Not boosting destructive nodes is not really related to how you do measurements, but it is important (and I have mentioned it several times earlier in tihs thread). You can do +dB fine, as long as you take care to not boost the deep and sharp dips.
But look at the picture below, this is 3 one-spot sweep measurements I did earlier, they are about 1 dm apart, the red one is where my head is on my listening positions, and the 2 others are where my ears is. Now tell me, which one should I use for room correction? I can tell you right now, that none of them will sound good if you flatten them out with filters.
And that is not all, if I tried to redo the measurements at the same spots, I would probably get different results on all 3. So in my opinion, an average of several sweeps (as for example Dirac does), or the moving-mic method I suggest in this guide, will get a more consistent result and in the end a better sound after correction.
If I had 1 ear, and it was 100% fixed in its position (head in a vice), then 1 measurement would be best.
After you did the multi-point measurement and calculated a correction and applied it, what did the three measurements look like?
My concern is that whether you base it on the red curve, or an average, you will have one correction and the differences remain. Did it work out that way for you? Or did you get a better unification? I acknowledge it is conceivable that a smart system can figure something out that is better than simple averaging, but I donāt know how that would work.
In any case, I donāt think your situation is bad. With the exception of three green dips, the curves are no more than five dB apart, which I think is a reasonable result for REQ. Thatās what I got. Smoother, but a few dB apart.
EDIT: this is the result I published in an Acourate thread, four seating positions in the same sofa, after correction. Pretty smooth, but five dB apart:
i am impressed that you could reach the same conclusions at your desk, as B&O did in their 13m square qubeā¦well done.
for 2.1 systems it is perhaps more important to do the phase/time correction than for stereo, so the sound delay is adjusted to allow different distances to the 3 speakers.
I agree about 2.1 system, especially if you have subwoofer placed far way (like corner) and listen to monitors nearfield. For me personally I have the sub below the left speaker so probably not needed. But this guide is meant for people new to room correction, so I have tried to stay away from becoming to technical (I am not a pro studio guy myself either).
Btw, for those who have tried this guide, let me know about the result. The result for me was phenomenal.
Finally, thanks to this great tutorial I managed to dive in world of room correction.
Read this topic twice and couldnāt find answers to my questions:
I have a problem with zip file. I have wav files for 48, 96 and 192khz. Zipped them on my Mac. When I try to upload this zipped folder with all the wav files Roon says āunexpected errorā. There is no problem with single wav files. What am I missing?
The second problem is connected with DSD files. I have a collection of about 100 DSD albums. Unfortunately, when I play DSD files my server (NUC6i5) is processing at 0,3 speed. In other words there is no horsepower and my music stops.
Is there any way to exclude DSD files from playing with REW filters applied? Any other solution?
Max frequency for REW filters is 192000. So there is no native filter for DSD frequencies?
In other words: what is the best way to use room correction and play DSD?
There is an alternative way to use .wav files for room correction, but it requires to manually input the numbers:
In Roon, open DSP settings for your output
Disable Convolution
Press āAdd Filterā and add a Procedural EQ
4, In the Procedural EQ, add 2 Parametric EQ
Make the first Parametric EQ apply to left channel only
Make the second Parametric EQ apply to right channel only
Open REW, and the measurement for your left speaker, open EQ window
Now press āEQ Filtersā button
In Roon, select the first Parametric EQ (the one for left channel) and press the EQ window
Now (very carefully) input the number in the Roon EQ from your REW EQ (same Hz, gain and Q)
Do the same for right measurement and right Parametric EQ:
Unfortunately Roon only supports 16 filters, and the āGenericā EQ used in REW supports 20, so if REW generated more than 16 filters you need to manually disable 4 filters in the window you get when pressing the āEQ filtersā button in REW-EQ, and then generate new filters in REW.
I havenāt tested this with DSD, but it should work no matter what the sample rate is.
I donāt know how to disable DSD from DSP, I guess thatās a question (and possible fix) for the Roon devs.
In meantime, something strange happened. During tests I found (listening Diana Krall āSuperstarā from Wallflower) some distortions in left speaker (highs). I had to make new measurement with small change - in step 4 I gad to lower volume from 75db to about 55db (my listening position is about 4m form the speakers).
