Keith_W
Major Contributor
I have been helping new users interpret REW measurements on ASR and I see many of the same mistakes being made all the time. Rather than repeat myself, I thought that a guide might help.
This is NOT a guide to take measurements or how to use REW. The REW manual is a treasure trove of information, and it is extensive. The only problem is that you may not know where to look if you are new to measurements.
PART 1: HOW TO TAKE MEASUREMENTS
1. Before you start ...
These items are essential before you start taking measurements with REW:
- PC/Mac capable of running REW.
- A USB or XLR microphone. If you choose an XLR microphone, you also need an interface with 48V Phantom Power
- Appropriate cabling for your microphone and interface
- A microphone tripod with a boom.
- Read the getting started with REW guide.
2. What microphone to use
Use an omnidirectional condenser type microphone. Microphones are not the same! Some microphones may be optimised for voice, some may have a cardioid pattern and reject reflections, etc. Omnidirectional condenser microphones work by varying the capacitance in the capsule membrane, and thus need to be powered.
There are two types: XLR and USB. XLR microphones require a microphone preamp that supplies voltage, usually 48V. This is known as "Phantom Power". It is built into interfaces. Common XLR microphones used by ASR members include: Behringer ECM8000, Dayton EMM-6, iSemCon EMX-7150, Earthworks M23/M30. USB microphones have the power supply and ADC built into the mic. Common USB microphones used here include: UMIK-1, UMIK-2, Dayton UMM-6.
3. Calibration
Microphones are manufactured to a tolerance and may have large variations in frequency response if not calibrated. The above shows the variance in response of a few dozen Behringer ECM8000 microphones - you can see that the variance is highest at both extremes of the frequency range. Make sure the microphone you buy comes with a calibration file. If not, send it to a calibration service. Also note that mics can go out of calibration after use, especially if the mic has been physically abused (subjected to temperature extremes, dropped, jiggled around, etc).
4. Why you need a mic boom tripod
Some microphones come with mini tripods (e.g. UMIK-1 and UMIK-2, Anthem ARC microphone, etc). These are good for landfill only. The first thing you should buy is a mic boom stand. This is because microphones need to be positioned away from reflective surfaces to take a meaningful measurement. Even the design of the mic boom can create reflections that will affect your measurement, but we don't usually worry about those because they are low in amplitude and can be ignored unless extreme accuracy is desired.
5. Microphone orientation
Microphones need to be pointed at the speakers, and not up at the ceiling. The reason is because "omnidirectional" microphones are not truly omnidirectional, they have a polar response and will measure a different frequency response if it is pointed in the wrong direction. Read this. The only time microphones can be pointed upwards is if they include a equalisation file for vertical orientation. Otherwise - (as a presidential candidate said) ... DON'T.
6. Moving furniture
Some guides might direct you to move furniture away so there are no reflective surfaces close to the speaker. Some tell you to leave furniture in place. What is going on?
The key is to understand that long and short wavelengths behave differently in listening rooms. The strategy to measure them is different, and the interventions we use to fix these problems are also different. The zones are defined by the Schroder frequency. It is very important to understand this concept. (As an aside: If you have high frequency problems, the best approach is to (a) buy better speakers, (b) consider repositioning your speakers, (c) try to fix anomalies with room treatment, (d) apply gentle DSP. In that order! If you have low frequency problems, the approach is to (a) consider adding more subwoofers, (b) DSP. Room treatment has a minimal role because it is less effective and excessively intrusive.)
When measuring long wavelengths, we want to include all the influence of room furnishings in our measurement because that affects how bass behaves. This includes: sofas, coffee tables, and other large furnishings. When measuring short wavelengths, we want to measure the output from the loudspeakers alone because that determines what we hear. This can be very difficult in a listening room, particularly if it is small. So we may decide to: (a) do the measurement anyway but interpret upper frequencies with caution, (b) choose not to correct the upper frequencies because we know we are not measuring them properly, (c) attempt to get a quasi-anechoic measurement by taking the speakers outside and away from nearby reflective surfaces.
