The Song Fight Bulletin Board

irwin wrote:Jast brings up an excellent point. The spectral analysis tools are great for helping you develop your ear, but the ultimate goal should be to learn to trust your ears. You don't listen to music with your eyes, after all.
.

Yes - this has all got me curious, and when I get the chance I will investigate more. I really don't trust my ears at all, so it will be interesting to see if my eyes can help them. (I actually have only one bodily organ that I trust completely, and even though it has a very finely calibrated knob - I am not sure if it can really help my mixing).

Caravan Ray wrote:

irwin wrote:Jast brings up an excellent point. The spectral analysis tools are great for helping you develop your ear, but the ultimate goal should be to learn to trust your ears. You don't listen to music with your eyes, after all.
.

Yes - this has all got me curious, and when I get the chance I will investigate more. I really don't trust my ears at all, so it will be interesting to see if my eyes can help them. (I actually have only one bodily organ that I trust completely, and even though it has a very finely calibrated knob - I am not sure if it can really help my mixing).

Umm, *looks both ways, whispers* it can help. I'm typing this handsss freeee.
...oops, forgot to lift.

(It should be noted that at this point, I'm not trying to convince anyone of anything, just trying to work this though in my head -- it's a lot harder for me to shrug off a disagreement with JB than it is for me to shrug off a disagreement with Sober.)

jb wrote:Now, if somebody went and took one of my songs, ran a spectral analysis and said "you see there, in the whosiwhat, it's too jiggermahootie." Then if they fixed it and did another analysis and said "now look at it. that's how it should look." And if it actually sounded better... then I'd be like "hey man, who's not using spectral analysis? Chumps!"

I'm just sayin'.

JB

Fair enough. For the record, I don't think there's any set way a spectrogram should look, because, obviously, every song is different. "When Doves Cry" doesn't have a bass guitar part, but there's no way I'm going to go up to Prince to say "Dude, you need some low end in there."

But when something doesn't sound right (which for me is usually the case), spectral analysis serves me quite well as a diagnostic tool, and it also makes it easy for me to highlight the offending areas specifically, rather than having to ballpark it with a normal EQ tool. To this end, it's also a highly-precise tool for selecting certain frequency bands (highlight the bright orange parts, and then you don't accidentally modify anything that you don't want to) and applying effects to them, be they volume adjustments (essentially EQ) or other effects (I haven't really gotten around to seeing what it would sound like if I tried, for instance, a chorus effect or compression on just a specific band of frequency within a track, but that's one of the things that's possible with spectral).

Finally, I'm also in Caravan Ray's boat; I don't trust my ears. You clearly do, JB, and that's why you've been getting along well for a decade without ever using spectral analysis (that's also why your music is so much better than mine). For my part, after six years of doing this stuff, I've completely failed to develop as a mixer/producer. I think I just have a tin ear for that stuff. So any tool that helps me to understand the connections between wonky sounds and the frequencies that create them is okay in my book.

What I do for "spectral analysis" is this: I take a very narrow-band filter, have it boost a lot (at least 10 dB) and move it across the entire frequency range. When I find a spot where it makes something that sounds kind of bad sound a lot worse in the same way, I change the filter to cutting that frequency instead. It tends to help.

okay, so let's have some show and tell. I couldn't get the ReaFir (spl?) to work with Krystal, but Audacity has a built-in "Frequency Analysis." Here are four different 15-15 second snippets from across my song for Spin the Bottle. What does these mean?

01

02

03

04


(sorry, they're chopped off at the far right, but you can read the whole graph).

I'm guessing, the graphs show that my song is bass-heavy. But that's not what my speakers say. If anything, I've worked to trim down the upper end (higher-pitch guitar chords, crash symbols, backing vox) so that it the whole thing breathe. And I like my bass line.

... or are we talking apples and oranges?

I'm not sure how Audacity's analysis works, nyjm, but there doesn't seem to be anything spectral about that analysis. It looks like an average of the concentration of frequencies over the highlighted portion of the file.

Mine looks more like this (screenshot isn't mine, but my results don't vary too widely from that). The X-axis is time. The Y-axis is frequency. Color represents intensity (volume): black is silent, purple is quiet, red is loud, yellow is very loud.

