This video tutorial provides an in-depth guide to digital room correction (DRC) methods, focusing on the latest developments and techniques. It covers essential aspects of setting up a stereo system, speaker placement, accurate room measurement, and the application of various digital filters for audiophile-grade sound. The tutorial is structured into steps suitable for beginners and offers advanced insights for experienced users.
Hi folks, obsessive compulsive audiophile here. In this tutorial, you'll be updated with the latest developments in digital room correction methods. Most of the revisions to the methods in the previous videos are minor, but the improvement in sound will be significant. There are also new shortcuts to previous techniques thanks to the recent developments in the Room EQ Wizard, which as we may know, is adding new tools and features almost weekly lately. Additionally, I would like to bring to your attention that, OCA channel has recently surpassed 2,000 subscribers. I guess it’s not bad in less than a year after its launch given the highly technical and quite unique focus and my tedious presentation. So, to extend my appreciation for your dedication, I will try and include as many new tricks and tips as possible in this one. The tutorial assumes the viewer doesn’t have any experience with system set up and digital room correction and has a total of 10 quick steps covering all aspects of digital room correction. If you follow all these steps, in less than an hour you will learn all that you need to know to achieve audiophile grade high fidelity sound in your home set up. The video is fully indexed and experienced users can skip straight to chapter 4. There are 6 different filters explained from chapters 4 to 9 and the optimal filter is a combination of all 6. In the final chapter, we will go over the techniques to create the optimal convolution files from this combination for various convolution engines.
To apply any kind of digital correction you first need a digital music library or a digital streaming service and obviously a computer to play them with. In the exceptional case that, you only play CDs and SACDs through a player, you will need an DSP capable interface between your player and amplifier. This could be a MiniDSP unit and even the most basic MiniDSP kits are capable of filters 6,7, and 8 which will get you 80% of the way to the optimal correction. But with a computer as your digital player you don’t need any of that. Even a Rasperry Pi is powerful enough to process high tap count convolution filters which can contain all of the 6 filters we will create in this tutorial. In the long run, you will want your music source totally silent for which you should consider investing in a fan-less PC. There are free music players for both PC & Mac like Foobar, there are free DSP programs like Equalizer APO and if you are on a Mac, Camilla DSP. For a little cost, JRiver is a very powerful media player with great convolution capabilities. For a little more investment, you can subscribe to Roon which can use your own music library, utilize high definition music streaming services, have a huge album information database, state of the art DSP engine as well as multi-room streaming capabilities. After being digitally corrected in the computer, the sound will need to be amplified before reaching the speakers. Just an analogue integrated amp will be enough if you want to use the DAC in your computer. A separate DAC or recent integrated amps with a DAC will be a better sounding solution and a pre/power amp combination will sound the best.
Your choice of speakers should be determined by a combination of factors: Budget, room size, ability to place speakers far or close from the walls and your room reverberation time. Some speakers are bright and doesn’t like reflective rooms while some speakers are designed to work better in reflective rooms and even floor carpets may ruin their sound. Some speakers need to be placed close to walls to produce enough bass while some only sound good when placed far away from walls. $100k high end floor-standers close to the walls in a small room will sound worse than a $1000 pair especially if the $1000 system has proper DRC applied. The amplifier choice is very much dependent on your speakers. An amplifier should be powerful enough to be able to deal with the lowest impedance loads a speaker requires and speakers can vary massively in that aspect. Unfortunately, speaker sensitivity as advertised by most brands today, is not a reliable measure of a speaker’s compatibility with less powerful amplifiers. Click on the link at the top right corner of the screen now if you’re interested to listen to famous speaker designer, John Devore’s excellent comments on this subject. So, how do you match a speaker with the correct amplifier? My solution is an audio database from 2021. It’s German like most good things in audio but you can translate it easily. The list compares most speakers out there according to their sound quality and required amplification power – they call this value: AK – which they also rate amplifiers with. The list also rates every speaker for ideal room size, preferred wall proximity and room reverberation. You can download the pdf file by clicking on the link at the top right corner of the screen now. It should be a good reference before you select your system components although ultimately you should do a listening test or at least search audiophile forums for views on the combination you’ve picked. Finally, not all amps and speakers are designed for neutral sound and flat response. Many popular amplifier and speaker brands embed their own sound into their designs. In that respect, it’s generally advisable not to match a bright amp with a bright speaker and an amp with a warm sound with a warm sounding speaker.
