High Fidelity Digital Room Correction with REW & rePhase
This video is a comprehensive, step by step tutorial for a state of the art digital room correction method.
* Don't forget to download the zip file in the link below which contains all the measurement and calibration files used in the video as well as continuous updates/revisions to the method.
Digital room correction is a process that involves analyzing the acoustics of a listening space and applying equalization to the audio signal in order to improve the overall sound quality.
The goal was to design a digital room correction method which would make this process more accessible and affordable for the average consumer by offering a solution using totally free tools. The techniques explained have the potential to significantly improve your listening experience.
Resolution: 1440p 60fps
00:00:00 Introduction
00:02:40 Microphone Alignment
00:05:50 Multiple Measurements
00:13:35 Filter 1 (Virtual Bass Array)
00:27:40 Filter 2 (Target Curve Inversion)
00:31:42 Filter 3 (Speaker Phase Correction)
Filter 4 (Excess Phase Inversion) video link:
• Video
_______________________________________________________________________
********IMPORTANT NOTICE*********
Remember to download the zip file below to access to the latest revisions and updates to the method, the VBA Optimizer Excel file, the "full collection" of REW, rePhase, txt and convolution files I've used during the full calibration of my system:
drive.google.c...
________________________________________________________________________
CITED SOFTWARE LINKS:
REW Early Access version used in this video:
www.avnirvana....
Accourate free trial:
www.audiovero....
Acourate Microphone Alignment Tool Instructions:
www.audiovero....
TEST TRACKS:
Speaker Cohesion ( & pre-ringing):
Bubbles - Yosi Horikawa
tidal.com/brow...
Pre-ringing:
Billy Jean - Michael Jackson
tidal.com/brow...
Clipping:
Sweet Jezabel - Turboweekend
tidal.com/brow...
Californication - Red Hot Chili Peppers (2014 Remaster)
tidal.com/brow...
Center Image Height:
Black Sabbath - Black Sabbath (2009 Remaster)
tidal.com/brow...
#roon "room eq wizard" "filter design" "phase inversion" #audiophile #impulseresponse #convolution #filters #rew #roomcorrection #calibration #fouriertransform #fourier #equalization #equalizer #hometheater #cinemasound #soundsystem #dirac #dolbyatmos #speakertest #impulse #phase #hifi #hifiaudio #bass #bassboost #graphicequalizer #hometheater #impulse #iir #fir #lowpassfilter #phase #rephase #sota #viral #audio
Пікірлер: 262
You know I did a lot of tests with the inversion so I know it quite well. Following the procedure you reported here I have a lot of post ringing (I know it comes from F2 because I have each filter separately reported in the convolution filters of Roon). I have to apply a frequency dependent window of 35 to R1 and L1 and then invert. Applying VBA before the inversion give a lot of benefits with respect to invert straightforward, the sound is a lot better with much many bass. Great job!
@ocaudiophile
Жыл бұрын
You're right, FDW has been excluded in F2 by mistake and I have actually added a note for that in the description section of the video. I'd recommend 15 cycles instead of 35 and no other smoothing.
@teb76
Жыл бұрын
@@ocaudiophile ok good. So we have to apply a FDW to L1 and R1 and then invert. I have another question regarding F3: should we repeat the process in rePhase for L2 and R2 or we do only on L2 and then apply to R2 too as per your former written tutorial?
@ocaudiophile
Жыл бұрын
@@teb76 It's better to do separetly for L2 and R2, there might be subtle differences between the two due to speaker placement and room shape anomolies.
@raymondwu9506
Жыл бұрын
@Obsessive Compulsive Audiophile Should I apply FDW of 15 and make the target curve then remove the FDW or should I make the filter with FDW still applied?
@joek6207
Жыл бұрын
Same question
These videos are awesome! Thank you for the work. My system already sounds 10X better than last week :)
@ocaudiophile
Жыл бұрын
Great to hear!
I would like to thank you again. the sound is keeping improve step by step, i'm discovering new details in records i know since 20/30 years. Thank you to share your knoledge with us.👍👍👍
OCD deserves worths his body weight in GOLD when it comes to tailoring response . Salute to you sir
ok just tried the target curve inversion pretty quick , because just picked umik-1 and don't have time , and i am speechless ! great channel with a lot of great information ! keep up the good work , greetings from greece !
@ocaudiophile
2 ай бұрын
Thank you!
Great tutorial. Thank you. Works on Raspberry pi4 + moodeaudio + camilladsp active open baffle 4 way speakers. Good job man. Sounds amazing now.
@ocaudiophile
Жыл бұрын
Glad it helped. I secretly adore those open baffles ;)
Hello, regarding the 50Hz harmonics, you could use a 220 / 220 V transformer to decouple. It could be used as as band pass filter around the 50 hz...
@ocaudiophile
5 ай бұрын
Thanks for the tip. I was thinking about this actually.
Very deep understanding of REW sir! Thanks for sharing. You correct the input signal right (2channels)? I correct the output (6 channels) This is how I usually do it in the activa filter. Set each output band delay. correct any standing waves and peaks on each band. Then clone all 6 x-over and filter settings in 'Rephase linear tabs' and swap with that .bin. In other words i do a full electronic Rephasing of my favorite setting, not acoustic.
@ocaudiophile
7 ай бұрын
Yes and yes!
Your tutorial are fantastic. Thanks
@ocaudiophile
Жыл бұрын
You are welcome!
Fantastic tutorial. I watch all of your videos although I don't yet implement them all. I am just trying to improve my Audyssey multeq XT32(with MultEQ-X app) based on your pointers you provide. Don't have the Minidsp. Thank you very much.
@ocaudiophile
Жыл бұрын
I need to clarify something. I have an atmos system with Audyssey and a separate hifi stereo system. The calibration procedure and application is very different for each although the theory behind is the same. There's almost nothing in this video that one can apply to a Marantz/Denon AVR because the filters produced are FIR filters. Although these receivers are capable of such filters (and Audyssey & Dirac use them), they're not open to the end user. Hopefully with enough of us being aware of the monopoly, they 'll need to reconsider soon. But you can take this opportunity to dive into the true audiophile's stereo world. These filters can be easily applied freely in a PC playing free youtube music and running your front speakers through your AVR (hdmi connection from pc, pure direct mode of course). Every filter in this tutorial will work with the free pc convolve "equalizer Apo". If you're a subscriber of this channel applying these methods, you're definitely also an audiophile and audiophiles listen stereo music. Some of them are all analogue even against digital correction but let's not go there.
Serkan, Thank you much for sharing your broad and deep knowledge of this. Much appreciated. One question regarding creating both left and right filters, do you have to time-align L0 and R0 before creating the 4 filters for each channel, or can we create them from the raw measurements?
@ocaudiophile
Жыл бұрын
Thanks. One of the speakers is the acoustic timing reference so it'll be at time zero anyway and the other speaker will need to be on top of it if you aligned your mic at the centre during the measurements. It'll not make an audible difference but for the health of all the calculations it's good practice to align (with cross correlation) everything to time 0. If your LP is not at the centre, than you should compensate for that in your convolver engine settings with the correct time delays applied to the closer speaker. Btw, do not forget to check video description area. I have made some revisions/updates/tips and added my complete measurement and calibration steps.
Thank you so much for your videos! This is the unique knowledge that you have given me! Could you do a detailed lesson for calibrating headphones?
@ocaudiophile
Жыл бұрын
Your welcome. I don't have any experience with headphone calibration :(
@ocaudiophile
Жыл бұрын
Here's a good starting point for you: github.com/jaakkopasanen/AutoEq/blob/master/results/INDEX.md
Around minute 16, the room length to put into the excel file is it the actual room length or the longer measure of the room? I ask this question because my room is wider than longer. Thank you
@ocaudiophile
Жыл бұрын
Longer side of the room.
