Skip to main content

How to get rid of frequency level up and down ?

Hello masters,

my problem is that when i see the frequencies of a professionally mixed and mastered studio recording in a frequency-anaylzer, the curve looks like this :
and it doesn't change most of the time, it always has this shape, during the whole song.
Now look at my curves :
this curves are from one song, but the frequencies are going up and down a lot and this is what i want to ged rid of. I want to have a basic shape of the frequency curve, that doesn't change a lot during the whole song.
How can i achieve that ?


hueseph Fri, 03/09/2007 - 11:27
I have one question. Why?

Why are you so concerned with the shape of a frequency analysis? Honestly, I would be more concerned if the frequency curve is static. It's likely that the "ideal" curves that you are looking at are that way because the song all compressed to hell and so all of the dynamics have been sucked out of the song.

Stop looking at data. Stop comparing with your eyes. Audio is about sound not diagrams and frequency analysis. If it sounds good, it is good. If it sounds bad it's bad.

Pro Audio Guest Fri, 03/09/2007 - 13:49
I tend to agree with hueseph, why? That being said, it is hard to believe that the curve you're after remains the same through an entire song as you said. I once did a project with someone who brought in a WAV file that looked as flat as Hitler's moustache. I was asked to match that WAV during mixdown. The WAV was that of a mastered cd by a popular artist. Every single drop of dynamic range had been sucked out of it for the sake of perceived volume. That has become the trend now-a-days. Everyone seems to want their finished product to be equal to, or louder than the cd the end user just listened to. Though this example is different from your problem, it still proves the point about listening be the most important element. If we all looked at the Mona Lisa with a blindfold on, could we judge the painting by what it sounded like? Just like the Hitler WAV, if my client would never have been able to visualize it he would only have the sound to go by. It was so obvious that the song had no dynamic range whatsoever, but he was blinded by his own ability to see!

juanito2000 Fri, 03/09/2007 - 14:09
Yeah you might be right guys. What I want is not loudness, not at all, I don't care about the loudness, because I am speaking about acoustic music, so it doesn't have to be modern sounding. But the thing is my mix sounds not right for me and with my ears I can't achieve the sound I am searching for, because if I think I have to boost the bass a little bit, suddenly the high frequency range needs to be boosted as well, then suddenly the mid-range is to low and at the end I have thousands of peaks and holes in my frequency spectrum. And this is also what the analyzer shows. When I see the frequency curve of a professional mix it sound very perfect, all is smooth and no frequency-range is dominating. That is what the analyzer shows,too. So I use the analyzer to "see" where the differences are between my mix and another. What I mean is, if the analyzer shows a smooth and equal frequency curve it also sounds like that and if it shows many peaks and valleys, it sounds like a bad mix. So what is wrong with using the analyzer for this ???
I also tried to use multiband and normal compressors as well as eq's but the up and down of the single frequencies (500hz,600hz,700hz etc.) doesn't change in the analyzer.

hueseph Fri, 03/09/2007 - 14:22
When you start using eq with a compressor, realize that if the eq is before the compressor it will affect how the compressor works. Try backing off on the compressor, raising the threshold or removing it all together.

What I think is happening is that you are boosting levels which now peak and trigger the compressor. So, although you want to boost the level of those frequencies, the compressor is actually attenuating them and so all the other frequencies are getting attenuated as well. Hence the perceived need to boost the highs, then the mids and it just gets worse.

Ditch the multiband compressor. Unless you know exactly what you intend to do with it, all it's going to do is cause you headaches.

