important?

The answer is that it depends. We want to avoid egregious errors in microphone centering because it affects timing measurements. A 10cm error in microphone centering is not much if your listening triangle is large (say 3m), but it is massive if your listening triangle is small (for example, if you are listening to bookshelf speakers at a work desk).

OPTIONAL MICROPHONE CENTERING PROCEDURE

An eyeball centering is usually sufficient, but if you want to go an extra step and make sure your microphone is really centered, I have come up with two ways to do that. The second method is easier, but I have included both.

1. Sweep both speakers and examine the ETC​

Set REW to sweep both speakers together (L+R, or a mono recording). After the recording is taken, right click on the graph and choose “Set t=0 at IR peak”. Then examine the ETC:

For this demonstration, I moved the microphone 30cm to the right of center and then swept both speakers together. I have zoomed in to the first few ms of the impulse. You can clearly see two peaks which I have marked 1 and 2, representing both speakers. I placed the crosshairs on the second peak, and you can see it is 908μs delayed compared to the right. The time discrepancy can be converted to distance with d=t/1000 c (with t = time in ms, c = speed of sound 343 m/s or 1125 ft/s), or 31.14cm.

Unfortunately there is no way to know whether you should move your microphone to the left or to the right. Just eyeball it and guess! Otherwise, move your microphone in one direction and see if two peaks diverge or converge. When the two peaks coincide, your microphone is perfectly centred.

2. Sweep one speaker using the other as timing reference​

In REW’s “measurements” tab, I set REW to sweep the left speaker using the right speaker as the timing reference. When the delta is 0ms (or as close to it as possible), the microphone is perfectly centered.

This is the result for two microphone positions - this particular measurement was taken at the exact same time as the previous method. As you can see, I was able to center the microphone to 0.5mm accuracy by eyeball alone! That makes me feel rather smug. When the microphone was shifted, REW recorded a discrepancy of 306mm - roughly the same as the previous method.

2.1.2. Multi-Point Averaged Measurement​

With this procedure, multiple sweeps are taken over a given listening area and the sweeps are then averaged in REW. It is worth remembering that the diameter of a microphone capsule is about 10mm, but the width of a human head is about 150mm or more. In addition, we do not sit still whilst listening, we tend to move.

Microphone positions for an MPA. Red = mandatory positions, pink = optional for a wider area. The microphone should be placed at ear height.

You may choose to average the measurements over a smaller area (to give an idea of what an individual sitting in the sweet spot would hear) or over a larger area encompassing several seating positions. You can choose to take more measurements or fewer measurements. However, no matter what you choose, a SPS should always be performed, and clearly labelled as such.

This is the procedure:

1. Take a SPS with a time reference. Examine the measurement for quality and accept or reject the measurement. Clearly label this as the MLP (Main Listening Position) measurement.
2. Move the microphone to a new position and repeat the sweep with a time reference. Name this something sensible, for example MLP-R30 (centimeters) or MLP-L30.
3. In REW, go to the “All SPL” tab. Select all the measurements, then right click and choose “Vector Average”.

2.1.3. Moving Microphone Measurement (MMM)​

With an MMM, pink noise is played, and the microphone is moved over a listening area and the output is recorded with the RTA (Real-Time Analyzer). Detailed instructions can be found here: https://www.audiosciencereview.com/...phone-method-mmm-for-dummies-using-rew.51333/

The advantage of an MMM is that it is easy to do, quick, and repeatable. It is very fast and very foolproof compared to an MPA! The disadvantage of the MMM is that all timing information (phase, decay, etc) is not captured, so it always needs to be accompanied by a SPS.

Microphone movement in an MMM. Red = over the listening area. Pink = over a wider area to accommodate more listeners. The microphone should be swept at ear height.

