 Hey, it's Anfa! And today I want to talk about an equalizer plugin that I'm using all the time. I've been using a few different equalizer plugins on Linux, and as you know me, they gotta be free and open source software as well. And I've settled on LSB Parametric EQ times 16 stereo. And that's what I'm using every day today. Whenever I need to EQ anything, I'm just using LSB EQ. So what is it? I've prepared a few test sounds for us to play with. Alright, let's insert the plugin, LSB Parametric. And as you can see, there is multiple versions of this plugin available. The one that I'm using the most is LSB Parametric Equalizer times 16 stereo. But there's also a mono version. There is a mid-side version, which lets you equalize the mid and the side signal separately. So for example, you can hypers your side signal, which will make your base perfectly mono. And that's an often used trick to ensure the sub-base is powerful and perfectly centered and has no phasing issues between left and right channels. There is also left-right version, which lets you equalize left and right channel independently. There is also a 32 filter version of this plugin. So the one I'm using is 16 filters and there's also 32 filters. And all the same versions are there as well. Let's insert the LSB Equalizer times 16 stereo. And you can see it provides us with a nice inline display of the equalization function. Let's double-click to open the interface. And this is the user interface. I'm going to make it slightly bigger, just not big enough, not as big. So I'm covering part of it. All right. So if I play something, you can see that we have input and output meters on the top, which help us identify what's going on. And first thing I do is enable analysis most of the times. So right here in this tab, we have an option to enable the analysis. And by default, it's disabled and we have two options to enable it. It can be post EQ. That means the graph we see here is already the equalized sound or it can be pre EQ. That means it's the sound before we apply our correction. Usually I use post EQ, but sometimes you may just want to use pre EQ. There's also two settings here, reactivity. Reactivity means basically it's the window time over which the signal is being averaged. So longer reactivity time, this is one second, not one millisecond, I believe. Longer reactivity time gives you a slower moving graph that's more averaged out. So it's easier to see like individual harmonics, otherwise they're kind of shimmery. But well, and shift lets you offset the vertical location of this graph. So if your signal is too quiet and you don't want to boost it with the input gain here or the output gain and you just want to boost it for the analysis, you can use this shift control. By the way, double clicking on the knob resets it to the default position. Maybe I'll talk about knobs in LSP plugins because LSP stands for Linux Studio Plugins and there's a bunch of excellent processors in this series, one of which is DAQ. There is also a graphic EQ, so I'm not using it because I think the parametric EQ is always a better answer to your equalization needs. And the thing about LSP plugins is that they give you a lot of options, usually a lot more than any other open source plugin you would find. For example, a commonly known suite of plugins is the KALF Studio Gear and I used to use them a lot. So I've learned that the DSP code, the digital signal processing is of questionable quality at times, especially in crossover plugins like the multiband compressor and the saturator, for example, also has a problem if you use the dry-wet control. It's creating a notch filter, which it shouldn't. So there's some sound quality issues with KALF plugins. That's why I've decided to try and use them as little as possible and see what else can I find to replace them with. And I found the LSP plugins to be a great replacement. And actually, the user interfaces are maybe less appealing in LSP plugins than in KALF plugins, but the features are better, usually. And I'm very confident in the sound quality of these plugins. OK, so how do you lose the knobs? If you click and drag up and down, you're going to change the value. I believe if you hold something like Control, Shift. Now, if you click with your right mouse button, you can fine-tune your values. Now, if you click on the color of the ring around the knob, you can set the value immediately. So if I click here, you see that the value jumped. And this is both useful and dangerous, because if you misclick and you click on this dial somewhere, instead of in the center of the knob, you can set your gain to plus 5 decibels or plus 20, when what you really wanted to do is just click here and make it a bit lower. And double-clicking resets the knobs to zero. All right, so we've talked about the analysis, we've talked about the input-output levels. Another thing is balance. Basically, for some reason, this equalizer lets you change the panning balance of your signal. I don't know why, really. And there's also pitch. This is very interesting, and we'll get back to that once we actually set up some filters, because right now, this equalizer is not doing anything to our signal. So let's do something to our signal. Below the graph window, we have the filter section. And here, you can see we have eight filters numbered from zero to seven, as programmers often do. And we have two sets of these, so we can select filters zero to seven or filters eight to 15. And you can see that you have different colors, so it's easier to see what you're actually doing right now. All right, let's enable a high-pass filter, which is something I do probably most often, is to control the lows. So let's click on this filter type. It's off. Select high-pass. And you see that here is our filter. I can click and drag to change the frequency. If this was a peak filter, I can also click and drag vertically to change the gain. Also, the mouse wheel lets me change the cue. And we can also change all of