Learning to EQ with purpose - response curves help?

[Ultfris101]

I'm trying to get better at applying EQ in post processing. My goal is not so much improving a recording (which can mean a lot of subjective things) but essentially getting as close to what being there sounded like and making the recording sound natural. If it gets "better" so be it, but I'm trying to stay in a somewhat objective realm.

So I've done a little experimenting with a recent recording and I'm starting to see an inverse correlation between what I'm doing in a parametric EQ interface in Samplitude and the frequency plot. By this I see that places where the frequency plot for the particular mic drops below 0 I am wanting to bump up a little in EQ and vice versa.

I'm probably over simplifying as I try to grasp some basic concepts, but assuming you're not recording in strange conditions to begin with and thus compensating for a lot of low level reverb or something does it make sense to shoot for a flatter frequency response? Meaning right off the bat if the recording doesn't happen to sound right do you start by correcting for the known variances for the mics you used?

Is this coincidental or is this a reasonable strategy to begin with? And is it what people who EQ keep in mind when they are processing recordings made with mics they are familiar with?

[Sloan Simpson]

In a very general sense, yes, the "hump" in the frequency response graph is usually what I aim to flatten out a bit. Of course this should be something that your ears tell you is improving when you change it, but the visual can definitely give you a clue of where to start.

It can easily be overdone however: if you keep going until the response looks perfectly flat, it won't sound good. Less is more and all that 

[page]

My goal is not so much improving a recording (which can mean a lot of subjective things) but essentially getting as close to what being there sounded like

A) I suspect you're working from memory then, as each mic will color the sound and it seems like you're trying to compensate for that.
B) I'm assuming that this is with the goal of creating a document more than something just to listen to?

[ScoobieKW]

One of my favorite tools in Reaper is RealFIR

It's default mode is EQ, with a large frequency graph and an overlaid eq curve. click to add a point, drag to move it.
you can see the original response as well as the eq'ed response.

Easy-peasy

[2manyrocks]

http://www.bluecataudio.com/Products/Product_FreqAnalyst/

Free plug in you can use to plot your favorite commercial recordings to visualize the end result of their EQ decisions.  You can use it to visually compare to your own EQ choices. 

[Ultfris101]

My goal is not so much improving a recording (which can mean a lot of subjective things) but essentially getting as close to what being there sounded like

A) I suspect you're working from memory then, as each mic will color the sound and it seems like you're trying to compensate for that.
B) I'm assuming that this is with the goal of creating a document more than something just to listen to?

Sort of. It's from memory but the point is more that I'm correcting for coloration the mics might have added and to do that, whether I ultimately like the sound or not one place to start is to compensate for what the frequency plot suggests.

So yeah, I guess I'm coming from the standpoint of capturing the event at least as a starting point.

Maybe this is kind of an obvious concept, but I hadn't thought much about it before. I was noticing in this recording that the guitar didn't sound right and I do know that electric guitar is in the mids so I started bringing it down a little and found a place around 500 Hz that I liked it brought down a few db. I looked at the frequency plat and there's a bit of a bump up around that frequency. Similarly I liked the way it was sounding (and was more closely matching an omni source) with some bump up around 5000Hz which is also suggested by the frequency plot.

I guess I like the fact that I can use both my ears and another more objective point of reference to make me feel like I'm on the right track.


2manyrocks,

I like the idea of that plugin. That's exactly the kind of thing I'm interested in.

I learn best by trying to recreate things I've seen others do and then go off on my own once I get an understanding of the basics.

[Gutbucket]

To get a bit philosophical first..

My goal is not so much improving a recording (which can mean a lot of subjective things) but essentially getting as close to what being there sounded like and making the recording sound natural. If it gets "better" so be it, but I'm trying to stay in a somewhat objective realm.

Stereo is subjective illusion.  Getting a recording sounding closer to what it sounded like to you when you were there for the live performance is making it subjectively better.

There are plenty of good reasons not to EQ or do any post processing, but when people claim their reasons for not doing any of that are because they want to faithfully and accurately record what it sounded like when they were there, I claim they are simply deluding themselves.  I don't mean to be snarky, and I realize that is not at all the position you are taking, actually asking for sort of the opposite.

For accuracy in an objective measurable sense, record using a single calibrated flat measurement mic referenced to an SPL meter.  If the goal is recreating the experience of being there, that's a subjective thing, often moving in the opposite direction from objective measureable accuracy.


