Hi everyone. This is my first thread in this forum …
Since we on a nearly daily basis see posts featuring graphs from REW or other measurement software with scales set as if it was my daughter (she´s one year old …) dictating the limits, I hereby provide you with a very short, but hopefully useful guide to what is considered useful. Please check it out before posting graphs, or the risk is overwhelming that you’re going to be asked to repost them with proper scales. If unsure about what scales to use and what graphs to include, why not simply upload the actual IR-files (either in REW format or exported as wav/aiff if captured in other software), so that others can open/import it into the software preferred and inspect it without limitations. Please label your measurement files/pics properly to avoid confusion.
First up, the one graph everyone knows about and unfortunately (since it’s the least interesting one), often the only one featured in a lot of posts: the Frequency Response curve
: Frequency range:
Well, not much to say about it really; 20 Hz (or 10-15 Hz if subwoofer in larger room) to 20 kHz (using a logarithmic scale) is normal (assuming the measurement mic is accurate to this range, more often; about 15 kHz is the practical limit) but the range can be limited in order to reveal more detail depending on the smoothing used (see below). Unless you know how to accurately center the measurement microphone exactly (millimeters count) between the L & R speaker *, plot L & R speaker separately in full range and plot the sum (L & R measured simultaneously) showing only the lower range (to about 300-500 Hz) since the range above it will suffer from comb filter effects due to the two signals arriving in slightly different time to the microphone. dB range:
Although there’s no right or wrong here, I personally prefer a 50 dB range (usually +20/-30 if amplitude normalized to 0 dB) since it will fit close to any measured response, even most untreated rooms, and it´s nice to be able to make comparisons without changing the scale. If you prefer another range, use one that fits the curve but not much more ( … unless you deliberately want to make your room look better than it actually is but if so, the next part will interest you even more): Smoothing (“resolution”):
A FR-curve without smoothing is often not very useful, especially in the midrange and up (and particularly if the room is untreated) since it is hard to read due to the extreme variation in amplitude and this is why we apply “smoothing” to make the curve more readable, but this can also be used to make the curve look a lot flatter than it actually is. A 1/24 octave resolution or higher (1/48 etc.) is what you want to use when doing measurements in critical listening rooms or you´ll miss out important information. Some like to also provide graphs featuring 1/3 octave resolution, and that’s fine as long as a high resolution (1/24 oct or higher) version is also included. A 1/3 oct FR curve alone is seldom very useful, especially if the modal region (below about 200-300 Hz) is the most interesting range. For this frequency range, you can actually turn the smoothing off completely (but do state that this is the case if it´s not clear by the pic).
As a side note; don´t freak out if your response is showing +/- 10 dB or worse. Even in very big, expensive control rooms, a +/- 5 dB response (assuming 1/24 dB resolution) is considered excellent. Also, the FR curve is not the only graph of interest (it´s actually the least interesting one). Waterfall / Decay / Sonogram
These plots show spectral energy decay. It is used to identify frequency regions that suffer from too long (or too short) decay time. Room modes for example, are easily identified using these plots. Frequency range:
The most common use of these plots is to identify modes and if so, the range below about 300 Hz is the most relevant but it is also useful to see a full range plot since it might reveal uneven decay times if for example a room suffers from overuse of broadband absorption in the high frequency range resulting in a room with very short decay in the highs compared to the lows. dB scale:
Again, no strict rules here but to only show a 30 dB range would not reveal the complete picture so I prefer a 60 dB range (since this is what is normally referred to when talking about decay times). By "range" I´m not only referring to the vertical scale but you actually want to see a 60 dB range of decay and if your graph starts halfway up, you´re only seeing the first 30 dBs of decay if the total scale is 60 dB. You want to see a 60 dB decay from the average level of your meassurement, so if normalized to 0 dB, about +10-15 dB (depending on the peak of the meassurement) to -60 dB or lower is a good scale. If your measurement is too noisy, try and redo them using multiple sweeps in order to increase the SNR (oh, and there’s no real benefits in restricting the frequency range when making measurements. Use 10-20 Hz to 20 kHz and a sweep length of about 5-10 seconds, 256k in REW is a good option) and also make sure to play the sweep(s) as loud as possible without pushing the speakers too hard casing them to distort, making any kind of noise except the sweep(s). Time range:
Again no right or wrongs, but I would suggest that you either choose a 400-500 ms range (and the problematic frequencies will be cut off but who cares; they need attention … ) or simply adjust the range until you see the end of the longest decay.