Those distortions scared me a lot. Is it possible that I have a problem with this speaker or is it known problem in room correction?
It might be because of clipping, turn on headroom management in Roon and enable the indicator, and see if the blue star (signal path) becomes red when you hear the distortion. If so, try -5 dB or so (or turn on volume leveling) to fix it.
You can also limit the range of room correction to say up to 800 Hz, if room correction for higher frequencies gives you problem. Generally speaking, room correction is most useful for lower frequencies. Thatās done in the REW EQ window, step 11, where it says āMatch Rangeā. After changing upper limit, regenerate the filters This need to be done for both speakers.
Got bored, so I decided to compare the way Dirac does measurements (9 sweeps) vs the RTA moving microphone method I describe in this guide (I did the Dirac sweeps manually in REW).
Dirac suggests doing sweep measurements at these positions, and does left, right and center sweeps on each automatically as part of the guided-setup, see this picture:
Below is the comparison for left and right speaker. Except for a little difference in volume, they are very similar, so if you prefer to do sweeps and averaging them out its also an option (but it takes longer to perform, in this case a total of 18 sweeps):
Nice comparison. A couple of questions: 1. Were theses measurements done with āa bodyā in position? and 2. How were the RTA measurements performed? Did you measure at each point for a specific time using the spiral pattern you previously employed, or a static measurement at each of the 9 points and move the mic on the fly between positions?
I sat in my listening position when doing both measurements. The RTA measurements was done in the spiral pattern I described, 2 spirals on each side of head, one up-down-left-right and one up-down-front-back. The start of spiral outside my ear, and then maybe 5 increasingly bigger circles until about 2 dm in diameter. I also did a left-right-up-down in front of my face. I probably looked like a complete mental case while doing this
I did each spiral for 15 averages (which is with my settings 15 seconds), so 75 seconds in total, and the microphone was never still, constantly moving slowly.
In the pictures, the higher volume curve is the measurement according to the 9 positions as done in Dirac, and you can see it suffers a little from localized readings at higher frequency, for example 1500 Hz and 3 kHz for right speaker (or so I suspect).
I should probably mention that Dirac does more than just averaging the measurement like I do, and a relatively simple guide like this one with free software can not quite reach Dirac (or so I believe, although I canāt hear any difference from my trial with Dirac).
@Magnus I managed to run a second set of measurements paying more attention to my mic movement and the Target level. I think the results are better, but I know I still have a lot of room for improvement for various reasons.
the good:
the overblown chestiness of some songs has been drastically cut down. This is likely due to the compromised position of my speakers ā they sit in compartments in my cabinet since they are a part of my 5.1 home theater setup. More thoughts on this below.
echo has been reduced
in one case, listening without the filter made it obvious that my right speaker was delayed comparing to the left. With the filter, the delay was eliminated. It was shocking. Iām still not sure which is more correct ā Iāll need to listen to that on my headphones without the filter as a reference.
what comes through is less muddled
the bad:
most songs seem to lose a lot of their energy with the filter enabled
the soundstage seems to collapse significantly and the sense of āairā suffers
Iām not sure if the instruments sound natural ā saxophones, trumpets, guitarsā¦ all sound thinner as if some of their body has been stripped away. This might be because I got used to the sound with the reverb and coloration that my cabinet adds. But I think it is more than that.
Thoughts:
I have decided to apply treatment at least on the inside of the cabinet surfaces where the speakers sit. I think REW was trying to compensate for this coloration and by extension went too far. Iām still looking for the right material to use, so this may not be a quick project.
since the filter was created using a reference level of 75db, the corrections are optimized for that level. That means that if Iām listening at a level significantly less than that as I always do, I will not hear the full benefits. any thoughts on this? Would it make sense to run the calibration again at a level of 60db for example?
although Iām using an i7 sonicTransporter as my Roon core, I donāt have the brute power to use up-sampling with convolution in the case of PCM or any convolution at all in the case of DSD. MQA is another problem as adding convolution negates the decoding in my Brooklyn DAC, but I would still opt for room correction at the cost of listening to folded MQA for now (until Roon supports any level of unfolding). Once I get my filter dialed in, Iād want to be able to use it across the board. But that seems that Iād have to upgrade my core to something more powerful than my sT i7 ā which I donāt want to do for now. I hope Roon manages to optimize the code further to help avoid having to upgrade. But of course I donāt know if that is even possible in any significant way.
