Sunday, March 28, 2010

Equalizer settings for the 2010 Subaru Outback

This article is in response to a question on the SubaruOutback.org forum. I recently got a 2010 Subaru Outback, and its Harman-Kardon sound system includes an equalizer controllable from the multifunction display. The manual says very little about how it works, and Subaru was not much help. It includes the cryptic comment "The larger the value of Q Factor (degree of acumination) is, the more the gain of the center frequency is enhance.” (P.G. 197, 2010 Legacy/outback Navigation System, book.)". Here's what I've learned and how I've set it.

Band & Frequency: The EQ has four bands, each with a setting for Frequency, Q, and Gain. This is known as a Param
etric Equalizer, which is different from the Graphic Equalizer that many of us are more familiar with. On a Graphic EQ, there may be many more bands, but the central frequency for each band is fixed. On this Para EQ, the central frequency can be selected for each band, but there are only four of them.

In Hz, the frequency choices
are:
Band 1: 60, 80, 100, 120, 160, 200
Band 2: 250, 315, 400, 500, 630, 800, 1k
Band 3: 1.25k, 1.6k, 2k, 2.5k, 3.15k, 4k
Band 4: 5k, 6.3k, 8k, 10k, 12.5k, 16k

Gain: This is the 'boost' or 'cut' that will be applied to frequencies in the selected range.

Q: What is this mysterious setting? It determines the width of the range of frequencies to which the Gain setting will be applied. As in a Graphic EQ, a bell-shaped range of frequencies, centered on the Freq setting, will be boosted or cut. Q is inversely related to the number of octaves affected, surrounding the center frequency. Here's the relationship:

Q Octaves
0.25 4.1
0.5 2.5

1.0 1.4

2.0 0.
7

An octave is defined as the distance from a frequency to double that frequency (or downward to half of it). Look back at the Freq choices, and you'll see that it takes three steps up or down to go an octave. Since the range is centered on the Freq selected, you can think of it this way: if Q=1, the frequencies affected will be about one and a half steps on either side, both up and down. If Q=2, barely one step each way will be affected. if Q=0.5, about three steps either way. If Q=0.25, about 5 steps either way. A good depiction of this can be found here.

So back to the question: how do I get rid of that excessive booming bass? Which frequencies should I cut? By how much? And what Q should I use? This is actually a fairly tricky question. It's hard to tell frequencies by listening unless your ear is well trained. If the spans affected by multiple bands overlap, then the effects are additive (boost+boost, or boost-cut, or cut+cut). One ends up boosting or cutting the ranges that are the most out of whack, then boosting and cutting adjacent ranges to try to mitigate the side effects and get a smoothly improved frequency curve. It's very tricky, and hard to do with music alone.

I happen to have a Graphic Equalizer with a frequency spectrum analyzer and a sound generator (I used to have it installed in my home stereo system.) One can use such a device to send a "known" sound (consisting of equal amounts of all frequencies, called pink noise) through the car's sound system and see what comes out the speakers. Fortunately the Outback has an AUX input, so this is pretty easy to do. This is much more precise than fiddling blindly with EQ settings. Here's how the sound input looks before the car sound system messes with it (note that my EQ's ranges differ from the car's):
Then I pumped the pink noise through the Subaru's sound system, plugged a microphone (held at the driver's head height) into the analyzer, cranked up the volume, and here's how it looked with no correction.
You can see that there's a peak at 125 Hz and one at 4 KHz. Actually in some samples the valley between them looked much worse. The 125 Hz peak is responsible for the booming bass that the other poster and I dislike. The 4 KHz peak and the dropoff of the higher frequencies contribute to a lack of clarity of some instruments and words. (The rolloff at the top and bottom may not really be as bad as they look - I could not find my good microphone and used a cheap one that was handy, and it may not be sensitive enough for the extreme ends.)

So what to do about it? We now know which ranges the Subaru H-K sound system overemphasizes, but futzing with the screen to set them is tedious at best. As mentioned above, tweaking a Freq actually tweaks its neighbors, so choosing a good set of settings r
equires solving several equations with many unknowns. I wrote a spreadsheet many years ago to do that math, but I can't find it now, so I hand-equalized it using the 10 ranges on my external EQ. Here are the settings I chose.