Now, on the subject of adding a subwoofer: Even the largest floorstanding speakers are not able to produce bass frequencies down to the lowest audible levels and almost no bookshelf speaker can go nearly low enough. A subwoofer not only extends the bass so you will hear more of the music the studio engineer has recorded but it will also take most of the hard work from the speaker [especially with two-way speakers]. It will decrease the overall distortion by helping the amplifier and the speaker produce mids and highs more comfortably. Although a subwoofer’s bass extension will increase with corner placement, optimal hi-fi systems usually prefer two subs placed right next to each of the carefully placed stereo speakers to keep the phantom image intact. Audiophiles will also prefer sealed subs to ported subs because these have faster bass response and no port phase distortions although they will require much more power to produce the same low frequency. So long as you measure your system with the subwoofers integrated and active, all the filters in this video can be applied in the same way but we will not go into any details of subwoofer integration to a pair of speakers.
Avoiding noise and hum: If you have dirty electricity, make sure every component is plugged in the same wall socket in your system chain. Any cable [including an irrelevant HDMI cable] coming from a source plugged to a different socket might cause electrical noise and hum. Don’t spend money on power conditioners or noise filters. Most are snake oil and every hifi component already have power conditioners built in. Don’t spend money on expensive cables, either unless you need to use very long cables. This is also true for RCA interconnects. If you need long analogue interconnects, prefer XLR cables. Speaker cables need to be thick [low gauge] and that’s about the only requirement. The resistance of the wire depends on the cable thickness and cable length. A longer thinner wire will have higher resistance than a shorter thicker wire of the same structure. An ideal wire should have the lowest resistance possible. So, keep them as thick as you can especially if they will need to be long. Some amps like Naim do not have extra inductance networks in their output and the speaker cable needs to provide the correct inductance and capacitance. Because of that, they should be at least 3.5m longs and equal length. But that’s an exception. Try not to fold or especially coil extra length of speaker cables and keep them away from power cables.
Now, on the subject of speaker placement. I aim to create a comprehensive video guide that covers all the essential aspects to accomplish exceptional stereo sound. Lately I've been creating convolution filters for numerous individuals who have been encountering difficulties with the techniques presented on the channel. While it does consume a significant amount of time, it enables me to truly test my methods across a variety of systems, speakers and their respective placements. Due to the significant number of measurements that exhibit fundamental mistakes in speaker placement, I made the decision to include these two additional chapters focusing on the basics. The listener and speakers should form a triangle, such that the angle between the LP and each speaker should be equal and between 22.5 and 30 degrees. So, not necessarily an equilateral triangle. Your speaker manual will often cite the minimum distance between speakers and this should be strictly followed for proper stereo separation. Speakers placed the optimal distance apart will produce a strong center image - also known as phantom image and a wide soundstage. The nearer the loudspeakers are to walls and corners, the louder the bass. You can reduce excessive bass by moving the speakers farther out in the room. How far into the room the speakers are positioned, also effects the clarity of the bass because certain speaker locations don't excite the room's resonant frequencies - resonant modes as strongly. You can reduce these resonances by following the rule of thirds, which states that the distance between the speakers and the wall behind them should be one third of the length of the room. This is often impractical, but one fifth the room length is generally the next best location. If you can't even manage one fifth try one seventh. A deep, expansive sound stage is rarely developed with the loudspeakers near the front wall. Pulling the speakers out a few feet can make the difference between poor and spectacular sound staging. LP should also ideally be one third or one fifth of the length of the room, but from the wall behind your seat. For example, putting the LP exactly in the middle of the room will sound quite bad no matter how you place your speakers. The distances from the speaker's acoustic center, which is usually the woofer's magnet, should all be different to floor, side, and rear walls. This is also quite important. Some people try to achieve golden ratio between these distances. Symmetry between left and right speakers is critical, but if you have an open plan room on one side, at least keep all the other distances equal between speakers. With digital room correction, the extra bass coming from the speaker at the closed side can be fixed. Toe-in means pointing a loudspeaker inward toward the listener rather than aiming it straight ahead. Optimal speaker toe-in amount is determined by tweeter directivity and dispersion, and manufacturer's recommendation should be closely followed. Some speakers need to be turned directly towards the LP, some need to be toed-in slightly while some need to be facing forward with no toe-in. Although toe-in will influence the strength of high-frequency response at the LP, depending on the speaker, its major effect is on the width and depth of image. Because this is purely a