If your room is large enough, you can move your speakers and microphone to the middle of the room to measure upper frequencies. You will be able to obtain a meaningful measurement to a lower limit if you window the measurement appropriately. The lower limit is defined by the distance between mic/speaker to the closest reflective surface (e.g. walls, floor, ceiling). See above diagram.
7. Choose the correct SPL for measurement
Measurement sweeps need to be: "not too loud, and not too soft".
"Not too loud": loudspeakers, and to a lesser extent amplifiers, are nonlinear devices. At high volumes they distort, compress, and clip. This WILL affect your measurement, for example high amounts of harmonic distortion will look like a rising frequency response. Loudspeaker compression will look like deviations from the frequency response depending on which driver is compressing.
"Not too soft": every listening room has a noise floor. Most of the noise in listening rooms is low frequency, so you may not be aware of the presence of a lot of noise because it is less audible and our brain filters it out. It will look like a rising bass response. The above graph shows a full range (20Hz - 20kHz) measurement of a tweeter (I was able to obtain this measurement without damaging the tweeter by choosing a very short pulse). All that bass you see is the room's noise floor, it is not tweeter output!
The goal of a measurement is to obtain a high signal to noise ratio. Noise can be minimised by choosing to do your measurement when ambient noise is low (evenings, weekends). Signal quality can be improved by increasing the amplitude of the impulse (i.e. turn up the volume) or length (time) of the impulse. A 45 second sweep has about 90dB of noise rejection. You could also take multiple short sweeps, reject those that look flawed or noisy, and average them. How to average graphs in REW.
You will notice I did not include "SPL Meter" in the purchase list, despite REW nagging you to calibrate your mic to an SPL meter every time you boot it up. Why? This is because knowledge of your SPL does not guarantee that your speakers will not behave in a nonlinear fashion. The best way to determine if your speaker is behaving strangely is to measure at different volumes (typically 75dB, 85dB, 95dB - but you can increase the volume in REW by 5-10dB each time). Then overlay the graphs and examine it. They should all look the same, but they won't. At some point the frequency response will start to deviate. You want to measure at a volume lower than where the deviation occurs.
SPL meters are good for determining the noise floor when interpreting spectrograms and waterfalls and for compression measurements. They have a number of other uses, but most of us do not need SPL meters.
8. Before and after measurements
You might want to know the effect of DSP, room treatments, different speakers, etc. The key is to repeat the measurement whilst keeping all variables the same, except the one you are testing. This means: don't move your microphone, or at least take steps to make sure you can re-position your microphone in the exact same position when comparing measurements. Some people hang a plumb line from the ceiling so they know where to place the microphone. Don't make multiple changes at once, e.g. one guy I am trying to help is measuring with one mic in five positions using his HT AVR and verifying with another mic using MMM and a different software program. One variable at a time!
9. Avoid speaker and hearing damage.
Be aware that speaker drivers can overheat and voice coils can melt. Do NOT play loud sine wave tones through your subwoofer for a long time. Do NOT allow your amplifier to clip - this sends high frequency tones to the tweeter and can burn it out. Sometimes, repeated loud measurements can cause drivers to overheat and behave in a nonlinear fashion - i.e. repeated sweeps can cause the frequency response to change!
Wear hearing protection, or exit the room while taking the measurement. You can use REW to delay the start of the measurement to give you enough time to leave.
10. Special measurement techniques
Special procedures are required if we want other types of information. For example:
- if we want to know the bass performance of a subwoofer or speaker independent of the room. Speaker designers and reviewers need to know this. Most hobbyists don't. What you need to know is how your sub or speaker is performing in your room.
- the average frequency response over a listening area. You will need to do an MMM or multi-point averaged measurement. I am not going to discuss that here to keep the length shorter.
- quasi-anechoic response of the loudspeaker and directivity measurements. You need to know this if you want to correct the upper frequencies with DSP. There are several techniques but I won't discuss them here.
- speaker compression.
- loudspeaker port measurement
- measurements of difficult speakers, e.g. dipoles and omnis.
PART 2: HOW TO ASK FOR HELP ON ASR
For us to help you (and we want to help!) we need:
1. Confidence that you have taken a proper measurement. Read the above and make sure you have not made any mistakes. This is very important. We don't know if the anomalies are real or due to an artefact from improper measurement! In your post, state that you have read this guide, and specifically state that you have used the proper microphone, pointed it at the speakers, used a mic boom tripod, and taken noise precautions.