Thanks a lot!
@ocaudiophile
Жыл бұрын
You're welcome!
i followed your REW video and was wondering if it's possible to make impulse response shorter (sub 200ms) without compromising on frequency amp correction? forgive me if it's a silly question as I'm not too knowledgeable on this thanks for what you do! what's your comment on sound id reference and their methods compared to what you show here?
@ocaudiophile
Жыл бұрын
The answer to shorter impulse is yes, you can. Watch the "Excess phase inversion" video, it's at the very end. These two were supposed to be one video but it would be too long so I had to split. I have no experience with Sound ID unfortunately. From their website, it's not clear what they use but probably some AI enhanced phase shifting techniques. The product might be more towards studio engineers as a plugin.
Thank you for reading your comments and replying! I hope you see this one I have a question about microphone placement when taking measurements: I watched a video i will link below and they said to move it around, please watch when you can and let me know your opinion on best technique; moving or static position. Also two quick questions 1) What is your background? You're smarter than anyone I've seen on the subject. 2) Is pink noise better than sweeping to measure? Thanks so much!
@satorizero5448
Жыл бұрын
Here is the video @8:35 is the specific technique but feel free to please share critiques or opinions on their entire method ❤ kzread.info/dash/bejne/aIad19mqqs3XnLQ.html
@ocaudiophile
Жыл бұрын
Moving mic measurement is a common and efficient technique and you'll get a good average frequency response free of local peaks and dips. But you'll have no phase information hence no phase correction will be possible. Still it's a good simple correction method.
@ocaudiophile
Жыл бұрын
Rew sweeps are extremely optimised for speaker measurement and are very efficient in stripping environmental noise and effects. I don't know any tool that can extract information of that quality with pink noise. It's more suitable for speaker volume measurement. I'm just a hobbyist and I work in finance but I've an aeronautical engineering degree ;)
@satorizero5448
Жыл бұрын
@@ocaudiophile Thats awesome! Thank you so much for your help I love the degree man! Kudos! I wish I went to school for engineering, but I learn from smart folks like you! Thank you again, from us all
AWESOME
@ocaudiophile
Жыл бұрын
Thanks 🤗
This video is pure gold! Finally I can follow along a room eq correction tutorial where things are well explained, make sense from a signal processing point of view and audible results start to appear in speakers setup! Thank you so much Serkan. I have some questions, though: - when convolving filter 2 to obtain the predicted response [at 31:03], wouldn't it be better to disable frequency dependent windowing on L1 before applying the multiplication? - it happens to my second stage filtered signal (after the target curve inversion filter) to have a phase slope inversion for extremely high freqs. This means frequencies from 23.600 Hz approx to 24000 seem to arrive later than freqs immediately before 23600 Hz. Do you have experience with this happening? Is it acceptable? Where shall I 0-center my rePhase phase response, then? - I don't have info about my Focal Shape 40 crossover specs. I wrote to Focal asking for the crossover tech specs but I'm not sure they will reply. Can I try to bypass this info and just go for a 0 phase linearization with whatever combo of reasonable xover parameters in rePhase (e.g. xover freq between 2000 and 3000 and 12 to 24 dB/octave slopes)? - I still miss a concept. After phase linearization of step 3, I would suppose the phase has already be minimized, since we zeroed the phase of the response. I can't understand why after this step the phase still exhibits an excess phase that we need (and, above all, are allowed) to further correct Kudos to you again!
@ocaudiophile
Жыл бұрын
I'll try to answer all. Without fdw the filter will cause some post echo in high frequencies due to lots of spikes in the response. Don't forget to download the zip file linked in the video description. I've added some extra info there in a word document as well as the.mdat file used in the video. Focal might get back to you but if they don't you can take phase discontinuties I rephase as xp points. We do excess phase correction on top of this because first correction is a very loose one with fdw:1 and excees phase by definition is just that : excessive phase to be removed. Phase shifts At very high frequencies are just speaker time alignment and should be ignored, you can see how fast they will change by playing with "time offset" in rephase.
@riccardorobertodelucia6060
Жыл бұрын
@@ocaudiophile thanks for your answers! I tried with your suggestions and project to move on and implement stage 3 and 4 of the filters, but there is something strange. First of all, I can create phase filters in rephase by equalizing down to 15 Hz, 180° and Q=2 with no pre-ringing artifacts. The final response is flatter in terms of phase, and the impulse response is very similar to the previous one (especially in the anti-causal part). I can barely hear any difference in transients, stereo placement before/ after filters. Also, when applying the filter, there are very little excess phase fluctuations, which makes it useless to compute filter 4. I don't know why it is so easy and at the same time useless to compute phase filters in my setup, it seems like everything already works without. Maybe is it because my room is properly acoustically treated with bass traps, and I'm in a mixing desk setup, with the listening position very close to the speaker monitors? Also, my setup involves two focal shape 40 and a subwoofer focal sub one. I did measurements with one main speaker on at time (while sub always on). In phase correction with rephase, I just applied one crossover linearization filter for the main speakers, since lower in frequency, although there is the sub, I don't have any other conventional crossover. The only 'crossover' is done by the LP+HP settings on my sub to split the signal between sub and main speakers, but since these settings are highly configurable, I'm not sure if it makes sense to try to apply a crossover linearization in that range, also because, from measurements, I cannot spot any phase inversion near the subwoofer cutoff frequency (it's around 200 Hz). Any suggestion on how to properly treat a setup with speakers plus sub? Thanks again!
@ocaudiophile
Жыл бұрын
I just checked Shape 40. They're active speakers meaning they have their own amplification and digital crossovers so no crossover phase shifts will need to be corrected. Also they're nearfield monitors hence little to none phase discontinuties to be expected by room reflections. I think that's why you don't hear any difference. Measuring them with the subwoofer included in the response is the right way but I think you should cross the sub one at 80Hz max to avoid localization of bass. This should be easy since your speakers can go down as low as 60Hz. So, you don't need crossover phase correction filter, you don't need excess phase inversion filter. VBA and target curve inversion filters along with a sub crossovered at 80Hz with your speakers will he optimal in your case imo.
@riccardorobertodelucia6060
Жыл бұрын
@@ocaudiophile makes sense! I'm relieved from the fact this is due to my setup, room and speakers and not to some mistakes during the procedure. Also, thanks for the tip about the focal crossovers characteristics. I ended up with the VBA filter and room linearization and everything already sounds much more flat and clinical (which is exactly what I need). You made my day!
Is it okay to use higher than 8% regularisation in the inversion filter if my dips are really bad? My current room is pretty bad so for now I'm just trying to do what I can as a band-aid fix before I'm able to move. I'd need somewhere around 10db boosts probably to sort out my dips, I do have the amplifier and speaker power to compensate for -10db preamp gain in dsp.
@ocaudiophile
Жыл бұрын
Boosting dips too much may degrade the overall sound but feel free to try. You need to decrease regularisation percentage to increase boost.
@ocaudiophile
5 ай бұрын
Yes!
Amazing amount of information in your videos! I've used a little bit of what I've learned and it has helped me a bunch already on my laptop and my cellphone. I have two questions 1.) A couple of my finished convolution files creates an echo that's quiet but noticable. What step did I miss? 2.) how would I do a 5.1 system? Seems your videos are mostly 2 channel. My PC is connected to my theater system and I would be using Equalizer APO if it's possible.