Here are a few points of detail that I feel require special emphasis:

• The best MMM is taken with your microphone mounted to your tripod with the legs collapsed, converting your tripod into a very long wand. The microphone needs to be pointed at the speakers with the appropriate calibration file loaded. With this setup, I am able to position myself well away from the speaker and to the side, thus avoiding reflections from my body from contaminating the measurement.
• The microphone needs to be moved SLOWLY to avoid capturing wind noise from whooshing the microphone around in the air.
• The cable to the microphone needs to be fastened securely to the wand to avoid it slapping around and introducing unwanted microphonics into the measurement.
• Run the MMM until the reading stabilizes, then stop it. The easiest way to do this is to use a wireless keyboard or mouse. Pre-position the cursor over the START/STOP button before starting the measurement.
• Finally, make sure you take an SPS so that you have obtained some timing information.

2.1.4. Noise Floor Measurement​

Calibrate the SPL of your microphone with an SPL meter. Then place the microphone at the listening position and start the RTA without playing any signal. Take about 100 averages, then stop the RTA recorder.

This is a noise floor measurement taken in my listening room with the doors open and the doors closed. Here are a few observations:

• There is a whopping 15dB improvement by closing the doors,
• Every room I have examined has a rising bass response, in my listening room there is a 25dB difference between 20Hz and 20kHz with the doors closed,
• The difference is widest at high frequencies, and the bass frequencies are barely attenuated. This is because rooms act as low-pass filters.

(There is a dip at about 5-6kHz due to early reflections from the listening sofa).

 

[JeremyFife]

Thank you

 

[Keith_W]

Part 3: Interpreting the Frequency Response​

This is sometimes known as the amplitude response or the magnitude response. It is the first graph we should learn to read, and the only measurement we will discuss in this document. When you have mastered how to read the FR, go read the other document to learn to interpret other measurements. We will use the “SPL & Phase” tab and “All SPL” tabs.

Please read the REW manual for more tips and tricks for viewing the frequency response:
https://www.roomeqwizard.com/help/help_en-GB/html/graph_splphase.html#top
https://www.roomeqwizard.com/help/help_en-GB/html/graph_allspl.html#top

This is what you will see after you take a measurement. Take the following steps in order:

1 - Click on “SPL & Phase”
2 - Select the graph you wish to view
3 - in the Graph drop-down menu, apply smoothing. Choose 1/6 or 1/12 smoothing.
4 - make sure that the display is set to “SPL” in the drop down box (it only appears when you hover your mouse near it).
5 - IMPORTANT choose an appropriate zoom.
6 - Camera button - captures a screenshot or you can save to file. Useful for posting graphs on forums.

Zooming is so important that I will single it out for special emphasis, because many people make the same mistake.

This is what we typically see when someone who is new posts a measurement on ASR. Look how flat that frequency response appears to be. But if you look closely, the vertical scale is ±180dB! This is like taking a photograph of a lady from a mile away and marvelling at her smooth complexion. The fact is, you can’t see it.

This is what we see when a more conventional zoom is used. Now we can see what is going on! In general, frequency response graphs need to be displayed on a 50dB vertical scale.

There are 3 ways to zoom in REW:

1. The first way, which is the most precise but least convenient, is to click on the “Set graph axis limits” button. This pops up a dialog box. Enter the values you want, then click “Apply”.

2. The second way is to hold down the Ctrl key and right click and drag. You will see an area of the graph selected as shown. Click in the shaded area to zoom.

3. The third way (which is by far the easiest) is to use the scroll wheel and right mouse button to zoom and click-drag your way around the graph. When you reach the vertical scale you want, use these controls on the bottom right of the graph to quickly zoom in and view the frequencies you are interested in.

Now that our FR curve is ready for analysis, we will create a guide to help us see what is going on.

Click on REW’s EQ button, then follow these steps:

1 - If your screen does not look like this, press the arrow button.
2 - In “Target type”, choose “Full range speaker”.
3 - Click on “add room curve” and use the settings I have shown. Alternatively, you can load a house curve. You can find a collection of house curves in various formats here: https://drive.google.com/file/d/1ziiH7NaGYxyIA1LFzUNiUWY6OOV-3phZ/view?usp=drive_link
4 - Click “Calculate target level from response” and adjust the target level with the up/down arrows if necessary. You want the house curve to pass through the middle of your measurement.
5 - Finally, click on “Generate measurement from target shape”, then close the window by clicking “X”.