function of side wall reflections. The beaming characteristics of the speaker's tweeter, which depends on the type, material, diaphragm size, et cetera. of the tweeter, will determine the area high-frequency sound is reflected from the sidewalls. For fine tuning of distance between speakers and correct toe-in, I strongly recommend experimenting with this song, "Her majesty" by Beatles. An only 23 seconds long song. The vocals and guitar start off in one channel and slowly pan all the way to the other channel. If the vocalist gets closer towards the LP as it moves to the middle, there is too much toe-in. If the vocalist moves away from the LP during the panning, there is too little toe-in. It should sound like McCartney's vocals roll smoothly and continuously across the sound stage. If it sounds like he jumps from right to center to left, the speakers are too far apart. Finally identical toe-in for both speakers is essential to realistic sound staging. You can experiment with the song "Amused to Death" by Roger Waters to achieve that. It starts with a conversation coming from a TV to your rear left with only two stereo speakers in front of you. You will only hear it coming from that location with precisely equal left and right speaker toe-in. This information should be enough to set up an enjoyable system, but for the dedicated obsessive audiophiles out there, I have a couple of more comments. Stereo imaging is an illusion and it's created in our brains from the direct and reflected sound waves coming from a pair of speakers, equal distance from our ears. If reflected sound waves arrive too soon after the direct sound, this illusion is spoiled. There are varying views on the required delay of the loudest first reflections to arrive, to avoid this. According to Toole, it's about five milliseconds. According to Linkwitz, it's six milliseconds. Also 8ft, which is 2.44m is the accepted minimum required distance between the speakers for proper stereo separation. Six milliseconds delay requires just over two meters of extra distance traveled by the reflected sound compared to the direct sound. To achieve equilateral triangle with 2.44 meter separation between speakers and 6 milliseconds of delay from every first reflective surface, you need a minimum room size of 5.19m x 4.33m (17.03ft x 14.21ft). In the link at the top right corner, you can download the paper in which this is calculated. It's quite interesting if you're into these things. I have also made an Excel sheet, which calculates all first wall reflections from your room dimensions and speaker properties and placement. It calculates the extra distances sound wave will travel compared to direct sound from all six wall surfaces. Then will calculate delays and the dips and peaks that will be caused in the frequency response. As you can see, I can not achieve 6ms delay in four out of six first reflections in my current setup. The room dimensions are simply not big enough to be able to do this. Click on the link above now to download the worksheet if you want to play with it.
Measuring your room correctly is the single most important aspect for optimal digital correction. You need a measurement microphone. A measurement microphone is very different to recording microphone. They have omnidirectional capsules designed to be sensitive to sound coming from all angles. They can be either USB or XLR types. USB types are simple plug and play devices, but have clocking deviations. XLR types will require an audio interface with phantom power and are harder to set up initially, but time alignment of measurements is easier. The difference between the best and worst ones will only be at the very low and very high frequencies. And as long as they come with a calibration file, a cheap one will be good enough for DRC purposes. Keep your measurement signal path as simple and direct as possible. A lot of measurements I receive from people suffer from large clock deviations, induced by unnecessary devices between the computer and the speakers. System generated large clock deviations, ruin the measured impulse response. Almost no electronic device would audibly change the correction required by your room. Let alone cause a change within the accuracy of your measurement microphone. So, just add them later after you complete the measurements. Use minimum latency drivers in REW. ASIO is the best, but Java EXCL drivers are usually easier to set up. For ASIO, you have to install either ASIO4All or FlexASIO. Having said that I never saw any measurable difference between Java exclusive drivers and ASIO to this day. Mathematically, you can only obtain optimal correction at a single point in a room. So rather than measuring it multiple microphone positions, I will suggest taking repeated measurements at a single central position, which is also your LP. Some of the measurements will almost always be off due to noise, clock anomalies and what have you and averaging repeated measurements after carefully eliminating faulty ones is the secret to robust correction. Tweeter height, your ear height at the LP and microphone height should all be the same. Always measured from zero Hertz to 24,000 Hertz. This is important for filter calculations with 48,000 Hertz sampling rate, which is the REW default. In general, to achieve better signal to noise ratio and resolution in your measurements, increase the measurement length AND/OR volume. For better timing, accuracy i.e. phase measurements, use shorter measurements. Keep your measurement volume at the level, you intend to listen to your music for the target curve to work as intended as our perception of low and high-frequency relative loudness varies significantly with volume. If you had to make multiple microphone position measurements, make sure you time align them with cross correlation before averaging.