@ocaudiophile
Жыл бұрын
Thank you! 1. Did you watch part 2 of that video for filter 4? I played a pre-echo sample there. Is your echo something like that? Then it's easy to deal with. 2. You can do a great surround calibration with a HTPC and EQ APO. You will need to create all 4 filters (combined) for every speaker in the system including the center speaker. For the subwoofer, filter 1 will be enough (or just use REW's auto EQ). Equalizer APO can be configured for surround systems. I think your limit is 7.1 channels: sourceforge.net/p/equalizerapo/wiki/Configuration%20reference/#convolution-since-version-10
@SkubaSam
Жыл бұрын
@@ocaudiophile I watched the the video and it seems likey echo is post echo. Hello.....hello.....hello. Like that and each is quieter than the last. My first convolution files dont do that. I went back to tweak them after learning the process a bit more and now the phone and the laptop do the same echos with the newest ones I made. I'll have to try again and see if I missed anything
@ocaudiophile
Жыл бұрын
@@SkubaSam Probably you're missing something. Are you applying MP to 1/A inversion? Anyway, I have now added a link in the descriptions for the actual .mdat file I used in the tutorial where you can see all the steps. It could be helpful.
@SkubaSam
Жыл бұрын
@@ocaudiophile maybe that's it . 🤷🏽♂️. I'll play around some more. Thank you again for all this hard work you've done on this stuff
@ocaudiophile
Жыл бұрын
Actually, you were right. I didn't test filter 2 for post-echo was too stuck with pre-echo testing. Lack of FDW is causing it. I added a note in the description as well: After filter 1 is created and applied, apply FDW:15 and Psychoacoustic smooting to both L1 and R1 and re-calculate Target curve with their RMS average before continuing to Filter 2. All the remaining procedure is the same. Sorry for that!
This tutorial and description are fantastic! Thanks for sharing your expertise and inside knowledge. I am not sure how to find the "order" of my crossovers. I have Klipsch Forte IV loudspeaker. While I know their crossover frequencies (HF: 5.2kHz, MF: 650Hz), I don't know more about them. If I can't find this information is there a way to guess? If not, can Filter 3 be skipped and then complete Filter 4 for use in conjunction with only Filters 1 & 2? So far, I have made the convolution files, including VBA and Target Curve Inversion (using the Harman Curve). The result sounds fantastic, but I'm eager to go further and apply all of these filters if possible. Many thanks in advance for your advice!
@ocaudiophile
Жыл бұрын
If you skip filter 3 and jump to filter 4, you will still be correcting the phase response but the corrections will need to be too strong and will cause ringing. The beauty of XO linear filteraztion tools is that they don't cause any ringing. For your XO orders, 24db/oct is a good starting point, very few XOs are below that. You can maybe see them easier if you adjust the very first impulse peak of the speaker's response (not the largest peak) to time 0 and export to rephase with no fdw or smoothing and start by applying 24dB/oct to 5,200Hz and try 36,48) and see how it goes from there.
Hello - amazing video. I am slightly confused, however, on where you export the target curve inversion filter on REW, as a wav file. After obtaining the magnitude response curve labelled as "L2" in your video, I cannot find a way to export the file as FIR coefficients. My DSP (BSS Blu-100) also appears to only accept .csv or .dat file type coefficients as well.
@ocaudiophile
11 ай бұрын
I couldn't find info on the exact format of the csv filters your DSP unit requiers but instead of exporting the filters as wav files, you can "Export impulse response as text" and then convert txt impulse reponse to csv either in Excel or with an online converter.
@eiddie4108
9 ай бұрын
@@ocaudiophile Thank you very much for the reply. This worked well. Second question. The speakers I use are Genelec 8030c, which have no type of analog filters crossing the woofer and tweeter over. How would you go about phase correction in this case ?
@ocaudiophile
9 ай бұрын
There's a lot of useful information on your speaker here: www.stereophile.com/content/genelec-g-three-active-loudspeaker-measurements There's only one crossover at 3000hz and tweeter is polarity inverted, port frequency seems to be around 45-46Hz You can find some additional info on how to fix speaker box/crossover phase correction in the more recent videos (workshops and speaker phase) @@eiddie4108
@eiddie4108
9 ай бұрын
@@ocaudiophile very great info. Thank you!!
Thanks!
@ocaudiophile
Жыл бұрын
Welcome!
This is top info. I watched your videos with a lot of interest. Great contribution to the community. Would you say that the filters you can create with this method are on par as the ones crested by Acourate/Audiolense?
@ocaudiophile
Жыл бұрын
Thank you. REW is now so powerful (there're also some new tools I'm aware of in the pipeline), I'd say you can do almost anything you do with pro software. I don't have any experience with Audiolense but DIY crowd loves it and they're a tough community to satisfy. In Acourate you can do a linear phase inversion which sometimes has advantages compared to min phase. Also it uses macros to automate tasks which can be more practical. However, with REW you can see all the details of what you're doing.
@joserafaelhernandezcarucci1324
Жыл бұрын
@@ocaudiophile thanks for the flavor, this is definitely worth to spend some time with
I'm curious if these steps are similar when tuning a speaker, not necessarily to the room, but to account for the natural distortion of the speakers themselves. The reason I ask is I purchased some generic T24n tops for my sound system and they don't come with any recommended information on crossover filters, so I'm playing by ear. I have found some settings that sound decent, but I wanted to perform some more detailed filtering. I have ordered a processor with FIR and all pass features and so I'm starting to look into methods to measure and adjust the speaker as a kind of default preset.
@ocaudiophile
7 ай бұрын
Try to invert (1/A) the minimum phase version of the speaker response at the LP. That should give you a filter which will bring out something close to the anechoic response of the speaker when convolved with the original response at the LP.
Hello, i've got a 2 way bookshelf speakers, and i noted a weird (to me ) behaviour in the phase while i work in the rephase section of your procedure: the crossover should be at 1900 ca 12 db/oct and it is quite linear at -180 degree till crossover point then it goes quite linear at 0 degree till 24kHz.If i click on invert polarity it seems much better. Could be the tweeter 180 degree inverted? How could i consider this fact in the procedure? Thank you sooooooooo much 😃😃😁
@ocaudiophile
Жыл бұрын
Very hard to tell without visuals but it's unlikely that your tweeter polarity needs to be reversed. Phase response after 1000-2000Hz is very volatile and dependent on the signal delay due to speaker placement so I'd take that part of the phase graph with a pinch of salt. Apply only xo phase shift correction and leave it at that. Btw, shift at 1900Hz is probably at least 24db/oct. 12 db/oct is quite unlikely unless your crossovers are made of just one capacitor and one resistor.
Wow this is amazing❤
@ocaudiophile
Жыл бұрын
Thank you.
Hi, very nice and detailed tutorial. What happed to part 4?
@ocaudiophile
2 ай бұрын
I automated the whole process instead. Check out A1 Evo!
Hello, Just One question, in VBA correction filter, we have to consider Always the Major lenght of the room, or the lenght of the room dimension where the LP Is ( between front and rear walls)? Thanx
@ocaudiophile
Жыл бұрын
It depends on a lot of factors in such a placement. You first peak frequency will tell you which room length you should be using.
Hi.. .I have a question: when it refers to the filter order of the divisor Does it refer to the electrical order of the filter or the acoustic order? I mean: in a 6.5 inch woofer cut with a low pass filter at 2300hz for example, the filter is 12 db per octave, but the effective slope is 24 db/oct. Which one do you mean to take into account in the RePhase? Thanks for your Effort!
