The No.1 Website for Pro Audio
best lowcut filter plugin
Old 2 weeks ago
  #31
Gear Maniac
IIR filters do it pretty well to kill DC, even 6 dB at minimum frequency 5 or 10 in Equilibrium. In FIR you should use maximum IR length, but you should know it depends on sample rate (for example, 2048 length at 48kHz is same as 2x times more IR length at 96kHz (4096), and is same as 4x times more IR length at 192kHz (8192)). And if you use maximum IR length in FIR mode, even at 44100Hz, it is not so accurate as IIR.
You can find info about IIR, FIR, FFT, length, and how it depends on sample rate. And even how to make linear phase with IIR. - this info on Meldaproduction site
Airwindows has DCVoltage plugin, it is free. It rotates the phase of DC, so you can remove or add more DC offset.
Old 2 weeks ago
  #32
Gear Maniac
 
Robb Robinson's Avatar
 

Quote:
Originally Posted by msmucr View Post
Otherwise comparisons will be flawed (and might lead to sensational results), just because of different filter steepness.
Very interesting even if almost entirely over my head. If I were to match their curves in PluginDoctor would that make it more of an apples to apples comparison? Or is the curve only telling us half the story?
Old 2 weeks ago
  #33
Lives for gear
 

Hi,

Quote:
Originally Posted by Robb Robinson View Post
Very interesting even if almost entirely over my head. If I were to match their curves in PluginDoctor would that make it more of an apples to apples comparison? Or is the curve only telling us half the story?
that's definitely good idea with PluginDoctor or similar analyzer, I'd say, it's kind of pre-requisite at least for initial idea, how filters in each compared EQ behaves.
Also if you have full paid version of PluginDoctor, it will allow you to load up two plugins simultaneously, so you can relatively easily match that as close as possible. Magnitude and phase curve will tells the whole story, a clean EQ is linear effect, so it can be fully characterized just by that. However you usually can match those just to certain extent. Be it because of rounding at control parameters (eg. Q of 0.41 vs 0.40742), or for example slightly different formulas for compensation of curve warping toward Nyquist, so there can difference.. but crux of that is to find out, whether such difference is audible and anyhow significant for perceived sound quality. But anyway, that's rather for some expanded writeup. Unfortunately lot of times, reported significant differences between clean EQs with comparable processing modes come down to poorly matched curves (that's always the most significant factor in perceived differences) or non-blind listening (HOFA 4U+ BlindTest is the friend). Of course kudos to everyone, who does that properly.

I essentially summarized main possible pitfall with MAAT Orange and Weiss EQ-1 LP in last paragraph of the previous post. Filters at those backward-forward EQs are steeper than at most other plugins due to inherent stacking. Also sometimes, that is misleadingly labelled (at Weiss, but there is that IMO because LP mode was additional firmware upgrade for hardware unit, which didn't change faceplate . And when they made its plugin version, they replicated behavior including that mistake)

So depending on what EQs you'd like to compare, it might be necessary to either switch HPF to 24 dB/oct slope (Equilibrium if I recall that right), or manually duplicate two identical 12 dB/oct filters (Pro-Q, Crave). Of course, there's lot of other plugins out there and I haven't tried all. So you need to verify that in PluginDoctor.

Michal
Old 2 weeks ago
  #34
Lives for gear
 
vze26m98's Avatar
 

@ musmcr Thanks so much for a great set of posts!
Old 2 weeks ago
  #35
You can demo apQualizr2 and see if it works for you. I have been using it for years. https://www.apulsoft.ch/apqualizr2/
Old 2 weeks ago
  #36
Lives for gear
 
thermos's Avatar
Quote:
Originally Posted by msmucr View Post
I'd just note one catch for proper comparisons of such linear phase EQs with forward-backward IIR principle.
It works block by block and essentially apply the same filter twice (once forward, then backward), this will compensate any phase shift.
But this naturally doubles amount applied magnitude change. You can rather easily compensate that for bells and shelves.. you enter +6 dB, but underneath it processes the audio twice with +3 dB filter.
However for HPF and LFPs, it's important to realize, such process essentially stacks filters. So for example, if you set it to standard 2nd order 12dB/oct filter with 3 dB cutoff point at say 50 Hz. The result is rather steep 24 dB/oct filter with 6 dB cutoff point (similarly like Linkwitz-Riley filter at your crossover) at the frequency.
That's kinda general characteristics of such type of linear phase filtering.. and applies to all of those EQs.. (Orange, Weiss in LP mode, old PLParEQ).