Some more useful but non-specific answers:
Play around with simple graphic EQs. I think simple graphic EQs are better for this self-education than more complicated parametric EQs.  Move to those later when you've learned to hear what you want to do better. Boost and cut each band enough on it's own to get an ear for what each frequency band sounds like.  Go up and down the spectrum making broad sweeping smooth corrective curves, and listen how even minor changes in the general shape of broad curves are very audible (to start throwing some terminology in, broad curves are low 'Q', single slider bands are high 'Q'). Listen for how making small tweaks in a single band within a broad curve changes is also very audible but in a different and more subtle way, often more so as things get closer and closer to being subjectively 'right'.  Revisit different frequency regions iteratively.  Regularly bypass the EQ to remind your brain what that sounds like.

All that is sort of like practicing and instrument, after a while you'll hear something and just recognize what frequency range needs correction, have a general idea of the shape of the curve, how broad it needs to be, how sharp and narrow, etc.  You'll still need to listen, adjust; listen, adjust again; listen, adjust something else; listen, re-adjust the first thing again, and over again..  but you'll begin to home in on things quicker and quicker, and get things were you are happy with them faster and easier.

When actually working on EQing music for release you'll need to make sure your playback system is well balanced in a timbral sense, but that doesn't really mater for this kind of experiential ear training stuff.

Spend a bunch of time playing around with adjusting your recordings or other tapers recordings until they just sound better and more 'right' to you.  Then compare with the EQ switched out.  Write down the settings or save the curve, then do it again another day and see how different and how similar the curves you end up with are.  Notice how you'll tend to EQ certain microphones or recordings made in specific venues, or different types of music differently.  Always suspect your choices and second guess them.  Notice when you just can't seem to get the timbre right for some recordings not matter what you try, and how some microphones seem to be able to withstand much more EQ adjustment than others.  It might be something other than your EQing ability or anything you could see on an RTA display that is the problem.

Comparision with good sounding, well recorded material which is similar is useful, both by ear and by eye is a powerful learnign tool.

Beware of the visual trap.  Displays can be useful tools, but can also just as often mislead you or hide things.  They are sort of learning a whole different angle on it, just as complex as listening.  Your ears and brain will mislead you too, and should also be suspect, but are generally more reliable in assessing sound than eyes, and are the ultimate target audience.

[Sloan Simpson]

Lately I've been using the Waves Scheps* 73 EQ plug as a starting point for broader "shaping" EQ. It's a 3-band with selectable center frequency for the Mid and Low controls, and the High is fixed at 12KHz.  No Q/width to mess with. In retrospect I wish I had started off with something like this rather than a full parametric, as the parametric can be a lot easier to go drastically wrong with.

If I need to do a "surgical" correction after I've shaped with this one, then I'll use a parametric with a narrow Q in a subsequent step.


* nothing to do with Schoeps, it's named for the engineer whose board they modeled this from

[macdaddy]

Marking thread...

[bombdiggity]

A few thoughts on some aspects of this: 

I'd not necessarily objectify what it sounded like there (unless that somehow was great).  That "there" may have been relatively better or worse and as we know varies greatly depending on where you are in a room.  If the ambient sound involved a PA you're actually starting with someone else's EQ (I doubt many sound engineers have ever run all the EQ pots flat on their mixers).  Likewise with instrument amps, effects, tone controls on pickups, and so on.  Presumably the musicians know what they want their stuff to sound like but that's not nearly a safe assumption when it comes to house engineers, especially ones that don't know the music well.  A garden variety example of the original premise are the many people on the trackers who think and say a rumbly, muddy, reverberant mess sounds "great" (because that's what concerts always sound like to them in the back of a giant arena). 

IMO EQ is actually most critical on a mixer in an amplified setting (before the house sound is recorded).  A few weeks ago I wound up running a pop up sort of stage at a fest we go to.  The musicians would arrive, plug in, and we'd roll.  I'd do a quick EQ adjustment on the elements I could based on what I was hearing when they started.  Though only three knobs (hi/mid/low) my friend who was helping instantly remarked (without being asked) "that sounds a lot better" after I tweaked to what I thought the musicians would want to sound like (if they were in the crowd rather than on stage).  With vocals and such there's no one setting fits all that works out of the gate, nor would I say most things sound best flat (no eq).  With more effects the value (and need) for EQ diminishes. 

Your ears should be your guide.  In that sort of live setting it's easy to hear the difference and with a limited set of options in some ways very easy to know when you've improved any channel of it. 

With 30 bands or parametric in post it's easy to get lost in the weeds.  When it is needed I favor as gentle as possible EQ to deal with tonal matters that approach is appropriate for. 