2. The intention of the measurement and the question you are asking. For example, "my left speaker sounds louder than my right, but the measurements look the same! Why?". A question like "please comment on this graph" is not so helpful.
3. Details of your room: height, width, length (people always forget about the height!). Where is the listening position, where are the speakers. A quick drawing is very helpful.
4. Details on your equipment you are measuring: how many speakers, subs, how many DAC channels, etc. and how you have connected them together.
5. Properly labelled graphs. Edit the name of each graph in REW so that the labels CLEARLY tell us what we are looking at, e.g. "L spkr + L sub" or "L sub - no DSP". You can rename graphs by right clicking on the graph panel on the left.
By far the best way to ask for help is to post the .MDAT file on ASR. To do this, you need to zip the .MDAT and use "Attach files" when posting your thread.
If you do want to post the graphs (and a few graphs are always helpful), please try to follow these guidelines:
1. Choose a proper zoom. If you are asking a question about bass, zoom in on the bass. REW has a number of options that help you scale the graph for easier viewing:
a) "fit to data" option. (see above)
b) Zoom in to an area of interest. To do this, hold down the CTRL key and right click on the graph to drag a box. Then click on the shaded area to zoom.
c) On the bottom right of your graph, you will see two buttons - "10 .. 200" and "20 .. 20k". One automatically zooms to show the bass, the other to full range.
2. Choose proper smoothing. Normally 1/6 or 1/12 octave smoothing.
2. If using a screenshot, make sure we can see the vertical / horizontal scales and labels CLEARLY. This is the best way to post a screenshot: click on the camera (top left) then click on copy to clipboard. Then go to ASR and type "Ctrl-V" to paste.
3. Do not post too many curves in one graph. It makes it hard to see.
Lastly, remember that ASR members are human! If you post a thread without sufficient information, too many mistakes in measurements, a poorly asked question, etc. - many ASR members will not bother to help. I can't blame them. I often help because everyone is ignoring the newbie and I feel pity. These guys are effing grumpy. So please, help us to help you.
This is NOT a guide to take measurements or how to use REW. The REW manual is a treasure trove of information, and it is extensive. The only problem is that you may not know where to look if you are new to measurements.
PART 1: HOW TO TAKE MEASUREMENTS
1. Before you start ...
These items are essential before you start taking measurements with REW:
- PC/Mac capable of running REW.
- A USB or XLR microphone. If you choose an XLR microphone, you also need an interface with 48V Phantom Power
- Appropriate cabling for your microphone and interface
- A microphone tripod with a boom.
- Read the getting started with REW guide.
2. What microphone to use
Use an omnidirectional condenser type microphone. Microphones are not the same! Some microphones may be optimised for voice, some may have a cardioid pattern and reject reflections, etc. Omnidirectional condenser microphones work by varying the capacitance in the capsule membrane, and thus need to be powered.
There are two types: XLR and USB. XLR microphones require a microphone preamp that supplies voltage, usually 48V. This is known as "Phantom Power". It is built into interfaces. Common XLR microphones used by ASR members include: Behringer ECM8000, Dayton EMM-6, iSemCon EMX-7150, Earthworks M23/M30. USB microphones have the power supply and ADC built into the mic. Common USB microphones used here include: UMIK-1, UMIK-2, Dayton UMM-6.
3. Calibration
Microphones are manufactured to a tolerance and may have large variations in frequency response if not calibrated. The above shows the variance in response of a few dozen Behringer ECM8000 microphones - you can see that the variance is highest at both extremes of the frequency range. Make sure the microphone you buy comes with a calibration file. If not, send it to a calibration service. Also note that mics can go out of calibration after use, especially if the mic has been physically abused (subjected to temperature extremes, dropped, jiggled around, etc).
4. Why you need a mic boom tripod
Some microphones come with mini tripods (e.g. UMIK-1 and UMIK-2, Anthem ARC microphone, etc). These are good for landfill only. The first thing you should buy is a mic boom stand. This is because microphones need to be positioned away from reflective surfaces to take a meaningful measurement. Even the design of the mic boom can create reflections that will affect your measurement, but we don't usually worry about those because they are low in amplitude and can be ignored unless extreme accuracy is desired.