So for example the Orange has only pretty abrupt 24 dB/oct filters.
Weiss LP filters set to 6 dB first order is really 12 dB/oct, and set to 12 is really 24.

Just keep that in mind, when matching and comparing to some other EQs with more common FIR filters. Say at Fabfilter, you need to create two identical 12 dB/oct filters to match the same response. I haven't played with Equilibrium for a while, but there I think anything higher than 12 dB/oct has 6 dB corner point, so one filter with 24 db/oct will suffice to match the Orange.
Otherwise comparisons will be flawed (and might lead to sensational results), just because of different filter steepness.

Michal
Interesting, I didn’t know Weiss was the same linear phase method as orange. Essentially orange was a hack to reduce cpu usage from my understanding, right? I think plugin designers shy away from this approach now because it is considered inferior by modern standards.
Old 2 weeks ago
  #37
Lives for gear
 

Quote:
Originally Posted by johannburkard View Post
I checked out your code and I think it's the same idea I had in mind: Computing a sliding-window average of both sides of the waveform plus difference removal.
In the case of my old JSFX it isn't a sliding window, it's averaged continously from start of playback. Just UI indicators and calclulated correction offset gets updated once per processing block (eg. working buffer size). As you stop playback, the offset is freezed and just statically applied to the audio, that was important bit for me, also the longer averaging is more accurate. It had also some specific usage context mentioned in the original thread.
There was some other old VST plugin for DC offset removal, which calculated new correction offset over every block (so similar to your idea), which was then applied to audio during next block... then updated etc.
But there is inherent problem with start of playback and renders, because the first block can't be corrected with realtime plugin, so you'll get always DC glitches there.

I guess, nowadays I could also write new version using ReaScript for true offline detection pass, which will detect offsets at clip by clip basis from respective start to end and then instantiate take based custom JSFX with appropriate correction offsets or use HPF instead, when offset will be changing (eg. add kind of "heuristics" there). ReaScript API was significantly improved from back then. But honestly so far I didn't feel so much pressure to do that
As I've mentioned, special DC offset treatment is rather very rare for me and usually I'm happy with either HPF or with current version of the script.


Quote:
Originally Posted by vze26m98 View Post
@ musmcr Thanks so much for a great set of posts!
Thanks, you're welcome.

Michal
Old 2 weeks ago
  #38
Lives for gear
 

Quote:
Originally Posted by thermos View Post
Interesting, I didn’t know Weiss was the same linear phase method as orange. Essentially orange was a hack to reduce cpu usage from my understanding, right? I think plugin designers shy away from this approach now because it is considered inferior by modern standards.
Yes it uses the same basic IIR approach. There was couple of similar LP EQs.. Algoirthmix Orange, PLParEQ and Weiss among hardware (or nowadays as a plugin).

Hard to guess, if resources were primary reason there, although 20 y ago it that was definitely big thing. Also in case of existing hardware EQ or plugin they definitely could use common existing minimum phase IIR formulas and modify those to make linear phase versions.
I don't see it as inferior in general, it's just different approach with this forward-backward processing.
Of course there are some disadvantages, when using common filters. Like previously mentioned inherent "stacking", so you get either 12 dB/oct (stacked first order filters without adjustable Q) or 24 dB/oct with adjustable Q.
Also if original IIR formula doesn't have compensation for magnitude warping towards Nyquist, so if you want to correct that, you have to use also additional oversampling for example (that's the case of all previously mentioned plugins). Also there are due to some attributes of processing it need for comparative higher latency compared to FIR.