While tone can be adjusted to at least some extent, instrument balance is even more important in the mix of what is being recorded.  That can almost never be fixed in an ambient recording since the ranges of nearly all (or all) the elements overlap. 

I think the "book learning" engineering principle is also generally to remove what is overly present rather than try to add what isn't there (which is sort of a futile endeavor when framed that way). 

I also tend to favor filters (rather than EQ) to deal with the real problems (where there is a clear spike of something unpleasant that should be reduced).  An FFT filter can really be targeted and work wonders on very specific sorts of issues. 

Having watched many sound checks I can for certain say that many musicians know exactly what frequencies need to be tuned.  The pros go through and call out to the engineer what frequencies to adjust.  Those who record music rather than play it rarely have so finely developed an ear, but the musicians learned it by experience (whether or not they are more sensitive to that skill to start with), so almost anyone else can too. 

In terms of the flatter lines element, what generally sounds best is on some sort of a diagonal starting from lows and declining toward the highs (in a linear FA view).  Most recordings naturally look like that.  The peaks and valleys are what give them character.  Some of those peaks and valleys are of a pleasing character, others may not be. 

Often a spectral display is of more analytical value than a linear frequency analysis view.  There are lots of great sounding recordings but the pictures of them can look considerably different from each other. 

[bryonsos]

There are plenty of good reasons not to EQ or do any post processing, but when people claim their reasons for not doing any of that are because they want to faithfully and accurately record what it sounded like when they were there, I claim they are simply deluding themselves.  I don't mean to be snarky, and I realize that is not at all the position you are taking, actually asking for sort of the opposite.

I agree and disagree with this. I never EQ my recordings before I share them (or at all really) because I have no idea what the downloader is listening with. I have nice Klipsch speakers, and reasonably good 7506 cans, and I've ended up with gear that makes tapes that sound pretty good through them. If the downloader is listening with soup cans connected with piano wire, I can't pretend to think my tapes sound good through them, or that any EQ I will do will make it sound right or good. Better to just release them into the wild and let the listener decide. Here's some gratuitous self promotion, Phish/Kevin Shapiro only did some mild EQ on this before they released it into the wild:

http://phish.com/tours/dates/sat-1992-03-28-variety-playhouse/

[Ultfris101]

To get a bit philosophical first..

My goal is not so much improving a recording (which can mean a lot of subjective things) but essentially getting as close to what being there sounded like and making the recording sound natural. If it gets "better" so be it, but I'm trying to stay in a somewhat objective realm.

Stereo is subjective illusion.  Getting a recording sounding closer to what it sounded like to you when you were there for the live performance is making it subjectively better.

There are plenty of good reasons not to EQ or do any post processing, but when people claim their reasons for not doing any of that are because they want to faithfully and accurately record what it sounded like when they were there, I claim they are simply deluding themselves.  I don't mean to be snarky, and I realize that is not at all the position you are taking, actually asking for sort of the opposite.

For accuracy in an objective measurable sense, record using a single calibrated flat measurement mic referenced to an SPL meter.  If the goal is recreating the experience of being there, that's a subjective thing, often moving in the opposite direction from objective measureable accuracy.

I get your point, but you are kind of making my point (or the point I was trying to make), if I don't have a calibrated, flat, measurement mic but a less expensive mic which is not flat, could I get closer to what the measurement mic would have picked up by starting with an EQ setting which compensates for deviations in the particular mic's frequency plot?

And yes, if I'm needing to do this at all then it must be because I didn't like the raw result and somehow want to improve it so I'm creating a bit of a paradox.

I have several top shelf mics now and they yield a faithful representation of what the venue and board master yielded. I am also experimenting with some less expensive (not "cheap") mics and they have a less flat frequency plot to the point that in a couple frequency ranges even my novice ear can spot it but a little bit of EQ seems to have "fixed" it. And by fixed I mean made it sound a bit more like it did for the audience and also what I hear when I compare to the DPA omnis that were sharing a stand. That's my real reference point since I don't trust my memory.

I plan to post a rig pic thread and I'll include some before and after samples once I get some more of the mixing done.

Thanks for all the input from everybody. This has been a very helpful discussion both practically and philosophically.

[Sloan Simpson]

There are plenty of good reasons not to EQ or do any post processing, but when people claim their reasons for not doing any of that are because they want to faithfully and accurately record what it sounded like when they were there, I claim they are simply deluding themselves.  I don't mean to be snarky, and I realize that is not at all the position you are taking, actually asking for sort of the opposite.