5. Microphone orientation
Microphones need to be pointed at the speakers, and not up at the ceiling. The reason is because "omnidirectional" microphones are not truly omnidirectional, they have a polar response and will measure a different frequency response if it is pointed in the wrong direction. Read this. The only time microphones can be pointed upwards is if they include a equalisation file for vertical orientation. Otherwise - (as a presidential candidate said) ... DON'T.
6. Moving furniture
Some guides might direct you to move furniture away so there are no reflective surfaces close to the speaker. Some tell you to leave furniture in place. What is going on?
The key is to understand that long and short wavelengths behave differently in listening rooms. The strategy to measure them is different, and the interventions we use to fix these problems are also different. The zones are defined by the Schroder frequency. It is very important to understand this concept. (As an aside: If you have high frequency problems, the best approach is to (a) buy better speakers, (b) consider repositioning your speakers, (c) try to fix anomalies with room treatment, (d) apply gentle DSP. In that order! If you have low frequency problems, the approach is to (a) consider adding more subwoofers, (b) DSP. Room treatment has a minimal role because it is less effective and excessively intrusive.)
When measuring long wavelengths, we want to include all the influence of room furnishings in our measurement because that affects how bass behaves. This includes: sofas, coffee tables, and other large furnishings. When measuring short wavelengths, we want to measure the output from the loudspeakers alone because that determines what we hear. This can be very difficult in a listening room, particularly if it is small. So we may decide to: (a) do the measurement anyway but interpret upper frequencies with caution, (b) choose not to correct the upper frequencies because we know we are not measuring them properly, (c) attempt to get a quasi-anechoic measurement by taking the speakers outside and away from nearby reflective surfaces.
If your room is large enough, you can move your speakers and microphone to the middle of the room to measure upper frequencies. You will be able to obtain a meaningful measurement to a lower limit if you window the measurement appropriately. The lower limit is defined by the distance between mic/speaker to the closest reflective surface (e.g. walls, floor, ceiling). See above diagram.
7. Choose the correct SPL for measurement
Measurement sweeps need to be: "not too loud, and not too soft".
"Not too loud": loudspeakers, and to a lesser extent amplifiers, are nonlinear devices. At high volumes they distort, compress, and clip. This WILL affect your measurement, for example high amounts of harmonic distortion will look like a rising frequency response. Loudspeaker compression will look like deviations from the frequency response depending on which driver is compressing.
"Not too soft": every listening room has a noise floor. Most of the noise in listening rooms is low frequency, so you may not be aware of the presence of a lot of noise because it is less audible and our brain filters it out. It will look like a rising bass response. The above graph shows a full range (20Hz - 20kHz) measurement of a tweeter (I was able to obtain this measurement without damaging the tweeter by choosing a very short pulse). All that bass you see is the room's noise floor, it is not tweeter output!
The goal of a measurement is to obtain a high signal to noise ratio. Noise can be minimised by choosing to do your measurement when ambient noise is low (evenings, weekends). Signal quality can be improved by increasing the amplitude of the impulse (i.e. turn up the volume) or length (time) of the impulse. A 45 second sweep has about 90dB of noise rejection. You could also take multiple short sweeps, reject those that look flawed or noisy, and average them. How to average graphs in REW.
You will notice I did not include "SPL Meter" in the purchase list, despite REW nagging you to calibrate your mic to an SPL meter every time you boot it up. Why? This is because knowledge of your SPL does not guarantee that your speakers will not behave in a nonlinear fashion. The best way to determine if your speaker is behaving strangely is to measure at different volumes (typically 75dB, 85dB, 95dB - but you can increase the volume in REW by 5-10dB each time). Then overlay the graphs and examine it. They should all look the same, but they won't. At some point the frequency response will start to deviate. You want to measure at a volume lower than where the deviation occurs.
SPL meters are good for determining the noise floor when interpreting spectrograms and waterfalls and for compression measurements. They have a number of other uses, but most of us do not need SPL meters.
