Stereo-to-Headphones DSP

LRN

Åñòü ëè äëÿ Kx DSP-ïëàãèí (ôèëüòð?), äàþùèé òîò æå ýôôåêò, ÷òî è ýòîò?

Chester01

ive been working on such things, I have been using points on the ISO 223:2003 hearing response curve to find reference points, then I am designing the eq within WinISD Pro so I can make sure my check points (I pick a point between each EQ point to make sure my slope is relatively accurate) are also working out properly. I will post info once my work is more complete.

I have also made a 'stereo simulator' chain with a delay, and a low pass filter, then 'swap channels' and reduce the gain by about 6 db. a while back I calculated a 19 sample (0.395833... ms) is about the delay with my head size (average id say). the lowpass filter is set around 2700 hz with a 3.6 Q factor. again, I will post images or something later

LRN

I understood only last sentence (about images later) )

Chester01

http://n.1asphost.com/Chesteta/DSP.JPG - a snapshot of my DSP
http://n.1asphost.com/Chesteta/ISD.JPG - my EQ filters I derived from the ISO 223:2003
http://www.nedo.go.jp/itd/grant-e/re...pdf/is-01e.pdf -source of hearing response curves (most up to date - 2003): I made a 3000x3000 bitmap of "Fig 2" (zoomed in alot) and have matched points on the chart to intensities - I also found points between each set of frequencies to make sure the slope of the lines was relatively accurate.
http://n.1asphost.com/Chesteta/80.JPG - this is what I used to find frequency to gain correlations (VERY LARGE (624k, 3167x2877 pixels...))

NOTE: this is the hearing response curve for 80 phons (percieved db @1000hz) , it does vary over intensity, im still working on other intensities... it takes a while to find a highly accurate curve with 7 gains and 7 Q factors to modify by hand

Looking at ISD.jpg, red curve is for use with an "UFX EQ P5" filter, it is less true to the original (yellow - most accurate) filter settings however I noticed that when running "Peaking EQ" filters with the 20 hz settings (see the blue curve - almost perfectly accurate) there is -60 db signal with no signal before the filters (the 20 hz portion generated some sort of noise), the EQ P5 does not do this however it does not have the range of frequency (only to 24.4 hz) of the other plugins so I did my best to find a similar curve (the red one) that would work.

What does this all do? if you set up a signal chain in your dsp with the frequencies, Q's, and gains shown in ISD.jpg, assuming your music is being played at ~80 db, and assuming your stereo is fairly linear (try a good pair of headphones) this will take out the response curve your ears add to sound. If you have an old stereo reciever with a 'loudness' button on it, this is a similar effect to that however it should be MUCH more precise.

if something does not make sense, feel free to ask.

I will write something about Head Related Transfer Function (that headphone filter you talked of earlier) in a bit, i have to go to a haircut now.

Peace!

EDIT: I forgot to mention that the PDF file (the third file) will probably ask you to install japanese characters or something when opening it, cancel that, only the first page has japanese, the rest (important stuff) is in english.

LRN

I'm not sure it worth the trouble, since both DSP.JPG and ISD.JPG looks SCARY to me

Chester01

all you NEED for the ear inverse curve are the things in the middle (between GainHQ and 20000)... those are peaking EQ's, set the gain HQ to atleast -20 db... -30.3 if you are going to be playing alot of 24 hz tones at +-0 db into the card
if you look at the box with the "red box" at the bottom, that will give you all of the Q factors (slopes), frequencies, and gains to enter into the parametric eq's. if you want i can make a screenshot showing whats in every parametric eq... its not that complex (IMO)

-Matt

DDRRE

Your image host doesn't resolve, currently.

Chester01

http://img201.imageshack.us/img201/5114/isdpx6.jpg

http://img150.imageshack.us/img150/4353/dspqa7.jpg

sorry bout that

LRN

http://lrn.no-ip.info/other/kx_dsp_default.png - here's my DSP layout. Where (and wich effects) i should insert?

Chester01

well what do you want to do with the dsp? what you have are the defaults...

Haos

Pls consider this. We like to experiment with DSP, but (at last me) think of your dsp/explanation as of Ye Black Magic (burn ye witches).

Could you pls drop that geek slang and explain to us, mortals, what needs to be done, step by step? And what would be the effect?

The best thing would be if you`d create a single plugin for dsp.... bcause my DSP is pretty crowded already... applying THAT STUFF u presented, would turn it into a nightmare

Chester01

if you are in the dsp (assuming you can get there), if you right click, all you need to find is
Add Effect/Plugin -> EQ & Filter -> EQ P5 (load one)
Add Effect/Plugin -> EQ & Filter -> EQ Peaking (load six)
Add Effect/Plugin -> Basic -> gainHQ (load one)

Place the gainHQ plugin first in the chain, set it to -24 to -36 db (depending if you are clipping you may need to set it lower than -24 db)

Connect all the plugins in a chain (draw lines from the blue output rectangles to the red input rectangles) after the gainHQ plugin, it doesnt really matter which order however I connected the 'effect' so that the lower frequency plugins are to the left and the higher frequencies are to the right...

If you wish you may rename the plugins (right click the plugin, click Rename) to the frequency it will be set to make them easier to tell apart.

Now, click on the first Peaking EQ filter, within my setup the first one is 860 hz: Change the three settings to Freq (hz): 860 - Width (Q): 0.70 - Gain (db): -2.4

After modifying this plugin you can close the plugin window and open the next one (1500 hz) NOTE: you will probably have to use the arrows on your keyboard to precisely set things if you cannot do so with your mouse (use the left and right arrows) just click on the parameter you want to set and press the arrow...

The Peaking EQ settings are as follow:
Freq (hz): 860 - Width (Q): 0.70 - Gain (db): -2.4 (should have done that one already)
Freq (hz): 1500 - Width (Q): 1.97 - Gain (db): +5.8
Freq (hz): 3070 - Width (Q): 0.71 - Gain (db): -5.6
Freq (hz): 9100 - Width (Q): 2.50 - Gain (db): +13.7
Freq (hz): 12334 - Width (Q): 4.50 - Gain (db): -3.8
Freq (hz): 19999 - Width (Q): 5.00 - Gain (db): +5.0

Now you can modify the EQ P5 plugin:
**You will need the first two filters of the plugin (there are five total)
Set the filter to a "Bell" filter type (next to the "IN" button)
Set the frequency to 24.4 hz (be sure to have the 'switch' next to the frequency selection nob to the bottom)
Set the "BW" (the Q) of the first filter to 4.00 and the second one to 0.30
Set the Gain of the filter to +18.0 db and the second filter to +15.0 db
In summary:
Freq (hz): 24.4 - Width (Q): 4.00 - Gain (db): +18.0 ("Bell" type filter)
Freq (hz): 24.4 - Width (Q): 0.30 - Gain (db): +15.0 ("Bell type filter)

Now all you do is place this effect chain between your input and your output... please do not use the EQ P5 part of the fitler if you do not have a good subwoofer that can handle the bass, I dont want anyone to blow their speakers however I take no responsibility if you do, use good judgement. Maybe try this with headphones...

Also, keep in mind that this is adding ALOT of gain to the bass end of the spectrum so you need to have the gainHQ plugin in there otherwise you will create alot of distortion by clipping, not much else. -24 db is a little less than 1/4 the initial volume so you will need to adjust your stereo accordingly (turn it up a bit more maybe).

I hope this helps!
-Matt

Chester01

Now: what is this doing: it is taking the inverse (opposite) of what has been found to be the average hearing response in humans. Consider this: you hear some frequencies (sounds) better than others, people turn up their stereos because they say that the bass sounds better when they do. Actually you hear bass more 'equally' with other sounds at louder volumes. This chain of effects is correcting for that so that you hear everything equally. Your ears are not a reference ('flat') microphone, 20 hz needs alot more db (intensity) to SOUND as loud as 1000 hz. http://www.nedo.go.jp/itd/grant-e/re...pdf/is-01e.pdf (this is the most recent study I could find of the phons scale)

Intensity (power) is in the db scale, where volume is in the Phons scale. This effect chain turns the db scale of the sound card into the phons scale of your ears. This is very similar to (but much more accurate than) a "loudness" button on some stereo recievers. Also, this is the phons scale at 80 db (remember how turning up a stereo makes the bass sound 'better'). The gains, and Q factors are different at different intensities; it took me about 3 hours to find the values for an 80 db/phons conversion so I have not done another one yet... (every time you modify a gain or Q, it changes the effect of all the other values slightly)

I hope this helps explain a bit what this effect chain is attempting to accomplish.