I agree and disagree with this. I never EQ my recordings before I share them (or at all really) because I have no idea what the downloader is listening with. I have nice Klipsch speakers, and reasonably good 7506 cans, and I've ended up with gear that makes tapes that sound pretty good through them. If the downloader is listening with soup cans connected with piano wire, I can't pretend to think my tapes sound good through them, or that any EQ I will do will make it sound right or good. Better to just release them into the wild and let the listener decide. Here's some gratuitous self promotion, Phish/Kevin Shapiro only did some mild EQ on this before they released it into the wild:

http://phish.com/tours/dates/sat-1992-03-28-variety-playhouse/

But wait  how did Phish know what people would listen on? 

Mild EQ is great! Nobody is suggesting radical moves. Perhaps it's my gear but it's exceedingly rare that I make a recording that isn't improvable by a gentle EQ move or two.

[bryonsos]

There are plenty of good reasons not to EQ or do any post processing, but when people claim their reasons for not doing any of that are because they want to faithfully and accurately record what it sounded like when they were there, I claim they are simply deluding themselves.  I don't mean to be snarky, and I realize that is not at all the position you are taking, actually asking for sort of the opposite.

I agree and disagree with this. I never EQ my recordings before I share them (or at all really) because I have no idea what the downloader is listening with. I have nice Klipsch speakers, and reasonably good 7506 cans, and I've ended up with gear that makes tapes that sound pretty good through them. If the downloader is listening with soup cans connected with piano wire, I can't pretend to think my tapes sound good through them, or that any EQ I will do will make it sound right or good. Better to just release them into the wild and let the listener decide. Here's some gratuitous self promotion, Phish/Kevin Shapiro only did some mild EQ on this before they released it into the wild:

http://phish.com/tours/dates/sat-1992-03-28-variety-playhouse/

But wait  how did Phish know what people would listen on? 

Mild EQ is great! Nobody is suggesting radical moves. Perhaps it's my gear but it's exceedingly rare that I make a recording that isn't improvable by a gentle EQ move or two.

Likewise. I admit it, I'm lazy.

Being an omni addict, I find I need to turn off the "bass boost" on my stereo to make my tapes sound good at home. On the other hand, I seldomnly listen to my own tapes, I listen to y'all's.

[page]

There are plenty of good reasons not to EQ or do any post processing, but when people claim their reasons for not doing any of that are because they want to faithfully and accurately record what it sounded like when they were there, I claim they are simply deluding themselves.  I don't mean to be snarky, and I realize that is not at all the position you are taking, actually asking for sort of the opposite.

For accuracy in an objective measurable sense, record using a single calibrated flat measurement mic referenced to an SPL meter.  If the goal is recreating the experience of being there, that's a subjective thing, often moving in the opposite direction from objective measureable accuracy.

This. I don't intend to be snarky or demeaning about it when I say that I see the reason most people choose gear is because they like how it sounds (not how faithful it is), but they don't want to EQ because they don't have the necessary skillset or knowledge to do it well instead of trying to be faithful. Sometimes we don't really understand our objective, or we respond to different cues and go in a direction without knowing it, or a plethora of other reasons. I work in education, an industry fundamentally built around eliminating ignorance (which is just not knowing something), so I approach this as a teachable task and not something to be embarrassed about. I came from an analytical background, and it took me a couple years to really grapple with the fact that audio creation (and it's various steps) is an art, not a science. There are scientific elements, but those elements don't play the major part in decision making. I'm not going to say that doing "documents" for live events is futile, but I think it takes a careful approach and an honest assessment as to what you want to keep and what is better to embellish given a value system (professed openly or subconsciously hidden) in people who will listen to it later.

Play around with simple graphic EQs. I think simple graphic EQs are better for this self-education than more complicated parametric EQs. Regularly bypass the EQ to remind your brain what that sounds like.

All that is sort of like practicing and instrument, after a while you'll hear something and just recognize what frequency range needs correction, have a general idea of the shape of the curve, how broad it needs to be, how sharp and narrow, etc.  You'll still need to listen, adjust; listen, adjust again; listen, adjust something else; listen, re-adjust the first thing again, and over again..  but you'll begin to home in on things quicker and quicker, and get things were you are happy with them faster and easier.