Go to the “All SPL” tab, and select your measurement and the target curve we just generated.

Now using REW’s helpful tonality scale, use it to guide your eyes and compare where your measurement deviates from the curve.

If we are comparing two measurements, for example before/after, it is useful to overlay the measurements on top of each other so that they can be directly compared. This is a comparison of two FR measurements with and without Dirac, showing deviation from the target curve. You can easily see the difference between the two graphs. One of them is worse than the other. Which one, and why?

A common mistake is to post individual graphs, like this. It is extremely difficult to see the difference.

Another common mistake is to do the opposite - post too many graphs overlaid on top of each other, like this. Remember: if we are comparing graphs, minimum two, maximum three. If you can’t easily see the difference yourself, chances are the rest of us won’t be able to see it either!

 

[Keith_W]

Interpreting the Frequency Response II: Basic Acoustics​

As wavelengths get longer, the way sound interacts with the room changes. This change is a function of the wavelength and the dimensions of the room. This divides the FR into several zones where sound behaves differently.

These zones can be approximately calculated using an equation described by Manfred Schroder, so it is known as the Schroder Frequency. Note that two expressions are required to calculate the Schroder Frequency (Fs): Topt (or RT60, or T30, or T20), and room volume. The Topt will be discussed in the “advanced” section. In the meantime, you can read more about the Schroder Frequency here: https://www.prosoundtraining.com/2021/10/14/divide-and-conquer-the-schroeder-frequency/

In the Pressure zone, a half wavelength of sound is longer than the greatest dimension of the room, including diagonal and oblique modes. Since modes can not be formed, the room is pressurized instead. In practice, the pressure zone either does not exist in larger rooms or is too low to be of any concern. It also requires that the room is sealed and there are no openings. A small-ish room of 4m x 4m x 2.5m has an oblique dimension of 6.2m, meaning that the pressure zone begins below 27Hz.

In the Modal Zone, defined as between the pressure zone and the Schroder frequency, wavelengths form distinct patterns of peaks and dips which represent reflections interacting with the direct sound, or with other reflections. This is normal in every room.

Note that the specific pattern of peaks and dips depends on (1) the dimensions of the room, (2) the location of the speakers and subs, and (3) the point where the measurement was taken.

A worthwhile exercise is to find the best listening position through measurement. To do this, position your speakers then take measurements in a straight line in 20cm intervals. Overlay all the measurements and look for the position that gives you the fewest or smallest dips. As you scroll through the measurements, you will see with your own eyes the pattern of room modes changing, depending on the position of the microphone.

In the Transition Zone, wavelengths get shorter and the behaviour becomes more chaotic.

In the Diffuse Field, defined as four times the Schroder Frequency, sound behaves more like “beams” and less like “waves”.

Credit: JL Ohl, loudspeakers.audio. Image used with permission.

This illustration shows the same speaker (a Behringer B2030A) in 9 different listening rooms. The author indicated that the measurement distance was “between 2m and 3m”. We can see that above 700Hz or so, all the loudspeakers perform almost the same - the curves barely diverge. Below 700Hz, we can see that the performance varies between ±30dB!

The take-home message is: in the diffuse field, loudspeakers will likely perform the same as published measurements. Below this, loudspeakers are heavily influenced by rooms.

Interpreting the Frequency Response III: Analysis of a loudspeaker measurement​

To illustrate this point, we will compare a home measurement of a well-known speaker with a professional measurement. This is an in-room measurement of a KEF R3 Meta loudspeaker posted on Audio Science Review.

We are fortunate that Erin has reviewed this speaker and published measurements. They can be found at: https://www.erinsaudiocorner.com/loudspeakers/kef_r3_meta/

I adjusted the vertical scale of the REW measurement to match Erin’s. I then copy-pasted both measurements into an image editing program (Gimp). Using the Magic Wand selection tool, I deleted the background of Erin’s measurement and then scaled both graphs and overlaid them for easy comparison.

