As @Magnus states you’ll need to use the Timing Reference option if you are playing external sweeps from your music server.
These sweeps must include the timing reference. This REW help page should be useful.
The options you used to generate the external sweep, eg. Frequency range, length etc., must also be used when performing the measurement to prevent frequency misalignment.
Initially, which speaker contains the timing reference is unimportant, but clearly, if your speakers are not equidistant from your listening position choosing one as a reference will greatly help determine speaker d3lay settings.
My recommendations: stick to the default settings 20Hz-20kHz 256k samples. Don’t start at zero hertz you can damage your speakers. Don’t use ERB for bass but VAR to see the room modes. If you post a link to your MDAT we can help more.
For those interested in room correction especially bass - I have shared a bit of my experience with corrections in acoustic and electrical domains in this post.
Additionally: don’t focus on SPL only but
ETC for early reflexions (cannot be fixed by DSP, absorber/diffusers required at early reflexion points)
RT60/Topt for room reverb (cannot be fixed by DSP, bass traps/absorption/diffusion help)
Spectrogram for 20-300Hz for the amplitude and decay of room modes (amplitude can be lowered by DSP, decay time requires bass traps)
EQing the overall response using a room curve that is not flat but slowly decreasing, and EQing with caution : no big boost, no need for a super flat target. Allow for high Qs for low frequencies, but not above 250Hz.
@Alandbush@Magnus OK, I understood how to create and apply the sweeps. But what do I do with the various measurements (3-4 each side) when I have them? Do I have to first calculate the average for the application with filters or do I apply the filters separately for each measurement? @alec_eiffel As soon as I have carried out the new measurements with sweeps, I am very happy to share the mdat file with you. Currently I only have the measurements based on the RTA.
The distance of both speakers to the listening position is exactly the same. The thing is that behind the listening position on the left side the room is twice as long as on the right side, where the distance to the back wall is about 140cm, on the left side it is 740cm (L-shaped room). From the listening position to the front (towards the loudspeakers) the room is symmetrical. For this reason, different room modes are produced or displayed during the measurements from the left and right.
You are slowly getting into the complexities of Room measurement and as @alec_eiffel has shown there are so many ways to use them.
REW can show a great deal of different interpretations of a sweep, unlike a RTA, each focusing on different aspects of room acoustics. Once you have some sweeps, become familiar with what is available. This Help index could be useful.
If you are comparing sweeps and especially if you intend to average then you must first time align. See this Help page for more information. Generally, though, they will look very very similar except at the lower frequencies where the UMIK1 is perhaps not too accurate.
Don’t expect a quick fix and the results to be perfect. Perhaps start with a single filter; remeasure to check its been applied and listen. If it sounds right try another filter or two; remeasure and listen and repeat.
@alec_eiffel So, this weekend had time to do the measurements. I made 6 measurements per page, where L1/R1 are in the center of the listening position. The other measurements varied up to 50cm laterally, to the front and back, as well as in height.
Then I had the average calculated for each page and experimented with various settings. I also created spectrograms and so on, but that’s where I quickly stop with the technical knowledge. What you can definitely see in the SPL is the massive drop on the right at 58Hz and certainly the other room modes. I didn’t get these deletions smoothed. Which is probably also good, because I better fix it with passive means, right? Is there anything else to consider with this measurement? BTW: I think the sweeps are time aligned because I did the measurement with time alignment. At least the setting in REW didn’t change to SPL.
First I have an issue with the recorded SPL level which is >120dB, obviously the calibration is wrong, otherwise you are probably deaf by now Have you properly applied the calibration file of your microphone? Or have you inadvertently used the onboard microphone of your computer (happens to me often…)?
Provided your recordings are good at least in relative terms, and looking at the bass mainly, here are my findings:
You actually have a dip on the left side 50Hz to 70Hz. It’s not too bad in my view as there is no dip in this range from the right speaker. The only way to limit the dip is to change your speaker placement or listening position, or your room .
Your room is way too reverberant for high-end hi-fi. Only passive treatment can address this. Start with bass traps with large bandwidth to cover up to the low mids. You won’t be able to lower RT60 to 0.4 starting from that high, but at least try to reach a somehow flat RT60. Please not that RT60 here is what is seen in the “overlay” tab of REW, which is by default Topt, roughly 50% of the RT60 as defined in the litterature.