Okay. Finally, we are starting the digital filters. The first filter is the speaker box and crossover phase correction. It's a minimum phase filter. It fixes phase shifts caused by passive speaker crossovers, Active speaker owners do not need to do this correction. I will also correct for the box phase distortion caused by ports or in sealed boxes, just the box. This correction, minimizes group delay and lifts the central sound stage because when you are time align all the drivers and the speaker the sound will be focused at the higher driver. So, the center stage, phantom image will be lifted. Let's check the preferences I use. Java drivers, not ASIO in this case. You can just as well use ASIO, but then you've to pick ASIO4All and then ASIO control panel and select your input and output here properly with buffer size and everything. This is usually easier. But instead of this, for example, to my DAC connected with a USB, you have to use the Wasabi exclusive Java drivers, which have much less latency. And also pick Umik here. There are two. The one with exclusive drivers. Output speaker left, timing reference output, right. This is mono. But, still whichever it picks. Treat 32 bit data as 24 bit. Use main speaker test calibration to check levels. And go to Sound settings for Windows/Playback. Your speaker output is the default. When you configure it, make full range speakers front left and right are ticked. And go to properties, levels, hundred, enhancement disabled. Advanced, 16 bit 48,000Hz. Allow applications take exclusive control, give exclusive mode, both ticked. And that's it. And for recording also, Umik needs to be properly configured. Listen to this device unticked. Levels hundred, enhancement disabled, and two channel 32 bits. Because this is 32 bits, here in preferences I have to tick this one. It's useful because internally REW works at 48 kilohertz sampling rate, I picked this one. Calibration files. No calibration for the output. If you use an XLR microphone with an audio interface, then you could add a calibration file for the audio interface here. REW can produce a soundcard calibration file by measuring its own. And, for the Umik, because it is on a vertical axis pointing upwards, I use the 90 degree calibration. These are all the other settings I use. Adjust clock with acoustic reference. This is important with USB microphones, nothing else ticked here. And use right window width for minimum valid frequency. Equalizer, we're not going to use. And View is up to you, really. These are the settings, make a check levels here. As you see, the microphone is working. I can see. Sometimes, when you don't see anything moving here, just do it one more time and the USB microphone comes to life. And then when you click next, you're going to start hearing also the output. Like, so! I dropped this from the default minus 12 to minus 16 because at -12, it's in the yellow area outside the green region. So, I decided that they should all be green. But, feel free to do whatever you like. This is also related to microphone sensitivity. This is a Umik-2 and still it's not very sensitive, but It's Input sensitivity is more than enough for calibration related measurements. Let's measure left speaker, from 0 Hertz to 24,000 Hertz. This is important for filter creation, all the calculations. Because we are using 48,000 sampling rate this has to be the range of the frequencies. This is not gonna damage your speakers at all. So, feel comfortable. Use acoustic timing reference. If you are using a sound card, XLR microphone, here you should use loop back as timing reference. But for USB, acoustic is your only option. Measurement length, because this is going to be a phase calibration, timing related, we want consistent and short measurements that precisely measure the phase response. 64K; the new addition to REW should be better, but in my system for some reason 64K measurements are not as consistent as the 128K measurements. When I say consistency, it's their impulse peaks. If they are not exactly on top of each other, then it means there are little clock deviations from measurement to measurement. But shorter measurements will normally give the best in terms of that. Sampling rate 48kHz. We are measuring my left speaker. Right speaker is the acoustic reference. When I'm measuring my right, right should still stay as the acoustic reference. And another new REW feature, you can do now repeated measurements. You can also measure all channels at once with squential channels. This is especially useful when you're measuring multichannel. But at the moment I need repeated measurements. normally I do five repeats, but for keeping this video short, I will do three. Do as many as you like, it's useful. Check levels. One more time here. Okay. 75dB, this is coincidental. This is the level I listen to my system. actually a single speaker being 75dB means when they are working together, it's like 80dB or more. But this is how I listen to my system all the time. So this is where I'm going to be measuring