One way to make use of Frequency Response Graph

[creatip123]

I'm sure most of you are already familiar with this one, use EQ to invert the Frequency Response Graph. Head-fi have an old thread discussing this. Anyway, here goes.

 

 

So this is a Frequency Response Graph, more precisely, this is the Frequency Response Graph for the Hifiman HE-400. What this is saying roughly, is when this particular headphone is given a flat tone (same volumes/loudness across the frequencies), it will distort that flat tone and give output mapped in the graph. This give us a general idea of the sound characteristic of that particular headphone. For example, from the graph above, the HE-400 is fairly flat on the lows and lower mid. Quite a big drop on the higher mid, and a spike on the high. I'd say, the HE-400 is generally a warm sounding headphone, with good (but not full) highs. This gives us the idea of the sound characteristic, NOT quality of the sound itself, though.

 

This graph was generated by Headroom, based on their tests. This is not the raw testing data, though. I think this is the adjusted graph.

 

 

How we test frequency response: To perform this test we drive the headphones with a series of 200 tones at the same voltage and of ever increasing frequency. We then measure the output at each frequency through the ears of the highly-specialized (and pricey!) Head Acoustics microphone. After that we apply an audio correction curve that removes the head-related transfer function and accurately produces the data for display.

 

 

You can try and find the graph of your headphone (if they're listed), here

 

Of course most good headphones are the result of months, if not years, of intensive researches, to get the best sound results. Meaning they are already factory tuned to give optimum sound output. Although I love the sound characteristic of the HE-400 as it is, I wanted to try something different. Even if you love burgers, sometimes you'd want to go for hot dogs, right?

 

So I got curious, and tried to fiddle with Foobar's EQ, based on the graph. I set my EQ to try to invert the graph above:

 

 

Of course it's just a rough invertion of the graph. The result is pleasantly surprising. I got fuller mids out of my HE-400. I can hear more details from the music. Of course the details were originally there this whole time, but due to it's frequency drop, it didn't really pop out. So this is something you might wanna try, if only just for variations/fun. 

 

Just remember, don't go crazy with adding the gain on the EQ bands, though. Your driver might be hardware limited to play certain frequencies at certain loudness. For example, in the graph above, the deepest valley is -15dB. To properly invert it, I'd have to add +15dB (maybe a bit less), right? Well, when I push the middle bands pass the +8dB mark, I got clippings. Clippings sound like 'click-click', or small pops. I think this is because the driver is just not designed to output that certain frequencies at that certain loudness. Clippings could potentially damage your driver, if played on a long run, so be careful there.

 

Once again, this is not about 'make your headphone performs better'. If anything, it might even performs worse. This is just something to do for fun, for trying a different 'taste' out of your headphone. 

[OmanGS]

Good information. I'll be checking it out for a few of my headphones and such.

 

Thanks for the great idea!

[BoredErica]

I tried this sort of thing with my personal HD800 FR graph. It added a veil to the sound for sure.

[creatip123]

I tried this sort of thing with my personal HD800 FR graph. It added a veil to the sound for sure.

 

Yeah, at some point, I got some sort of a robot sound (from a female vocal). Turned out I was adding too much gain on the wrong frequency. 

[BoredErica]

Yeah, at some point, I got some sort of a robot sound (from a female vocal). Turned out I was adding too much gain on the wrong frequency. 

It's a lot of trial and error and fine tuning. I've spent less than half an hour on tuning, so maybe I need to spend a bit more time screwing around with EQ. I started out by using the Foobar EQ but some people are saying it's better to use something like Electri-Q although I don't know how to use it.

[Uber Hamburgler]

Does anyone know of any software or possibly pre-written python or matlab/scilab/octave scripts for generating frequency response curves?  I have the V-MODA Crossfade LPs and I've only been able to find curves for the LP2s, which should be close, but obviously aren't the same since Katy Perry (my guilty pleasure) sounds more like a squealing cat than a singer after trying this, and this is actually doing less than balancing the curve with somewhat bass-heavy headphones.

[ShearMe]

Does anyone know of any software or possibly pre-written python or matlab/scilab/octave scripts for generating frequency response curves?  I have the V-MODA Crossfade LPs and I've only been able to find curves for the LP2s, which should be close, but obviously aren't the same since Katy Perry (my guilty pleasure) sounds more like a squealing cat than a singer after trying this, and this is actually doing less than balancing the curve with somewhat bass-heavy headphones.

 

Not all headphones respond well to EQ, and I've heard people claim that V-Moda's are in that category. I don't want to sound all negative and tell you not to even bother, but you might want to consider not even bothering.

 

+1 for Katy Perry as a guilty pleasure

[SSL]

Does anyone know of any software or possibly pre-written python or matlab/scilab/octave scripts for generating frequency response curves?  I have the V-MODA Crossfade LPs and I've only been able to find curves for the LP2s, which should be close, but obviously aren't the same since Katy Perry (my guilty pleasure) sounds more like a squealing cat than a singer after trying this, and this is actually doing less than balancing the curve with somewhat bass-heavy headphones.

 

No. Frequency response is measured using a mic'd dummy-head in an anechoic chamber. Not something that you can whip up at home in 15 minutes.

[xyza]

Would it be wrong to use the raw frequency response instead?

[ShearMe]

Would it be wrong to use the raw frequency response instead?

 

No, not morally wrong.

 

If you're human, it'd be silly to do it for anything more than experimentation though.