ETC is not bad at all, always good if you can absorb first reflections a bit more.
You have some important modal peaks before and after 100Hz. Bass traps will help here but in the meantime a little EQ can improve your listening experience. I quickly built a stereo correction file with REW EQ. Note that the filters are very different from Left to Right.
The associated stereo impulse response is here and can be loaded into Roon convolution engine in a few clicks. The master data file is here, including time aligned averaged L and R responses.
Your statement with the SPL level also astonished me with the measurements and the setting with the SPL meter. The volume was my normal listening volume. I used the calibration file. So it must have been the switched on microphone of the laptop. Next time I have to pay attention to it.
Otherwise first thousand thanks for your super evaluation. I really didn’t expect that. I am really enthusiastic, this forum and the helpfulness is really great.
I have read through your explanations and there is probably no way around it that I seriously deal with passive treatments of my room. Especially for the bass and lower midrange. The reflections left and right, where I have windows (like behind the speakers) I have already tried to optimize with special acoustic curtains. A carpet is available, as well as several plants and two fabric sofas. Maybe I could work on the ceiling with some acoustic panels.
I have seen that the target volume on the left is quite different. Isn’t that a problem when merging into the impulse response file or during playback?
Thanks also for the specially created impulse response file. At what sample rate did you export the Stereo Impulse response? I tried to import the wav file into Roon. There was a message that it cannot be parsed.
The link to the Master File (mdat) unfortunately doesn’t work, but also leads to the impulse response file. Can you please tell me the correct link?
Sorry I fixed my post with the correct snapshots, files and links. It’s very visible you have already addressed the early reflections as they are quite low.
The IR sample rate is 96kHZ but Roon shouldn’t bother as it will resample it at your current rate. Maybe try again with the new link. I’ve tried and it works on my system.
Faced exactly the same and also ended up adding 10dB headroom. @Magnus it may make sense to add this as note / last step in your (excellent) tutorial if this happens to be kind of common, not everyone will think of headroom management.
Its there, step 17. For my corrections its enough to turn on volume leveling in Roon zone control (I have mine set at Album and -20LUFS and -15 when unknown). That sorts out the clipping issues as well as making all music about the same level. Internet Radio is unknown and therefore set to -15 dB.
I don’t know the technical details, but something about convolution/FIR and how it’s handled means that even if you only lower frequency response, you can still get clipping. My suggestion is to put the target at average around 1khz, and then handle clipping either with automatic volume levelling or -10 dB (or so) headroom.
From my experience, putting the target to low and only adjust down will make the music sound flat and less dynamic.
BTW I experienced that right channel was measured higher than left by a couple of dBs. I assume I should take opportunity to manually set the EQ volume targets to the same value versus let kept different with the auto setting, right?
Clipping can occur even if no frequency has a magnitude higher than 0dB, not intuitive but true, due to “intersample peaks” when applying filters (resampling and/or convolution). That’s why a -3dB headroom is recommended.
Regarding EQ between Left and Right channel: personnally, I am balancing them at the time of EQ, setting the same value in the target curve.
As you mentioned in an earlier post Step 0 (Step 1 being what this tutorial covers) being speaker positioning (+ acoustic treatment), do you know a REW-based way to help speaker placement? I believe I read somewhere to use RTA to see real-time how moving speaker creates/remove frequency dips… any opinion ?
I realize many are using their ears for the placement exercise w.r.t stage/image, now I am not sure how one can systematically comprehend frequency dips across the range to « best help step 1 » thus wondering if a REW-based method could assist…
Hi, and thanks for this very nice step-by-step guide!
I’m doing some measurements with REW and the UMIK-1 mic. The below measurements are for my left and right speaker, red one is left and green one is right. It seems a bit odd to me that the graphs are in the 30-50dB range, while the SPL meter in REW indiacted 75dB (step 4). Furthermore, I find it very strange that my speakers have such a huge drop in dB from 1kHz onwards. Did I do something wrong with my measuring? BTW, my speakers are Tannoy XT8F floorstanders.
Have you properly applied the calibration file of your microphone? Or have you inadvertently used the onboard microphone of your computer (happens to me often…)?