This may look kind of "all over the map", but if you draw a bell curve around each setting, and add them up, it becomes a smooth correction curve.

When I turn on the pink noise, apply the external EQ settings, and capture the sound with the mic, the results are not half bad. There's still a dropoff at the top and bottom, but that might be the mic. This sample still shows some peaks, and loss of the 63 Hz range, but some samples were quite a bit flatter, and it sounds less boomy already.

Now we need to translate this to the H-K EQ settings, with only four Freq bands available but a variable band width (the Q). My strategy is to use as narrow a range as possible for the problem bands , so I started with a Q of 2. For the peaks we need to seriously cut, I chose:

Band 1: Freq=125, Gain=-5
Band 3: Freq=4k, Gain=-5

To offset the side affects and level out the middle, I chose Band 2: Freq=500, Gain=2.
I'm not sure what's happening with the top end (music sounded OK to the ear), so I chose Band 4: Freq=12.5K, Gain=0, in effect doing nothing at the moment.

When I turn on the pink noise, apply the para EQ settings, and capture the sound with the mic, the results are pretty flat - which is what we want!

Now I switched to music instead of pink noise to see how it sounded. The first thing up on my iPod was "Listen to the Music" by the Doobie Brothers, which sounded like excellent advice! Never having worked with a Parametric EQ before, I played around with getting the low end flatter. I switched Band 1 from 125 down to 80 and even 60, and spread the range out by using a Q of .25. I had to cut the Gain as far down as -9 to get rid of the booming bass and make it sound like a real drum. I didn't write down any more results from the analyzer (music is much more dynamic than pink noise), but I could tell from the display that it was much more balanced.

The H-K sound system has different EQ settings for each audio source, so I had to go enter the settings into each one. I began to notice differences right away:
  • FM radio didn't sound as clear. Well, radio transmission does not have the same fidelity as a digital source, so that made sense. I had to boost the top end quite a lot to get clarity. (Remember I said the top end sounded OK to the ear? That was probably with digital sources.) Settings:
    Band 1: Freq=80, Q=.25, Gain=-9
    Band 2: Freq=500, Q=2, Gain=3
    Band 3: Freq=4k, Q=2, Gain=-3
    Band 4: Freq=125, Q=2, Gain=7

  • iPod, XM Satellite Radio, and CD were pretty similar to each other, being all digital sources, except the top end.
Now, this was all done in my driveway, with no road noise, so it was time to hit the freeway. You can't adjust the EQ settings when the car is in Drive, you can only switch between Default and Presets. So I stopped a few times to make adjustments. The bass was much improved. But the high end was very harsh with digital sources, almost painful. By experimenting, I settled on Band 4: Freq=16k, Q=.25, Gain=-3. In other words, lower a broad range of the top end, but only by -3. I think this confirms my suspicion about the quality of the mic I used for the measurements: it had poor high-frequency response, misleading me to think I needed to boost the top end, when really I needed to cut it a bit. (The mic was from a home karaoke machine, so since it is for picking up human voice, it doesn't need to be good in the upper ranges.) To summarize the settings for digital sources:

Band 1: Freq=80, Q=.25, Gain=-9
Band 2: Freq=500, Q=2, Gain=3
Band 3: Freq=4k, Q=2, Gain=-3

Band 4: Freq=16k, Q=.25, Gain=-3

So I cut the top end for all the digital sources, and now it sounds pretty darn good! I played several songs from my iPod that I know very well, and found some familiar tunes on the XM, so I'm pretty happy. I won't be able to listen in detail for a while - my wife is taking the car on a weeklong trip - but I look forward to hearing the results in the future, and will probably do some more tweaking. Maybe I can record pink noise onto my iPod and a CD, so I can analyze them in more detail. And I need to find my good mic!

This is quite a long post, probably much more than the forum posters were looking for. But I wanted to show the process one goes through to actually analyze the system and what I learned about choosing the parameters for a Parametric Equalizer. It's not a simple process without a frequency analyzer!