We can easily see the differences. The home measurement and published measurements are remarkably similar between 150Hz - 4kHz. The responses diverge above 4kHz, and there are some room modes below 150Hz. Since the room modes are expected, the question is: why do the two measurements diverge above 4kHz?

The typical reasons are:

• Using the microphone in the wrong orientation or with the wrong calibration file. As we learnt earlier, microphones oriented vertically with the wrong calibration file loaded will show a typical treble drop-off above 5-6kHz.
• This may be an off-axis measurement. The KEF R3 is a well behaved speaker, but its response narrows in the treble if it is measured off-axis.
• There may be inappropriate DSP processing.
• The speaker may be faulty (a quick comparison with the other speaker indicated that both were the same, so this was not the reason).

Quite often, professional measurements are not available, which means we have no comparison. We are then even more responsible to make sure that the measurements have been taken correctly.

Here is another measurement from someone else at ASR. See that rising treble response above 10kHz? Typical reasons are:

• Wrong microphone orientation / calibration file (again!).
• Measuring the speakers too loud. Remember that harmonic distortion produces high order harmonics which may affect treble response. Tweeter breakup modes may do the same.
• Measuring in close proximity to a reflective surface. A mini tripod perched on a sofa will certainly produce a measurement that looks like that.
• Inappropriate DSP.
• Poor speaker design. This turned out to be the culprit.

And here is another measurement that I took myself, so I know it was taken properly (ahem). I was flabbergasted when I saw the measurement, I could not believe it. This is not some DIY hack speaker, it is actually being sold to paying customers!

As you can see, both left and right speakers have the same problem. I tried measuring at various distances, and I wanted to try things like invert the polarity of the tweeter, which I was unable to (since I had no access to the crossover). I wondered if it was floor bounce producing cancellation - this would go away if I elevated the speaker. But I did not have time to do that. Given it was a dipole, I tried moving the speaker further into the room to see if it would go away. It didn’t. In the end, some speakers are just faulty by design!

The take-home message is this: if you see something odd, make sure that you were not the cause! I can tell you from personal experience that by far the majority of funny looking measurements are a result of user error. Once this is excluded, then we can blame the designer.

 

[Keith_W]

Interpreting the Frequency Response IV: Bass Frequencies​

All room measurements of bass frequencies (i.e. below the Schroder frequency) will have peaks and dips in the response. This is normal. However, identifying the cause of these anomalies may be difficult. I like to divide the causes into 3 categories:

1. “Local” phenomena. The peaks/dips are caused by reflections from the chair or your body. If you move your microphone a short distance, or if you remove the chair, the pattern changes radically and may even disappear. The solution to this problem is to ignore it.
2. “Room” phenomena. The peaks/dips are caused by room modes, i.e. sound interacting with its reflection, sometimes known as the Allison effect. If you move your microphone a short distance, the pattern does not change much. You only see big differences if you move your microphone a large distance relative to the wavelength (about ¼ wavelength, so for e.g. if the dip is at 50Hz, a quarter wavelength is 1.72m!). The solution will be discussed below.
3. “Speaker” phenomena. The peaks/dips are due to interaction between speakers, for example crossover phase cancellation. These changes tend to stubbornly persist no matter where you move your microphone, although some positions in the room may cause a peak in that frequency and remove a null due to XO cancellation. You can diagnose this by taking measurements at 1m, 2m, 3m closer to the speakers and observe what happens. If it stays the same, it is likely XO phase cancellation. If it changes by a lot, it is likely a room mode. The cure is better time/phase alignment of your speakers and subwoofers.

As you can see, diagnosing the cause of peaks and dips in the bass region involves moving your microphone to new positions and taking measurements for comparison. This is why I prefer the MPA (multi-point averaged) measurement when taking room measurements. Half your work is already done!

It also helps to think about bass frequencies in a structured way and follow these general principles:

1. Bass peaks are more audible and objectionable than narrow bass dips. Broad bass dips are another matter entirely.
2. However, bass peaks are more easily manipulated with DSP. It is simple to chop off peaks with equalisation.
3. Bass dips can be improved with DSP, but the process is much more difficult. This is an advanced DSP discussion.