At FIRs counterparts you can essentially do any filter shape or phase response you want, so it's much more flexible. That includes some more exotic shapes - from brickwall to super gentle, from minimum through linear up to completely reversed phase response during some trippy creative evenings.. (like Equlibrium in free mode or Ozone EQ in surgical mode). Also you can directly generate internal IR, which has perfect analog like magnitude or phase response towards Nyquist frequency and thus it's not necessary to implement additional oversampling or some corrective filter.
On the other hand design and implementation of FIR based EQ, which is relatively efficient, stays responsive to realtime filter adjustments, handles continuous automation well and without glitches is rather non-trivial and rather very advanced task. Not to mention dynamic FIR filters.

Michal
Old 2 weeks ago
  #39
Here for the gear
 

Quote:
Originally Posted by msmucr View Post
Yes it uses the same basic IIR approach. There was couple of similar LP EQs.. Algoirthmix Orange, PLParEQ and Weiss among hardware (or nowadays as a plugin).
Interesting - I remember seeing that the Sonoris Mastering Equalizer also has an IIR linear phase mode. Does anyone have any experience using the Weiss / Orange / Sonoris - do they have similar sound profiles? It seems like a lot of people have written about how the Orange manages to maintain depth better than many FIR linear phase eqs - is this a trait that the Weiss / Sonoris / PlParEQ / etc. also share because of this IIR implementation? Does this older, less-efficient IIR linear phase actually maintain depth better than FIR linear phase which can seemingly mess with transient definition and image?
Old 2 weeks ago
  #40
Lives for gear
 
vze26m98's Avatar
 

Quote:
Originally Posted by bqstudio View Post
Does anyone have any experience using the Weiss / Orange / Sonoris - do they have similar sound profiles?
You can search here for the threads that chronicled their initial release. Makes for interesting reading, as folks’ concerns were pretty different back when. PSPaudio’s Neon and MasterQ2 got a fair bit of respect...
Old 2 weeks ago
  #41
Lives for gear
 

Quote:
Originally Posted by bqstudio View Post
Interesting - I remember seeing that the Sonoris Mastering Equalizer also has an IIR linear phase mode. Does anyone have any experience using the Weiss / Orange / Sonoris - do they have similar sound profiles? It seems like a lot of people have written about how the Orange manages to maintain depth better than many FIR linear phase eqs - is this a trait that the Weiss / Sonoris / PlParEQ / etc. also share because of this IIR implementation? Does this older, less-efficient IIR linear phase actually maintain depth better than FIR linear phase which can seemingly mess with transient definition and image?
Well, you likely find lot of various reports ranging from "It's all the same" up to "It changed my life, never heard anything like that before." Some super fantastic adjectives/
And for example some of MAAT's descriptions and previous claims were story by itself to an extent, that I wondered, whether mathematics could withstand that
Also lot of times, one has to read claims and user comments in some broader context. Of course a good algorithm works still the same way and it doesn't "rot" over years. But if you read some quotes from 2002 or so, when some famous guy compared it to other common tools back then, it could be a bomb.. however that couldn't necessarily apply to current offerings. As you have bunch of super talented developers around, with well implemented and refined products with good usability (that bit is often overlooked and directly affect work with such tool).
So I personally always take that with grain of salt.

But I see, where you are coming with your question. That's basically over years, how I think about comparisons from technical point of view. First to classify an EQ (or its operational mode) according to its processing properties.. like FIR/IIR, magnitude warping compensation (that can be also just for certain filter types), phase warping compensation, oversampling.. etc.
Not surprisingly the ones with close or same processing modes was possible to match pretty much perfectly (up to extent of controls range and its precision) and also I typically haven't found any perceivable difference.. be it free, super expensive EQ or even some filters in my programs.

On the other hand I believe, there can be perceived and reported (ID'd in blind tests with certain filter combination and material) difference among processing approaches and modes. From obvious ones (like linear vs. minimum) through much subtler ones (minimum with magnitude compensation vs. additional compensation of phase towards Nyquist).. up to some splitting hair territories.
But in general to me (sorry for adjectives).. the IIR zero latency modes are roughest, most forward sounding.. it gets tad smoother overall (even if you don't do manipulation with highs around Nyquist), when you enable oversampling or FIR modes.