Pro Audio Guest Fri, 03/09/2007 - 15:29
I completely agree with hueseph. Forget about the multiband compressor. Also, try to always use subtractive eq before you go adding any. The 500Hz to 700Hz range is also a very tricky range. Most people refer to it as the "honk zone". I surely wouldn't worry about this area showing up much on the analyzer. Being that you are recording acoustic music, I would like to know more about your signal chain. Of course, the kind of mic you're using is the most important element. Have you tried various mic's and placements? Give us the lowdown on the instruments you're trying to record, the mic's you're using, pre's, etc.. Often times eq problems exist because the freq's you're trying to chase down don't exist with any viable abundance in the first place.

juanito2000 Sat, 03/10/2007 - 03:57
I am using a Rode NT-2A with a DMP3 of M-Audio. Soundcard is the Delta 1010LT from M-Audio. I am recording my nylonstring-guitar at about 10-15 inches away from the guitar, with the mic pointing towards the area between neck and soundhole. I really like the microphone, so it shouldn't be the reason for the peaks and valleys in the frequency curve. I think it is the room as well as the natural sound of the nylonstring-guitar that produces the peaks and valleys. So the neutral recorded sound of the guitar (without any EQ, Compression etc.) is ok but not perfect and also always doesn't sound near to any professional guitar recording you can hear on professional cd's. That should proof that the guitarsound you hear on a professional cd is always processed a lot with eq etc. So I usually do that as well. But I don't get rid of the peaks an valleys. For example I would have a piece where during the whole song the 1khz frequency range (very narrow) is always at -40 db (that would be what the analyzer shows) but suddenly a few times during the piece this 1khz rises to -20 db or something like that. You know what I mean ? I don't want narrow frequency ranges to vary so much, it is no problem if the range between 100hz and 500hz or 800hz and 3 khz varies sometimes, because these are broader ranges, but in my recordings every single narrow frequency varies up and down all the time.

hueseph Sat, 03/10/2007 - 08:49
You have to stop thinking like a technician. Get rid of that spectral analyzer. It's obviously an obstacle for you. Just listen to your track. Is there really something missing? Use a parametric eq and boost the gain a bit then sweep the band and listen for any offensive or complementing sound. Is there something that could be boosted? Is there something that could be taken away? forget how the analysis looks. Listen adjust and listen again. You can't listen with your eyes. If you don't learn to use your ears, you'll never get better at recording.

ghellquist Sat, 03/10/2007 - 09:37
Hmm. I think you are on the wrong track. The peaks and valleys in the spectrum are there for a reason. The reason is that we are talking music, not white noise.

Try this. Play only one tone for a long while. Look at the peaks. There is mathematical relationships between the peaks. The fundamental tones is a the bottom. The over-tones are at two times the fundamental tone, at three times, at four times and so on. As an example take the open low string A. It has a fundamental frequency of 110Hz in standard tuning. The overtones will be at 220, 330, 440, 550,660,770,880 Hertz and so one. The higher tones are getting weaker and weaker.

So simply stop thinking about trying to make the frequency spectrum flat with only one guitar.


Pro Audio Guest Sat, 03/10/2007 - 14:09
Each instrument has its own frequency range. You can't expect a single instrument to display it's essential freq's across the entire frequency spectrum. On the other hand, if you are trying to emulate a mastered commercial cd recording of a single acoustic guitar, it is curious as to why it would present across the freq spectrum the way you say it does. Though I hate to play the spectrum analysis game, it's difficult to believe that something else isn't going on that's causing the readings that you're getting. Also, keep in mind that a well mastered recording typically enhances that recording using mastering tools that are well beyond the scope of this thread. But, there is obviously something that you have overlooked. That being said, get away from the technical mumbo-jumbo and try to make your recording sound as close as possible to what you desire by simply using good technique. Afterall, what we are really talking about here is an effort to fix something in the mix. Not to imply anything about you personally or your recording, the old adage about trying to polish a "turd" is definitely at play here. After reading your equipment post, I can almost guarantee that the primary cause of your problem is the Rode NT2A. Years ago I tried to record an acoustic guitar track with the original NT2. I tried it on a Gibson, Ovation, and a vintage Martin acoustic. All it produced was low-mid sludge and boominess. The NT2A is still a large diaphram condenser microphone. It's not capable of producing the sound that you're after, especially with nylon strings. As I said in an earlier post, you need a solid quality small diaphram condenser. In fact, two would be ideal. For years, the go to mic for recording acoustic guitars in major studios was the SM81. It's still not a bad place to start. Another option that is less expensive would be a pair of Rode NT5's. Many people hate these mic's, but I have heard excellent results using them on acoustic guitars. Once again, you'll never be able to eq out the bad elements of the NT2A, and eq in the elements that the NT2A doesn't provide in the first place.