it. Let's start and it's going to do all three measurements one after the other automatically. I just click one start button. I'm not moving the microphone. I'm not changing anything. Here it goes. All done!. Let's go to AllSPL. As you can see, there are differences between measurements. I mean below 20 Hertz, it will always be very different because neither the microphone nor probably the amplifier or the DAC are not designed for anything below 20Hz. Now copy everything to overlay graphs and go to overlays. And check the impulse responses. Zoom in. It's useful to have the "Show points when zoomed in" ticked. So you see all these dots. Well, there doesn't seem to be any anomalies in any one of the measurements. Sometimes you will see some with short peaks, a measurement, and you can just delete that one. Eliminate if you have enough measurements at hand. Although, I am tending to get rid of this one, this measurement. Because here there are weird peaks on this one and there's a huge dip here. I will eliminate that. And as you see, it looks a lot better. So this is the importance of repeated measurements. Everything was the same, but that one measurement turned out and there wasn't really any external noise in this measurement. That's the importance of multiple measurements. So I'm going to average these two measurements, but before you vector average, them, you have to always align everything with cross correlation. This faulty one is not ticked. So it's going to only align these two. Cross correlation align. And as you can see now, they are all on top of each other. You don't need to click cross correlation multiple times anymore. This has been fixed. With just one click, it gets to the final aligned position. Once they are aligned, you can vector average them comfortably. You won't lose any frequency response. If they're not properly time aligned, this vector average would deviate in the high frequencies or elsewhere from the original measurements. Go to overlays one more time. Go to phase response. Fit to Data. This is the final phase response that I'll be dealing with. It can be seen that it is in line with every other measurement that are not anomalies. This average seems to be showing at least up to 20 Hertz, proper phase response. Now we are ready to export it to rePhase. Export measurement as text. Use range of measurement, resolution of measurement, smoothing of measurement. And this one is ticked and you can tick this one, although it doesn't make a difference just to get rid of the headers and comments. And save it as left speaker, let's say. Then time for rePhase. Measurement, import from file, LS.txt. And here comes my measurement, okay? Also with this one. Otherwise, I can't reach my auto hide task bar because of screen recording. Now, Focal Kanta 2, review. here. if you're lucky and your speakers are tested, you go to specifications. Crossover frequencies. It's a 3-way speaker, so it has two crossovers. One at 260 Hertz. And one at 2,700 Hertz. Crossover frequencies, you will almost always find. It's usually listed even in the manual of your speaker. So what I have is the frequencies. When you go to filter linearization and the first crossover is here. And the lower frequency is here. What I don't know are the filter orders but it should not be less than 24dB/oct. You can be almost certain for that with every speaker today. And this yellow vertical line, temporary line just shows 2,700Hz in the graph. 12 dB/octave, this is second order. Every order is 6dB per octave. This is fourth order. This is 6th order and eighth order. Anyways. I know it's a fourth order. Then comes the 260 Hertz, which is an additional two more orders 12dB/oct. So even this crossover is 36dB/octave, which is sixth order. But because you already dialed in 4 orders here, it's the additional. So in fact, no crossover will be less than four orders in today's speakers. What I see from this graph right now is nothing is in the middle where they should be, zero degrees phase. This is because, I can see them, but also if I go to measurement site, it says that somewhere here. In the time domain, Kanta No: 2's step response on the tweeter axis reveals that it's tweeter and mid-range drive units are connected in inverted acoustic polarity. The woofer's in positive polarity. So, this is also quite common. These two crossover filters are inverted. So, I invert them here also. Polarity inversion. But from here, not from here, this doesn't affect the Filter the rePhase will generate. It only inverts it on the display. Here you can invert it. And once inverted, you can see that 260 Hz now is almost at 0, 2700 is almost at 0. It doesn't look like it's at 0 because of the time alignment now, because of all these filters, if you do that: as you see, this is around zero now. It's properly aligned. This also has no effect on the filter generated. It's just for display. Anything in this area you do, doesn't have any effect on the generated impulse from rePhase. Everything you do here will go into the filter. I wouldn't