I tried the new Orange pretty briefly, had used cheaper three band PLParEQ for years and have some days on iLok for EQ1 plugin. I plan to do proper shootout during next week (it really takes time and energy to do that properly in studio and I'm either busy or tired ). But according to some of my previous experiences, I think most people react to different sound signature of Orange (eg. powerful bass for low end shaping), when it has defeated oversampling, which is rather rare thing to have with LP eq.
So for example, if you try also Sonoris or Neon HR (good call @ vze26m98 I almost forget on that), which are also linear phase IIRs with option to defeat oversampling, then there is IMO quite high chance, that reported "magic" of energetic sound with linear phase might be also there.
Similarly if you happen to match curves, then those plugins at minimum phase mode with enabled 2x oversampling at base rates could be very close to EQ1.

Finally I will return back to usability. As I've described those raw (zero latency), smoother something, linear phase.. kind of "flavors". That's exactly what I like to have available at single plugin and without any distracting jumps (like in filter slopes) for instant comparison during real work. And why I like at modern plugins (like Pro-Q, Crave or Equilibrium), because there is that option pretty much instantly available for judging in the context. That's I also applauded to update of last version of Slick EQ M, which added zero latency mode besides previous oversampled one. So I'm personally pretty happy with what I have, because I feel, it works for me.
Similarly other people, might have preference for something else, and although certain EQ doesn't sound anything better or worse than other similar EQ (although they might claim that ), its UI works for them and feels good under their hands (can be something as overlooked as good mapping of controls steps to mousewheel), so they can achieve their sonic goal faster.

Michal
Old 2 weeks ago
  #42
Lives for gear
 
vze26m98's Avatar
 

@ bqstudio From the PSPaudio Neon manual:

Quote:
To overcome this problem, mathematicians and DSP engineers came up with very clever mathematical tricks which significantly reduce the number of required operations. One of them is frequency-domain fast convolution, which we use in PSP Neon and PSP Neon HR. The other approach, found in some commercial products is IIR filtering done backward and forward in time. We chose frequency-domain fast convolution because it reduces the accumulated calculation error as well as allows for perfect phase linearity without needing windowed overlapped processing to diminish the effects of truncation.
Old 2 weeks ago
  #43
Here for the gear
 
JordanTailored's Avatar
 

I remember my all-linear-phase-Waves mastering chain when I first started out. Humbling memories.

Then Neon-HR for quite a while, at least the UI was larger and I made smaller adjustments because of it (though they looked the same as on the Waves plug, lol!)

Many many years later...

I can appreciate how a full range system that's giving you honest info down there (or low-end excelling headphones like Audezes) will let you choose a variety of low end options like a 20hz shelf, gentler filters (6db, 12db), targeted bell, and leaving it with no cut.

I like the DMG for these and extreme filtering if you need, and if you want to deviate from IIR you can build your own filter with a lot of options, though I prefer to not.
Old 2 weeks ago
  #44
Lives for gear
 

Quote:
Originally Posted by vze26m98 View Post
From the PSPaudio Neon manual:
Ahh.. good catch, I always though Neon was also backward forward IIR in LP mode. Maybe it's time to download demo, I haven't found just manual at their site.
Anyway its interesting. Because they employ FIR filter for linear phase mode, but generated IR produces cramped output, hence there is switchable additional oversampling. That's rather unique, because in most cases that IR already produces decramped response. So another step with oversampling (with another short anti-aliasing FIR filter) isn't really necessary.
Whereas for IIR filters when standard bi-quads, oversampling is more common.
Anyway I was confused by that, that's also why I though it's IIR based.
Honestly Neon somewhat always passed around me, although I demoed it many years ago, all I can recall is, I wasn't exactly excited from its UI with LCD like graph and tiny controls and thought PLParEQ was superior even just for usability reasons.

Michal
Old 2 weeks ago
  #45
Here for the gear
 

Wow, some greatly informative answers! I too tend to go for zero latency minimum phase EQs like FabFilter's but I've found linear phase to be a good option when working on orchestral recordings or jazz recordings where phase shift introduced when cutting resonant peaks (especially from live recordings) feels weird enough that I'm willing to take LP with the trade-offs... hence why I'm interested in this "magical" IIR LP that seems to be so hyped up around here as of late with the orange
Old 2 weeks ago
  #46
Lives for gear
 
vze26m98's Avatar
 

Quote:
Originally Posted by msmucr View Post
Ahh.. good catch, I always though Neon was also backward forward IIR in LP mode.
I was comparing a few parametric EQs today using a 24dB HPF at 60Hz: MAAT Orange, PSP Neon & MasterQ, Equilibrium in IIR & FIR modes, FF Pro-Q.