Iāll report back if I have any more significant developments. Thanks again.
Room treatment is always good!. The soundstage should not shrink though, I experienced the opposite (wider sound, better stereo imaging). 75 dB sounds fine, you should at the least stay 20 dB over the sound floor (which in my apartment is 40 - 50 dB). You can easily check in REW with the SPL meter.
Here are some things you can try:
Point the microphone towards the ceiling during the measurements and use the 90 degree calibration file. This will make the measurement include more of reflections, and give a more total correction.
Do 16a as well, to get the final fine tuning in place.
If the above still produce bad results for higher frequencies, then do room correction on lower frequencies only, for example up to 500 Hz.
But if you plan to treat acoustic in your room, do that first. There are several guides on the net, for example this.
Finally, after room treatment and correction, save the Pink PN noise to a wav file, and play that from Roon with all correction in place and do the spiral-measurement with both speakers. The result should now be very close to the target line (see my last measurement picture).
Btw, I just tested without room correction, and even on music without bass (which my room required lots of correction), the difference is huge. Guitars and singing voice sound so muffled and dead in comparison. Its a little like comparing a pair of $50 speakers with $1000 speakers.
I canāt do full room treatment due to aesthetic and usage reasons. But I want to at least try to address the transference of vibration/resonance between the speakers and the cabinet. Iāll check out that link - thanks.
One acoustic treatment that helps for higher frequencies, and is cheap and can look good, is to buy a nice looking thick cloth and hang like a very much folded curtain on the back wall, covering as much as possible of the wall. This wonāt help against bass, but will help with flutter echos and higher frequencies. And done correctly it will look very nice also. I heard from a studio setup guy that he always did like this. Then he used bass traps and normal absorbents for first reflections behind and to the left/right of the speakers.
Also, check out some speaker and listening setup guides, for example this: speaker placement. It might not be possible to follow it to 100% due to how the room looks and furnitureās (and wife tends to frown at big rearrangements), but it could help.
the overblown chestiness of some songs has been drastically cut down.
echo has been reduced
in one case, listening without the filter made it obvious that my right speaker was delayed comparing to the left. With the filter, the delay was eliminated. It was shocking.
what comes through is less muddled
the bad:
most songs seem to lose a lot of their energy with the filter enabled
the soundstage seems to collapse significantly and the sense of āairā suffers
Iām not sure if the instruments sound natural ā saxophones, trumpets, guitarsā¦ all sound thinner as if some of their body has been stripped away. This might be because I got used to the sound with the reverb and coloration that my cabinet adds. But I think it is more than that.
First, I have found that getting speaker distance (or delay) set up correctly down to the 1-inch level matters a lot, particularly with overall imaging. Good news is that you can set the appropriate delays in most software (HQPlayer, Roon, etc.) without further changing the sound. If you fix this first you will then be even more sensitive to the issues you mention that I respond to below.
Second, each time I have played around with REV or Dirac-based frequency adjustments, I encounter exactly what you describe (loss of energy, loss of transient bite, loss of air, instruments sound thinner). There are equipment manufacturers (Bryston being one) who donāt offer equalization/room adjustment in their equipment for just this reason. Dirac goes to great efforts to say that its software ācompensatesā for this by both adjusing frequency and retaining time coherence, but my experience says you canāt really do both. I think that is why fixing the roomās own response, through curtains, carpets, diffusers, absorbers, etc. is almost always a better place to start and then to resort to REV or Dirac only to fix the very last bit. Some choose to only let these programs adjust frequency response below say 200Hz so that the time coherence, transients of everything above that remain untouched.
What Magnus has created here is hugely valuable as neither the software nor the testing methodology has previously been laid out so clearly. But as we learn to use the software better, I think we will also discover some of these tradeoffs and the limitations they impose.
in this case its hardly rocket science.