these parameters with the knobs below. So the left top knob in the set is the frequency. Underneath that, we have the cue. And to the right, we have gain, which is irrelevant for a high-pass filter. Let's reset it. And hue, which is the color. I'm not going to touch the hue because the default color coding of the filters is perfectly fine. All right. So this is a very basic high-pass filter. What if we want to make it steeper? This is a very delicate filter curve. There's not a lot of attenuation per octave happening. So we can change the slope from times 1 to times 2, times 3, or times 4. And this is still not as steep as we can go, as you'll see in a moment. It is much steeper. Now let's talk about the filter modes. If I click on this dropdown here, you can see there's a bunch of different modes. Now every filter in LSP equalizer has multiple different implementations that the equalizer can use, and you can select from. There is RLC, which is, I believe, resistor loop condenser. I don't know what loop. It's a coil. Okay, resistor coil condenser, I don't know. And there's also two types, BT and MT. Now what I've learned is that the BT and MT versions differ with how they respond in the high frequencies. I'm going to touch upon that after in a while. But we have RLC, which are the default filters. There is BWC, which is Butterworth to Chebyshev, I believe. If we select that, you can see this filter is much steeper at times 4 slope. At times 1, actually we can see and compare it to RLC. Yeah, it is a bit steeper. You can see that the knee is a little bit smaller on that filter. And they do sound different. Now if we turn it up to times 4, we can pretty much remove a single, just the fundamental from our note and everything else stays in place. But this is still not as sharp as we can go. So also I believe that the BWC filter, it starts like this, what is it? If we go up with Q, it can become much more steep, but it also has a lot of rippling. It's probably not a very useful filter for clean processing, but if you want to make some interesting moving equalizer curves, then I think it's great because it introduces a lot of peaks. All right, so this is BWC, Butterworth and Chebyshev. I think it's like crossfading between Butterworth at Q0 and the Q above 0 is the Chebyshev filter being morphed into this. I don't know. The Q or the steepness goes up really quickly with the rippling being pretty minimal, so you can do some very sharp filtering with this. This sounds almost like a brick wall filter. Okay, let's go further. So BWC and there is LRX. I believe this is Linquid's Relay, Riley, I don't remember the names, and this is pretty much as sharp as you can go. That's pretty amazing. And this is at Q0. If we go up, you see we have these familiar ripples. So we can kind of like select the trader between the flatness of the frequency response and the sharpness of the filter roll off. Let's see how these two differ. Yeah, that's so steep, my goodness. And there's also APO. I have no idea what that stands for, but this filter doesn't respond to the slope option and it's very smooth. If I reset the Q, you can see that this high-pass filter is like, it has an amazingly slow roll off. Like, I've never seen a filter with such a smooth soft roll off, such a gentle roll off. I think the APO filter is probably going to be great if you want to work with acoustic instruments and you want to have the extremely natural high-pass that, like, you can't tell it's there, never. Okay, let's try some different types of filters. We have high-pass, let's see low-pass, and APO works the same way. It is a resonant filter, though. You know what? Let's talk about this pitch right now. I'm going to switch this back to high-pass and we're going to do a funky thing. Let's enable the second filter, or the first one. I wish the numbering would start from one because people who are not programmers get very confused by this, numbering from zero. And it's not very natural for, like, not programmers to count from zero. And also it's difficult for me to talk about in videos, like, the zeroth filter, what, zeroth? Okay, I don't know how to talk about it. Let's change this to low-pass, and now we have two filters. Let's give this some Q, or not as much Q, please. Okay, oh, I clicked. Yeah, you see, I find the knobs a little bit difficult to work with sometimes because I can be a bit sloppy and sloppy and I misclick and I click on the ring when I wanted to click on the center part. This is a clever design, I think, but it requires precision and it's not always the case with me. All right, let's play. So that's two filters. And now with this pitch, look what you can do. We can offset all of the filter frequencies by a given amount of semitones. So basically, like, we can automate, because this can be automated, we can automate moving, shifting the whole filter curve up and down. Now I've tried doing, automating this with some really fast motion and there is some zipper noise, so it's not perfectly transparent. But for slower movements, and if you're not, like, going with super low frequencies, it's automatable, so we can use this for sound design, like, you can put a bunch of peaks and froths and automate this to create, like, vowel movement or something. All right, let's reset the pitch. So we have a high pass, we have a low pass. The filters I use often are also bell filters, of course. So, you know what? Maybe let's switch to a different, let's switch to a different sound, because I'm using the same sort of patch I've been using. So let's go to the pad now and insert this parametric EQ. So something I use a lot is a bell filter, which allows me to find problematic areas and also find interesting areas in my sound. So we could make it a bit more narrow, up the gain and sweep. Yeah, I'd like a bit more high, so I'm gonna use my mouse wheel to reduce the Q factor. Now we can mute this filter so that it's no longer making effect. This way we can preview what's going on. Now for highs, often