what is the width of the "q" parameter in Rockbox EQ?

[shady_slim]

I have Sansa Clip+ with RB v. 3.7.1 and Phonak PFEs with green and gray filters.

Knowing the characteristics of Phonak`s gray filters (the diagram is available on many sites) I would like to make it as linear (or rather sloping a bit down as the frequency goes up) using the advanced/graphical EQ. Actually - I have already done some EQing, but am not sure how accurate it was - just because I do not know how the "q" parameter works at different frequencies...

My guess is the width of the "q" changes with frequency - e.g. q=1 at 100 Hz means 100 Hz +/- 10Hz (so we can @ 100 Hz it is 20 Hz "wide"), while q=1 at 4000 Hz means 4000 Hz +/- 400 Hz (so we can say @ 4 kHz it is 800 Hz wide).

In the example above q=1 width is equal to 20% of the frequency it is applied at both for the frequency of 100 Hz and 4 kHz. So we could say it is a constant ratio parameter. Of course - it is just my assumption - it does not need to be constant and it may be calculated in a totally different way. But even if it is constant - the original question returns - how wide is it?

I will appreciate your help!

[Multiplex]

I'm no expert but there's an article on Wikipedia about  Q http://en.wikipedia.org/wiki/Q_factor

[shady_slim]

I'm no expert but there's an article on Wikipedia about  Q http://en.wikipedia.org/wiki/Q_factor

thanks a lot! I will try to "digest" it - it seems quite complex....

[Datman]

Knowing the characteristics of Phonak`s gray filters (the diagram is available on many sites)

I haven't found it. Where is it?
Post Merge: February 28, 2011, 07:20:43 AM

My guess is the width of the "q" changes with frequency
e.g. q=1 at 100 Hz means 100 Hz +/- 10Hz (so we can @ 100 Hz it is 20 Hz "wide"),
while q=1 at 4000 Hz means 4000 Hz +/- 400 Hz (so we can say @ 4 kHz it is 800 Hz wide).

No. Q=f/B, where f=frequency and B=bandwidth, then B=f/Q. In your examples Q=5:
100+/-10Hz@-3dB => B=20Hz; Q=100/20=5
4000+/-400Hz@-3dB => B=800Hz; Q=4000/800=5.
Keep in mind that -3dB points are valid for a high gain or attenuation filter; if gain/attenuation is only a few dB you can't consider -3dB points...

E.g.: if you have a 4dB gain you have to consider the -3db of the 4dB over the unity-gain: 4dB=1.585, -3dB are 0.707 then 0.585*0.707=0.414; 1+0.414=3dB!

If you have a 2dB gain:
2dB=1.259 then 0.259*0.707=0.183; 1+0.183=1.46dB.

In few words it's similar to multiply dBs to 0.7:
4*0.7=2.8
2*0.7=1.4

[shady_slim]

I haven't found it. Where is it?

e.g. here: http://sonusmobile.pl/data/include/cms/Phonak/Audeo_PFE_Bass_-_Frequency_Response_550.jpg


as you can see there are 3 peaks which need to be flattened:
1. @ about 2550 Hz which occupies fairly wide bandwidth and which needs to be reduced by some 8-9  dB (just by guessing I applied q = 0,7),
2. @ about 9600 Hz - very narrow, needs to be reduced by some 7-8 dB (similarly as above, I applied q = 5,5),
3. @ about 15 kHz - also narrow, needs to be reduced by some 4-5 dB (q = 5).

Sorry, but I am not sure if I understand your maths

You wrote Q = f/B

Assuming the first peak is @ 2550 Hz and it starts @ about 700 Hz I have calculated  B/2 = 2550 - 700 = 1850. this needs to be multiplied by 2  - i.e. B = 3700.

Does it mean Q = 2550/3700? It makes 0,69 - and this would means I was very close in my guessing  - provided you will confirm my calculation

Similarly f = 9600 Hz, B estimated at 2200 --> Q = 4,36

Am I right?

[Chronon]

The bandwidth should correspond to the FWHM (full width at half maximum).  For a given peak, take the difference of the two frequencies where the height is half of the maximum.  Looking at the image you posted, 700 Hz looks much too low for the lower frequency at half-maximum.  Also, the frequency at the peak looks to be greater than 3000 Hz to me.

[Datman]

It's very hard to equalize mechanical/acoustical resonances using an electrical filter. Earphones have to be chosen for the right sound. Strong resonances also produce distortion. It isn't simple, I know, but it is the same as for microphones: every singer chose his brand and model. No one thinks he could by a cheap microphone, then ask to the sound engineer to equalize it...

I've tried Shure E2 and E3 and Etymotic ER-4s (I've found good prices for new or used ones): ER4s are the best. Much better and cheaper are AKG K518LE foldable headphones, if you modify them with glass wool, resistor and capacitor. Now I use them outdoors with my Rockboxed Sansa Fuze.

[cpu98]

I did flatten my etymotic mc5 with 2 shelving filters and 11(!) peaking filters.
Now I'm with a 20~20KHz-within-1dB earphone for mere 79 bucks.
I hear neglectable ringing, phase issue and group delay at sub bass and near ultrasonic region but it is comparable to a studio mointor.

I used electri-q equalizer(Bandwidth based) to get necessary values for filters.
(They are compatible with rockbox after you convert bw to q factor. Confirmed with a FFT analyzer)

First I made equalizer window transparent and let 'upside down response graph' below it and scaled the image.
And I fine-tuned equalizer so that those graph look like only one graph.
After correcting that get those value.
There is a q factor<-> bw converter (http://www.sengpielaudio.com/calculator-bandwidth.htm) you can use to determine q factor.

High range >= 8KHz needs care because filters become asymmetric.(Highs need broader q and lower cutoff than expected)

After that you measure that monster dap thing.
And say 'shit there is a line not a mountain on the monitor'

Extremely flat earphone is ultrabright as if it is made for dolphins so introduce gentle roll off after final correction.

[Datman]

How did you get 11 peaking filters in Rockbox???

[cpu98]

15band patch I used to equalize mine.

http://pastebin.com/vAXRd1zy

13band enabled + meier crossfeed eats 70+MHz on ipod video.
Luckily mine overclocks to 100MHz.
See #8668 if you use ipod video.

[soap]

(Late to the party, but)

Why are you trying to flatten the frequency response of a IEM?  A flat response for them is wrong as they bypass the outer ear (and to a lesser extent the head).

[cpu98]

The peak around 2~3khz is external ear resonance.
And that image seems raw response graph. You shouldn't bulldoze it. That's normal.
http://www.etymotic.com/ephp/images/er4pt_compliance_graph_480g.png (response curve of er4pt)
If you like er4p sound reference to this image.

What I said flat was ~99% diffuse field referenced.