[k00zk0]

I recall hearing that equalizers should never be set higher than 0db and should only be lowered because they can remove data well but not synthesize it. I plan to perfect a parametric curve in Electri-q but I find it weird that 0db is in the middle of the screen instead of at the top. Are they saying that this eq can do it correctly? Even if the preamp is lowered by at least the amount the EQ raises it at most, doesn't creating signal where there was none make it bad to use it above the 0 line?
 
Another thing is the peak type. There is "Type 1" used everywhere in the examples and "Orfanidis" which upon checking out is used above about 1/2 Nyquist due to better quality in this range with increased resource usage. But, even adding 50 Orfanidis points, the cpu usage is relatively nothing. Is there a benefit to making all points Orfanidis type, or is type 1 actually better below 1/2 Nyquist frequency?

 

[LFF]

When using EQ it is generally better to cut rather than boost because boosting also raises the noise present in that band. It's not a rule set in stone and sometimes you have to boost but in general, it's better to cut rather than boost.
 
The peak types usually means how narrow or how wide the frequency band will be. A narrow band may only target a couple of frequencies like 10,000cps to 11,000cps. A wide band might move a lot more like from 2,000cps to 8,000cps. A wide band will usually sound more natural and more smooth while a narrow band is useful for eliminating certain key problems.

 

[RPGWiZaRD]

Electri-q are one of those EQs that simply work a lot better by lowering frequencies below 0dB rather than boosting because at least for me it started distorting very quickly otherwise whereas on my soundcard's EQ I can boost the whole range by +12 no problem and it won't distort/clip and even start sounding better in my ears (more forward/aggressive/powerful/attacking/"dynamic" sound and fine details start jumping out more distinctly which is what I prefer it to sound like) whereas cutting provides the opposite effect giving a more laid-back sound so it will depend on the EQ you use what works best (possibly also on amping as in the stronger amp you use the more likely reducing is the way to go)

 

[xnor]

Quote:

I recall hearing that equalizers should never be set higher than 0db and should only be lowered because they can remove data well but not synthesize it. I plan to perfect a parametric curve in Electri-q but I find it weird that 0db is in the middle of the screen instead of at the top. Are they saying that this eq can do it correctly? Even if the preamp is lowered by at least the amount the EQ raises it at most, doesn't creating signal where there was none make it bad to use it above the 0 line?

 
Signal processing is done with floating-point numbers that range from -1.0 to 1.0 where -1.0 usually is mapped to the lowest and 1.0 to the highest integer before it's sent to the DAC.
 
Now imagine a sine wave with sample values like [0, 1, 0, -1, 0]. If your EQ configuration happens to boost that frequency you'll get sample values above 1.0. The signal clips.
 
EQs can cut as well as they can boost. Electri-q is not special in this regard. If the input signal provides no headroom for boosting certain frequencies then you simply attenuate the whole signal first. If you're not sure then it's safest to keep the EQ curve below the 0 dB line.
 
 
@rpgwizard: I guess your soundcard's EQ dynamically compresses the signal further (reduces dynamic range) to prevent potential clipping noise. If you want to keep the sound quality as high as possible then don't do this.
 

 

[NecroNeo]

I have read the tutorial/wiki on equalizing with pink noise, and I have to say it was very helpful and a great introduction to the concept of head-related transfer function. I've tried, with moderate success, to use this approach in my rig, but I find that because I'm using a subjective measure (my perception of peaks and valleys in response when listening to sine files or singen), I fiddle more than I want to and I doubt I've got it right. Since I can't afford to get my HRTF measured, I decided to try a different approach.
I listen to my 280 pros using fb2k with George Yohng's VST Wrapper and Electri-Q (posihfopit edition), and I'm building a CMOY that will soon be in my chain (pots arrived today!). I grabbed the HeadRoom frequency response graph and painstakingly made an inverse of the FR in Electri-Q, with the goal of a flat response. I made the inversion centered around 0 dB and then dropped it till the curve was only cutting. Then I upped the fb2k preamp by the dB drop in Electri-Q to make comparisons at the same level. With only about 10 minutes of switching back and forth between no EQ and the inversion EQ, I got the impression that this was an accurate but possibly fatiguing route to a flatter response.
I am guessing that the three valleys in the 280's response between 1 and 6 kHz are average ear HRTF corrections intentionally built into the phones by Sennheiser. If I find my new EQ causes copious listening fatigue, I may allow those valleys to occur and see if it helps.
 
Has anyone else tried this with their phones? If so, what did you think of the sound? I'd love to hear from Head-Fiers with more EQ experience about what works and doesn't work.
 
P.S. - I just listened to a sine sweep and I'm very impressed – never by listening and tweaking did I get a sine this clean and even. It's not perfect, but the midrange especially has benefited.
 

 
 
Note the direction of dB change in either graph is opposite that of the other.  I screencapped the EQ's curve and turned it upside down so it would be easier to compare them.  I didn't bother with anything above 10 khz or under about 30 Hz.

 

[RPGWiZaRD]

Well IMO this isn't necessarily the optimal way to go but it could possibly be an easier way of doing it if you're not experienced with EQing. Because having a perfectly flat line resembling around 0dB line might not sound the optimal in your ears even if it technically should perhaps do it (but I don't trust the EQ that the "10dB" on the EQ actually provides an accurate 10dB boost etc but that's another topic of discussion). 
 
The most important instrument for me is my own ears, I trust them more than anything and simply go by what sounds "best" which is usually a more balanced sound except for bass that I like to pronounce a bit more being a basshead. How I do the EQing is taking one "band" or frequency point at a time on a graphical EQ and quickly both drag it to the very bottom and to the very top to get a quick glimpse of which side it should be more closer to, if it sounds worse (as in more excessive) bottomed out then I know it has to be more closer to the top and vice versa. Then the following step is to determine whereabouts it will start "overpowering" the rest of the frequencies and when I find that spot I then I lower it slightly to make it more in-line with the rest followed by finetweaking by critical listening and jump to the next "band" or frequency point and do the same procedure and eventually I get a custom shaped curve for the particular headphone. Now I do like to compare with frequency graphs as well and most of the time it reflects very well what's seen there and if I get uncertain which frequencies to adjust how (which doesn't usually happen these days as I've tweaked EQ for so long now I can pretty much tell which frequencies sounds like what by now and call tell quite easily what should be changed for a better experience) but my EQing doesn't always perfectly counter the dips or peaks on the graphs and still sound "perfect" the way I EQ'd it and that's why I've come to the conclusion to trust my ears the most, they tell me how I want it to sound like, not the graphs.
 

 

[xnor]

rpgwizard, it's quite easy to verify if an EQ provides an accurate 10 dB boost if you call for it.
 
NecroNeo, the problem with headroom's 0 dB line is that it's a compensated version of what the dummy head recorded. The size and shape of the dummy head, his outer and inner ear etc. is designed after some mathematical average.
If the seal was okay during the measurement (problematic with some on-ear headphones), the lower frequency measurements usually are all right (as far as I can tell), but in the kHz range things get tricky. Compared to the dummy head your ear can react completely different in the treble range.
 
If you want your EQ curve to be as accurate as possible you don't get around doing measurements.
 
Ideally you set up your stereo speakers properly (and EQ them) to measure flat at the listening position. Then you sit down and switch between speakers and headphones and EQ them until they sound close enough.

 

[RPGWiZaRD]

Here's s how in my ears it sounds perfectly balanced using minimasticly EQ'd settings for my XB500 and Sennheiser HD 212 Pro compared to the headphone.com's FR graph:
 


 
But I also use another setting usually for my headphones being not a flat/neutral listening freak but forward/warm/bassy freak but so far only kX Audio's 10-band EQ have been able to handle such boosts without distorting or simply not sound like crap xD like most other EQs would with such settings, yet adding a more aggressive/powerful/forward sound I like (sounds more like speakers):
 

 
How weird it may look/sound like I think even this setting sounds very great even if it goes slightly against what most people here would say. I know it looks ridiculous but even if it's bassier at same time if there's no bass in the songs I feel the detail jumps out even better and clearly, in bassheavy passages the bass might take a bit more space than optimal (I like bass tho so that's ok) but it's not as ridiculous as it may look like.
 
(Should be noted that I seriously doubt +1.0 on this EQ stands for 1dB as I can notice difference between a lot smaller adjustments and in dB that would be quite difficult to distinguish if not impossible)
 
Just saying, use your own ears what sounds best to you.

 

[xnor]

I wonder why do you cut around 250 Hz and boost at 2 kHz with the HD212? Have you tried out boosting at 200 Hz?
 
The cut makes sense on the xb500, but is the sub bass boost really necessary? Also, did you notice a peak around 5 kHz with the xb500's?
 
I don't think that +1 isn't one dB but maybe the bands are weird or the x-axis labels are off? You could also take a loopback cable to record the soundcard's EQ'd output and analyze the spectrum to see what's really going on.

 

[RPGWiZaRD]

Assuming you're talking about that 2nd graph, yes it's very necessary to keep the 250Hz region a bit lower or else it starts bleeding into the mids, with a steep low-bass curve it doesn't color the mids as much, at the minimalisticly tweaked EQ setting I can keep 250Hz slider higher in comparision as bass isn't boosted and it won't color the mids but if I did the same thing on the maximal setting then mids would suffer a lot from "muddiness". When starting to boost the low frequencies as much as I did then you'd realise such a bass curve works a lot better, it even resembles the Arrowhead 12HE amp with bass boost and this amp is concidered like having among the best bass boost for an amp and it doesn't suprise me with the way I usually boost the bass on my EQ. 
 
Oh and I only use a 10-band EQ (that's why the x-axis features those frequency values) because it's the best working EQ I've used, works a lot better than the parametric EQ Electri-Q even if I lose the precision of a parametric EQ, it hasn't been any problem achieving wonderful results yet at least.
 

 
I didn't notice any peak at 5kHz on XB500 but around 7~8kHz it's more "in-line" with other popular headphones while the rest of the high frequencies are a bit veiled.

 

[k00zk0]

Well we can see that different people's ears definitely say different things!
 
Too many factors would change why a certain person would need to boost or cut a certain area.. perhaps you heard some noise as a kid all the time and the brain/ears dropped some sensitivity in that frequency range (or boosted it, a pleasant noise was present?)
 
I'm surprised how far a flat-sounding curve of sine-tone adjusted volumes is from one that actually sounds natural to the music... It definitely needs something different than sines. I think narrow-filtered pink noise is the correct thing. There are some generators for that, namely one included as an option in the head-fit crossfeed vst plugin. Something about the width of the band it filters, correlating with the way audio works in nature, versus a synthetic sine tone of one bandwidth.
 
There is definitely clipping if one raises the equalizer above 0db, without lowering the preamp. Any signal near the max height already will get flattened to the max. What I meant was that the actual function an EQ uses creates some kind of extra products or distortion when raised, rather than lowered; at least this is what I think I've read. In this case, After making a curve that has the highest peak at +5 and then setting the preamp to -5, the better thing to do would be to lower the entire eq curve by 5 to bring the max peak to 0. I am not sure of this, I was hoping someone could clarify whether it actually ends up as exactly the same signal. EQing should always yield a lower power signal else you are taking the chance to go into clipping.
 
Any word on peak type (not Q/width, but Type 1 vs Orfanidis type?

 

[xnor]

I'd prefer a pre-amp feature over lowering the entire EQ curve as for example in a graphical EQ by lowering every slider. With some EQ implementations the result might be the same but with others this is not the case. In other words, with a graphical EQ I'd leave as many sliders as possible at zero and set the preamp to -x dB where x is the highest boost / slider position.
Why? Because with some graphical EQ's you don't get a flat frequency response if you lower all sliders uniformly.
 
Parametric EQ's usually have some kind of pre-amp or allpass function built-in. Use it to lower the curve like described above.
 
edit:
On peak types, have you read e.g. the Electri-Q manual? Choice depends on where you place the peak and what frequency response you want to achieve, and other variables.

 

[k00zk0]

Resurrecting -
 
Yeah, the manual gets into a little detail. Orfanidis method avoids artifacts with greater than half the nyquist at the cost of more resource usage, but with 30 points all Orfanidis type, I end up with negligible CPU usage anyway (5%)
 
Here's the new case:
 
I bought Electri-Q. There are some new engines that are apparently better than "Digital" based on what I can find, namely the Linear Phase, Linear-IIR, and FIR-min. Using any of these also adds new types of peaks, such as S-plane. From some discussions which were had closer to the release of these new function types, before FIR-minimum though, users reported S-plane peaks along with the Linear Phase was the best solution, save for a little delay it put in the music to deal with the phase shift.
 
At the moment, what is the best configuration of peak type, and engine to use? I assume FIR-min may be best as it is the latest thing added which "closes the gap between linear phase and digital mode", whatever that means. :S Linear-IIR uses double to triple the CPU of the other modes hinting that it is somehow better.
 
Go head-fi!