a better idea is to use a high shelving filter. So I'm gonna mute this bell filter and I'm gonna use a high shelf. Now when I'm moving it around, there is some zipper noise, you can hear some clicks. So this is not like, it's not automatable. Maybe this could be corrected so that the filters don't produce any clicks when you change the cutoff frequency or the order parameters. Yeah, so high shelves. We can also use different implementations of the high shelves. The BWC version is a bit steeper. The LRX version is even steeper. I wonder what happens if we, whoa, if we up the Q. Okay, that filter gets pretty nasty when we up the Qs because it also attenuates what's below the cutoff frequency. I mean that's interesting. It's not something I would usually want. And the APO filter, whoa, at zero Q it's a very interesting curve. Okay, and Q0.5 is like the sigmoid function. It looks familiar, probably closest to RLC, can go a bit steeper. And at Q0.25 it's like a linear slope. So I think this is like the softest, smoothest filter you can use like have the highs be amplified very, very smoothly and naturally because like there's not a single point where it just jumps up and you know, like with the LRX filter, like here, there's a single point when just plop. So the RLX filter works best at Q0 in my opinion because otherwise it's going to, it's getting unwieldy, it's getting weird. I think a nice thing would be to talk about the EQ mode, by the way. There is one more setting that we haven't touched upon which is pretty interesting. Now there's three settings we can choose here. So by default all the filters are working as infinite impulse response filters, IIR. And this basically means minimal phase distortion if like this is the most common type of filters. What that does is it has zero latency, but it introduces minimal phase distortion. That means the waveform is going to change. So these aren't like the IIR filters are not suitable for using them in dry wet scenarios. Like you should not mix this with the original signal or because of the phase shift, there's going to be a bunch of weird things happening, like you're going to get notches and different stuff going on. But there's also finite impulse response and fast Fourier transform versions. Finite in both of these, as you can see right here, we've got latency introduced. So let me switch back to IIR. So infinite impulse response, zero latency. Finite input response, 85 milliseconds of latency. So we know what's the downside, it introduces latency. What's the upside? The FIR version has linear phase. That means the frequency content is changed, but the phase relationships between harmonics remain the same. And this is often used in mastering, where you want to introduce as little change as possible. However, for this sound, like this pad, let's maybe do a bunch more stuff, like enable this thing here, maybe a high pass, let's make it steep, because I like steep high passes. And now if I change the mode, it sounds exactly the same, like except for the introduced latency of 85.33 milliseconds, there is no audible difference between the two. But this is a sound without short transients, or fast transients. Now I'm using the FFT mode, which is fast Fourier transform. And I don't really know what's the difference in like, it's also linear phase. That's what I read, but I'm not sure how it's different from the FIR version. I don't know, maybe it works by instead of like synthesize, like filtering the original silence, like re-synthesizing the signal from individual harmonics. I don't know if you guys know, let me know, like to learn, what is the difference between FIR and FFT modes in this equalizer? Because I guess there is some scenario in which you would prefer to use the FFT over FIR or the other way around, but I don't know of such a thing. So yeah, now I'm going to show you a percussive sound. And percussive sounds are special as they like expose the problems with, they expose the problems that happen when you use linear phase equalizers. Let me enable the POSTQ. So let me use a high pass on that. Now let's change it to the FIR mode, which is linear phase. Okay, I can't actually hear a difference except for the latency. But what about a bell? All right, let's do a bell, maybe an RLC bell. Let's lower the output high shelf. Huh, interesting. I can't really hear the difference, but the problem with linear phase equalization is that it can smear transients. That means often the punch of a kick sound or a snare sound or any drum sound comes from the fact that there is a fast impulse and we could actually measure or visualize that by using an oscilloscope. Let's go with a mono version. We don't really need to see anything more than one thing, thingy. Okay, let's sacrifice the keys or maybe I'm just going to move them here. Let me make this little oscilloscope show us the transient every time. Okay, so we are in the infinite impulse response. So minimum phase, let's switch to linear phase and see. That's a bit different, so it could be just, well actually it is quite different. Look at this. This is minimum phase EQ. This is linear phase EQ now. You can see how different the layout of the peaks is, like that's because in this mode we are retaining the relationships between different harmonics and if I bypass the EQ, of course it's a bit too loud, you can see that the original sound is resembling a little bit of a square wave. So if I use the infinite impulse response or minimum phase EQ, you see that this square shape is being lost because the filters are naturally introducing phase shift and you would think, oh no, that means they are destroying my sound, actually no because our ears are very insensitive to phase relationships and most of the time the minimum phase EQ, that means IIR, sounds perfectly natural and the only problem is phasing when you mix the original signal with the process signal and then the phase changes actually become a problem because they cancel each other out in different