The signal was a 60Hz sine wave run through Waves Submarine, default preset, which generated subs at 15 & 30Hz, with the "-2nd" harmonic louder than the first. So the HPF made a dramatic change in the sound of the signal.

Later, I was reviewing the rendered files in RX6 and was startled to see that the Orange and Neon wave forms were identical! I thought I had made some rendering mistake, but no. I thought about posting them here for curiosity's sake, but I hadn't been that careful about gain match; laxness allowed about 1.25dB creeping difference in amplitude across the group.
Old 2 weeks ago
  #47
Lives for gear
 
vze26m98's Avatar
 

Another HPF nobody's mentioned is the one in DMG's free TrackControl: 18dB going down to 1Hz. MP of course.
Old 2 weeks ago
  #48
Lives for gear
 
vze26m98's Avatar
 

Quote:
Originally Posted by msmucr View Post
I haven't found just manual at their site.
PSP has re-worked their site recently, and it seems you can't Google for the manuals anymore:

http://vze26m98.net/gearslutz/PSPneon.pdf

Some other stuff worth reading in there.
Old 2 weeks ago
  #49
Lives for gear
 

Quote:
Originally Posted by vze26m98 View Post
I was comparing a few parametric EQs today using a 24dB HPF at 60Hz: MAAT Orange, PSP Neon & MasterQ, Equilibrium in IIR & FIR modes, FF Pro-Q.

The signal was a 60Hz sine wave run through Waves Submarine, default preset, which generated subs at 15 & 30Hz, with the "-2nd" harmonic louder than the first. So the HPF made a dramatic change in the sound of the signal.

Later, I was reviewing the rendered files in RX6 and was startled to see that the Orange and Neon wave forms were identical! I thought I had made some rendering mistake, but no. I thought about posting them here for curiosity's sake, but I hadn't been that careful about gain match; laxness allowed about 1.25dB creeping difference in amplitude across the group.
I can believe, that waveforms were very similar visually. However common testing with pulses or periodic noise in analyzers like PluginDoctor or older VST Plugin Analyzer is much more accurate than simple tests with sines, as you can reveal much subtler differences than when you visually compare waveforms.

Another common approach is simple nulling between tracks directly in DAW. You can do it in realtime, no need for renders.
I typically use one source track with white noise generator (It's likely at all DAWs or you can use looped generated white noise clip), it's suitable for that because it contains all frequencies uniformly distributed over whole spectrum.
This is simultaneously routed to two auxes with reference and tested plugins respectively. Aux with reference plugin has also flipped polarity (either directly at mixer, or via some built-in plugin like in PT or Logic). Then those two auxes gets summed, either to master bus or another aux.
Finally I put some realtime spectrum analyzer like Voxengo SPAN to this bus.
It's important to set the analyzer correctly, so avoid any tilted slope (set it to 0 dB, so white noise produce flat spectrum) and possibly increase its FFT resolution (say to 16k).
That way you can easily see differences between source and reference track, and thanks to FFT analyzer instead of just single figure RMS level, it's also much more detailed, because you can easily reveal, where that difference lies.
For example according to RMS level, you'd get some plugins null to -40 dB (eg. quite off), but on FFT, you can see that 20-20k is at -120 dB, and the difference is just above 20k, due to different anti-aliasing filters in oversampling section.
Also if you prepare such four tracks to some template, you can load it to whatever project you like, so for example you can tweak some favorite EQ over particular music material, then simply copy such plugin to the "reference" aux and try to match some other plugin via nulling. As you make closest match, you can copy the other EQ back to music track and compare those with real audio.
Just during such null tests, be aware, that if latency of some plugin changes, say after switch modes or quality from medium to high at some EQ, you need to start and stop DAW playback. Most of DAWs recalculate offsets and change its latency compensation just after next playback, otherwise nulled track would be out of phase.