have to do this time calib ration here. If I exported the measurement properly from REW. Let me show that to you quickly. I forgot to do this. This is our average response. You have to move the first peak of the response to exactly time zero. So in our case, This is the average response. And this is its first peak. Here. 53.1 microseconds arriving t oo late. To fix that, you go to impulse, controls, Offset t=0. And because it's arriving late, you have to apply a delay, but 0.0531 multiplied by 1000, it's going to become milliseconds. that is the correct way. Here you enter milliseconds and it is 53.1 microseconds. Apply! Now you can see, it's exactly at time 0. If you exported it like that then you wouldn't have to do these time alignments in rephase and it would be easier to see the correct filter orders. Let's call this LS aligned. Save, and now with rePhase. Which I guess is here. If you import this new measurement. As you see, you don't have to give this offset. It is already properly aligned. This is the easy part. I have now corrected the phase shifts caused by both of the crossovers in my speakers and the phase is now a t around zero up until the low bass. Here we have port phase deviations, phase shifts caused by the port. And I also know from this test here they have the anechoic measurement for my two ports and the woofer separately where it says: the minimum notch in the woofer's output occurs at 41 Hertz coinciding with the ports peak output. So, my port's peak frequency is 41 Hertz. 41 Hz, because it's a port, it should be vented. Closed Q's are for sealed speakers. So let's try vented low Q. vented standard Q and vented high Q. As you see these all come to the middle to zero axis. So all of these are working, but which one is the best? This one. This one. Or this one? Ignore the area below the bass your speakers can produce. So mine are around 30 Hertz. It is not trying to put here. On zero, but everything after it, Well, unfortunately, there is no way to find out visually, at least in my case which one is the best. So, what you do is you start creating the filters for each and every one of them. I'm using 131,072 taps just to match with REW's measurements. It's not necessary, even 8,000 taps will be more than enough for accuracy. Although with the box filter, which is at low frequency, you might need higher. And for example, if you're going to use a mini DSP for phase correction, the most primitive MiniDSP 2x4 HD. You are limited per channel to 1024, maybe 2048 if you use only two channels. I'm not sure. But if this is your limit, then your only option is to pick rectangular and for centering 99% . And use no box correction because it's not going to be capable of doing box correction. But with this one, you have a chance. It properly replicated. If you see the magnitude response, it's also probably okay. As you see. If you did the normal, for example, Hann with middle centering. Look, what you would get. By the way for MiniDSP this has to be binary. Like that. And everything else the same. See, it can cope with this one, but when you do box correction, And try. See that you will change the SPL graph massively and the phase correction cannot really cope below 70 Hertz, with normal. If you do rectangular with either 99 or a hundred percent, you can try. That will do a better job. What about a hundred percent? The same. So with MiniDSP, these are the tricks you have to improvise. But, if you're using your PC as your convolution engine, you're unlimited. just make it middle, 32bits IEEE mono wave file. Generate it and to desktop. Call this one XO low Q. Generate one of this. As you see with this many taps, it's the perfect replication. And standard Q. Call this one standard Q. Generate. And a high Q. Call this one high Q, generate. That's all. Done with rePhase! Now. Drop all these three crossover phase corrections to REW. They came with a 109dB SPL offset. So, they are up here. This the value changes from day to day. But it's always the same on the same REW session. All 3, as you see are here at 109dB. You don't need to do that for this calculation, but in general, it's a good idea to offset their SPL, -109. Done! You can see now it's at zero here. And for this one also. And also for this one. Now, why did we import all three into REW? Because we want to see which sounds best with my speakers, now Trace Arithmetic.. Our average response times this crossover. We are convolving the filter with the speaker response. And generate. Call this: Llo, generate another one. and call this Lhi And the last one it's standard Q. Generate. L standard. Select the original one and three different filtered responses. Copy selections to Overlays, go to overlays. And check for the best phase response. This is axis zero. I can kind of see that this low one is the best one here, comparing all these. I mean, each and every one of them is much better than this, as you see. Original. But, between these three. This one seems to give the best response. Although, it's not very clear to see from here. So