I don’t think a different mic will solve the problems we are encountering here.
Open baffle speakers shouldn’t be responsible for them either.
I know, because I had DIY standard multiway, Magnepan Tympani4A, Quad ESL57 and Audio Static ES100, while dabbling around with room acoustics measurements in different environments, designing and modifying x-overs, applying graphic and parametric equalizers and more.
I’m playing around with these sorts of things for over 25 years.
Room layout and placement do play a major role, though, but still, I can make our living room sound great without the need to dress it up in dedicated acoustical treatments.
Strange, it doesn’t show in your file, see comparison screen shots between mine and yours.
It took me a few days to get back to this. I had moved my Ominmic Calibration file while I was trying to “stay organized” so that’s what happened there. Oops. At any rate (no pun intended) when I reset REW to defaults and try to start again using the OCA tutorial I cannot get ASIO4all to see the mic. The control panel for that program is not taking changes when I reload using the button in REW and available inputs show up as just 1 and 2. Looking online there are others that have had this problem and said it “fixed itself” but didn’t know why. I tried restarting all to no avail. I take it it’s not possible to use Java instead?
To circumvent the issue, you can use the alternative to export the measurement signal to a file, import that into Roon watched folders to play from there.
That’s actually the way I always do my measurements.
Hope that gets you going again.
Hi @Marin_Weigel, I finally took another crack at this with the correct Omnimic cal file this time and with all the settings correct to the best of my knowledge. The regularization setting mentioned in step 19 in the pdf created by @OCA was not available for the 1/A operation (software version issue) so I used a gain setting that provided 8dB. That is the only place I deviated. I played the test tracks through Roon-that’s the way I’ve done it from the start. The final 48k stereo WAV file still has attenuated volume when applied in Muse as a convolution filter. See files here if you care to: GitHub - BelakScientific/REWNew
Thanks!
My judgement is to be taken with a grain of salt, being face value, broad brush, but let’s just look at the left channel as an example:
Red = measured response as uncorrected
Blue = correction response
Green = measured response as corrected
The measured response as uncorrected still looks broken - weak, peaky lower frequencies below 250 Hz, seemingly a crossover phase problem between 2 to 5 kHz, and highs missing in action above 11 kHz.
To be of more help, we’d likely need to look at a sketch of the room floor plan with some dimensions, your speakers’ make and model, maybe even have some photos of your measurement mic in relation to the speakers to oogle at.
The correction response does look bad also, with abrupt and steep slopes, something you’d actually want to avoid at any rate.
To me, something still seems wrong in implementing OCA’s guide, but I’m not versed enoiugh with it to guide you appropriately.
To explain, OCA’s thread initially got my attention to the possibility of using the inversion method, but I have since adapted the procedure in large parts for my special use case.
In the time domain, we’re looking at your speakers being inverted in absolute polarity somewhere along the chain, the midrange response seemingly arriving earlier at the microphone than that of the tweeter, the correction starting 500us late, just to name a few things that look awkward to me.
I wouldn’t know where to start helping out without more background info and maybe also @OCA’s take on the matter.
Ok, here’s a drawing of the room and a few pics. My speakers are Lii Song Fast 15 drivers in home made open baffles. They have a ragged response, mid peaks (and dips), and are suspect in many ways but there’s something about the immediacy of percussion and the realism of certain instruments that I really like. The subs I use are not getting signal in in any of the measurements but they get the Muse modifications too so that’s why I fit the house curve to make it extend lower than the full range speakers can go. The subs go through a separate bridged amp with subs connected in series and a minidsp to adjust.
A few more mic details in case it’s useful. Measurement was taken with mic at ear level on the couch. For this set of measurements the small Omnimic tripod was perched on a stack of pillows. To simulate me sitting there, possibly a bad idea. Previously I had used a spotting scope tripod sitting on the couch at ear level. I’m not sure what’s ideal and how that might have affected the measurement. Measurements were all taken with a horizontal mic because that’s what the omni mic calibration file is intended for.
One more thing: when I tried to export the final.wav file I got an error that the left and right impulse responses were not time aligned. I used the tool in REW to do that but this was not specified in The instructions from @OCA. It was the only way to get the convolution file so I did it.
Re: my last message the mdat files in the zip file that don’t have trimmed on the end of the filename have not been trimmed. These would not export as a stereo wav.
Heya, finally found time to tinker some more in REW with your measurements.
This is my best effort - download the zipped correction files from my Nextcloud and get back with your findings, please … link was wrong, edited now
Thanks for your work on this! I’m not sure how to use this output, though. Do I need to enter the values at the end of the text file into the PEQ manually to get the correction?
Thank you so much again, Marin! The zip file loaded fine. Most obvious difference when I switch it in and out is the huge loss of gain when it’s enabled. I can get usable volume when it’s maxed out. Low gain setup: I am currently using an Orchard Audio PecanPi+ DAC/streamer with no preamp into a Starkrimson Duo amp by the same manufacturer. When I compare I need to adjust volume subjectively when I go back and forth. I’ll need to listen more to come up with any opinion but will be doing that and will report back. My audiophool friend is coming over is a few minutes so I’ll get his take on it too.
Of course, DRC always comes at a price, and you can’t raise gain above 0dB without causing clipping.
If you look at the graph I provided, you see only negative gain of up to -35dB!
Of course, see above.
The spectral balance of the correction should sound more balanced and thus quite different compared to the peaky uncorrected response, so rather listen to full tracks instead of trying to do quick switches.
I wonder about my mic cal file and whether making a minimum phase cal file like @OCA suggests is important. Also in general if the mic is accurate enough or out of whack somehow (Omnimic V2)
Can‘t remember that recommendation, and since I’m suffering from TLDR, could you kindly give me a pointer?
Maybe if you store it in extreme climate conditions and keep dropping it.
But I take it, that you and your friend were underwhelmed by the corrected room response?
Could you courteously spare some comments for us, please?
You could even remeasure with the correction in place, for me to get some more insight into possible problems.
That said, there still seem to be deviations from the procedure in your measurements, such as starting at 20Hz instead of at 0.
Maybe try one more time, more closely following the procedure.
I well understand your listening area‘s placement and general purpose restrictions, but your current layout is challenging, especially for an open baffle design.
Have a nice Sunday, maybe even enjoying some music!