Finally it's IMO good to remind some proportions of results from those null tests between EQs.. The level you see on the analyzer bands depends also on chosen FFT length - number of bins there (due to phenomenon called FFT gain).
Anyway for instance with 16k length, if your EQ curves differs by 0.1 dB (pretty much widely accepted threshold of perceivable level difference with very quiet room and revealing system), then at affected band area, you'll see residuum level reaching somewhere to -90 dB. With 0.01 dB it will be -110 and naturally with 0.001 dB (pretty much insanely close match), it will be close to -130 dB of residuum level there.
Just to have couple of reference points.

As a bonus, if you will be bored and have some stereo analog parametric EQ with non-detended controls, then you can try to match left and right channel to see what differences are there Or for some mono EQ, do two complete recalls with respective aligned recordings of common white noise source clip.

Quote:
Originally Posted by vze26m98 View Post
PSP has re-worked their site recently, and it seems you can't Google for the manuals anymore:

http://vze26m98.net/gearslutz/PSPneon.pdf

Some other stuff worth reading in there.
Thanks, I'll check that more thoroughly. But briefly looking there for cited paragraph in context, it's basically what I wrote in last post.

Michal

Last edited by msmucr; 1 week ago at 11:03 AM..
Old 1 week ago
  #50
Lives for gear
 

Quote:
Originally Posted by bqstudio View Post
Wow, some greatly informative answers! I too tend to go for zero latency minimum phase EQs like FabFilter's but I've found linear phase to be a good option when working on orchestral recordings or jazz recordings where phase shift introduced when cutting resonant peaks (especially from live recordings) feels weird enough that I'm willing to take LP with the trade-offs... hence why I'm interested in this "magical" IIR LP that seems to be so hyped up around here as of late with the orange
You'll see, whether it could bring something to table for you. Personally I feel, I'm pretty much set with what I already have for linear phase, but just for curiosity I also plan to do some more indepth tests later.
But if you trial Orange, Sonoris (just checked, he prolonged normal trials to 90 days during this pandemic) and compare it to Pro-Q, you're used to, it can be possibly interesting shootout for you.
Of course curve matching is essential and PluginDoctor or so is your friend (and IMO well worth of asked money for any plugin comparisons IMO).

Michal
Old 1 week ago
  #51
Lives for gear
 
vze26m98's Avatar
 

Quote:
Originally Posted by msmucr View Post
I can believe, that waveforms were very similar visually.
Thanks very much for your detailed post. It's gone into my digital clippings file along with a number of other things you've penned for future reference.

I am very careful to match my EQs and compressors in Plugin Doctor and lately, Ilya Orlov's CMT. Most of this preparation ends in a blind listening session rather than trying to null or further examine the signal with metering.

I realized also that so much of this assessment I've done involve compressors, which perhaps because of their non-linearity, I've never really bothered to look at their individual waveforms in any detail. So I was surprised that a quick glance evidenced such similarity, but I guess with equalization being a linear affair, I should expect that degree of commonality.

Thanks again!
Old 1 week ago
  #52
OMU
Lives for gear
 
OMU's Avatar
 

Quote:
Originally Posted by msmucr View Post
Similarly other people, might have preference for something else, and although certain EQ doesn't sound anything better or worse than other similar EQ (although they might claim that ), its UI works for them and feels good under their hands (can be something as overlooked as good mapping of controls steps to mousewheel), so they can achieve their sonic goal faster.

Michal


So true.
Old 1 week ago
  #53
Lives for gear
 

I just like to use the Shallow Low-shelve in TDR Slick EQ M and cut -8dB (and increase to 200%) set it at 10hz and move up + adjust till it sounds good.
Old 1 week ago
  #54
Here for the gear
 
Mindtree's Avatar
Not mentioned enough in my opinion, but the low cut filter on the Flux Epure v3 is superbly transparent and, yet, somehow remains musical. It is limited, though, to just the 12 dB/octave slope. For steeper slopes, I've had some success with the IIR filters in EQuilibrium.
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump