Digital EQ Fact & Myth.

[alias_9]

Originally Posted by rhythminmind

I would like to get some technical examples of EQ algorithm differences. I understand compression algorithms. These can be drastically different. How one digital EQ is actually better then another gets fuzzy. I'm not talking about features or interface but audio integrity.

I have recently switched to sonar 7PE. This has made me reevaluate my plugin arsenal.
I have been using Christian's plugin analyzer to compare my EQ's
Christian’s Blog » Programs

I have been comparing the
Duende channelstrip
urs CSP
Sonars Native mixer EQ
VC-64 (kjaerhus audio GAC-1 port)
& LP-64

The EQ variables that i am aware of are
frequency response
filter slope / Q
& phase

I have been able to recreate the the slope, phase, & sonic characteristic of the Duende & CSP with built in native EQ . Or with any fully variable parametric EQ for that matter.
As i said compressors are a different story. Duendes program dependent comp is it's own beast

Besides linear phase, internal bit resolution, or to free cpu with external DSP i'm now questioning why anyone needs anything but there native EQ.
I can't find technical reasons.
Please inform.

I'm coming late into this discussion, and I'm a bit new to all this, but what did you use as a sound reference. Was it just White noise?

[princeplanet]

Originally Posted by WildCowboys

Has anyone ever thought that if digital EQs can be matched very well, maybe ANALOG EQs can be matched with digital EQs just as good in the frequency and phase domain?
imho, they can. the prove is impulse response stuff like Q-clone and liquid mix.

If you sample an analog EQ´s freq and phase response with Q-clone and then phase cancel it with the analog EQ, the remaining signal can only be THD from the EQ´s analog circuit. How loud will those remains be? How significant will they be for the listener?

I do not think it is a "scandal" that n extremely flexible digital EQ can be set to sound like another digital EQ - if it is flexible enough to dial in the same curve (which means same phase alteration as well) why should it sound different? At the same time I am sure if this thread was about matching analog EQs with digital ones, the results would not be very far from the results we have here.
Yes, everyone says "I hear a huge difference between my hardware eqs" and it makes sense, because their flexibility is very small. the shelves for example are fixed in 99% of all cases of analog eqs - with the SSL-X-EQ you can however change from a resonating shelf to one that looks like a flat line starting at 1kHz.....no wonder a HW 1073 10kHz boost sounds different than 10kHz on another hardware EQ, the 1073 starts affecting frequencies starting from around 2kHz while others start at 5kHz and the next at 8kHz....they all have different looking slopes.

imho EQ is 95% about the frequency curve and maybe 5% (sometimes less) about other factors like THD. Most digital (maybe even all) do not emulate these 5%, but apart from this they can simply sound like any other EQ, digital or analog.

It´s not a scam, it´s quite logical to me that any EQ can be matched with an almost infinte flexible parametric eq (and those happen to exist in the digital domain, not the analog).

Rock on!
Pat

The "scam" is simply that the plugin marketeers claim they faithfully reproduce HW staples. As yet this can not be scientifically substantiated, as people above have shown. At the saturated non linear end, HW still eludes the various forms of emulation, although Nebula's volterra kernelling shows promise....

[WunderBro Flo]

Originally Posted by princeplanet

The "scam" is simply that the plugin marketeers claim they faithfully reproduce HW staples. As yet this can not be scientifically substantiated, as people above have shown. At the saturated non linear end, HW still eludes the various forms of emulation, although Nebula's volterra kernelling shows promise....

Well I think your definition of "scam" is a little strict then. I remember UA officially saying in an interview that no emulation will ever be able to replicate all aspects of a hardware 100%, so they "admit" that software will always sound different. And I agree with you that the saturation part is the biggest one missing even in the state of the art emulation plugins. However the other aspects like frequency&phase response in EQs and GR behaviour are already emulated at an extremely high level. And to be honest, those aspects are a much bigger part of what the original units are famous and useful for than the last 5% that are missing in non-standrad uses. Actually most people use the hardware in it´s linear range anyways, so given you do not use let´s say a 1073 EQ in the "spot" that saturates but does not distort ugly (yet), then the differences between the HW and the plugin(s) will be so small barely anyone will be able to tell them apart. There is a video online of I think Eddie Kramer setting up a 1073 eq and then duplicating the sound with the Waves "V" 1073 emulation, coming to the conclusion that the plugin sounds exactly the same on that source to him. So this illustrates that for most "real world" situations the good plugins (like UAD, SSL, some Waves, LM eqs etc...) come close enough to the real deal that the users do not hear any benefit from using the hardware.
So even when not every aspect is covered 100% by the plugins yet, my definition of "scam" is different.

Now it would be most interesting to see/hear if a saturated 1073 can be emulated by running a signal through a 1272 (saturating) and then through a good 1073 plugin....if I ever find the time.....

Best,
Pat

[zmix]

Saturation isn't something that you hear much of in a BA283 based Neve design (1272, 1064, 1066, 1073), and when they do get pushed into saturation the sound is pretty ugly ( listen to the guitar on U2's song Discoteque ).

I've done plenty of listening to the plugins combined with the hardware and it's still not much of a substitute for the hardware EQ alone.

This discussion reminds me of the Drummer vs Drum Machine arguments of the mid 1980s. Back then everyone thought that live drums were 'sloppy' and machines could emulate a real drummer by adding randomness to the timing. Wrong.
In much the same way, 'saturation' gets thrown around here in this discussion, but really saturation is a pretty extreme state for a circuit. The difference between the UAD plugins and the actual hardware is pretty obvious when you are attempting to get a very specific tone happening, but they may be close enough for the purposes of the end user. Note that in the low end automobile market, vinyl seat covers are more 'popular' than leather, and the owners of the vinyl seats will be the ones arguing that they 'look' and 'feel' just like leather, and never does the argument go the other way round.

[WunderBro Flo]

to me your last post shows the point:

the differences between an 80ies drummachine beat and a real drummer are so big anyone can spot them, the average listener can, and it can be done even without having the real drumrecording as a comparison.


the difference between hardware and good emulation plugins has reached a point where an experienced engineer like eddie kramer can not detect any quality difference between the two in a high quality studio environment even when AB comparing the two.

concerning the leather example.....you are right with this. but both guys arrive at their destination. once they have arrived, nobody is able to tell which one was sitting on leather and who on vinyl.

Rock!
Pat

[ericwall]

Bump!


Are all digital EQ's still the same?


Has anyone done null-tests on other plug-EQ's yet? UAD perhaps?


Let's keep this thread alive folks. If someone would care to sum things up regarding facts so far, I'd be greatful.

Sorry I ain't contributing much myself.

/EW

[rhythminmind]

Originally Posted by ericwall

Bump!


Are all digital EQ's still the same?


Has anyone done null-tests on other plug-EQ's yet? UAD perhaps?


Let's keep this thread alive folks. If someone would care to sum things up regarding facts so far, I'd be greatful.

Sorry I ain't contributing much myself.

/EW

Everything has been explained throughout the thread. Read back to answer any question you might have.
What have i learned from all the information?
Simply put i'm not going to be buying any plug-in/digital EQ's anytime soon.

[ericwall]

So people keep spending $$$ on bottled water though it's obviously nothing else than tap-water inside...

Well, most people don't seem to care anyway, wich is a mystery. Maybe the visual aspects of a cool GUI gives us the same boost as sugarpills does in certain methods of treatment. It is a fact that placebo works, and if there's a classic look on the nobs on our screens and we feel good about it... You perform better if you feel better, right?

Maybe that's what we pay for, so far.

[LPK]

Yeah, but you could still use Freeware:
sticky: latest versions / downloads « Variety Of Sound
Try BootEQ and NastyCS

[Kenton]

THE PSP sQuad EQ's seem to be something more than ordinary Bi-Quad EQ's.

Interesting results when you try and null them - even with no EQ boost or cut.....
I don't know exactly what they're doing from a technical POV but they don't sound bad to me.

[LPK]

Probably they influence the frequency-respoonse-curve even without using a band and/or you had the saturation turned on. If I remember correctly someone in the KVR Forums already analysed it ... seemed to be nothing special (like any other EQ plugin).

[zboy2854]

Well, I'm not sure whether to declare my mind blown or not yet.

I just tried an experiment myself and it almost worked. I came really, really close to nulling, but not quite.

In a 96k session I set up a pink noise generator, fed to two Aux tracks. On one I inserted the Waves SSL channel, and on the other the Sonnox EQ. I checked to make sure there were no delay compensation issues or other anomalies, which there weren't, since I was able to hit the polarity reverse button on the Waves plugin with both the Waves aux and the Sonnox aux on and the plugins active (with no EQ boosts or cuts yet), and got a perfect null. So far so good.

Then I tried boosting 15db @3.5k on the Waves channel with a Q of 1.5 (with the polarity still reversed), then with the help of a frequency analyzer and my ears, boosted a similar range with the Sonnox with the goal of getting the sound to disappear as much as possible. The closest I could get was +15.55db @4.604kHz with a Q of 1.99 using Type-3 on the Sonnox. The remaining signal difference peaked at -53.4 dbfs. Pretty impressive, if not a perfect null, and certainly when listening to each channel by itself with each plugin's respective boost, I couldn't discern any audible difference. So I guess on that count alone it was an eye opener, since there was clearly no greater "color" being applied by the SSL EQ than the Sonnox.

The question is, is it possible to get a 100% perfect null? Obviously it can be argued that the cancellation I achieved here is well below any audible difference that could be heard in the context of a mix, but then again it could also be argued that it's the very low level information differences that make a cumulative impact.

Thoughts?

[jdtrbn]

Originally Posted by Kenton

THE PSP sQuad EQ's seem to be something more than ordinary Bi-Quad EQ's.

Interesting results when you try and null them - even with no EQ boost or cut.....
I don't know exactly what they're doing from a technical POV but they don't sound bad to me.

In the manual there is something about processing. I think it's pre-saturation - biquad - post-saturation -chain, also the transformer sim affects the frequency curve at lows and highs.

[rashman]

Originally Posted by ericwall

...Maybe the visual aspects of a cool GUI gives us the same boost as sugarpills does in certain methods of treatment. It is a fact that placebo works, and if there's a classic look on the nobs on our screens and we feel good about it... You perform better if you feel better, right?

Im not with you there. Iprefer to close my eyes and let the music make me feel better. I think if plugins all start to use very simple practical (mouse friendly and VGA display friendly) look, maby in a few years hardware manufactureres start simulating plugin faces on their boxes to imply they are of top precision quality!

Originally Posted by Kenton

THE PSP sQuad EQ's seem to be something more than ordinary Bi-Quad EQ's.

I was just going to ask that...on their site:
PSP sQuad is a bundle of four high-quality equalizer plug-ins...
I mean by high quality they mean that they use oversampling, or 64bit float, or saturation, or that transformer simmulation or the Q widths? I dont mean to bash them. I respect their great work on some of the finest plugins any way...

Originally Posted by zboy2854

Well, I'm not sure whether to declare my mind blown or not yet.

I just tried an experiment myself and it almost worked. I came really, really close to nulling, but not quite.

In a 96k session I set up a pink noise generator, fed to two Aux tracks. On one I inserted the Waves SSL channel, and on the other the Sonnox EQ. I checked to make sure there were no delay compensation issues or other anomalies, which there weren't, since I was able to hit the polarity reverse button on the Waves plugin with both the Waves aux and the Sonnox aux on and the plugins active (with no EQ boosts or cuts yet), and got a perfect null. So far so good.

Then I tried boosting 15db @3.5k on the Waves channel with a Q of 1.5 (with the polarity still reversed), then with the help of a frequency analyzer and my ears, boosted a similar range with the Sonnox with the goal of getting the sound to disappear as much as possible. The closest I could get was +15.55db @4.604kHz with a Q of 1.99 using Type-3 on the Sonnox. The remaining signal difference peaked at -53.4 dbfs. Pretty impressive, if not a perfect null, and certainly when listening to each channel by itself with each plugin's respective boost, I couldn't discern any audible difference. So I guess on that count alone it was an eye opener, since there was clearly no greater "color" being applied by the SSL EQ than the Sonnox.

The question is, is it possible to get a 100% perfect null? Obviously it can be argued that the cancellation I achieved here is well below any audible difference that could be heard in the context of a mix, but then again it could also be argued that it's the very low level information differences that make a cumulative impact.

Thoughts?

I did some test too after reading this thread (including waves ssl and, sonnox, urs and others), and got simlar resaults. Most EQs nulled with Nuendos built in channel EQ! The quality of wich is questioned and disgussed by many users forever! The Nuendo Q, however didnt null in high freq boosts with others. Nowdays I mostly use the Electri-Q for its very practical interface and when I want to boost highs, I go for Q.

[roadsweeper]

In terms of pure sonics I don't think that there is any argument that different plugins have different qualities. If any of you have used the API 550B plugin you will notice a huge difference compared to, for example the SSL channel. You could of course argue that this is down to the crossover frequencies etc. that are inherent however, I have never got results from any other eq plugin that are similar to this. Then of course there are things like the Abbey Roads brilliance plug-ins which could be classed as eq's and these definitely have a character all of their own.

However, if you are talking predominantly about parametric's such as the Sonnox eq's then I would agree that there is much less difference between them.

[rashman]

Originally Posted by roadsweeper

... If any of you have used the API 550B plugin you will notice a huge difference compared to, for example the SSL channel. You could of course argue that this is down to the crossover frequencies etc. that are inherent however, I have never got results from any other eq plugin that are similar to this.

The point is all that sound and character is easily achievable using a full featured parameteric Eq. Its like we are sold some presets of a simple parameteric Eq, integrated in a sexy GUI. Or a zoomed region of a big picture which is available in full size and resolition, with ability to zoom!
The fact-for example that waves puts a button which adds a simple white niose and labels it Analog and don't bother even to add a word "niose" befor that label says a lot about what attitude the market people have developed towards their buyers, who they know well how easily fall for buttons and meters be it real or virtual.
Just thoughts. Hey any way take it easy...!

[DSpec1]

And now I feel really dumb for owning a UAD 2 DUO with a bunch of EQ plugs, The SSL Duende and TC Powercore with EQ's, not to mention the URS Strip Pro...

[rhythminmind]

Originally Posted by DSpec1

And now I feel really dumb for owning a UAD 2 DUO with a bunch of EQ plugs, The SSL Duende and TC Powercore with EQ's, not to mention the URS Strip Pro...

No need to feel dumb.. I own the Duende myself.. Great tool.. External DSP to free CPU is a great thing. Nevermind the compressors are top notch & i havent been able to clone those. Same goes for the Powercore & UAD.. The TC verb is great.. & more importantly the functionality is 2nd to none & interface is useful & inspiring. They all offer advantages that only they provide.
At the same time it's never good to be naive to what you use & how it works.

[WunderBro Flo]

Actually this thread is a little on the wrong track. It is no big revelation that almost all digital eq curves can be matched with a very versatile parametric eq. The only reason why there is no thread called "all analog eqs sound the same" is that analog eqs are not as flexible and versatile as many digital parametric eqs, they stick to "their" curves and that´s it.

Now I do not think that owning many digital EQs does not make sense anymore. Simply because of the way that engineers use EQs. We remember how specific EQs sounds like and when we come across a signal that would benefit from that character, we strap that eQ across it and apply the character. Could we just use a flexible parametric instead? Yes, in most cases, but not even that would guarantee that we dial in result that is as fast, easy and good sounding as if we use the right EQ plugin to start with. For example the UAD 1073 EQ, it sounds amazing on bassdrums. I tried to match it with a flexible eq but did not get there. I am sure it would be possible, but it would take 2-3 additional bands with small tweaks and lots of time. So reaching for the 1073 gets you faster and most often better results. Same goes for the others, the SSL EQs, the APIs, whatever. Their limitations are their strengths. They are classic curves that have been sucessfully proven useful on thousands of hit records. Sure those can be matched pretty close with a flexible eq, but really, why try hard to imitate something that you long for when you can just grab the right emulation eq it and it gets you there without trying, checking, readjusting....and ending up with something not quite as good many times.


Rock on!
Pat

[Lyonel]

Originally Posted by WildCowboys

Actually this thread is a little on the wrong track. It is no big revelation that almost all digital eq curves can be matched with a very versatile parametric eq. The only reason why there is no thread called "all analog eqs sound the same" is that analog eqs are not as flexible and versatile as many digital parametric eqs, they stick to "their" curves and that´s it.

Now I do not think that owning many digital EQs does not make sense anymore. Simply because of the way that engineers use EQs. We remember which EQ sounds like and when we come across a signal that would benefit from that character, we strap that eQ across it and apply the character. Could we just use a flexible parametric instead? Yes, in most cases, but not even that would guarantee that we dial in result that is as fast, easy and good sounding as if we use the right EQ plugin to start with. For example the UAD 1073 EQ, it sounds amazing on bassdrums. I tried to match it with a flexible eq but did not get there. I am sure it would be possible, but it would take 2-3 additional bands with small tweaks and lots of time. So reaching for the 1073 gets you faster and most often better results. Same goes for the others, the SSL EQs, the APIs, whatever. Their limitations are their strengths. They are classic curves that have been sucessfully proven useful on thousands of hit records. Sure those can be matched pretty close with a flexible eq, but really, why try hard to imitate something that you long for when you can just grab the right emulation eq it and it gets you there without trying, checking, readjusting....and ending up with something not quite as good many times.


Rock on!
Pat

Great post Pat !
the importance is the work flow.

[Firechild]

First of all , my apologies I haven´t read all the previous posts...
I assume all the arguments is based on 88.1 kHz and higher samplerates because at lower frequencies plugins with upsampling will of course sound better at high frequencies.
Anyway I did a test with the three "high end" SSL 4000 emulations outthere @ 44.1 / 24bit with pink noise and a EQ curve for "male vox" cutting low end and 6 dB top end boost.

Duende SSL channelstrip in G-mode ( becuase it is sounding better and more like the E-type emulations in Waves and UAD ) "Audio 11"
UAD 4k channelstrip ( E-type ) "Audio 13"
Waves 4000 collection ( E-type, waves G-type is not sounding really good actually) "Audio 14"

As you can see both UAD and Waves have troubles in the 16 k band.
The Duende has a an audible higher bandwith and therefore is sounding more "expensive" than both UAD and Waves, more like an "analog" EQ.
At 88.2 the results will be completely different of course but I will only point out that at 44.1 kHz there is a HUGE difference between PEQ, especially at the high end frequencies.

( The differences in the screenshot at the lower frequencies is due to the fact that the meters has more activity there , even when setting the response to slow, but at the high frequency bands there are not much activity, therefore the screenshot is showing an acurate picture regarding the top end )

[WunderBro Flo]

Firechild, thanks for doing tests but I doubt they say much. Better or worse cannot be shown with a frequency graph. If you had the Duende in G mode and the others in E mode no wonder that the graphs do not look alike by the way, they simply have different curves. Besides, the highest band of the Duende is maybe 1dB (max) different than the same band on the Duende. The Waves is rolling off earlier (maybe because it does up and downsample and therefore needs filtering before going back to 44.1 ?) but it does not say anything about quality. A signal with an ugly 16kHz region will sound best on the Waves (which measures and sounds the "worst" according to you) because it does not put the ugly 16kHz to the front as much as the others.
Anyways, just my opinion.
The thread is about something else though imho, not how good or bad certain plugins are compared to each other but if it is posible to match any eqs curve with another one that is ultimatively flexible. I think so, but I till think different eq plugin make a lot of sense...see my post above.

Best,
Pat

[rhythminmind]

Thats one of the issues.
The Q & Gain numerical value settings are a developer decision. You can’t compare EQ’s by matching dials & numbers. You have to match curves.
I wish AES would finalize that standard.

[Firechild]

I had to boost a few more dB:s on the Waves to reach a similiar curve but the top frequencies are still rolling off. if you press "bell" on the Waves E channel, a few dB.s more can be achieved.
I still think there are differences at 44.1 kHz between plugins due to the difficulties in calculating the highest frequencies when not using upsampling technologies first introduced in the old Q-metric from Steinberg in the late 90:s.

[TheMarqueeYears]

Many years ago I found I could match the sound of my Millennia SST-1 EQ with my Sonnox EQ.

I would get my friend to double blind test me and I and many others failed to pick the SST-1 over the Sonnox. I used the STT-1 as you can set the channel very clean with no input xfmr and all solid state.

I was amazed at software = hardware so perfectly.

I have therefore concluded that digital EQ properly desinged IS very high quality. However I have still gone on to buy UAD's Neve EQ's.

Why? Well as others have said, for the work flow - I'm paying for the GUI's interaction with the software. The curves and the gain dependencies have been mapped for me by someone else. They've done the work and I've payed for that service.

I suppose this thread does point towards someone - Paul Frindle comes to mind - creating the ultimate "can do all EQ" that has on the back end a perfect parametric EQ and on the front end dozens of GUI's and curves and distortion mappings that can emulate ALL hardware models.

BUT - it's commercial sucide to give the whole farm away in one deal.

So either make your own mappings and settings combining saturation plugins and a sprectragraph, or just pay someone else to do it, and don't begrudge them giving you a little slice of the cake at a time - never begrudge someome trying to make a living.

UAD,WAVES, DUENDE are just saying .... "and here's the curves/mappings/saturation/ GUI for an SSL.

UAD,WAVES, DUENDE are just saying .... "and here's the curves/mappings/saturation/ GUI for a NEVE

UAD,WAVES, DUENDE are just saying .... "and here's the curves/mappings/saturation/ GUI for an API

etc etc.

I don't think anyone said it was rocket science - it's just very convienent.
Convienence cost money!

Now compressors are TOTALLY different, IMHO harware always sinks plugins.

My Roll RMS755 is so different and in a different leauge to a SSL bus compressor plugin. It's sick.

Fantastic thread though -

Rob

[Iggy Poop]

Excellent thread. I think it it all boils down to the same thing. What sounds best. I use Sonalksis SV-517. They sound better than everything else I've used but I'm sure its just the way they designed them to work. Ultimately I could probably achieve the same desired settings on alot of other plugins, just not as easily. I think that Sonalksis makes it easy to find the best settings.

[JP11]

Originally Posted by TheMarqueeYears

UAD,WAVES, DUENDE are just saying .... "and here's the curves/mappings/saturation/ GUI for an SSL.

UAD,WAVES, DUENDE are just saying .... "and here's the curves/mappings/saturation/ GUI for a NEVE

UAD,WAVES, DUENDE are just saying .... "and here's the curves/mappings/saturation/ GUI for an API

etc etc.

I don't think anyone said it was rocket science - it's just very convienent.
Convienence cost money!

If only these companies said so in the first place, as opposed to saying virtually nothing about what they're plugs are made of...could've saved a lot of people a lot of inconvenience!!

Of course, maybe convenience doesn't sound like it's worth all that much compared to the mythical nature of classic hardware emulation. I'm not sure someone would pay much for a hardware Neve EQ if it was just a mock up with a cheap Behringer EQ inside....would they?

[piano]

Wait ...... so a Rupert Neve hard ware EQ is not going to really improve my sound all that much since I already have API and Waves EQ .... and a few more?

Hardware EQ is a waste of money now if you have the digital versions?

[piano]

Anyone?

Is hardware EQ a waste of money considering that plug ins can now duplicate it?

[rhythminmind]

Originally Posted by piano

Anyone?

Is hardware EQ a waste of money considering that plug ins can now duplicate it?

You have it completely backwards. Analog hardware is by no means a waste of money. It's the plethora of plugin/digital EQ's that are the potential waste of money. There are many issues at hand.
Did you read the post? I recommend going over this thread to get a better understanding from all sides. Read the compiled version here. Digital EQ Fact & Myth. - signaltonoize.com