There appears to be some low-frequency drone in the analyzed file here, which stays constant throughout (that bright yellow band along the bottom), as well as some odd but not very intrusive harmonic interference (those three pairs of purple bands, one at 5000hz and 6000hz, one at 12000hz and 13000hz, and one at 18000hz and 19000hz).

As far as the actual song: you can see the low-end of the music starting up before everything else (about :10). The rest of the instrumentation kicks in about :40, and stays pretty constant until it drops back down around 3:35 (from 3:35 to 3:45 is practically silent except for that aforementioned ground hum and those periodic but very thin vertical lines... maybe a hi-hat?), after which the bass part comes in first before the rest of the song kicks in again at 4:10 and stays that way until one last drop in activity about 5:20, and even this ends shortly before the file ends.

I pulled this screenshot from Google Image Search, and I can't find the context, so I have no idea what song is being analyzed here, but I'd be willing to bet (based on the high levels overall) that it was something recorded this decade, and it wouldn't surprise me if it were something hard (or at least heavy), with a very strong bassline. Possibly something like Interpol.

You'll notice that the low-end is nearly white from about 3:00 to 3:30, which might be causing some clippage in the actual song. If this were my file and I thought that section sounded too oversaturated or otherwise problematic, I would look at the spectrogram and see that there's a ton of sound in the bass. this would help me figure out where to reduce volume.

Obviously, and as per JB's comments, if nothing sounds wrong, then a visual imbalance in the spectrogram alone is no reason to tinker. But as a diagnostic tool, it's quite useful.

In Audacity (at least in the version I have, 1.2.6), pick "spectrum" from the track drop down menu, and you'll get what Jon describes.

Generic wrote:I'm not sure how Audacity's analysis works, nyjm, but there doesn't seem to be anything spectral about that analysis. It looks like an average of the concentration of frequencies over the highlighted portion of the file.

Same idea, but Jon Eric's display shows the frequencies over the whole song using those colors, while the other is a smaller snapshot. JE's is probably more useful.
-bill

Okay, take 2. This is the mix I submitted for "Spin the Bottle." Lots o' white down there at the bottom.

I've since remixed it to push the percussion and bass full-left George Martin style and spaced everything else out. Since I wasn't going for a full-on surround-sound stereo vibe, I was able to mute a few tracks and now the whole thing breathes easier.

At least to my ears it does. Take a listen.

It's true, the new mix is a lot less muddy to my ears as well, though I wonder whether that had more to do with the stereo differentiation than the lowering of the concentration of bass frequencies. Most likely a combination of the two.

Glad to see the experimentation is helping, though. Keep it up!

Oh wow, I should probably check these subforums sometime.

Anyway, there are two Reaper plugins that do spectral analysis - Spectro and ReaFir. I like using Spectro as a spot-check on my master track to make sure I've hit all the frequencies. I turn the glow setting all the way down to get the most accurate representation.

To start out with: if you want a full-sounding song, you have to have a relatively complete spectrum (which usually isn't a problem with decent recording gear and instruments).

If you find a hole, fill it with an instrument, EQ a track to be more prominent using that hole, or raise the track that should be filling that hole. Don't worry too much about the really high frequencies since they're mostly inaudible anyway.

The second piece is a little more complicated. By switching on and off your various tracks (or adding a spectrum analyzer to all of your tracks individually) you can find which frequencies are being competed for. This is what I understood your original complaint to be about. You just need to lower that frequency band on the track(s) that are competing with the one that's being muddied. That will allow, for example, your vocals to come through clear at a lower volume without being muddied by guitar.

Another thing I've heard of people doing, but which I have not done, is running the reverse phase of the vocal track on the competing track. This is how "vocal eliminator" plugins work. By adding the left channel of a recording to the inverse of the right channel, you negate anything that's panned to the center. So, in theory, if you add your guitar track to the inverse of your vocal track or somesuch, that would 'make room' for the vocal track.

Caveat: I am not a professional sound engineer. I am mostly self-taught. And I usually don't EQ any of my tracks (but I do do a lot of spectrum checking to ensure 'fullness').