Because of the above points, I suggest this approach to addressing bass problems: the first step is to find a speaker/subwoofer and listening position that has as many peaks and as few dips as possible. The next step is to remove all the peaks with DSP equalisation, leaving you with residual dips. After this, decide how objectionable you find these dips and whether you want to do something about it. This may involve advanced DSP techniques or the purchase of additional subwoofers.

You will note that I did not say “use acoustic treatment” to address bass problems. This is because it is impractical for most rooms (they need to be very large and a large surface area needs to be treated), or difficult to tune (Helmholtz resonators or membrane absorbers), and because you will need DSP anyway.

 

[3ll3d00d]

Point your microphone at the speaker, and not at the ceiling. Only point the microphone at the ceiling if you have the appropriate calibration file.

Nice writeup. I am curious why you recommend this for a beginner doing room measurements though unless you mean to move the mic for each individual speaker? I think the conventional approach is point at individual speaker for speaker measurements and point at ceiling for room measurements as that eliminates the problem of varying off axis angle the mic measures at. For stereo setup this might be the same degree off axis but it will still likely mean some roll-off at high frequency which then has a risk of over correction.

 

[Keith_W]

Good point. I normally point my mic between the speakers, but then I have a rather large listening triangle. I'll amend it.

 

[janbth]

Interpreting the Frequency Response IV: Bass Frequencies​

All room measurements of bass frequencies (i.e. below the Schroder frequency) will have peaks and dips in the response. This is normal. However, identifying the cause of these anomalies may be difficult. I like to divide the causes into 3 categories:

1. “Local” phenomena. The peaks/dips are caused by reflections from the chair or your body. If you move your microphone a short distance, or if you remove the chair, the pattern changes radically and may even disappear. The solution to this problem is to ignore it.
2. “Room” phenomena. The peaks/dips are caused by room modes, i.e. sound interacting with its reflection, sometimes known as the Allison effect. If you move your microphone a short distance, the pattern does not change much. You only see big differences if you move your microphone a large distance relative to the wavelength (about ¼ wavelength, so for e.g. if the dip is at 50Hz, a quarter wavelength is 1.72m!). The solution will be discussed below.
3. “Speaker” phenomena. The peaks/dips are due to interaction between speakers, for example crossover phase cancellation. These changes tend to stubbornly persist no matter where you move your microphone, although some positions in the room may cause a peak in that frequency and remove a null due to XO cancellation. You can diagnose this by taking measurements at 1m, 2m, 3m closer to the speakers and observe what happens. If it stays the same, it is likely XO phase cancellation. If it changes by a lot, it is likely a room mode. The cure is better time/phase alignment of your speakers and subwoofers.

As you can see, diagnosing the cause of peaks and dips in the bass region involves moving your microphone to new positions and taking measurements for comparison. This is why I prefer the MPA (multi-point averaged) measurement when taking room measurements. Half your work is already done!

It also helps to think about bass frequencies in a structured way and follow these general principles:

1. Bass peaks are more audible and objectionable than narrow bass dips. Broad bass dips are another matter entirely.
2. However, bass peaks are more easily manipulated with DSP. It is simple to chop off peaks with equalisation.
3. Bass dips can be improved with DSP, but the process is much more difficult. This is an advanced DSP discussion.

Because of the above points, I suggest this approach to addressing bass problems: the first step is to find a speaker/subwoofer and listening position that has as many peaks and as few dips as possible. The next step is to remove all the peaks with DSP equalisation, leaving you with residual dips. After this, decide how objectionable you find these dips and whether you want to do something about it. This may involve advanced DSP techniques or the purchase of additional subwoofers.

You will note that I did not say “use acoustic treatment” to address bass problems. This is because it is impractical for most rooms (they need to be very large and a large surface area needs to be treated), or difficult to tune (Helmholtz resonators or membrane absorbers), and because you will need DSP anyway.

Regarding point 2. "Room" phenomena, I think that you might me mixing up room modes and SBIR (the Allison effect). Room modes are the standing wave patterns in a room and exist independently of where the speakers and listeners are located. They are different from SBIR, which is a cancellation effect.

