This is part 1 in a series.

1. Recording Through the Decades - Frequency Range
   visual EQ comparisons of recordings through the decades.
2. Recordings Through the Decades 2 - Frequency Range (again)
   audio samples so you can hear the differences

A while ago one of my friends asked me how to get that "old recording sound" for something he was working on. I sent him a graph of the frequency content of an old Robert Johnson song so he could follow the EQ curve in his recording. It's not exactly a substitute for using the right equipment and recording techniques to capture the right sound to begin with, but it is instructive in just how recordings have changed through the decades.

This article analyzes the frequency content of songs from the 20's through today. In later articles I'll look at the overall volume level of these songs, and supply some mp3's so you can tune your ears to and hear what's going on in these diagrams.

Methodology

I went to my CD collection, and took recordings from different time periods and encoded them to .wav format using CDex (EAC didn't work for me). I brough these into Cool Edit Pro (CEP, now Adobe Audition), and found a 10 second sample of a loud section of song - usually in the chorus. Then I used CEP to show me the frequency content of the full 10 seconds of song. For you tech heads out there, I used the Blackman-Harris window and an FFT size of 65536.

Note that these graphs are logarithmic, and I assign the same horizontal width to each octave rather than for each 100 hz of frequency.

Limits and Caveats

All of these recordings are from my CD collection. This means they've been re-mastered by somone recently while encoding them to CD, so the frequency content won't reflect exactly what would be on the original master tapes, or how someone would process them for that day's listening equipment.

Also, playback technology has changed significantly over the decades. What I hear listening to a CD isn't the same as what I would hear on an old Victrola record player.

1920's and 1930's

This graph shows the frequency content of 3 songs from the 20's and thirties. The lines are color coded. None of these songs have drums.

As I was listening to these songs, trying to think of ways to describe the changes from era to era, the only obvious way I could think of was "mood." These recordings simply invoke a different mood than modern recordings do. I'm not going to try to name this mood because to name something is to limit it. If you're not familiar with recordings from this era, you owe it to yourself to check some out.

There's more space between the notes and the frequency range of the recording is limited, but they're no less powerful than modern recordings. Both the music and recordings were very different back then.

Speech is contained largely between 300hz and 2khz, and human hearing is tuned in to this range. That's the range that's emphasized in these recordings. They really peak around 300 - 500hz, and again, that's where human speech peaks.

1940's and 1950's

This graph shows the frequency content of songs from the 40's and 50's. All of these songs have drums.

These songs already sound fairly modern in comparison to those from the 20's and 30's. You see a strong emphasis on the 60 - 100 Hz range. These songs were still probably recorded live to 1 or 2 tracks, but the recording of the bass range is much better.

This was a real era of innovations in recording - multi mic recordings were introduced where different mic's where used for different instruments, probably mostly to seperate the vocalist from the rest of the band. Electric instruments were invented during this period, though they're mostly thought of as novelty instruments.

I'm going to guess, though, that a lot of that 60 - 100 Hz stuff was mixed in for the modern mastering of these recordings and not naturally that strong in the original recordings.

1960's and 1970's

This song shows the frequency content of songs from the late 60's and early 70's. Note the scooped mids.

It should be little surprise that these recordings extend the upper and lower frequency range as the recording fidelity improves. What is interesting is the dip in the 300 - 500hz range that was so important before. This gives the music a bit of a hyped, unnatural feel. This also contributes to the percieved loudness of the recording. (More on this later.)

During this era two obvious and important innovations contributed to the quality of these recordings recording - the wide acceptance of electric instruments, especially the electric bass, and multi track recordings.

1980's and 1990's

This graph shows the frequency content for recordings in the late 80's and early 90's.

By now multi-track recording is the norm, and each drum often gets it's own mic. The electric bass has basically replaced the upright bass for most popular music. Drum machines invaded during the 80's and the infamous TR-808 kick drum shook dance floors all over the world. People have been ditching their LP's for CD's. Sub woofers are becoming more popular, and cassettes have replaced 8-tracks in cars.

This is about as close as this survey comes to pure pink noise.You'll notice that Nirvana's Smells Like Teen Spirit dips more in the 250 - 500 midrange than the other recordings. This should translate to a greater percieved loudness when played, even at low volumes.

Early 2000's

This graph shows music from the late 90's and early 2000's.

As I said, the previous recordings were as close to pink noise as we got. By now the bass content keeps increasing well beyond the point of pink noise, which is close to naturally occuring noise like rain and traffic. I'll refrain from commentary about the effect this has on people, but as you can see, this music, especially the Madonna track, demephasize the 300-500hz speech range quite a bit.

This isn't wildly different from what we had during the late 60's and early 70's, so it's probably no coincidence that popular music on the radio doesn't typically go back farther than that, and when it does, there's a station dedicated completely to music before that era.

Era Comparison

This graph shows 3 songs seperated by 25 to 30 years each.

Here's an overview that should give you a good idea how things changed over the years. The Edith Paif track from 1936 peaks in the 300 - 2000 range and is especially strong in the 300 -500 range (which is partially obscured by the other tracks, refer to the earlier diagram). By 1971, there was a greater focus on the bass range, it's sleightly scooped in the 300 - 500 range, and peaks again around 500 - 1000. The modern recording really peaks around 60 - 100hz.

If you ignore the sub 100hz range, you can see that the content of these recorings has changed very little from the 70's through today.

The Fletcher-Munson Curve and Perceived Loudness

Also known as equal loudness contours, the Fletcher-Munson curves show us how we hear sound differently at different volumes.

This chart is a bit hard to read, the lower the line, the more sensetive we are. Using a 1kHz tone at different decibel levels, they charted a graph to show us how we hear that sound. Let's take a look at the bottom line, which is the minimum audible line for most people.

Take an 80 decibel sound - find 80 dB along the left side. That straight line represents a real-world 80 dB sound. At 20 Hz, we percieve that sound to be (follow the curve) 10 dB. At 3 kHz, we percieve that sound to be 90 dB.

In other words, our ears aren't mathematically linear, and the quieter things get, the less bass and treble we hear. The louder things get, the more we can hear the bass and treble. You can experiment with this yourself by putting on headphones and listening to one of your favorite rock and roll tracks. Lower the volume and you probably won't hear the bass or kick drum anymore. The more you raise the volume, the louder the bass seems to get.

The upshot of this is that if you EQ in more bass, it will seem louder at lower volumes. This is the reasoning behind the Bass Boost button in many stereos. Songs EQ'd this way will sound a little unnatural and hyped.

You can also see by the chart that we're more sensitive to frequencies in the 200 - 500 Hz range, and again in the 3 kHz to 6 kHz range. This is the range needed to understand langauge. Most voices are in this range, and if we lose this range, it becomes more difficult to understand people. One range hears the voice, the other hears any sibillance, which helps define the consonant sounds. This is also why we're so sensitive to scratchy noises, even a quiet noise in the right frequency will seem louder.

Early recordings emphasized this frequency range, and it's not a coincidence. Early recording and playback equipment was tuned (by ear) to make the most of the limited frequency range they could capture and reproduce. By reproducing this area predominantly, it could be assured that we could understand what was being said.

Later recordings eventually got more full spectrum so that listening to a recording sounded more like having a whole band in front of you. Eventually, we went past this point and started to exploit the fullness of modern equiment to make recordings sound unnatural and hyped, all in an effort to grab your ear when you hear the song on the radio, or stimulate some of the excitement that loud, low noises create even at lower volumes.

Continue on to Part 2 - Recordings Through The Decades 2 - Frequency Range (again).

Discuss this article in the Forum.