[creatip123]

On the other note, for mobile use if you like to play with EQ and high resolution audio, and have an Idevice, try the Onkyo HF Player. It can play FLAC, WAV, even DSD (not natively, but using PCM or DoP), and go as high as 192khz/24bit. You gotta buy the upgrade, though, for $9.99. The free version can only play MP3s stored in your Idevice's library.

 

The best feature I like is the EQ. 

 

 

* Free HF Player Includes 16,000-band FIR Equalizer with Memory Presets

 

In reality, who could squeeze 16.000 dots on the EQ to that screen, even on Ipad? But it is very good, smooth and responsive. 

 

Too bad, they don't make the android version, as of yet...

[creatip123]

It's a lot of trial and error and fine tuning. I've spent less than half an hour on tuning, so maybe I need to spend a bit more time screwing around with EQ. I started out by using the Foobar EQ but some people are saying it's better to use something like Electri-Q although I don't know how to use it.

 

Some references on the net suggest to use VST EQ plugin, that are supposed to be really good. Dunno though, I haven't tried it yet.

 

 

Does anyone know of any software or possibly pre-written python or matlab/scilab/octave scripts for generating frequency response curves?  I have the V-MODA Crossfade LPs and I've only been able to find curves for the LP2s, which should be close, but obviously aren't the same since Katy Perry (my guilty pleasure) sounds more like a squealing cat than a singer after trying this, and this is actually doing less than balancing the curve with somewhat bass-heavy headphones.

 

Try going easy on the bands' gain increments. I found that I got better results by adding little increments on more bands in its vicinity, compared to adding big increments on one or two bands only. 

[Stefky]

The problem is the compensation curve they use is wrong. It's always been wrong.

Adjusting for the dip in treble that every single set of headphones -even treble cannons - exhibit will lead to an unpleasant experience.

Use your ears to adjust with EQ if you want to use it.

[Uber Hamburgler]

Try going easy on the bands' gain increments. I found that I got better results by adding little increments on more bands in its vicinity, compared to adding big increments on one or two bands only. 

 

I'm actually only went about 2/3 opposite of the curve, but playing with it and some Googling (actually Bing because it was at work, but whatever, its not that bad),  I've found that just going straight off the frequency response curve won't actually give a flat sound to the human ear, there are some somewhat complicated transformations that must be done to it.  It is similar to the reason that pink noise actually sounds more even than white noise.

[creatip123]

I'm actually only went about 2/3 opposite of the curve, but playing with it and some Googling (actually Bing because it was at work, but whatever, its not that bad),  I've found that just going straight off the frequency response curve won't actually give a flat sound to the human ear, there are some somewhat complicated transformations that must be done to it.  It is similar to the reason that pink noise actually sounds more even than white noise.

 

2/3? I didn't even go past the +8dB mark, for the -15dB valley....

[Uber Hamburgler]

I used this curve that I had found while looking for one for the original LPs.  It doesn't register the valley that was on the HeadRoom one.  I guess I had just assumed that they would be the same.  Also, the bottom of the valley is at 6 kHz, and I'm using the Creative equalizer for my DAC/ADC/AMP that doesn't have a setting there, it has 4 and 8 kHz.

 

Even so, it is the treble that was overblown, so increasing it more would make that even worse.  Tweaking it, I'm getting closer to something desireable, so I'll post back when I get a good one. 

[creatip123]

I used this curve that I had found while looking for one for the original LPs.  It doesn't register the valley that was on the HeadRoom one.  I guess I had just assumed that they would be the same.  Also, the bottom of the valley is at 6 kHz, and I'm using the Creative equalizer for my DAC/ADC/AMP that doesn't have a setting there, it has 4 and 8 kHz.

 

Even so, it is the treble that was overblown, so increasing it more would make that even worse.  Tweaking it, I'm getting closer to something desireable, so I'll post back when I get a good one. 

 

That's what Stefky said above, use your ears as the main tool to adjust the EQ. Because in the end, it's your ears that you're trying to satisfy, not some graph out of nowhere Many people (me included), when presented with an EQ, especially the ones that have a lot of bands, wouldn't know what to do, or where to start. I found that inverting the graph gives a good general idea of what I'm going for, in terms of experiment. Add and subtract them further to suit our preference. 

 

The original post in this thread is more of a guideline, a starting point, if you will. I like to think of it this way: say the inverted graph is a cooking recipe, and it says, "add 5 tablespoons of sugar". Now when I do that, the dish turned out to be too sweet for my taste. Do I keep on adding 5 tbs of sugar to stick with the recipe, or should I experiment with less tbs of sugar, maybe 3-4 tbs? Because in the end, it's my tongue that I'm trying to satisfy, not the recipe, right?

 

*can't wait for the other's comments on me using food analogies..... :lol:

 

Edit: just saw your curve, it's quite different than the one at headroom and innerfidelity

 

The one from headroom: http://graphs.headphone.com/graphCompare.php?graphType=0&graphID[]=4021&scale=30

[SSL]

To make the best use of EQ you should be be able to differentiate octave ranges by ear. This allows you to identify which EQ bands to adjust, no need to use a frequency response graph as a starting point. Of course, this takes time and practice if you don't have perfect pitch.

 

As Stefky points out, these graphs are useful mainly for differentiating between different headphones, NOT necessarily for determining what they actually sound like across the whole frequency range. Many headphone manufacturers deliberately apply what is known as "diffuse field" equalization to their headphones; attempting to reverse this may, again, cause unpleasant results.