@ocaudiophile
9 ай бұрын
It's one order per capacitor/inductor in the analogue filter
Thanks so much for being willing to share the efforts of your hard work! I have a question following up on the process when one has a separate sub in addition to the two mains. You suggested below to first time align the sub with the mains and then follow your process with the sub actively involved. That's all clear and makes perfect sense. However, I've found there is a lot of inconsistency in terminology across the internet concerning time aligning subs. Would you be so kind as to provide a little more detail on how to do that in this context, or point me to another youtube resource you think explains it well? There's a lot of information on aligning multiple subs with each other, but I haven't found anything that appears relevant to this situation.
@ocaudiophile
Жыл бұрын
kzread.info/dash/bejne/mZVmx7GwerXaiNI.html
Thanks for the tutorial. Have you tried to compare the REW filter, with a filter from Acourate, Audiolense or DRC Designer?
@ocaudiophile
Жыл бұрын
I am familiar with all of them but I am most confident with REW to produce my filters. Some of the vector calculations REW can do lately are not available in any of these platforms.
I need to downsize to a 2.0 full range system. Do you recommend using the UMIK-1 pointed at the speakers or at 90 degrees? Does the VBA filter create meaningful corrections for a 2.0 system? Thanks!
@ocaudiophile
Жыл бұрын
İf you'll take measurements at just one central position or multiple positions very close to each other pointing forward is OK and will give slightly more accurate HF response but otherwise stock with 90 degrees. As long as you use the correct calibration file, messuemrdnts will be accurate enough. Mic direction only effects very highfrequencies. VBA works best with 2.0 especially if the room is symmetric.
Hey, that's a great tutorial! Thank you so much for all the work you've put into this to make it so detailed and clear and for replying to our questions and comments. I reached to the end, but have a couple of questions and would appreciate your help: 1. When I multiply all filters and convolve with the initial response, I get a new response that is not at all that different from the initial. It still has peaks (smaller than the initial - but existing), is not flat against the target and has all the inherent flaws of the initial response in terms of Step, ETC, RT60 etc. So, something is not working right with all the multiplications I do.. On the other hand, if I just convolve L3 with the excess phase filter, I get a perfect L4, from which I can then export the impulse response for Jriver. Is there anything wrong with this approach? 2. Even with the final response, I have the second mode peak in the bass area visible in Spectro, Decay etc and evident when listening. Is absorption the only way to manage this? Thanks again!!
@ocaudiophile
Жыл бұрын
Thank you. 1. Try putting the filter with the widest right hand window length first (as in "A" in "A x B" multiplication). You will have slower filters but accuracy will increase. And you can of course only use certain filters, they are independent of each other. 2. FIlter 2 should be able to remove any peaks from the response. Try removing any smoothing from "1/A" before generating its MP version.
@jimaekgr
Жыл бұрын
@@ocaudiophileThank you, I'll try to play around with the multiplications as you suggest. Filter 2 has removed the peak from the frequency, but it's still visible in the spectro and I can hear it. However, after L4 I tried correcting the phase once again (like new, additional filters 3 and 4) and I managed to smoothen it. On the verge of pre-ringing in the graphs, though I can't hear anything yet.
A quick question Serkan. After filter 3 , L3 has shifted a bit compared to L0 and L2 copy, even though I did cross align L0 and L2-copy prior to phase correction. Can I just realign the final L4 response at the end, before generating the convolution files? Thank you again for all your help!
@ocaudiophile
Жыл бұрын
some shifting is expected during trace arithmetic operations and has no effect on the filters. Just keep adding the filters on the produced results and ideally don't change the window ref time figures in the filters' IR windows. It's not carried over to the convolution file but it helps rew calculate vector operations accurately
Just discovered your channel. I want to use REW and Roon for room correction. Do you recommend this video or the EW (Room EQ Wizard) Top Tricks? Thanks and great job!
@ocaudiophile
Жыл бұрын
REW Top tricks is a quick and efficient correction and it comes with a written guide so you can start with that. This one (Supreme..) is more complex and the first video method is only one fo the 4 filters in it.
This is very interesting. Based off the comments this cannot be used on Denon AVRs? Even with miniDSP? Since I am so new maybe I am missing something... When I finally build my stereo music setup. How would you apply these changes to the amp you are using? Do you connect a PC and then use a EQ software to implement all of these changes?
@ocaudiophile
Жыл бұрын
Your digital music library needs a PC/Mac to be stored and played anyway. And you can use these filters in most media player software. There're also free convolution engines for PCs like Equalizer APO.
Hello again, i would like to know what filter 2 actually does. I know filter 1 is to correct standing waves in the room, filter 3 and 4 correct speaker crossover anomalies, but what about filter 2? To me is the filter thaI has the most impact improving the sound in my hifi, is it for room correction? thank you very much.
@ocaudiophile
Жыл бұрын
Filter 2 works like an an equalizer dimming peaks and boosting dips in the response.
Amazing video, yhank you for your efforts. When I click for the excess phase video, I can not see it because is telling that the video is Private. How can I do to see the next filter video? Thank you in advance!
@ocaudiophile
11 ай бұрын
I removed it because the techniques were updated after certain REW developments. Watch this new video for ultimate excess phase correction : Phase Match Your Speakers with AllPass Filters kzread.info/dash/bejne/dZyE15KuY9HAqMY.html
I tried to make measurements like you and my measurement lm is exactly at 0 like the right spekaers measurement , is it normal or have i done something wrong ?
@ocaudiophile
7 ай бұрын
If right speaker is chosen as the "acoustic timing reference", it will always start at t=0.
Really interesting one. Just a question: are the filters built with this toolchain compatible with CamillaDSP for use in Moode Audio, at your knowledge?
@ocaudiophile
Жыл бұрын
I think they're because I heard Camilla DSP a lot in forums but I have never used. If it's able to process WAV files, this is all that's needed.
@francescodesantis9746
Жыл бұрын
@@ocaudiophile Thank you! Yes, it supports WAV (both mono and stereo) so I will try this setup as soon as I have the time to take the measurements :)
Great tutorial! Do you provide this service for audiophiles?
@ocaudiophile
Ай бұрын
No but I have automation scripts. Look for A1 Evo
Hello @ocaudiophile, I plan to implement this type of work in a 4-way active installation in the car. There are DSPs of different brands in the car. First of all, as far as I understand, in order to use these filters, it is necessary to use a processor with DIREC support. I would be very happy if you could share information about loading the filters into the processor. Thank you very much in advance.
@ocaudiophile
11 ай бұрын
İ don't know what direc support is but most DSP engines with FIR filter capacity will accept wav or bin files. It's possible to save REW filters in these formats.
@Bayezit2000
10 ай бұрын
@@ocaudiophile Thank you very much
@vortexor1
3 ай бұрын
@@Bayezit2000 Which DSP are you using please? Did you sucessfully implemented it in your DSP with the help of REW?
Hi do you have a video on "RT 60" measurements in rew for offline measurement method? (Is there any way?)
@ocaudiophile
Жыл бұрын
Unfortunately no but REW has a lot of explanation in its manual regarding RT60: www.roomeqwizard.com/help/help_en-GB/html/graph_rt60.html
OCD, I am relatively new to room correction although i am long time audiophile. I recently purchased Roon and want to use REW to correct for my room/speaker anomalies. You have so many videos...is this the video i need to fully understand and follow? Please advice which one i should follow to use convolution correction technique for room correction, adjusting phase/timing and aligning with the Harmon curve. Thank you so much for your service.
@ocaudiophile
5 ай бұрын
This is a good one overall showing all steps of room correction but some of the methods in this one are replaced after new REW additions like excess phase inversion. This even older one is a good start if you lack REW experience because it comes with a written manual: kzread.info/dash/bejne/Z42XqpmjYpCyYbw.html Its results are quite good. For phase correction, this one is the latest and beyond anything you can find online: kzread.info/dash/bejne/dZyE15KuY9HAqMY.html
@Guitarkaran
5 ай бұрын
@@ocaudiophile Thank you so much!
hi, this indication "after 1/A operation, decrease window left & right sizes to 1000/1000" is in ir windows left and right width?
@ocaudiophile
Жыл бұрын
Yes indeed!
@lucasantilli6510
Жыл бұрын
Thank's
We are in the obsessive compulsive audiophile channel right? To find peaks and dips for F1 I've exported the measures as text and I read the value to report to the excel tool. Am I obsessive compulsive enough? 😂
@ocaudiophile
Жыл бұрын
That's a new high :) Btw, it will be possible to export txt files without the header in the new REW.
@teb76
Жыл бұрын
@@ocaudiophile Have you ever considered to open a discord channel where we can discuss about audiophile stuffs?
@ocaudiophile
Жыл бұрын
@@teb76 Not a bad idea. I've never used Discord but will have a look.
How can we import these results to a mini dsp or the avr Marantz 6015?
@ocaudiophile
Жыл бұрын
You can't unless you have a HTPC connected to your Marantz with multi channel analog inputs
@ocaudiophile
Жыл бұрын
You might have a chance with minidsp 10x10 but I don't have it and wouldn't know for sure.
@hiramb7956
Жыл бұрын
@@ocaudiophile I Have a few spare PCs not sure if powerfull enough for an HTPC… Could you provide more info (or a video) on HTPC Recomendations and how to connect these to an AVR? It will be nice to see how it all works. Thx
@ocaudiophile
Жыл бұрын
@@hiramb7956 I use an Intel NUC for stereo music and I am quite happy with its performance. But I am not up to date on HTPC tech at the moment to recommend one.
Hi, thanks for the videos. I've been through them a couple of times and created the filters for my system. Filter 2 is certainly working correctly as it successfully gets rid of the huge bump in the bass response from my room. Not sure I can really hear much difference from the other filters, perhaps my implementation wasn't quite right. One thing I noticed was that if I pause a track with the filters on, there's a few seconds of garbled electronic noises. Is this a feature of FIR filters? Or perhaps indicates a problem with my filters?
@ocaudiophile
Жыл бұрын
No, it's not normal. Phase corrections aren't immediately audible but you should hear a heightened and more accurate sound stage.
@oliverfreeborn8944
Жыл бұрын
Ok thanks. I'll have another crack at building the filters, otherwise I guess there's something going on with EQ APO/Qobuz
@ocaudiophile
Жыл бұрын
@@oliverfreeborn8944 EQ APO has some latency inconsistencies but generally behaves well with these filters. Do not normalize IR peaks while exporting filters to EQ APO if you are...
Hello, tell me please, how to make a 0-24000 sweepton for 6 channels (5.1)? In your archive, the sweeptones are 20-20000Hz, but if I set 0-24000Hz in the measurement settings, then the section from 0 to 20-25Hz is measured incorrectly (I have a generator in REW that allows me to select only the L or R channel for saving)
@ocaudiophile
Жыл бұрын
You don't need to use played files for 5.1 surround channels. REW ASIO or Java EXCLusive drivers can produce sweeps up to 7.1 channels. After you select correct inputs and outputs in the Preferences / Souncard settings, just run a sweep from Measure window with "From REW" selected instead of " From file"
@user-rk8pc6rj3w
Жыл бұрын
@@ocaudiophile Yes, I know. But if I choose RAW ASIO or Java EXCLusive, then the sound goes directly to the speakers. And I have it to go through the crossovers/bass management configured in EQ APO. If I use a file from your archive "256kMeasSweep_20_to_20000_-12_dBFS_48k_Float_L_refR" , but set 0-24000hz in the settings, and use files for different channels, for example ""256k_20_to_20000_48k_C", then the measurement occurs with errors on low Freq. I tried to generate a file in REW "Left" 0-24000 and "256kMeasSweep_0_to_24000_-12_dBFS_48k_Float_L_refR" to use it for measuring - everything is fine, but REW can export only the left and right channels, but I need all 6
Really impressed by the quality and all the ideas you come up with ! Maybe just a little question, why are you using add FDW "1 Cycle" in order to smooth the curve before using it in rephase ? Because of this smoothing there is still an excess of phase (hence why you made the part 4 haha) after correcting with rePhase. So I'm wondering what would happend if: -you try to directly correct the non smoothed curved in rephase in order to try to avoid part 4 ? -you directly use the excess phase inversion process and skip part 3 ? Is it only to properly flatter the phase response at the crossover points and then to treat separatly the rest of the phase excess ? I'm trying to understand a bit of the theory behind. Thanks again anyway !
@ocaudiophile
Жыл бұрын
Thanks. Smoothing with fdw:1 is to avoid ringing. Phase filters with a bandwidth (Q) above 1 will cause pre-rigning in the bass frequencies and a phase filter with a Q at or below 1 will effect a large area which can only be useful for a very smoothed down response curve. If you watch the excess phase inversion video following this one, I demonstrated what happens if you fully inverse the unsmoothed phase response.
@-_-3698
Жыл бұрын
@@ocaudiophile Indeed thanks for your fast answer !
This might be obvious to some folks, but I thought I would note it here if someone else has older speakers and no technical specs. I have Paradigm Studio 40v3 mains, purchased way back in 2003, and while I still have the original paper documents, they didn't include the xover info. I couldn't find any site with reliable info for them. I managed to find a site from years ago from a fellow who decided to change the tweeter and crossover, but only his modified xover schematic was shown. Not much help! These are "2.5 way" with a woofer that according to all descriptions is allowed to roll off on its own, plus a midwoofer that is crossed over to a tweeter. One site noted a 1.5 kHz xover to the tweeter, but that seemed crazy low. I then recalled that as they have two sets of terminals for bi-wiring, I could simply disconnect the jumpers and run full sweeps on both the woofer and tweeter taps. SPL traces on REW confirmed the mid-woofer to tweeter xover frequency was 2.5 kHz with 18 db/octave roll off on either side. Of course, I see nothing for where the woofer actually rolls of, and where the mid-woofer picks up. For such a configuration, is the only xover for which phase correction can be applied the mid-to-tweeter? Thanks!
@ocaudiophile
Жыл бұрын
You can try to correct the phase with this alternative method: with a Left + Right speaker measurement, no FDW applied & no smoothing; look at the impulse response and manually adjust to align the FIRST peak of the impulse to t=0. If this peak is negative then you should also invert polarity. Then export the response to rePhase. You should see your phase curve gently and asymptotically reaching 0° (or a 360° multiple if looking at an unwrapped curve) at the Nyquist frequency (24kHz). This means the global LP filter of the system is ignored. Then start with a 24dB/oct linearization at 2.5KHz and move down from there adding filters at phase discontinuties (the dottted vertical lines). It's very unlikely that an old speaker will have less than 24dB/oct XO shift so I'd go higher (36,48) rather than lower (12). As long as you have a fairly flat phase response, it will do the job. When you convolve the resulting XO filter (vector multiply the wav file produced in rePhase with the speaker response), if the "plot not normalized" impulse peak magnitude is higher than the original (it might reverse and this is ok) then you've done a good job.
Hello, when i try to built FILTER3 i have improvement only on one channel (the Left one) compared to filter 2 results ( clarity i.e. is some dB worse than filter 2). Is that possible?, i tryed to repeat at least ten times , but the result doesn't change. ANY IDEA ? an obsessive compulsive audiophile DEPRESSED at the moment.😮😮😮☹
@ocaudiophile
Жыл бұрын
Clarity is meaninglessly too high in typical household rooms, it's a measure developed for auditoriums and halls.
@robertolusa6350
Жыл бұрын
@@ocaudiophile So I'll noit consider that, thankyou very much
Hi, do I need to take the measurements only from the exact listening position or can I take them from a larger area of the room?
@ocaudiophile
Жыл бұрын
If you're the only one watching, just one central measurement point can be enough.
@andreioprea6068
Жыл бұрын
@@ocaudiophile I am not the only one listening in the room and I often move around, this is what I get stuck at - the way you have to take the measurements
@ocaudiophile
Жыл бұрын
Measure at every seating position
First of all, thank you for your great video. I am a beginner Can you explain in more detail the centering of the microphone at the beginning of the video and what are the parameters in REW which allow the right speakers to be aligned at Zero Second, at what point in the video should we apply an FDW to L1 and R1, then reverse. And what do you mean by fdw must be applied? thank you again for your videos
@ocaudiophile
Жыл бұрын
You can use Acourate trial software mic alignment tool for perfect mic centering. There're links in the description. Alternatively you can compare left & right speaker impulse peaks 8n overlays graph and move the mic accordingly until it is centered. The speaker you chose as acoustic reference during measuremrjts will always have its impulse peak at time zero. Frequency dependent windowing should be applied as shown in the tutorial.
@spencerjohnson8026
Жыл бұрын
@@ocaudiophile - I've been a huge fan of your convolution and inversion video. In that video, we offset/align both the R/L to t=0. I did not use Acourate (on OSX) to adjust the microphone for this tutorial. I took measurements as instructed, with R as the reference channel. When I look at the impulse graphs, Rm is not aligned to T=0, and Rl and Rr are not aligned to Rm. Am I missing a setting to get the R reference channel to align to T=0? Or is this due to microphone placement and I need to manually offset/align R/L to t=0 as instructed in the convolution and inverstion video? Thanks for all the knowledge sharing and time you put into educating the masses!
@ocaudiophile
Жыл бұрын
@@spencerjohnson8026 Thank you. Acoustic reference speaker measurements' impulse peaks should normally always be at t=0 regardless of the mic position. In REW/Preferences/Analysis tab, make sure "Adjust colck with acoustic ref", "Align IR peak", "Align t=0 to a smapling instant" are all ticked (and others untiecked). If you're still seeing deviations from t=0, just manually align them until every impulse peak sits on t=0. You can also use "time align feature at some stage. It will not make an acoustic difference from your listening position. It's only for the vector calculations to behave nicely.
@spencerjohnson8026
Жыл бұрын
@@ocaudiophile - Thanks! I have those exact settings configured in the Analysis tab but I still need to manually adjust both the R/L channel to t=0 (Impulse, Offset=t=0) so the R/L aligns to 0u on the Impulse Overlays. Basically, exactly what you show in your Convolution and Inversion Video. Adjustments are very minor and hopefully not missing something major. Additional notes, running 5.20.14, OSX, UMK1 I'm hearing some great (best yet) results taking additional measurements, cross-correlation and vector alignment in this video and applying the Harman House curve and steps in the Convolution and Inverstion video. These are also steps detailed in the chapter after you configure the Virtual Bass Array.
hi, for the linearization filter i have a problem with my speakers with this data Acoustical Active 130 Hz - 4.5 dB/oct and Electrical 1.45 kHz - 6 dB/oct broadmann vc1. how do it? many thanks
@ocaudiophile
Жыл бұрын
I couldn't find a test online for the VC1s but as far as I understand the first crossover at 130Hz is an active one (rather than passive) so it will not cause phase shifts and will not require alignment. The second one at 1,450Hz is even harder to understand as it says just electrical. If it's passive it must be causing some phase shift but at 6db/octave that must be first order so too small to fix with rephase linearization filters. Best way is to check the phase response graph of averaged, multiple measurements and try various rephase filters for 1450Hz to yield the flattest response if required.
@lucasantilli6510
Жыл бұрын
the first filter is a resonator panel tuned to the speaker cabinet
@ocaudiophile
Жыл бұрын
Do you see any discontinuties in the phase graph at 130Hz?
@lucasantilli6510
Жыл бұрын
I did a rough study but it seems so. Unfortunately or fortunately it is a particular speaker
@lucasantilli6510
Жыл бұрын
Hello how is it possible that L3 is -3 db compared to L2? L3 has no Spl info
First question... When doing the crossover linearization, should one do the crossovers and box type linearization also when using a sub and use the complete frequency response including the sub? The prediction in REW shows a lot of preringing this way, but i don't know if the preringing will be real. Second question... When i do crossover and box linearization including the sub in the frequency response and then use a phase filter at 100Hz with +45°, but a bandwith of 0.4 or 0.3, will this cause preringing? If i do it this way the excess phase is naturally hovering around 0° (plus minus 5°) so i basically could skip the excess phase correction in my opinion.
@ocaudiophile
Жыл бұрын
Apply the spec XO phase correction of your speakers, even the lower band will be above the subwoofer crossover frequency. Box/port correction will have to be about the subs though. Excess phase correction below 100Hz is inaudible and will almost always cause ringing.Box phase linearization tools is your best option in that area. You can do mild phase inversion between 100Hz-1000Hz
@Cathul
Жыл бұрын
@@ocaudiophile i already applied xo correction for sub to front (and for all crossovers in the front speakers. It's for a car btw. and it works really good). Question was if i should add the box correction also. It's a sealed sub crossed at 60Hz acoustically to the front system. I would add the box correction for a Q=0.7 and the -3dB point of the subs response, would that be correct? For xo correction i always chose the acoustical crossover, not electrical. Prediction says i get massive preringing from the box filter, but it may be a fault in the prediction. I did the excess phase correction only between 100Hz and 1000Hz, but when adding the box filter for the subwoofer i basically could get rid of that step as the excess phase would then be around +-5° according to my measurements/predictions.
@ocaudiophile
Жыл бұрын
@@Cathul I am not sure of the correct parameters for the sealed box correction but in my home speaker experiences, ringing will only be introduced if the frequency is very low (
@Cathul
Жыл бұрын
@@ocaudiophile this is a problem in a car as you basically sit inside of the box. Therefor i wanted to try to just do the crossover linearization as you cannot eliminate the reflections from the direct sound in a car. Boundaries are so close inside a car that the first reflection is sometimes as early as fractions of a millisecond (imagine a tweeter located right next to the side window or windshield with a distance of 10-15 centimeters). With just the crossover filters and some slight excess phase corrections i was able to achieve several wins in sound quality competitions against competitors which were slightly ahead of me before, and this with a good advantage in regards to the score. That's why i wanted to try the box filter to try to get rid of the excess phase filter. I don't know if i should use the free field -3dB point or the -3dB in the car. In the half field it would be around 40Hz with a typical 12inch sealed sub that is put in a car while in the car it might easily be around 20Hz or even lower, depending on the equalization and actual dimensions of the car (it's a pressure chamber in sub frequencies that can add up to +12/octave starting at around 60-70Hz or even higher depending on size of the interior of the car). I will probably test this out with both options (free field and pressure champer frequency) as i can switch presets rather easily in the Raspberry PI with CamillaDSP.
@ocaudiophile
Жыл бұрын
@@Cathul You sure know car acoustics a lot better than me ;) In the subbass area, the calibration microphones also deviate quite a bit in their phase response and the calibration files lack phase correction. You are usually flattening an inaccurate phase response if you are only doing visual work. WIth very short (5-6 ms) right windowing of the reponse, you can see the phase response at the source but you will lose low frequency resolution which wouldn't help your case with the subs either.
Worderful tutorial! I have a question… I've already done several inversions and I've always done them from about 60 hz more because I'm afraid of breaking the woofers... The VBA acts like an EQ?? does it increase the db and can it put the woofers under strain?
@ocaudiophile
Жыл бұрын
No digital correction can break your woofers. VBA is a maximum phase filter and simply cancels out standing waves by producing a counter signal. There is a more optimized VBA method in this video: kzread.info/dash/bejne/o22VrNOKoLHffLw.html
@BarileTixxoFilms
Жыл бұрын
@@ocaudiophile thank you!
@BarileTixxoFilms
Жыл бұрын
@@ocaudiophile I don't have the access at xlm file....I sended request to you
@ocaudiophile
Жыл бұрын
@@BarileTixxoFilms I granted it hours ago?
@BarileTixxoFilms
Жыл бұрын
@@ocaudiophile ok ok
Hi, very interesting. Struggling a bit too implement. I have a very reflecting room also due to a sloped roof. I have 2 questions please on the vba part to start: 1. Is it necessary to measure my sub apart from my mains and time align and then create an averaged response for left and right speakers to start creating filters, ór can I just make measurements with main and sub together for left and right and ignore any possible timing differences for the sub and main speaker. (Or do the lower sub frequencies not cause timing differences? 2: Is the crossover frequency álways 108Hz, or what determines this? My room sim gives (of course) different standing wave frequencies and I guess due to the sloped roof they differ quite a bit when trying to match with the actual measurements... Appreciate your advice and comments!
@ocaudiophile
Жыл бұрын
1. I would first time align the sub and the front speakers, cross the sub over around 80Hz or below with the fronts and then measure and correct the mains with the sub actively involved. 2. The crossover frequency for VBA is determined by the frequency of the optimized first peak frequency ( x 3.5). The XL file provided will calculate it for you.
@MrLugt
Жыл бұрын
@@ocaudiophile Hi, great, thx. But for 2: the XO frequency now calculates 141Hz for me... Is that not too high? Roomsim peaks expected at 45, 91 and 136 for a rectangular (but wider) room without of course the sloped roof into account: I observe peaks at 47,.9 81,9 and 117 (I guess ...)
@ocaudiophile
Жыл бұрын
@@MrLugt Where's your first peak and first dip (around x1.5 of first peak) at? If you write down first peak&dip frequencies for left and right, I can send you a short impact VBA due to your tilted roof.
@MrLugt
Жыл бұрын
@@ocaudiophile first peak observed is at 47,9 first dip at 65,4. This is my what my first try gave: combined sub with main. Need to do reruns with time aligned sub maybe... Is there a formula for a tilted roof?
@ocaudiophile
Жыл бұрын
So we're talking about a room with around 3.6m length. Create a lpf with XO at 68.60Hz. The polarity inverted MP version will have impulse dip at 11.52ms. Give it a -10.3379 time offset and move the impulse dip to -21.8579. That's your VBA filter for whichever speaker that is. Do the same process for the other one. There's no formula for tilted roof, at least not one I have developed but I am just cutting the filter quickly to ignore second and third reflections. This should help shave the worst peak and dip.
Dumb question: what do I do after I complete the inversion filter (filter 2)? Do I generate a Wav and use it as a convolution filter in Roon or Jriver or EqualizerAPO?
@ocaudiophile
Жыл бұрын
Yes you can but better combine all 4 filters in one and use that one.
@joek6207
Жыл бұрын
@@ocaudiophile woah!? So these aren’t individual methods but methods to combine them? Did the video cover that?
@ocaudiophile
Жыл бұрын
@@joek6207 There are 4 different filters working together. The 4th one, excess phase inversion didn't fit in this video, it plays after this one.
@joek6207
Жыл бұрын
@@ocaudiophile I saw that. So we add all of the filters to Roon or EQAPo? The accept multiple filters?
@ocaudiophile
Жыл бұрын
When you multiply (trace arithmetic) them with each other, you create one filter which contains all of them and you can use this filter (.WAV file) in Roon or EQ Apo.
Is it possible to do active crossover by this? Do we only need to change target type in EQ target setting tab? Thanks! Just realized from this video that cross correlation is based on the one at the top......I need to do the measurement again now
@ocaudiophile
Жыл бұрын
In the REW EQ window, there are two crossover filters available (last two filters) and you can create FIR filters (wav files) from them. rePhase has even more flexible low pass and high pass filters.
@andreasheiden7122
Жыл бұрын
Excellent Video... again!😊 I'm still struggling with the phase correction of the speaker(?): As per measurement I have 4 phase shifts in the high frequencies. After uploading the txt.file (fdw 1!) to rePhase I need to apply a time offset of -288 and then I do the Phase corrections (after I Inverter polarity) with the filters linearization tab and the paragraphic phase EQ. When I upload the generated wav file to REW then, and so the a*b convolution with the base measurement (the one I uploaded) and do time alignement and then look at the phase it still goes zick Zack in the high frequencies. This only disappears, when I Offset t=0 to -288... Is this correct? What does this mean? The step response after time alignment looks much better though (First Peak upwards due to inverse polarity of filter). Nevertheless under SPL &Phase, the Phase does only Look better with time offset -288.... Maybe I'm just to stupid to understand, what's correct, but I'm confused....
@ocaudiophile
Жыл бұрын
@@andreasheiden7122 The phase response at high frequencies is very much dependent on the mic position (thus measurement timing and your results being dependent on Offset t=0 parameter) and is very volatile. You don't really need to flatten phase outside of 100-1000Hz except fro crossover phase corrections. Watch around minute 5:00 of this video: kzread.info/dash/bejne/gaSus6iHcpuoc5M.html There's also more up to date info on crossover and box phase correction in these videos: kzread.info/dash/bejne/opZ3rMyHp5bIeps.html kzread.info/dash/bejne/q5iBzdKJe7zbkaQ.html
@andreasheiden7122
Жыл бұрын
@@ocaudiophile I now did new measurements with 1Mio/48khz in different positions. This now not only gives me very strange results when I calculate vector average, but also when importing in rephase there are now so many phase shifts to the right side (highs) that I can't get rid of them. I also have to time Offset to the opposite direction than before (now + more than 1000micrs) with my previous measurements that I took with 256k.... I have no idea ...🙄 That's frustrating...😭
@ocaudiophile
Жыл бұрын
@@andreasheiden7122 you must be vector averaging before you cross corr align them...
Should you have ‘decimate IR’ selected in the ‘Analysis’ tab of the preferences pane in REW?
@ocaudiophile
Жыл бұрын
Very rarely like if you measure a subwoofer for a limited freq. range only, it will decrease the sample rate. For example, if you measure with a lower high frequency range like 0-250Hz instead of 0-24,000Hz, the sampling rate will drop from 48kHz to 3kHz and operations like "Cross corr align" will not work between measurements of different sampling rates. So better untick it.
@ts6640
Жыл бұрын
@@ocaudiophile yea…..that is what I thought. Many thanks for confirming
@ts6640
Жыл бұрын
Did you mention why you didn’t use Frequency Dependent Windowing? Sorry if I missed it
@ocaudiophile
Жыл бұрын
It's in the description under the video.
@ts6640
Жыл бұрын
Apologies but I do not see the explanation. perhaps you can copy and paste the answer? sorry for the aggravation
Wouldn't it be better to do the inversion step with a dB average instead of a vector average given that vector averages usually give less volume than a dB average? I know that this is due to the vector average also putting phase into the calculation, while dB averages obviously do not. As we don't correct phase above 1-2kHz anyway this wouldn't really matter, right? Anyone tried this?
@ocaudiophile
Жыл бұрын
I used to use RMS average (root mean square of dB is more accurate than just dB average) for the exact reason you've explained but you'll see that target curve levels produced for both are almost always nearly identical so I don't bother anymore and with vector average I have additionally the phase response to play with.
@Cathul
Жыл бұрын
@@ocaudiophile nevertheless i've tried it. ;) Love to tinker with stuff. With dB averages i get a lot more preringing after inversion than with the inversion of the vector average. I don't know why as i don't see any reason as the inversion of a frequency response is independent of any phase information and with variable or psychoacoustic smoothing the corrections aren't stronger/larger than with the vector averages and a FDW of 15.
@ocaudiophile
Жыл бұрын
@@Cathul If you remove smoothing from the final filters you produced with dB average (and this is how they really go into the convolution engine) you will see sharp peaks and dips which cause the ringing.
@Cathul
Жыл бұрын
@@ocaudiophile yes, but i also have sharp peaks and dips in the final filters from vector averages. ;) Still vector averages filters produce less ringing here.
I just went through your FIR video... is this worth doing instead?
@ocaudiophile
Жыл бұрын
This is more complicated but even better. Make sure you download the zip linked in the description.
@-_-3698
Жыл бұрын
Hello, I've watched a lot of videos from this channel, but I'm not sure about what video you are talking about, can you please tell me exactly ? Thanks
@ocaudiophile
Жыл бұрын
@@-_-3698 the one above...
@-_-3698
Жыл бұрын
@@ocaudiophile My bad I didn't understand the comment
Hi! Thank you so much for your effort! **SOLVED! Works fine in v 5.20.13** Possible bug or change of methodology in latest beta release. When I do a L1/Target calculation, the resultant measurement has an average of -40db or so, when I do a 1/A inversion of this, it generates at around 0db (target level reads at -34.7db). I have verified over and again, and to this point all else looks good. Any idea? Thanks!
@ocaudiophile
10 ай бұрын
dB offset levels of trace arithmetic A/B operation have been recently revised in REW early access versions. The offset you need to apply to 1/A might have changed. As long as the result of AxB operation with the original response and the filter (1/A MP) is on the target curve, you should be fine.
@nitinsidhu
10 ай бұрын
@@ocaudiophile It seems to be ignoring the target level, but follows the target shape. Irrespective of the set target level, 1/A-mp * L1 results in a level which is the avg level of L1. Does it matter that I used loopback as my timing source rather than the right speaker. I have aligned all impulses to time 0. Are you available on any forums, or do you recommend any where I can further understand and educate myself. I Thank you so much again!
@ocaudiophile
10 ай бұрын
Although I don't think there's much relvance for the probelm you mention, loopback as timing reference is very tricky to get right. I would suggest to try with acoustic timing reference. If you share your mdat here I will have a look. You can post google drive links here.
@nitinsidhu
9 ай бұрын
@@ocaudiophile You are a hero! Thank you! I find loopback quite accurate to measure timing delays and use it with SMAART all the time. Although i find the REW implementation very unintuitive. mdat file attached. Input 1 and 2 denote left-right measurements. I have tried to have a measurement for each step. I get stuck generating filter 2 as it is not following the target spl. I suppose that is because I am unable to calculate the db offset required to bring the 1/A inversion to 3db. Googledrive link below. Apologies for the large size. These measurements are in an almost 10ftx10x10 room with no treatment. I monitor near field and obviously while these being extremely poor acoustic conditions, my aim is to learn. Thank you much once again! drive.google.com/file/d/1KVYCZh5-zCDoOMO43RhzQsjboU_876pe/view?usp=sharing
@nitinsidhu
9 ай бұрын
**SOLVED! Works fine in v 5.20.13** Possible bug or change of methodology in latest beta release.
Hello, I have watched several of the videos and some few times over but I feel incomplete in my understanding. Its me, not you. As a relative newbie, all I am trying to figure out is how to calibrate by subs and speakers including the process on how to calibrate subs (steps in sequence) and how to calibrate and integrate subs and mains (steps in sequence). I find the information sprinkled in places and find it hard to follow. Can you tell me if there is a specific video that answers these or is it all over the place? Perhaps a suggestion would be to have an overall calibration video without going into tech details. Just a suggestion and what I found for myself and by no means a criticism. I am just not smart enough...:)
@ocaudiophile
Жыл бұрын
You're right. I will make a new video soon.
@Rayman-expertsoundpro
Жыл бұрын
@@ocaudiophile Thank you for the response. Just a little bit more detail on what I am thinking and then of course you can decide what works best for you: 1. Before starting calibration clear out EQ settings in minidsp etc and turn off REQ (Audyssey etc), zero out speaker distance, change trim level to zero etc 2. Then calibrate subs. Steps to follow to set up subs correctly. Impulse response between sub to sub 3. Integration of main to subs. Steps to follow in sequence - Impulse response between sub to speaker and speaker to speaker? , crossover setting 4. Loading of filters into minidsp as needed 5. Audyssey run required? If so your recommended steps including any hacks etc 6. Use of RT60, spectogram and waterfall etc I am sure I am missing several steps but wanted to show what I had in mind. Question: I have been setting up speaker crossovers based on what had the best frequency response. If I go with a 80hz crossover it does not measure better than a 100hz one. Would not go with what measures the best and likely sounds the best too?
@ocaudiophile
Жыл бұрын
Ideally the crossover frequency should be chosen according to the actual frequency response of the speaker in question in anechoic chamber. You can find the frequency where the bass starts to roll of in the tech specs of your speaker. The summed response with the sub will very much depend on the sub and that speakers distance settings though due to phase cancellations.
@Rayman-expertsoundpro
Жыл бұрын
@@ocaudiophile That is the kind of stuff I am requesting in the overall video....thank you for all that you are doing
@ocaudiophile
Жыл бұрын
I have read your suggested steps and they make sense, too (although there's a lot more to optimal correction than you believe!). But the reason I don't post a new tutorial with the best possible correction is not that I can't decide on the steps. It's the variety of all the systems out there and the difficulty in coming up with a simple standard method which fits to all (try and read all the comments here or any other video if you don't believe me, to have an idea about all the different problems people can come up with). I can make a video for the best possible method for your system in an hour but 3 people will watch it, majority will stop the video within the first 10 seconds when they hear subwoofer phase alignment. YT will stop promoting the video seeing people dash out in 10 seconds so eventually even the 3 people who would benefit from the video will not be able to ever see it. What I have laucnhed so far is quite impressive in terms of compatibility, quality and ease of use.
Is this better than Dirac Live?
@ocaudiophile
Жыл бұрын
By a margin 👍
@Smecerul08
Жыл бұрын
But Dirac says they also improve impulse response. This does the same?
@ocaudiophile
Жыл бұрын
It does more!
@Smecerul08
Жыл бұрын
If this is doable also its genius. I am looking to band pass a channel in 3-4 way, and by convolution alone can the phase be aligned if its not? I ve read somewhere that time-phase alignment its the ultimate thing that can be achived by dsp, represented with that spectrum which should be symetrical from the middle, i dont know how its called. Do you know how to do that? Do you have a video on it?
@ocaudiophile
Жыл бұрын
Check the "workshop" videos. The methods are slightly more optimised and they also make use of recent REW improvements. Anything can be done to the sound signal with FIR filters.
"hi guys" means "hi everyone". Don't listen to the whiny commies.
@ocaudiophile
11 ай бұрын
:)
Thanks!
@JP-ds3lg
Жыл бұрын
I know this isn't about money at all but your guidance has opened the next door in understanding my next level of learning. I really apprciate your time and knowledge being shared.
@ocaudiophile
Жыл бұрын
Very nice to hear that from you and many thanks!