frequency regions differently and you have notches and you have a comp filter then, which is usually not what you'd want. Now the infinite impulse response or linear phase EQ retains the original shape of the waveform but that comes at a cost and the cost is called pre-ringing. If we zoom in even further, maybe move this a bit, let me bypass the EQ, you can see that in here our waveform starts pretty much dead on, it just starts. If I enable the EQ, you can see that our waveform starts bending a little bit backwards before it goes up and this is the pre-ringing effect. It's not very, like, it's not very bad in here, like this won't be, isn't an issue but if you have very tight drum transients, this starts to become an issue because in order for the filter to keep phases aligned, it has to pull some of them back and they start resonating before the transient hits so you get something like instead of dush, you get lush, lush, dush, lush. The transient is lost to some degree because of this pre-ringing. And this is also why the EQ has latency. The infinite impulse response means, like, if I'm correct, I'm not a DSP expert by any means or even an electronic man, that means that the function which is, like, convoluted under the signal is infinite in length. So there's no point in delaying the signal to recover anything because you'd have to delay it by forever. Finite impulse response, I believe, means that the convolution function used is of fixed length and for the lowest frequencies that length is 85.0 milliseconds. And because the length is known, or maybe it's half of that, I don't know, we can process all the frequencies separately and then align them together. But to align them, we need to delay some more and some less and to be able to make sure that they are all aligned, that they are all delayed by the same amount by the end, we need to delay the whole signal by the maximum delay we need to apply to be able to align them. I think I'm getting into deep into that. All right, never mind. Let's see what FFT will do. It looks pretty much identical to FIR. All righty, is there anything left to say? I think there isn't. We could maybe try. Yeah, there's one more sound I'd like to show you. And we can also, like, talk about the difference between this BT and empty filter type, because there's two of them and they are different. So let's see what we can do with that. Are you guys there? Can you see me? Let me know if the stream is working for you guys. All right, here's a little FM patch I made in Surge. Now let's maybe use a, oh, there's an all pass filter as well. Damn, haven't tried this one. I think I should. We can save it for the last. All right, so I want to use a high pass. Maybe make it very, very, very gentle. I should be able to see what I'm pressing on my keyboard. I can see. I thought that was working. Sorry about that. Now let's use a low pass filter and let's compare the BT and empty versions. So low pass of BT and empty. You can see that it has, like, BT is a bit, like, steeper. The empty filter lets more high frequencies through. It's like the, the roll-off is, like, limited to a certain slope. Let's play a higher note and see what happens. Maybe I'm going to test it like that. You see the difference is most pronounced on the highest frequencies. So it's like the, the BT filters kind of, the BT filter, like, snaps to make sure we're cutting off everything at the 20 kilohertz mark. While the empty filter just doesn't care and it just keeps the same curve. The same slope. Yeah, so you can see that the BT filter snaps and makes sure that we hit maximum attenuation at the highest frequency. Let's see how that works in high pass. Let's also do the same thing. No, I don't see any difference in the high pass. Check some other low pass. I'm testing BWC now. BT and empty. All right. Yep, BT snaps, curves down to attenuate at 20 kilohertz. And empty doesn't. It's the LRX. Yeah, BT is snapping. Empty isn't. Yeah, so pretty much that's the difference between the empty and the BT implementation. Now, there is an extensive, there's extensive user manual on the LSP website. And if you have any questions about this equalizer, feel free to consult it. I'm going to link it in the video description. There's also a video by LSP themselves, demoing this and talking a little bit about the different implementations and types of filters. So the video was recorded for an older version, so the interface is a little bit different. But I'm sure we're going to figure it out if you watch that. All right, that's all for today. Thank you for watching. Thanks to everyone who joined me live. Thank you for like going through all the trouble we've been through with the problematic YouTube streaming and yeah, internet not working. So yeah, thanks for watching. I hope LSP Parametric EQ is going to be a great tool in your toolbox. I'm using it every day. Like it's a real workhorse in my toolbox. And I think it's a great, a great plugin. Yeah, also huge thanks to everyone who is supporting me on Patreon and Liberapay. Recently, there's been an influx of new people there and I'm really happy to see that. You guys are letting me do this, even though life is sometimes doing its best to get in the way and make it hard. So I really appreciate that. And it also gives me hope that one day I might be doing this full-time and there's so much stuff to talk about, so many awesome open source plugins, programs that I haven't even had a chance to even had a chance to try out and experiment and learn in the in-depth to make a video about them. Yeah, there's a lot I'd love to do, but time is limited. So I'm going to do the most important stuff first. Yeah. Oh, if you're the reviewer would like to help me out and support this I'm going project of mine of educating about free and open source audio production, please go to patreon.com slash ANFA or to liberapay.com slash ANFA where you can support me with a monthly donation. Thanks for watching and I'll see you in the next video. Bye.