Wow, thanks for taking such a deep dive under the hood! Allow me to explain some conceptual decisions behind Nova. From the beginning on, Nova was designed as a dynamic EQ. Most of the structural decisions went into optimizing the dynamic EQ aspect. This is where the largest differences appear. With a dynamic EQ, the filter topology has great effects on the final curve. Instead of setting Gain to say, -6dB, adjust Threshold to have GR reach -6dB. Then switch the precise/musical mode, they offer two distinct dynamic EQ curves and interesting sound flavours (whereby static EQing is largely equal over both modes). With dynamics, you have a non-linearity in each band. In series this distortion would multiply. In parallel, it only adds up, never spreads into the other band. Similar applies to oversampling vs non oversampling. As a dedicated dynamic EQ, Nova of course puts great emphasis onto anti aliasing. Something a mainly static EQ concept doesn't have to care much about. Fabfilter's static EQing is technically more "to the point" than Nova's static EQing, because dynamic EQing has never been its main conceptual priority. Reducing complexity/CPU load. On the other hand, Nova's dynamic EQing is likely cleaner than Fabfilter's dynamic option (especially with regard to antialiasing, say, in a deesser configuration). This aside, Pro-Q's way of defining Q is quite smart imho. The world needs all of them ;)
Great products, and great insight. It's good to see a developer jumping in a discussion about one of their products and offer real info and not drama ;)
The end "I have absolutely no idea" made the whole video worth it. To have something taught to you so thoroughly and beautifully detailed for NO REASON at all is what makes Dan a Master Jedi level UA-camr. Hilarious. Thank you for this video.
Spoiler alert for some, but I was about to stop watching this video about 20% of the way in. This comment got me to watch the vid to the end to hear the quote and the lead up - and earned a channel sub!
I think this video in particular exemplifies why I love music production so much, you can do all the science and engineering you want to try and understand what's going on, but at the end of the day it's your emotions and spirit that determine what really matters when it comes to making decisions. ....a bit like real life really!
There's very few audio You Tubers out there who can't seem to resist showing you how well informed and clever they are. Dan bucks that trend in spades: pragmatic, thorough, rational and humble. Long may he continue to be the scourge of the irresponsible urban myth-making that passes for education on the internet
The definitive word..."What was the point of this video? I have absolutely no idea." Another enjoyable and informative episode, thank you, Dan. I really enjoy using both eqs as well-- in many ways due to your other excellent tutorials.
Back in the early 2000s, my job had me meeting with the head of DSP (name lost to time) for a leading pro audio company. In our meeting, he was excited to tell me that they had created a parallel topology EQ in software. At the time, standard DSP practice was to run the digitized signal through one EQ band, then through the others in series (serial topology). He said the parallel EQ curves behaved and sounded much more like an analog EQ, which typically has multiple bands built as filters in parallel. What he told me was that with parallel EQs, the band gain interactions summed, and with serial EQs, the band gain interactions were the product. My days of filter theory and supporting math are too far back in the rear view mirror to prove it, but instinctively it makes sense. Combine this with any unwanted distortions multiplying for series processes (TDR's comment), and it would be interesting to analyze what happens with multiple plugins in a chain.
I've always felt like the TDR Nova has more of a glue-y effect than other EQs, and I apparently correctly identified which one it was even just listening on my phone's speakers, but I'm also cognizant of the fact that most EQs sound good and that this was probably a lucky guess combined with a fine slurry of pride and confirmation bias that tricked my stupid meat brain. Anyway, this was really interesting. Thanks for the video.
I've been using the Nova GE for a while now. Was contemplating getting Pro Q3 but for my money, the Nova product works plenty well enough. And this video helps to reinforce my purchase decision...at least for now. Thanks Dan!
"Nova's wideband covers everything that isn't covered by another band." So, in other words, if I want to compress everything but a bell centered on 700hz I just make a static bell on one band and a dynamic wide band? That's pretty neat.
Something about nova and Slick EQ for that matter, sound better, get me the desired results faster. They're also crazy good value for the price and the free versions Rival any eq out there. Well done TDR.
Sound quality aside, a lot of time I'm going for Nova instead of other EQ plugin is almost always because of the analyzer. Can't really explain it, but Nova gives "better" representation of what is going on for me doing subtractive EQ work compared to, let's say Pro-Q. Im guilty of not knowing it actually have different topology (parallel) compared to Pro-Q, I know it sounds different than the Pro-Q especially in the dynamic mode, my guess was the more controls it offers (att, rel, ratio, etc), apparently it was so much more. Thanks for the video, I actually learn something new today. :)
Great video! Some comments: Usually, digital filters in DAW EQs are implemented using IIRs in the form of biquads in the Transposed Direct Form II (TDF-II) (/ cascades of them) or FIR filters (way less common). When using that topology, it doesn't really make sense to combine multiple parallel EQs, as you'd unnecessarily waste computing resources, seeing as you parametrize the filter coefficients to produce the exact filter type you want. Another approach is to use a filter topology that has multiple outputs, such as an SVF (there's other topologies you can use, usually these filters are linear discretized versions of real-life filters), where you have seperate Lowpass, Highpass and Bandpass outputs that all get calculated in one "pass". WIth these filters, you usually get the different filter types (bell, shelves, etc) by combining the outputs in various ways. If parametrized correctly, the output of such a parallel combination of these SVF taps and the output of a TDF-II are exactly equal (minus numerical errors) _for the time invariant case_. Once you start modulating the filter, so for example change the Q or the cutoff in real time, you can't prove that the TDF-II Biquad will remain stable, while for the SVF it is proven the filter is stable for all configurations and any arbitrary modulation signal (audio rate, nyquist, doesn't matter). Check out Aaron Wishnick's 2014 paper "Time-Varying Filters for Musical Applications", it goes into some detail there if you're interested.
Maybe this is obvious but while doing the null tests with the bell curves, no matter what frequency they were at (150hz vs 500hz for example), there was indeed a bit of noise being let through *around* that frequency, but the exact frequency chosen would always null perfectly, as if there was a notch filter right at that frequency. It's probably due to the EQs only varying slightly at the slope portion of the bell curve, while the very top (or bottom for a cut) of the curve stays generally the same.
Well observed. The behaviour at the centre of the bell is well defined, assuming no modelling of "analogue" inaccuracies: its the specified amount of boost, at the specified frequency, with zero phase shift. The differences (including phase shift) are all in the curves either side.
SSL X - EQ 2 manual…. Parallel Passive EQ Parallel EQ exhibits quite different sonic properties to the familiar serial parametric EQ. We are generally used to hearing the effect of one EQ band superimposed on another, as opposed to the band interaction inherent to a parallel EQ. Because the bands are placed in a parallel configuration, phase cancellations and re-enforcements happen that are not always obvious when first encountered. Passive EQ is something that is found in old equaliser units and is generally known for its transparent and natural sound, but has some problems associated with it. However, in the digital domain these shortcomings do not have such an influence. A passive EQ does not have any gain elements, but can still have controls to seemingly boost frequencies as well as cut. What actually happens is that the entire signal is cut by an amount, but the frequencies that are apparently ‘boosted’ are simply not cut as much. Therefore the unit must attenuate either the input, the output, or both to allow enough headroom. Unfortunately, in the analogue domain, a 20dB reduction in signal level produces a 20dB increase in the noise floor. Luckily in the digital domain, with a 64-bit floating point DSP, these issues do not remain. In X-EQ 2 - when the ‘parallel’ button is engaged - you are presented with a parallel passive EQ model that the original designers of these devices could only have dreamed of. The noise floor can be disregarded due to the huge resolution available in SSL Native plug-ins. You may find yourself entering this mode more and more as you become familiar with the sonic signature. Larger gain changes are possible without colouration, and boost starts to become something that is useable to a significant degree in a digital EQ! Parallel EQ does exhibit asymmetry in its boost and cut characteristics. But this is not such a bad thing as most engineers would agree that boost is best done with low (wide) Q values and cut with a higher (narrower) Q.
TDR make some of the best sounding bang for the buck plugins you can get IMO. I use Slick EQ, Molotok, Kotelnakov, and Nova as my go to choices on my travel laptop along side Valhalla Delay and Classic Reverb. I would be perfectly happy if I could only use those plugins and nothing else. The Free TDR plugins are fantastic, but the GE versions make them total powerhouses. Great plugins.
In your A/B test, I can hear (A) preserve transient better than (B). Maybe I'm wrong, but does anyone else sound like me? As always, thank you for such a great and deep video, Dan!
My hypothesis is that if you could set the Q in Nova more precisly to 1/root(2) = 0.7071... rather than rounding to 0.71 those small differences would go away
Almost certainly the underlying value is actually precisely sqrt(0.5), and simply rounded to two decimals for the GUI. The internal values used in the processing and the numbers displayed in the interface are almost always decoupled from one another.
@@Friedeggonheadchan That still requires it to be set to this precice value. Nova starts up with other Q values, so if you want 0.7071 you would need to type it in manualy. But we can't see Dan doing that in this video.
I did a fast null test, and it makes absolutly no difference between typing in 0.71 and 0.7071. So either Nova rounds the value you type in before setting the underlying parameter or the parameter itself is stepped to 0.01
In the A / B there was definitely an 'air' difference waaaay up towards the limit of my ears 17k-ish, A sounded more open but I'm on adam ribbon tweeters in a controlled environment so I doubt its a big deal unless there's lots in a project.
I have a set of old battle scarred ears and in that first A B Listening Test I got 'Brighter' HiHats from A than B. The differences in those double bell curves is intriguing ..
I never listen to WHAT you say. Just listen to your voice for sonically enjoyment :) Seriously: GREAT CONTENT, every time ! Thanks !!! Pleeeaaaase take a look at the ToneBooster EQ and whats about the (growing) hype about the Kirchhoff-EQ
Another amazing video. I've started taking notes as I realized that if you took the time to make a graphic to show us, it's probably a pretty dang important concept to understand. I drew out the pictures and wrote notes and noted a few places for further exploration. I downloaded plugin doctor and the other thing required to run multiple plugins through it from the phase video just to mess around and have some science. I've switched from Nova to Pro-MB for dynamic EQ on my latest album and I'm gonna go do some experiments to see how Pro-MB stacks up against Nova and especially Pro-Q3's dynamic mode. My guess is that Pro-MB would be parallel just due to the visual representation of the graph and the fact that Q3 already is serial. But I guess we'll see. Again, great video. I get super excited every time you post and have shown these videos to people who have no idea what's going on, just to show that music engineering is an incredibly complex and sciency thing. And maybe for some bragging rights... Have a good day y'all. Hope everyone's doing well.
The surgical and musical curves could be described a Gibbs fenomena, and sorry for the interruption but how often someone came across then the plugin the use the standard window to grass the results
I hear a difference; even using headphones. Listen to the top click on the kick drum, claps, and stereo width/image. On "A" you can here that tiny top end click sound on the kick drum more than on "B". Also, the information on the sides also seems brighter. This gives the illusion of a widened stereo image. It's just oversampling can cause hi end information to seem brighter. Therefore, if hi end information is on the sides, the sides will appear slightly more pronounced. Also there is a slight quality difference. The clap 👏 tails are more detailed on "A" than "B". Honestly, after listening again, I hear "A" offering and entirely more detailed picture than "B" and the 808 boom is more controlled on "A" than "B". The differences are extremely tiny but noticable. The real question is, which one do you think fits the job? That will forever be subjective and the beauty of music and music creation! Tip: You have to be very careful with the application/deployment of oversampling and how much. If you use too high of an oversampling rate your music or tracks will start to lose their natural emotion. This becomes most evident in the impact of low end the higher you go up in oversampling.
I've been using FF Q2/Q3 for ages and it's my go to EQ for cleaning up, analyzing etc. But for boosting I prefer Pultec like EQs (Wesaudio Prometheus or as a plugin e.g. SPL Passeq) and love my 2 Clariphonic 500 especially on anything that need to cut thru the mix. For me it sound more musical/ less harsh. Interestingly or rather fittingly I preferred A and will give Nova another try (own the GE but forgot about it 🙄😂). So thanks it was (again) a joy to watch and helpful! 🙂
@@MusicChannel-rf5zz no, different. Since my comment I've been using nova ge a lot, especially for cleaning up too long decays in the bass region, or sidechaining /ducking bass to BD. It's excellent for that
Nova more like an RMS compressor. Thats the part a lot ppl here on YT making videos about Nova they dont get. They think its a Peak detector but its totally different. Its great that Dan pointed that out since it seems like ppl tend to listen to him more than anybody else 😄
Well, today I learned something, thanks Dan. The interaction between two bands varies in more ways than just the "different types of parallel eq" btw. But now I have part of the explaination. PS: I can't be the only one watching these videos on 1,5x speed, right?
Oddly enough, I heard a difference. Maybe I imagined it, but the high end on Nova sounded sharper/crispier to me, and I preferred option b for sounding smoother/silkier
Same observation here. But, if Nova would have musical curves option selected, it would be the other way around. When you need it silky with a tiny bit of haze - musical, when you need it crystal clear - surgical.
If you want to look for the Point!! This video helps a lot that's the point...this is what more people need to do...make it easier to understand and make it easier to try to understand something as big as this...sharing knowledge makes us all smarter🧠
Thanks for showing this plugin! ~ In your sound test, I listened with Bose Sound Sport Free earbuds and I could hear subtle differences in the mid to low end range when A/B'd. I thought "A" (Nova) sounded extremely clean. I assume that may be mainly due to the differences in frequencoes added compared with Pro-Q. Thanks again for an interesting comparison, Dan. ~ Cheers
Hi Dan, Thank you for making a great point that I've been making for decades. 61 years of recording and no I can't hear a difference. And if I could hear a difference, what difference does it make ? Thanks for taking the time to point this out. Bill P.
Nova had slightly sharper transients in my ears. But it didn't amount to much honestly. Loved the conclusion though, maybe one day you'll do a follow up :)
nice video thanks dan. in the AB comparison at 17:38 the difference i hear is clearly in width and A seems definitely somewhat wider than B curious about your opinion on this?
I don't perceive any difference in width, nor can I think of any reason why there would be one: both plugins process left and right channels exactly the same. Also stereo differences aren't that significant in the highest frequencies: width is really all in the midrange. If you have both plugins and you want a definitive answer, try an ABX test.
I can't hear a difference between A and B. Your assumption is correct with me at leat. :) As for how Pro Q adds two peaks, it seems that FabFilter made a conscious choice of flattening the sum to allow for the case where you want to boost one frequency region. I believe that KiloHearts EQ does that also. It certainly makes sense to me but what is strange is that they didn't communicate on this feature.
may I ask where the 0.1 comes from? I've not coded in a good while but I thought 6db was literally the result from an x2 in the math, I even remember reading a paper arguing that technically a 6db boost would be cleaner than any non integers as it wouldn't result in as many rounding errors of the 'digital voltage' (not accounting for interpolation of course) so that .01 is really throwing me haha.
Anyone knows whats happened with the no longer available (in the last 2.1.5 version of Nova GE) processing modes, "precise -" and "insane - " ? The Nova (2.1.5) "precise" and "insane" is the new "-" ? Clean? With the "+" (nonlinear, some subtle saturation) option?
Late comment and love both of these plugins and own them both. Your video's are always so informative it's awesome! Would you then know if Pro-Q3 is different between the AU/VST3/AAX compared to the iOS AUv3? My assumptions would be there is none but thought I'd ask 🤘💀🍻🔥
Always interesting Mr Worrall...ive learnt a lot form you in the last couple of years....what would actually be really interesting is a video from you talking about why (and why not) where and what Eq you would use in certain situations. There are so many Eqs on the market and all are marketed with musical/surgical hype terms...when i use plug in doctor on the Eqs I have from expensive mastering ones to bog standard they all seem to exhibit similar curves on the screen ..Waves F6,ProQ,MAAT Blue Eq, to Plugin alliance AMEK 200 and their Soma and Digital V3 all of which I have seen on high quality mastering chains.. .what gives Eqs character, what is that character and why would you chose that and when. Please consider this ..I would love to know your opinion.
In theory, direct EQ and parallel filters are equivalent. Let's say the direct EQ has transfer function T(f). We want the parallel filter blended with the unprocessed signal to have the same response, i.e. T(f) = 1 + P(f), where P(f) is the frequency response of the parallel filter. There always exists a solution to this equation, and thus for any direct EQ, there exists an equivalent parallel filter. In practice, most EQ plugins probably don't have enough degrees of freedom to match any direct filter to a parallel one (this part is just my guess)
Aww yeaa Edit: Have you done or will you do any videos on DMG plugins? They're so dense and full of teaching opportunities, I think you would be great at articulating all the different possibilities. Anyways, great vid! :)
For those who have worked with physical rc filters, which software EQ are models of - regardless of how their approach varies, the parallel approach is much more intuitive to use. I don't expect one band to preclude the other's existence, that seems beyond magical to me. The plotted curve of one bell should not be an additive activity, unless there were a simulation of am op-amp before the output, but account for its own nature only. Perception is one thing and there may be an imparted addition in a graphic EQ between two bells which are within each others' Q but not in a hardware parametric EQ, however, perception isn't what we are aiming at. Sure, going back to Q function the NOVA Q can be somewhat widened and their centers brought closer in order to match Pro-Q's but, from that *perception*, which is correct? We already know that Pro-Q took a certain liberty with its rendering of 0.71 as 1 (when that is NOT what is really happening) - why should we assume that it also didn't take a liberty with its output of additive Q function between filters? Maybe they assumed that if a user wants the normal hardware parametric function of real rc filters they are free to add a 3rd bell between them and set it to cut (maybe you should investigate that). Personally, I don't like it when anyone begins to round off reality in such ways. I prefer the TDR method, and that might be just the nature of the parallel implementation but I don't see it. Unless the series Pro-Q approach really does mimic a graphic EQ relative to how the parallel NOVA approach intentionally mimics a hardware parametric, the folks at FabFilter just got caught taking artistic liberties with the nature of Q function. Is that the end of any world? Nah. But, it could prove to be annoying if one were not aware of it, coming from an analog environment (which many of us have). Great video, Dan.
It's the nature of the parallel implementation. In a serial design the input to each filter is the output of the previous filter. In a parallel design the input to each filter is the original input signal.
@@DanWorrall Sure, now explain the high frequency anomaly if Pro-Q isn't really just cascading a sequence of set graphic bands with a width parameter that they forgot to extend beyond the graphically displayed range. Any node between bands can be accounted for when there IS a node to even be counted. Below 10hz there's not much to read but above 20khz is another beast.
@@DanWorrall It's not important. I was suggesting Pro-Q has fixed centers and a mechanism for moving between any two or more of them to fake a parametric Q function which, when placed in a series, is fine until they run out - then you're left with a fixed singular ''last bell'' that has no other bell to partner with (which could explain the non-nulling you found at the highest frequencies in the tests).
Thank you for this! I love your videos because of your nerdy technical approach. But one question I have is what's the difference between this two EQs phasewise. I mean there have to be some but I don't have plugin doctor or ProQ3 either... Could you make a video on Kirchhoff EQ?
Ok, so both useful, both ok. I'm glad I spent all that time confirming that one ;-) Lucky I was doing something useful (animating cogs) at the time :-)
20:15 seems exactly like the difference between the Motown EQ emulations by UAD and Kit Plugins, UAD being the parallel approach and Kit being the serial. Might be different iterations of Motown EQ modelled, or Kit just did a very bad job.
Hi,Daniel! What you thing about - "Drawing eq curves" in a flip polarity of 2 fx tracks? It make sense if i draw at very low level of db to rich more clear sound and less artifacts of eq?
At first,sorry for my English! it;s many ways to approach eq! Let;s say - I have 2 indentical tracks with the same audio material! I flip the polarity to one of them and i insert a eq plugin like Voxengo Curve Eq! Next,I set the scale view to 3db! Ten i draw a eq curve to my needs in subtractive way! Finally, i summed this 2 tracks to a group to compesate the volume! In terms of quality,it's better way to eq something? because if nedd drastic mooves i can achieve in limit of 3db! Theoriticly,should be better to preserve quality!
@@cezarciucium5580 ok I think I understand. And no, I don't think you're gaining any extra quality. What quality problems are you trying to avoid anyway? People are much too worried about 'phase smear' and such: don't be. Get the frequency response right, that's unusually all that matters.
I try to avoid phase as much as possible when it comes to hp filtering and get rid of narrow resonances! Both reduce the transient response! You sad in one video that the phase is cause by how much db and q factor are increased! Also everybody talk about that subtactive sounds better then boosting eq! Thank you very much!
Why there's no spectrum analyzer plugins that utilizes Gammatone filter banks or sliding DFT constant-Q transform? I mean the IIR Gammatone filter bank is fast as CQ-sDFT with ERB frequency resolution
Could it be that, for a developer, choosing between either topology would come down to what feature set they are looking to offer? I was thinking about the example you gave with the dynamics processing features in both plugins since you mentioned that, in theory, it's easier to do with parallel bandpass filters. Or maybe I misunderstood something
Wow, thanks for taking such a deep dive under the hood! Allow me to explain some conceptual decisions behind Nova.
From the beginning on, Nova was designed as a dynamic EQ. Most of the structural decisions went into optimizing the dynamic EQ aspect.
This is where the largest differences appear. With a dynamic EQ, the filter topology has great effects on the final curve. Instead of setting Gain to say, -6dB, adjust Threshold to have GR reach -6dB. Then switch the precise/musical mode, they offer two distinct dynamic EQ curves and interesting sound flavours (whereby static EQing is largely equal over both modes).
With dynamics, you have a non-linearity in each band. In series this distortion would multiply. In parallel, it only adds up, never spreads into the other band. Similar applies to oversampling vs non oversampling. As a dedicated dynamic EQ, Nova of course puts great emphasis onto anti aliasing. Something a mainly static EQ concept doesn't have to care much about.
Fabfilter's static EQing is technically more "to the point" than Nova's static EQing, because dynamic EQing has never been its main conceptual priority. Reducing complexity/CPU load.
On the other hand, Nova's dynamic EQing is likely cleaner than Fabfilter's dynamic option (especially with regard to antialiasing, say, in a deesser configuration).
This aside, Pro-Q's way of defining Q is quite smart imho.
The world needs all of them ;)
You made a really great tool! I don't know how exactly but every time i use it, i feel like, its well balanced, clear and clean, like smooth magic.
Great products, and great insight. It's good to see a developer jumping in a discussion about one of their products and offer real info and not drama ;)
I love Nova and really appreciate you responding here
You are doing great products. Thank you for sharing some insights!
And what about Slick EQ curves without any addded distortion? Would they be more like Q3 since it's not designed as a dynamic EQ?
The end "I have absolutely no idea" made the whole video worth it. To have something taught to you so thoroughly and beautifully detailed for NO REASON at all is what makes Dan a Master Jedi level UA-camr. Hilarious. Thank you for this video.
Yup sometimes you just gotta feed that curiosity
Spoiler alert! lol
@@RafaelBernatto My bad 🤣. But you know not to read Dan's comments before watching the video.🤷
Spoiler alert for some, but I was about to stop watching this video about 20% of the way in. This comment got me to watch the vid to the end to hear the quote and the lead up - and earned a channel sub!
@@Avant5 Dan is a true master. Check out his older stuff and his fabfilter tutorials on their site. Love ‘em
I think this video in particular exemplifies why I love music production so much, you can do all the science and engineering you want to try and understand what's going on, but at the end of the day it's your emotions and spirit that determine what really matters when it comes to making decisions.
....a bit like real life really!
Yep. How the end result comes out is all what matters. You can have the cleanest eq and you music can still be trash
There's very few audio You Tubers out there who can't seem to resist showing you how well informed and clever they are. Dan bucks that trend in spades: pragmatic, thorough, rational and humble. Long may he continue to be the scourge of the irresponsible urban myth-making that passes for education on the internet
The definitive word..."What was the point of this video? I have absolutely no idea."
Another enjoyable and informative episode, thank you, Dan. I really enjoy using both eqs as well-- in many ways due to your other excellent tutorials.
Back in the early 2000s, my job had me meeting with the head of DSP (name lost to time) for a leading pro audio company. In our meeting, he was excited to tell me that they had created a parallel topology EQ in software. At the time, standard DSP practice was to run the digitized signal through one EQ band, then through the others in series (serial topology).
He said the parallel EQ curves behaved and sounded much more like an analog EQ, which typically has multiple bands built as filters in parallel.
What he told me was that with parallel EQs, the band gain interactions summed, and with serial EQs, the band gain interactions were the product. My days of filter theory and supporting math are too far back in the rear view mirror to prove it, but instinctively it makes sense. Combine this with any unwanted distortions multiplying for series processes (TDR's comment), and it would be interesting to analyze what happens with multiple plugins in a chain.
Thanks for the analysis. I remain convinced that it's my musical aptitude that needs development and not my plugin collection.
I've always felt like the TDR Nova has more of a glue-y effect than other EQs, and I apparently correctly identified which one it was even just listening on my phone's speakers, but I'm also cognizant of the fact that most EQs sound good and that this was probably a lucky guess combined with a fine slurry of pride and confirmation bias that tricked my stupid meat brain.
Anyway, this was really interesting. Thanks for the video.
I've been using the Nova GE for a while now. Was contemplating getting Pro Q3 but for my money, the Nova product works plenty well enough. And this video helps to reinforce my purchase decision...at least for now. Thanks Dan!
"Nova's wideband covers everything that isn't covered by another band."
So, in other words, if I want to compress everything but a bell centered on 700hz I just make a static bell on one band and a dynamic wide band? That's pretty neat.
🤔🤔😮😮
Yes
Something about nova and Slick EQ for that matter, sound better, get me the desired results faster. They're also crazy good value for the price and the free versions Rival any eq out there. Well done TDR.
Sound quality aside, a lot of time I'm going for Nova instead of other EQ plugin is almost always because of the analyzer. Can't really explain it, but Nova gives "better" representation of what is going on for me doing subtractive EQ work compared to, let's say Pro-Q. Im guilty of not knowing it actually have different topology (parallel) compared to Pro-Q, I know it sounds different than the Pro-Q especially in the dynamic mode, my guess was the more controls it offers (att, rel, ratio, etc), apparently it was so much more.
Thanks for the video, I actually learn something new today. :)
The amount of happiness I experience when Dan uploads a new video! Thanks
Oh man, I missed that "Hi, and welcome back."
Great video! Some comments: Usually, digital filters in DAW EQs are implemented using IIRs in the form of biquads in the Transposed Direct Form II (TDF-II) (/ cascades of them) or FIR filters (way less common). When using that topology, it doesn't really make sense to combine multiple parallel EQs, as you'd unnecessarily waste computing resources, seeing as you parametrize the filter coefficients to produce the exact filter type you want. Another approach is to use a filter topology that has multiple outputs, such as an SVF (there's other topologies you can use, usually these filters are linear discretized versions of real-life filters), where you have seperate Lowpass, Highpass and Bandpass outputs that all get calculated in one "pass". WIth these filters, you usually get the different filter types (bell, shelves, etc) by combining the outputs in various ways. If parametrized correctly, the output of such a parallel combination of these SVF taps and the output of a TDF-II are exactly equal (minus numerical errors) _for the time invariant case_. Once you start modulating the filter, so for example change the Q or the cutoff in real time, you can't prove that the TDF-II Biquad will remain stable, while for the SVF it is proven the filter is stable for all configurations and any arbitrary modulation signal (audio rate, nyquist, doesn't matter). Check out Aaron Wishnick's 2014 paper "Time-Varying Filters for Musical Applications", it goes into some detail there if you're interested.
I'm glad you ended on that conclusion because that's how I feel about most things in audio. 🙇♀️
Maybe this is obvious but while doing the null tests with the bell curves, no matter what frequency they were at (150hz vs 500hz for example), there was indeed a bit of noise being let through *around* that frequency, but the exact frequency chosen would always null perfectly, as if there was a notch filter right at that frequency. It's probably due to the EQs only varying slightly at the slope portion of the bell curve, while the very top (or bottom for a cut) of the curve stays generally the same.
Well observed. The behaviour at the centre of the bell is well defined, assuming no modelling of "analogue" inaccuracies: its the specified amount of boost, at the specified frequency, with zero phase shift. The differences (including phase shift) are all in the curves either side.
16:36 This is impressive. FabFilter achieved the same result of over sampling without over sampling.
This channel is a big thing in audio world. Thank you Dan !
Fell out of my seat at "I have absolutely no idea." Brilliant and knowledgeable with laughs to boot
SSL X - EQ 2 manual….
Parallel Passive EQ
Parallel EQ exhibits quite different sonic properties to the familiar serial parametric EQ. We are generally used to hearing the effect of one EQ band superimposed on another, as opposed to the band interaction inherent to a parallel EQ. Because the bands are placed in a parallel configuration, phase cancellations and re-enforcements happen that are not always obvious when first encountered.
Passive EQ is something that is found in old equaliser units and is generally known for its transparent and natural sound, but has some problems associated with it. However, in the digital domain these shortcomings do not have such an influence.
A passive EQ does not have any gain elements, but can still have controls to seemingly boost frequencies as well as cut. What actually happens is that the entire signal is cut by an amount, but the frequencies that are apparently ‘boosted’ are simply not cut as much. Therefore the unit must attenuate either the input, the output, or both to allow enough headroom.
Unfortunately, in the analogue domain, a 20dB reduction in signal level produces a 20dB increase in the noise floor. Luckily in the digital domain, with a 64-bit floating point DSP, these issues do not remain.
In X-EQ 2 - when the ‘parallel’ button is engaged - you are presented with a parallel passive EQ model that the original designers of these devices could only have dreamed of. The noise floor can be disregarded due to the huge resolution available in SSL Native plug-ins.
You may find yourself entering this mode more and more as you become familiar with the sonic signature. Larger gain changes are possible without colouration, and boost starts to become something that is useable to a significant degree in a digital EQ!
Parallel EQ does exhibit asymmetry in its boost and cut characteristics. But this is not such a bad thing as most engineers would agree that boost is best done with low (wide) Q values and cut with a higher (narrower) Q.
i dunno if it's good or not :) I have no idea now XD
I laughed loudly at "what a rollercoaster ride we're on" :D
Amazing deep-dive, thank you so much, Dan!
I have been checking your channel for 2 weeks now. Let's go :)
Hi Dan, really great video! The world needs part II, showing phase response plots!
TDR make some of the best sounding bang for the buck plugins you can get IMO. I use Slick EQ, Molotok, Kotelnakov, and Nova as my go to choices on my travel laptop along side Valhalla Delay and Classic Reverb. I would be perfectly happy if I could only use those plugins and nothing else. The Free TDR plugins are fantastic, but the GE versions make them total powerhouses. Great plugins.
I've been wanting to know this for years - thanks for exploring it :)
In your A/B test, I can hear (A) preserve transient better than (B).
Maybe I'm wrong, but does anyone else sound like me?
As always, thank you for such a great and deep video, Dan!
My hypothesis is that if you could set the Q in Nova more precisly to 1/root(2) = 0.7071... rather than rounding to 0.71 those small differences would go away
Good point I have to agree, it would seem the developers chose to round off the calc.
Almost certainly the underlying value is actually precisely sqrt(0.5), and simply rounded to two decimals for the GUI. The internal values used in the processing and the numbers displayed in the interface are almost always decoupled from one another.
@@Friedeggonheadchan That still requires it to be set to this precice value. Nova starts up with other Q values, so if you want 0.7071 you would need to type it in manualy. But we can't see Dan doing that in this video.
I did a fast null test, and it makes absolutly no difference between typing in 0.71 and 0.7071. So either Nova rounds the value you type in before setting the underlying parameter or the parameter itself is stepped to 0.01
In the A / B there was definitely an 'air' difference waaaay up towards the limit of my ears 17k-ish, A sounded more open but I'm on adam ribbon tweeters in a controlled environment so I doubt its a big deal unless there's lots in a project.
I have a set of old battle scarred ears and in that first A B Listening Test I got 'Brighter' HiHats from A than B.
The differences in those double bell curves is intriguing ..
I never listen to WHAT you say. Just listen to your voice for sonically enjoyment :)
Seriously: GREAT CONTENT, every time ! Thanks !!!
Pleeeaaaase take a look at the ToneBooster EQ and whats about the (growing) hype about the Kirchhoff-EQ
The conclusion made my day. I dont know why but thank you Sir, always a pleasure!
Another amazing video. I've started taking notes as I realized that if you took the time to make a graphic to show us, it's probably a pretty dang important concept to understand. I drew out the pictures and wrote notes and noted a few places for further exploration. I downloaded plugin doctor and the other thing required to run multiple plugins through it from the phase video just to mess around and have some science. I've switched from Nova to Pro-MB for dynamic EQ on my latest album and I'm gonna go do some experiments to see how Pro-MB stacks up against Nova and especially Pro-Q3's dynamic mode. My guess is that Pro-MB would be parallel just due to the visual representation of the graph and the fact that Q3 already is serial. But I guess we'll see.
Again, great video. I get super excited every time you post and have shown these videos to people who have no idea what's going on, just to show that music engineering is an incredibly complex and sciency thing. And maybe for some bragging rights...
Have a good day y'all. Hope everyone's doing well.
Great stuff! The punchline at the end made me LOL!
I’ve been using this kind of eq in the ssl x eq for a while.
One of my favourites.
By my calculation, doubling the amplitude means adding 20 * log(2) decibels, which comes to 6.02.
I caught that too. I'm one of the people not satisfied to simply say 6dB, I say 6.0206 dB.
i love nerding the heck out with these videos
Also Dan, that tune sounds sick. Seriously.
i personally found it very interesting, another quality video to make me think
haha the ending cracked me up 😂 awesome video like always Dan!
EQs' behaviour aside... impressive video editing skills, matching the narration script!
The surgical and musical curves could be described a Gibbs fenomena, and sorry for the interruption but how often someone came across then the plugin the use the standard window to grass the results
best of the best, always pure, full of knowledge informative video, from someone with a passion! I have learnt so much from you Sir!
“Let’s shelve that idea for a moment”
Goddamnit Dan
I hear a difference; even using headphones. Listen to the top click on the kick drum, claps, and stereo width/image. On "A" you can here that tiny top end click sound on the kick drum more than on "B". Also, the information on the sides also seems brighter. This gives the illusion of a widened stereo image. It's just oversampling can cause hi end information to seem brighter. Therefore, if hi end information is on the sides, the sides will appear slightly more pronounced.
Also there is a slight quality difference. The clap 👏 tails are more detailed on "A" than "B". Honestly, after listening again, I hear "A" offering and entirely more detailed picture than "B" and the 808 boom is more controlled on "A" than "B".
The differences are extremely tiny but noticable. The real question is, which one do you think fits the job? That will forever be subjective and the beauty of music and music creation!
Tip: You have to be very careful with the application/deployment of oversampling and how much. If you use too high of an oversampling rate your music or tracks will start to lose their natural emotion. This becomes most evident in the impact of low end the higher you go up in oversampling.
I've been using FF Q2/Q3 for ages and it's my go to EQ for cleaning up, analyzing etc. But for boosting I prefer Pultec like EQs (Wesaudio Prometheus or as a plugin e.g. SPL Passeq) and love my 2 Clariphonic 500 especially on anything that need to cut thru the mix. For me it sound more musical/ less harsh. Interestingly or rather fittingly I preferred A and will give Nova another try (own the GE but forgot about it 🙄😂). So thanks it was (again) a joy to watch and helpful! 🙂
Is FF q3 a better buy to Nova GE?
@@MusicChannel-rf5zz no, different. Since my comment I've been using nova ge a lot, especially for cleaning up too long decays in the bass region, or sidechaining /ducking bass to BD. It's excellent for that
Both eq have benefits that's the real fact. But pro q 3 is still my first choice. Thanks for your time Dan
Kush Audio Clariphonic is my personal favourite!
Nova more like an RMS compressor. Thats the part a lot ppl here on YT making videos about Nova they dont get. They think its a Peak detector but its totally different. Its great that Dan pointed that out since it seems like ppl tend to listen to him more than anybody else 😄
Well, today I learned something, thanks Dan. The interaction between two bands varies in more ways than just the "different types of parallel eq" btw. But now I have part of the explaination. PS: I can't be the only one watching these videos on 1,5x speed, right?
Oddly enough, I heard a difference. Maybe I imagined it, but the high end on Nova sounded sharper/crispier to me, and I preferred option b for sounding smoother/silkier
Same observation here. But, if Nova would have musical curves option selected, it would be the other way around. When you need it silky with a tiny bit of haze - musical, when you need it crystal clear - surgical.
Please take a look at Kirchhoff-EQ by Three-Body Technology. it's new and it has some very unique and competitive features.
Did you know that you can change the db scale in TDR Nova from +-12db to +-24db by clicking on it on the left side of the window?
If you want to look for the Point!! This video helps a lot that's the point...this is what more people need to do...make it easier to understand and make it easier to try to understand something as big as this...sharing knowledge makes us all smarter🧠
Thanks for showing this plugin! ~ In your sound test, I listened with Bose Sound Sport Free earbuds and I could hear subtle differences in the mid to low end range when A/B'd. I thought "A" (Nova) sounded extremely clean. I assume that may be mainly due to the differences in frequencoes added compared with Pro-Q. Thanks again for an interesting comparison, Dan. ~ Cheers
Hi Dan,
Thank you for making a great point that I've been making for decades.
61 years of recording and no I can't hear a difference.
And if I could hear a difference, what difference does it make ?
Thanks for taking the time to point this out.
Bill P.
..and Bob's your uncle ! Dan for president...phenomenal stuf
"What a rollercoaster ride we're on." made me LOL suddenly and frighten the cat
Im guessing all this music is your own, its good stuff. Inspired.
I was just watching your older videos because I had a severe lack of Dan Worral content :D
The ending made me laugh out loud while I'm alone....
Top content, Dan is the man!
Nova had slightly sharper transients in my ears. But it didn't amount to much honestly. Loved the conclusion though, maybe one day you'll do a follow up :)
nice video thanks dan. in the AB comparison at 17:38 the difference i hear is clearly in width and A seems definitely somewhat wider than B
curious about your opinion on this?
I don't perceive any difference in width, nor can I think of any reason why there would be one: both plugins process left and right channels exactly the same. Also stereo differences aren't that significant in the highest frequencies: width is really all in the midrange. If you have both plugins and you want a definitive answer, try an ABX test.
I can't hear a difference between A and B. Your assumption is correct with me at leat. :)
As for how Pro Q adds two peaks, it seems that FabFilter made a conscious choice of flattening the sum to allow for the case where you want to boost one frequency region. I believe that KiloHearts EQ does that also. It certainly makes sense to me but what is strange is that they didn't communicate on this feature.
I'd love to hear examples of specific use cases where either is better than the other in the dynamics prosessing that you touched on.
may I ask where the 0.1 comes from? I've not coded in a good while but I thought 6db was literally the result from an x2 in the math, I even remember reading a paper arguing that technically a 6db boost would be cleaner than any non integers as it wouldn't result in as many rounding errors of the 'digital voltage' (not accounting for interpolation of course) so that .01 is really throwing me haha.
Anyone knows whats happened with the no longer available (in the last 2.1.5 version of Nova GE) processing modes, "precise -" and "insane - " ? The Nova (2.1.5) "precise" and "insane" is the new "-" ? Clean? With the "+" (nonlinear, some subtle saturation) option?
Thx Dan, awesome as usual.
‘A’ has a bit more of a quicker transient response to my ear so I liked it a bit more. Not a deal breaker though either way.
Late comment and love both of these plugins and own them both. Your video's are always so informative it's awesome! Would you then know if Pro-Q3 is different between the AU/VST3/AAX compared to the iOS AUv3? My assumptions would be there is none but thought I'd ask 🤘💀🍻🔥
I only run them on one platform, but I feel fairly confident saying the audio processing is the same for them all.
Always interesting Mr Worrall...ive learnt a lot form you in the last couple of years....what would actually be really interesting is a video from you talking about why (and why not) where and what Eq you would use in certain situations. There are so many Eqs on the market and all are marketed with musical/surgical hype terms...when i use plug in doctor on the Eqs I have from expensive mastering ones to bog standard they all seem to exhibit similar curves on the screen ..Waves F6,ProQ,MAAT Blue Eq, to Plugin alliance AMEK 200 and their Soma and Digital V3 all of which I have seen on high quality mastering chains.. .what gives Eqs character, what is that character and why would you chose that and when. Please consider this ..I would love to know your opinion.
Great and entertaining video 😊
Such a detailed video. Appreciate your work, Dan! 💯
I laughed so hard at that conclusion. Brilliant!
In theory, direct EQ and parallel filters are equivalent. Let's say the direct EQ has transfer function T(f). We want the parallel filter blended with the unprocessed signal to have the same response, i.e. T(f) = 1 + P(f), where P(f) is the frequency response of the parallel filter. There always exists a solution to this equation, and thus for any direct EQ, there exists an equivalent parallel filter.
In practice, most EQ plugins probably don't have enough degrees of freedom to match any direct filter to a parallel one (this part is just my guess)
Aww yeaa
Edit: Have you done or will you do any videos on DMG plugins? They're so dense and full of teaching opportunities, I think you would be great at articulating all the different possibilities. Anyways, great vid! :)
DMG Compassion is my goto comp. takes a while to master but can do pretty much anything (except colour as it's clean).
Cool video Dan!
I’m interested in your expertise regarding ToneBoosters Equalizer 4 vs Fabfilter Pro-Q3.
For those who have worked with physical rc filters, which software EQ are models of - regardless of how their approach varies, the parallel approach is much more intuitive to use. I don't expect one band to preclude the other's existence, that seems beyond magical to me. The plotted curve of one bell should not be an additive activity, unless there were a simulation of am op-amp before the output, but account for its own nature only. Perception is one thing and there may be an imparted addition in a graphic EQ between two bells which are within each others' Q but not in a hardware parametric EQ, however, perception isn't what we are aiming at. Sure, going back to Q function the NOVA Q can be somewhat widened and their centers brought closer in order to match Pro-Q's but, from that *perception*, which is correct? We already know that Pro-Q took a certain liberty with its rendering of 0.71 as 1 (when that is NOT what is really happening) - why should we assume that it also didn't take a liberty with its output of additive Q function between filters? Maybe they assumed that if a user wants the normal hardware parametric function of real rc filters they are free to add a 3rd bell between them and set it to cut (maybe you should investigate that). Personally, I don't like it when anyone begins to round off reality in such ways. I prefer the TDR method, and that might be just the nature of the parallel implementation but I don't see it. Unless the series Pro-Q approach really does mimic a graphic EQ relative to how the parallel NOVA approach intentionally mimics a hardware parametric, the folks at FabFilter just got caught taking artistic liberties with the nature of Q function. Is that the end of any world? Nah. But, it could prove to be annoying if one were not aware of it, coming from an analog environment (which many of us have). Great video, Dan.
It's the nature of the parallel implementation. In a serial design the input to each filter is the output of the previous filter. In a parallel design the input to each filter is the original input signal.
@@DanWorrall Sure, now explain the high frequency anomaly if Pro-Q isn't really just cascading a sequence of set graphic bands with a width parameter that they forgot to extend beyond the graphically displayed range. Any node between bands can be accounted for when there IS a node to even be counted. Below 10hz there's not much to read but above 20khz is another beast.
Sorry, you've lost me..?
@@DanWorrall It's not important. I was suggesting Pro-Q has fixed centers and a mechanism for moving between any two or more of them to fake a parametric Q function which, when placed in a series, is fine until they run out - then you're left with a fixed singular ''last bell'' that has no other bell to partner with (which could explain the non-nulling you found at the highest frequencies in the tests).
No, that's not how it works.
Jet another smile on my face. thank you Dan ;)
I'm a simple man...I see new Dan Worral video in my feed, I press 'Like', then I actually watch the video.
This was like the National Geographic of EQs haha great video! got a sub from me!
Thanks Dan for reminding me that I still have NOVA installed. I almost forgot about that EQ. :D
such a goldmine, thanks dan
Enjoyable as always.
Thank you for this! I love your videos because of your nerdy technical approach.
But one question I have is what's the difference between this two EQs phasewise. I mean there have to be some but I don't have plugin doctor or ProQ3 either...
Could you make a video on Kirchhoff EQ?
Ok, so both useful, both ok. I'm glad I spent all that time confirming that one ;-)
Lucky I was doing something useful (animating cogs) at the time
:-)
your videos are solid. well done, i'd like to see info on multi-rate filters, like polyphase filters . thanks keep up the great videos.
Transients sound clearly clearer to me on A (TDR).
Funny enough Pro Q and NOVA are both one-key shortcuts on my REAPER, absolutely invaluable tools.
Здравствуйте. Подскажите где сохраняются пресеты у TDR Nova GE
What a rollercoaster we're on
I'm a TDR Nova guy. I just wish it was a bit easier on the CPU.
yes.
If I had to choose only one plugin to use for the rest of my life, it would be TDR Nova. AND ITS FREE!
20:15 seems exactly like the difference between the Motown EQ emulations by UAD and Kit Plugins, UAD being the parallel approach and Kit being the serial.
Might be different iterations of Motown EQ modelled, or Kit just did a very bad job.
Hi,Daniel! What you thing about - "Drawing eq curves" in a flip polarity of 2 fx tracks? It make sense if i draw at very low level of db to rich more clear sound and less artifacts of eq?
Sorry, I don't understand the question?
At first,sorry for my English! it;s many ways to approach eq! Let;s say - I have 2 indentical tracks with the same audio material! I flip the polarity to one of them and i insert a eq plugin like Voxengo Curve Eq! Next,I set the scale view to 3db! Ten i draw a eq curve to my needs in subtractive way! Finally, i summed this 2 tracks to a group to compesate the volume! In terms of quality,it's better way to eq something? because if nedd drastic mooves i can achieve in limit of 3db! Theoriticly,should be better to preserve quality!
@@cezarciucium5580 ok I think I understand. And no, I don't think you're gaining any extra quality. What quality problems are you trying to avoid anyway? People are much too worried about 'phase smear' and such: don't be. Get the frequency response right, that's unusually all that matters.
I try to avoid phase as much as possible when it comes to hp filtering and get rid of narrow resonances! Both reduce the transient response! You sad in one video that the phase is cause by how much db and q factor are increased! Also everybody talk about that subtactive sounds better then boosting eq! Thank you very much!
Don't be scared of phase shift, first of all, its fine. And I don't think you're avoiding any phase shift by doing it your way.
Why there's no spectrum analyzer plugins that utilizes Gammatone filter banks or sliding DFT constant-Q transform?
I mean the IIR Gammatone filter bank is fast as CQ-sDFT with ERB frequency resolution
Could it be that, for a developer, choosing between either topology would come down to what feature set they are looking to offer? I was thinking about the example you gave with the dynamics processing features in both plugins since you mentioned that, in theory, it's easier to do with parallel bandpass filters. Or maybe I misunderstood something
Nice review, can you compare TDR NOVA, PRO Q 3 and Kirchhoff EQ ?
Lol... Oh, Dan, you're killing me, dude! Great analysis as usual, however.