Spend a bunch of time playing around with adjusting your recordings or other tapers recordings until they just sound better and more 'right' to you.  Then compare with the EQ switched out.  Write down the settings or save the curve, then do it again another day and see how different and how similar the curves you end up with are.  Notice how you'll tend to EQ certain microphones or recordings made in specific venues, or different types of music differently.  Always suspect your choices and second guess them.  Notice when you just can't seem to get the timbre right for some recordings not matter what you try, and how some microphones seem to be able to withstand much more EQ adjustment than others.  It might be something other than your EQing ability or anything you could see on an RTA display that is the problem.

Beware of the visual trap.  Displays can be useful tools, but can also just as often mislead you or hide things. 

If I need to do a "surgical" correction after I've shaped with this one, then I'll use a parametric with a narrow Q in a subsequent step.

While tone can be adjusted to at least some extent, instrument balance is even more important in the mix of what is being recorded.  That can almost never be fixed in an ambient recording since the ranges of nearly all (or all) the elements overlap. 

Those who record music rather than play it rarely have so finely developed an ear, but the musicians learned it by experience (whether or not they are more sensitive to that skill to start with), so almost anyone else can too. 

In terms of the flatter lines element, what generally sounds best is on some sort of a diagonal starting from lows and declining toward the highs (in a linear FA view).  Most recordings naturally look like that.  The peaks and valleys are what give them character.  Some of those peaks and valleys are of a pleasing character, others may not be. 

Mild EQ is great! Nobody is suggesting radical moves. Perhaps it's my gear but it's exceedingly rare that I make a recording that isn't improvable by a gentle EQ move or two.

This is solid advice (the entire post is actually, I can't disagree with it. The above is just bits that need to be reinforced). Sometimes I just wait and let Lee et al do my work for me. 

Specifically about decreasing volume in frequencies that bombdiggity notes, i've seen the rule of thumb that you're sub 2khz. should be relatively flat, and above that it trails off. I think for room recordings, it's best to drop that to around 1.3khz and then trail off. My rationale is that low frequency reverb sounds more spacious and pleasing compared to high frequency reverb which I find I want "real" details in, not reverberation.

In general, it's a skillset as I mentioned earlier. Getting good at it takes time and a methodical attention to what you're doing. Some tips on getting started I think would be to pick a couple recordings, and listen to them on 3 or 4 radically different setups (your home speakers, your best headphones, your car playback with the engine off, office speakers, etc). What stands out? Make notes as Lee says. Now learn how frequencies sound. I used to take a low Q value (that gradually grew the higher up I went) and amp something by 10DB and do a gentle sweep so I could hear what 400hz sounds like, what 1k, 2k, 2.6k, etc. Learn these. Now listen to the recordings in those environments again and make more notes using your new knowledge of frequencies. Over time things will get better and the various environments will show you something different compared to each other. I have an excellent set of headphones (in terms of flat tonal response and transient response) and a plugin to simulate speakers. Thats well and good for the workhorse of adjustments once I got a feeling for it, but it doesn't replace those other environments. Once you do this enough, it becomes easier to anticipate what you'll hear in other environments. Thats mixing with anticipation as Sloan points out.

It's not easy, or fast (took me a solid year or two to really get as far as I am now) but it's very rewarding to get good at it. I'm going to whip out one of Lee's favorite quotes:

"At the core is the illusion vs document debate. I look at it more as an "effort/ROI" question." —page

How strict of a document do you want? Who are you trying to fool into thinking it's a document? (and will they care) How much do they need to be fooled? If you do the job well enough, they may not even care (or know) that they aren't hearing "reality" so much as a view. You can create a view that sounds like reality (and that sort of sounds like what you want), but if you chuck that requirement out the window, you have more freedom to create something even better.

I get your point, but you are kind of making my point (or the point I was trying to make), if I don't have a calibrated, flat, measurement mic but a less expensive mic which is not flat, could I get closer to what the measurement mic would have picked up by starting with an EQ setting which compensates for deviations in the particular mic's frequency plot?

You can on the assumption that you have a mic which has a consistant polar pattern through the entire frequency range (so that eliminates LD mics and a ton of SD mics), OR you are EQing to return just one segment of the experience back to "flat" (the PA, the room, etc). In my experience, in beautiful rooms, it's less of an issue, in crap rooms though, the off-axis of your polar pattern will exacerbate this decision. It's inherently a trade off.

And yes, if I'm needing to do this at all then it must be because I didn't like the raw result and somehow want to improve it so I'm creating a bit of a paradox.

correct.

Phish

completely off topic; the Gin from Randalls night one is a sick jam. I forgot what I was typing a couple of times during it. And thats my post for the day.