The same as for FR-curves applies to these plots in terms of L and/or R speaker.
A note on decay time: It´s easy to assume that the decay time can be sourced from these plots but unfortunately, it’s not that easy. The shape of these plots will vary with the setting called "Window" in the control section (try changing it to 1500 ms and see what happens). If decay times are of interest, use the "Topt" option (RTxx or EDT values are close to useless if small acoustic spaces like control rooms is considered) in the "RT60" section of REW. This option ("Topt") checks the slope of the filtered Schroeder integral looking for the most linear fit and this can sometimes be useful above the modal range but sourcing accurate decay times in the modal region is not trivial. Use an appropriate scale. If your highest reading is 0,5 sec, don´t show a 3 second range.
Finally, the perhaps most useful graph but unfortunately, seldom spotted in posts in this forum; the ETC (Envelope Time Curve / Energy Time Curve)
This graph shows the energy plotted over time. The highest spike at 0 dB (assuming normalized) is (hopefully … ) the direct sound from the speaker (only use one speaker at the time when checking the ETC). The rest are reflections from your room (or stuff within it). The use of the ETC is obvious; tracing early reflection points, monitor the ISD-gap (and the termination of it if LEDE/RFZ) and the (semi) diffuse field hopefully following it (unless NE design) etc. dB range:
30-40 dB (0 to -40 dB for example) is normal, depending on how much energy remains after the direct sound. Time range:
Usually about 80-100 ms is enough. If there´s no energy above -40 dB after say 40 ms (an extremely “dry sounding” room in other words) , then you might just as well zoom in and make it 40 ms in total for more detail. Unless there´s considerable energy before the direct sound (yes, it can happen, a resonance caused by a speaker on a desk for example), only show a couple of ms before the direct sound.
Note that you can filter the ETC in order to see the energy content of reflections for different frequency bands..
I hope this simple guide will help reduce the amount of graphs posted, featuring … “interesting” scale settings.
Sincerely Jens Eklund
Oh, and all frequency scales should be set to logarithmic, not linear.
EDIT 2: *
Unless you´re doing multiple measurements (perhaps in order to locate a good initial position) in different locations (this is in my opinion not very important if before treatment analysis, since any improvements due to acoustic treatment is likely to be observed in others locations as well but the actual sweet spot is the most important location), it´s a good idea to make sure that the measurement microphone is positioned exactly in-between the left and right speaker. If not, you´ll see the effects of constructive and destructive interference in the frequency response when recording both speakers simultaneously, especially in the higher range. Checking for symmetric response between L & R speaker might be tricky if the microphone is not centered properly.
Setting up the mic centered with accuracy using distance measurements can be fiddly and often leads to less than perfect results so another way is to use acoustics to position the mic in the right position. Generate three sound files consisting of a single period of a sine wave with the frequencies; 2 kHz, 6 kHz and 20 kHz followed by about 50-70 ms of silence. Position the mic as centered as you can and also make sure that the speakers are angled symmetrically towards the mic and play (in loop mode) using both speakers (and make sure they play exactly the same level) the first sound file (2 kHz) and move the mic sideways back and forth (preferably by moving the micstand so you’re not too close to the mic, causing strong reflections from yourself) while observing the level meter and find the position that results in the highest reading. Repeat the procedure with the 6 kHz file, and finally with the 20 kHz file (you might need to increase the level of playback to generate enough SPL for the 20 kHz file). Your measurement microphone is now centered exactly in-between the L and R speaker. You can confirm this by making some measurements and zooming in on the direct sound in the ETC and compare L, R and the combined measurement.
Finally, make sure that you know what type of microphone you´re using in order to aim it in the appropriate direction. If diffuse field type, you should aim it towards the ceiling (the exact angle should be stated in the user manual but if not, try strait up for starters). Note that some mics are labeled as free field, but they actually behave more as diffuse filed. The Behringer ECM8000 and Neutrik 3382 being a few examples. For more info on this topic, see this thread: Why point the mic at the ceiling?