 

[Keith_W]

It is true that the original "Allison effect" referred to SBIR. But I have heard people use the term "Allison effect" to describe any phenomena where the direct sound interacts with its reflection, not just SBIR. To me, making the distinction between SBIR and room modes is less important these days. Playing with the REW room mode simulator tells me that you can not isolate the effect of moving the loudspeaker closer to the wall in an enclosed room. Maybe the distinction might be important if one end of the room is open to the free field. But in an enclosed room ... no.

 

[goryu]

MODS: I wanted to create this thread in "Audio Reference Library" but I have no permission to do so. Would you be able to move this thread there if you think it is appropriate and delete this message. @amirm @RickS

A few years ago, Amir wrote Room Measurement Tutorial for Dummies Part 1 and Part 2. The third part has never materialized despite many requests. In the meantime, there is a clear need for clear guide on how to take room measurements because we are seeing more and more people take and post their measurements on ASR and making a lot of mistakes in the process.

LINK TO DOWNLOAD EBOOK

I have written two documents on taking and interpreting room measurements and posted it in my Google Drive. You can find both documents in the link above. You may find other free eBooks or resources in that link that I have written and decided to share.

The first document, which I will reproduce below, is an extremely basic guide to how to place your microphone and interpret the frequency response. I have tried to keep it simple and approachable. Note that as time goes on, the eBook will be updated, whereas I can not edit old posts on ASR.

The second document, which I am still writing, is a far more extensive guide that covers some speaker measurements and how to interpret other measurements in REW. At the moment you will be able to see it in Google Docs format, but when it is finished it will be replaced with a PDF.

When I click the link all I am seeing are two accourate files and a target curve file...????

 

[FrantzM]

Thank you

Can I multiply this by a million?

Thanks Keith!

 

[WeekendWarrior]

This looks really useful, many thanks for taking the considerable time to do this.

Can I just check the content in the link you included. This is what is available when I follow it, which doesn't seem to match the description.

 

[janbth]

To me, making the distinction between SBIR and room modes is less important these days. Playing with the REW room mode simulator tells me that you can not isolate the effect of moving the loudspeaker closer to the wall in an enclosed room. Maybe the distinction might be important if one end of the room is open to the free field. But in an enclosed room ... no.

SBIR and room modes are different phenomena. SBIR typically shows up above 100 Hz, while the lowest lying modal peaks appear well below. Moving the speaker will of course change the way the modes are excited in the frequency response, but it is not due to SBIR.

If someone who is starting out doing measurements sees a peak at, say, 45 Hz, and wants to interpret that peak, he or she must learn about room modes (standing waves). In my opinion, the most useful thing a beginner can do to combat resonance peaks is to map out the standing wave pattern in the listening room. This will provide information about where to position speakers and listener, and whether to get more subs.

 

[Keith_W]

When I click the link all I am seeing are two accourate files and a target curve file...????

Sorry guys. I have changed the permission of the files, it should appear now.

(Edit) and I did suspect my use of the "Allison effect" might upset some pedants. I'll go and change it and hopefully you'll be happy now.

 

[mosco]

Thank you for the initiative. I suggest you keep an editable PDF version of the book in addition to future editions so it can be translated for non-native English speakers.

 

[goryu]

Sorry guys. I have changed the permission of the files, it should appear now.

(Edit) and I did suspect my use of the "Allison effect" might upset some pedants. I'll go and change it and hopefully you'll be happy now.

Thank you!

 

[janbth]

(Edit) and I did suspect my use of the "Allison effect" might upset some pedants. I'll go and change it and hopefully you'll be happy now.

I guess the pedant is me, but it's not about what to call it - SBIR or the Allison effect. The point is that these two are different from room modes.

The peaks/dips are caused by room modes, i.e. sound interacting with its reflection...

No. Peaks and dips can be caused by either room modes or by SBIR/Allison effect, but they are not the same. Room modes has to do with resonances, not reflections.

 

[AudioJester]

Thanks so much Keith!