then you go here. To impulse. And make sure. Normalized. It's not ticked. You can also do this for your crossover frequencies. It's easier to see. Now, this is the original response. All three of them improved. Look at the height of the impulse peak. That's a cleaner sound than the original, which is quite rare with digital correction. But, the only difference between them because it's a low bass port correction, okay? With crossover phase corrections, it could be easier to see because there would be height differences. Their Heights are all the same. Not easy to see in this one. So we go to step response. Un- normalized. Fit to data. which one is the least deteriorating? This is the original. It's going to go as low as minus 839 and as high as 790. But look which step response goes as the highest peak. Negative, but still. This one goes all the way down to here at -1,200. The high Q phase correction goes hardly to 900 and standard is a little bit better to 1,000. So, clearly this one is correct. And also, if you just highlight zero axis here, you can see that this step response is better than the others. Okay. It goes to the highest and also comes closest to zero when it is calming down, let's say. So, this is how you determine your perfect crossover and box phase shifts. Now, you can get rid of the standard and high ones. this is your crossover correction, and this is your speaker response after crossover correction. The difference... let's apply 1/48 smoothing to both of them so that the phase graph looks better. The original response, here is zero axis where the phase should be. Okay? Well, you were seeing that for example, there is a drop after 1000 Hertz in both of the responses, no matter what. And there are like weird rises around here. This is because of the microphone calibration file. It doesn't have a phase calibration. Although there is definitely a phase distortion in a calibration microphone. So don't take this too seriously, but the area between 1000 and a hundred Hertz, even from 20 to 1000 is not only audible, but also it has to be in the middle, in the zero axis. With no phase shifts. So, this is the original. Look at that here. And this is the phase corrected one. And to compare some properties of the original measurement and after phase correction is applied... Overlays, group delay. This is the original one. This is the corrected one. And this is zero. You can see that there are. Improvements all the way although minor. RT 60, again, The corrected one has less reverberation. And even clarity... it's better than. And this is C80 for music, better than the original. Almost everywhere. As a last point with this filter, you're not going to be always as lucky as me to find the port frequency of your speaker. Then you will have to go for near-field measurement or measure outdoors. If you can, but near-field measurement does enough good job. Actually near-field measurement is quite complicated. You have to measure each woofer and each port, and then correct them according to their diameters and what not. It's quite complicated. But it gives results as good as the outdoor measurements. But, here you're interested only in your port peak frequency. The way to do this is to put the microphone inside the port and take a measurement just like you see in this figure. Once you measure it like that, you will get this response, okay? this is the measurement I did before. And here you can easily see the port peak is at 41 hertz. You can even read from this graph that the first crossover frequency is 260Hz here. But it's not that easy to see. And it is quite similar to the anechoic chamber measurement that stereophile magazine did for the ports of this speaker. So, you can find it out yourself. And that's it for this filter. Now we are going to move to the second filter. The second filter is a virtual bass array filter. It's a maximum phase filter. not a minimum phase or linear phase filter. It eliminates the peaks and dips caused by room modes. And extends the lowest produced bass frequency by your speakers. To produce this filter, we have to first determine the room, resonant frequency from the room dimensions. And then we create a Dirac, perfect pulse and a low pass filter. then we invert the polarity of this low pass filter, time align it so that it counteracts the resonant signal, the bass wave. Also adjust the strength of it based on the reflective surfaces of our room. I have done an optimized VBA filter video which is linked on the top right corner at the moment. It's a 15 minute video and it explains everything in detail, how to create this filter. So I'm not going to go into details of that. But make sure you watch and use this filter. It's a very interesting filter. You will not find this filter elsewhere and it is a game changer in terms of digital correction. Now, assuming that you have watched the tutorial on the virtual bass array filter and you created your filter, let's open Room EQ wizard. And start from scratch. My original left and right speaker measurements