Pro Audio Guest Sat, 03/10/2007 - 14:27
A quick thought on what ghellquist said. You must also consider the "zero gain" aspect when analyzing your freq response. There are so many different variables that can affect the way that your analyzer reacts including the analyzer itself. When you play an entire song on your acoustic guitar, can you definitively say that every time you play a note on the D string that it is at the same volume or timbre as the previous time? If you were capable of taking a snapshot of the freq analysis at the exact same time that a certain repetitive movement of the song ocurred, those snapshots would still not be identical.

Pro Audio Guest Tue, 03/13/2007 - 00:39
I believe there is an important point here. The frequency analyzer has a very impoortant function. It can show you frequencies that are beyond the human listening range. U can use a notch filter to tame out this to avoid frequencies that are just there to drain resources and have your music play freely to 'human beings ears' if someone needs to correct me on somethig here please feel free, im a learner FYI

Pro Audio Guest Sat, 03/17/2007 - 13:54
I do not see the problem with assessing music visually as a cross reference for your ears in a synesthesic manner it can be quite useful. Having a tecnical aspect to your approach to music does not preclude a creative one either. If something sounds wrong to your ears and doesnt look right either when compared to similar proffesional recordings then theres probably a reason for that and tracking it down as you are seems to be the right path. However are the proffesional analysis graphs you are comparing to the same instrumental line up and style/octave range as your music ?
I know a lot of engineers split the song into 4 frequencies and then widen and boost each one to increase the volume etc . It is true that the loudness competition is crude and loses dynamic range but there are reasons proffesional studios are doing this and there is no reason you shouldnt do the same if you want to achieve the same results as proffesional studio. Hweve rit would not seem to appropriate for purely acoustic guitar music.

Michael Fossenkemper Sun, 03/18/2007 - 07:59
most mastering places don't use a multiband and widen each band. please stop saying this cause it's not true.

Also, there is very little need to use a spectrum analyzer during mastering. I have a couple, I only use them if I'm testing a piece of gear. I use other analyzers if I'm setting up a piece of gear. But after that's done, I don't refer to them while mastering. There are so many variables involved with a recording that it's going to be useless to visually reference it to something else. Also if your room is setup properly and you know what you are listening for, then a spectrum analyzer isn't going to tell you anything that you can't hear.

drumist69 Sun, 03/18/2007 - 14:08
Three full length CD's I've played on (none of them went anywhere, so FWIW...), and an EP just sent out for mastering, and all they wanted was a stereo track (choose your format, these days stereo WAV files seem to be the norm). I think maybe what you're talking about is sending the mastering lab stem mixes? Some places will take stem mixes, say one mix of drums, one of bass, one of guitars and one of vocals. That way they have more control if there was a mix/eq issue they need to fix. Personally, I'd rather send a stereo mix, and have the mastering lab tell me if I need to fix something in the mix and redo it. That way I might learn something instead of him/her just charging me more money to attempt to fix it for me! Just my way of thinking and my understanding of the process. ANDY

Michael Fossenkemper Sun, 03/18/2007 - 17:39
IMO, this is the correct way to deliver a mix to a mastering house. Deliver a stereo mix that everyone is happy with. Develope a relationship with the ME so if there is something that can be done to the mix that will improve it, have the ME tell you. Correct the mix and send them another. Sure it's a bit of a hassle the first time around, but I guarantee that EVERYONE learns how to improve a mix and the product and subsequent mixes turn out better in the end.