8. Before and after measurements
You might want to know the effect of DSP, room treatments, different speakers, etc. The key is to repeat the measurement whilst keeping all variables the same, except the one you are testing. This means: don't move your microphone, or at least take steps to make sure you can re-position your microphone in the exact same position when comparing measurements. Some people hang a plumb line from the ceiling so they know where to place the microphone. Don't make multiple changes at once, e.g. one guy I am trying to help is measuring with one mic in five positions using his HT AVR and verifying with another mic using MMM and a different software program. One variable at a time!
9. Avoid speaker and hearing damage.
Be aware that speaker drivers can overheat and voice coils can melt. Do NOT play loud sine wave tones through your subwoofer for a long time. Do NOT allow your amplifier to clip - this sends high frequency tones to the tweeter and can burn it out. Sometimes, repeated loud measurements can cause drivers to overheat and behave in a nonlinear fashion - i.e. repeated sweeps can cause the frequency response to change!
Wear hearing protection, or exit the room while taking the measurement. You can use REW to delay the start of the measurement to give you enough time to leave.
10. Special measurement techniques
Special procedures are required if we want other types of information. For example:
- if we want to know the bass performance of a subwoofer or speaker independent of the room. Speaker designers and reviewers need to know this. Most hobbyists don't. What you need to know is how your sub or speaker is performing in your room.
- the average frequency response over a listening area. You will need to do an MMM or multi-point averaged measurement. I am not going to discuss that here to keep the length shorter.
- quasi-anechoic response of the loudspeaker and directivity measurements. You need to know this if you want to correct the upper frequencies with DSP. There are several techniques but I won't discuss them here.
- speaker compression.
- loudspeaker port measurement
- measurements of difficult speakers, e.g. dipoles and omnis.
PART 2: HOW TO ASK FOR HELP ON ASR
For us to help you (and we want to help!) we need:
1. Confidence that you have taken a proper measurement. Read the above and make sure you have not made any mistakes. This is very important. We don't know if the anomalies are real or due to an artefact from improper measurement! In your post, state that you have read this guide, and specifically state that you have used the proper microphone, pointed it at the speakers, used a mic boom tripod, and taken noise precautions.
2. The intention of the measurement and the question you are asking. For example, "my left speaker sounds louder than my right, but the measurements look the same! Why?". A question like "please comment on this graph" is not so helpful.
3. Details of your room: height, width, length (people always forget about the height!). Where is the listening position, where are the speakers. A quick drawing is very helpful.
4. Details on your equipment you are measuring: how many speakers, subs, how many DAC channels, etc. and how you have connected them together.
5. Properly labelled graphs. Edit the name of each graph in REW so that the labels CLEARLY tell us what we are looking at, e.g. "L spkr + L sub" or "L sub - no DSP". You can rename graphs by right clicking on the graph panel on the left.
By far the best way to ask for help is to post the .MDAT file on ASR. To do this, you need to zip the .MDAT and use "Attach files" when posting your thread.
If you do want to post the graphs (and a few graphs are always helpful), please try to follow these guidelines:
1. Choose a proper zoom. If you are asking a question about bass, zoom in on the bass. REW has a number of options that help you scale the graph for easier viewing:
a) "fit to data" option. (see above)
b) Zoom in to an area of interest. To do this, hold down the CTRL key and right click on the graph to drag a box. Then click on the shaded area to zoom.
c) On the bottom right of your graph, you will see two buttons - "10 .. 200" and "20 .. 20k". One automatically zooms to show the bass, the other to full range.
2. Choose proper smoothing. Normally 1/6 or 1/12 octave smoothing.
2. If using a screenshot, make sure we can see the vertical / horizontal scales and labels CLEARLY. This is the best way to post a screenshot: click on the camera (top left) then click on copy to clipboard. Then go to ASR and type "Ctrl-V" to paste.
3. Do not post too many curves in one graph. It makes it hard to see.
Lastly, remember that ASR members are human! If you post a thread without sufficient information, too many mistakes in measurements, a poorly asked question, etc. - many ASR members will not bother to help. I can't blame them. I often help because everyone is ignoring the newbie and I feel pity. These guys are effing grumpy. So please, help us to help you.
Last edited:
