CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS
Lobby => Headphone, IEM, and Other Audio Related Discussion => Topic started by: Bill-p on January 29, 2015, 07:48:08 PM
-
I thought this was an interesting discussion... as it would give quite a bit more insight into what's going on in those frequency response and CSD graphs that we have access to, and it may give way for more research later.
Personally, I hear frequency ranges like this:
<20Hz: felt as puffs of air
20-60Hz: rumble, theater/battlefield atmosphere stuffs
60-150Hz: impact
150-300Hz: bloat when resonant or distorted
300-500Hz: warmth in vocal, or some may call it "bass bleed"
500-1KHz: cuppy, vocal emphasis, boosting this range will most likely cause a forward/in-your-face sound
1-2KHz: what some refer to as "upper midrange". Audio Technica headphones in the past typically boost this range quite hard core (can be seen on FR graphs)
2-4KHz: "shouty", as some would say. Having a healthy amount of this range will give "texture" to vocals, but too much and it's harsh/annoying
4-6KHz: "edgy/sharp", I think this range specifically boost "definition" and "resolution" per se. But too much of it can be grating (a la HD800)
6-8KHz: sibilance... plain and simple.
8-10KHz: peakiness, sparkle, some call this "high frequency extension", but I hear it more as "limit of my ability to withstand high frequency sounds"
>10KHz: just gets more and more piercing until it dies off. Some call this region "air", but I don't think so. I think the misconception stems from the fact that when people force an EQ for these frequencies, their headphones sound more "open". Personally, I think the phenomenon is more closely related to how DSP (either in software or hardware) is shaping the impulse in order to match the EQ settings, and as a result, the impulse decays faster than usual, thus giving this impression of "air". This is just a guess though, from simulating square wave and looking at their forms when EQ is applied. I'm not sure this is the case in reality.
In my opinions, what many refer to as "air" is simply a faster, or more "natural" decay of mid to high frequencies. This can be achieved pretty easily by open-back headphones because they have very little enclosure resonance as opposed to closed-back headphones. Most closed-back headphones are very poorly damped, and this prevents them from being "airy".
On the other hand, being "airy" alone doesn't necessarily mean a headphone will have great soundstage. I think "air" aids in accurate reproduction of sound imaging and "separation", but "soundstage" is an entirely different ballpark that probably has more to do with the acoustic space in between the ears and the diaphragms. From what I can tell so far: distance from ear to diaphragm determines distance to sound source, and thus aids in creating "depth" and "width", and angle of the driver/diaphragm relative to the listener's ear would determine the "shape" of the soundstage. But I haven't been able to pinpoint exactly what would create a good "height".
Anyway, please discuss, and be merciless in your refutation/agreement of my opinions.
ahoy
-
I agree with a lot of that. <60Hz is primarily rumble and some impact. 60-150Hz or so is mostly impact and a sense of thickness and body to the sound. 150Hz to somewhere around 300-500Hz also adds a sense of thickness but with more of a "bloat" to the sound or, if the response is high enough and centered more around that 300-500Hz spot, it makes things sound wooly (think everything sounding like it's going "wow" or "whoa"). Anything up through 1KHz or so seems to give certain instruments and vocals a sense of body and presence (i.e. not thin).
1-2KHz or so can sound shouty. Best way to describe it. An emphasis in the 2-4KHz or 5KHz range starts to sound "hot" to my ears. Enough of it starts to give the sound a rough, grainy quality (almost like, uh, loud sandpaper).
5-8KHz or so primarily has more of a raspy sort of rough quality.
9-11KHz or so is less offensive when elevated and starts to give the impression of definition. Too much of it can start to make things sound thin or tinny. Or too sharp.
Generally elevation above 11KHz can be a mixed bag. Usually it's not so much of an issue, but sometimes it can just me feel uneasy in ways that are hard to explain. Like a light piercing.
I do think frequency response plays a large role when it comes to airiness and soundstage, but it's not the only criteria by any means. And I think there's more to air/soundstage than upper frequencies. For example, with certain instruments and vocals, having good midrange quality from top to bottom can give them a better sense of lower end reverb and body that assists with a different sort of airiness and placement. I've also heard headphones with certain treble peaks that make things sound closer and more congested than they would otherwise. And some headphones with depressed upper treble response do sound less open and generally more compressed. There are always counter examples, though, so it's a loose correlation. I think it's best to have a sort of balanced response from top to bottom to capture everything well.
All just my opinion, of course...And in terms of what I find pleasing or not, that's a whole different story. I've found I'm OK with some extra bass (probably inner basshead and the fact I started my audio journey with a bass-heavy DIY car system), though I don't care for an emphasis centered around 300Hz or so, but anything in the upper-mids or treble that sounds subjectively elevated or peaky, even slightly, is the fastest way to kill something for me (even when I'm not listening to my go-to genres or at low volumes).
-
Good effort to build vocabulary.
Described in terms of peaks (harder to hear nulls for the most part):
125
Punch
250
Warmth
400
Fullness
Muddy
600
Nasal
1k
Honk.
2-3k
Vocal shrillness.
Guitar crunch.
Shouty.
4-5k
Adds a bit of edge, bite, and excitement to the sound.
Detail (macro / coarse).
6k
Generally tolerable, but unpleasant.
Similar to 5k in that it adds some bite
Hardness bordering on glare.
7k
Glare galore with some sibilance.
Sharpness.
8k
Very bad. Nasty. Painful.
Sibilance with some tizziness.
9k-12k
Adds definition.
More tolerable, but still unpleasant if too much.
Tizziness and sizzle.
-
MORE - give or take a 1/2 octave
Also, "working" terms, so Nazi's need not apply.
Sub Bass < 40Hz
Low Bass 40-100
Mid Bass 90-180
Upper Bass 180-300
Lower Mids / "Fundamentals Region" 300-700
Middle Mids 600-2000
Upper Mids 2000-6000
Lower Treble 6000-8000
Mid Treble 8000-11000
High Treble / "Air" - 11,000-20,000
-
From my personal experience through the years:
<40hz: sub-bass. felt more than heard, not as big of a factor on headphones because they don't pressurize much air
40-80hz: low bass. a healthy amount adds a solid foundation to sound. too much muddies the sound.
80-125hz: mid bass. emphasized mid-bass can lead to added sense of impact, too much can lead to muddiness
125-250hz: upper bass. a healthy amount that isn't bloated leads to warm sounding qualities. bloated upper bass can lead to dark sounding headphones. is more thickness than impact.
250-700hz: lower midrange; fundamental region. a healthy amount leads to warm and natural sound, too little leads to cold sound, and 'closed cup' coloration.
700-2000hz: middle-mids. presence region, voice recognition and start of many instruments' harmonics. too much can lead to shouty, plasticky qualities. too little can lead to hollowness and thinness. I prefer slightly less than neutral in this area.
2000-4000hz: upper mids: think guitar crunch and upper brass and string energy. Too much can lead to edgy and hard sound. Too little leads to dullness.
4000-8000hz: lower treble, adds mixture of edginess, snap and breathiness to music. Too much leads to sibilance and glare; primary sibilance area.
8000-11000hz: mid-treble, definition area. A lot of headphones have a peak here to add a false sense of 'hi-fi'ish'definition. Small peaks are tolerable, but too large a peak is a disaster. Especially hard to control resonances in this area. Tizzy.
11000hz and beyond: upper treble-- is felt as pressure more than heard-- just like sub-bass. adds 'air' to sound. Too much is piercing.
-
Made this picture a while ago
(https://diyaudioheaven.files.wordpress.com/2013/03/graph-coloration.png)
-
Dubstep girl agrees.
-
Thks for the thread. This is really a topic I'm interested in.
Can we to sum up something as a collective work we can make reference to ? At best could we associate some useful track tests well suited to analyze each FR ? And examples of headphones that can illustrate peaks and dips in their sound sig?
popcorn
-
This chart sill nails it, more or less, for me...
http://www.independentrecording.net/irn/resources/freqchart/main_display.htm (http://www.independentrecording.net/irn/resources/freqchart/main_display.htm)
-
Another thing that really helped me was listening to a playlist of ascending test tones, while looking at such charts. I do not recommend this as a tool to actually analyze your headphones, since it is completely artificial, but it provides mental anchor points for each of the relevant frequencies. Its one thing to look at a chart and wonder if you are envisioning the correct type of discomfort. Its another thing entirely to listen to an 8kHz tone and say "ah, yes."
I find that doing this periodically helps me listen more critically, and helps me make more understandable notes when I am listening to new / different headphones.
-
Dunno if anyone here tried the Philips Golden ear challenge but at last stage, there's an exercice where the goal is to detect peaks or dips at different frequencies : 63hz, 125, 250 , 500, 1khz, 2khz, 4khz, 8khz 16khz. I really would like to have the training tracks to help my brain to remembre the effects of dips or peaks a those frequencies.
-
Simply play your test music through a player with a multiband equalizer and play with the sliders.
This is how I 'tought' myself when I was young (used a real equalizer then which was hugely popular in those days.
Nothing beats those kind of experineces.
use a 'flat' headphone for this... HD600 or DT250-250 if you want 'flat' as a reference and play with colorations.
-
Simply play your test music through a player with a multiband equalizer and play with the sliders.
This is how I 'tought' myself when I was young (used a real equalizer then which was hugely popular in those days.
Nothing beats those kind of experineces.
use a 'flat' headphone for this... HD600 or DT250-250 if you want 'flat' as a reference and play with colorations.
You're right indeed... I did that exercice sometimes bit I forgot. Sorry ::)
-
Thks for the thread. This is really a topic I'm interested in.
Can we to sum up something as a collective work we can make reference to ? At best could we associate some useful track tests well suited to analyze each FR ? And examples of headphones that can illustrate peaks and dips in their sound sig?
popcorn
Will try to consolidate impressions into the first post as a collection of frequencies. I'll take the "broadest" width of a certain frequency range and then relay the impressions for that range specifically. That way, anyone who is interested in learning about it can just read those and cross-reference them.
Also a nice bonus would be that one only has to search the first post for certain "words" that people use to describe a certain frequency range to get a ballpark of what it means when they go "this sounds 'grainy'" or "vocal is shrill!"
Still not sure on what I should do for test tracks and all. Some things can only be noticed in certain tracks at very certain locations, and sometimes even just for a brief moment.
-
The problems I see here are a few.
1: dips are a lot harder to hear than peaks but lots of headphones have peaks.
2: sometimes sharp peaks preceded by a dip are hard to 'detect' or 'put a finger on'
3: headphones usually have lots of peaks and dips often exceeding 3dB and thus quite audible. It is difficult to 'break' those apart and one peak may mask dips and vice versa
4: resonances may not be visible in FR plots but can affect the sound, CSD and/or impedance must be included.
5: sometimes wildly varying HF can be caused by the measurement technique while in reality it is smoother. This could lead to an assumption of grainy highs where it may not be the case.
6: a sharp drop-off above 13kHz isn't detected as lacking 'air' all the time. A peak that isn't too high may be masking it.
7: Some people find LF roll-off not objectionable and say bass is tight yet another one, who knows what should be there, may find it lacking bass.
8: a peak in the 100Hz area with a steep drop-off in the subbass may still be perceived as a bassy headphone while it has no extension at all.
In short.. It is pretty difficult (if not impossible) to characterise a headphone's sound by looking at graphs.
Certainly not when graphs are taken by different people.
I have given up characterising it and know what to look for by now.
I like MY headphones to measure as flat as possible from 20Hz to 20kHz within +/- 3dB and this MAY get you a tonally balanced headphone but NOT necesarily a headphone everyone likes nor will it always sound good with any recording but it will with good ones.
The least possible resonances and 'echos' from the rear of the driver seems essential and is hard to see in FR and interpret from CSD.
I like the idea of characterising headphones myself and is why I made the FR plot where you can find substantial peaks at those points providing it is just one or 2 peaks that F up the sound.
Then there is subjective impressions that differ. One mans bass shy may be another ones 'real bass'. One mans 'good bass' will be another ones 'bloat', one mans 'good treble' may be another ones 'I hear no highs' or 'piercing s#it'.
This too (perception) may be the biggest problem of all.
So my motto has become (for myself only) have an educated guess from plots when available while keeping in mind WHO posts the plots.
Read reviews and look for common traits in it or find one from a person who's hearing you trust.
But the best one of all is to audition one AT HOME for a while and compare it (one gets used to a presentation really quick).
Sell it on and regard the money loast as 'educational fees' or return it if possible and you don't loose out too much.
Just my POV
-
Hmmm... I can't seem to track down the old practical EQ guide for recording and mixing. It was surprisingly precise about certain frequencies such as the wall between sibilance and vocal body (8kHz down, 9kHz up).
Anyway, a good read from the recording perspective is here: http://www.dummies.com/how-to/content/home-recording-eq-frequency-guidelines.html
With so many variables, iit is hard to get a recording to sound its best. It gives credence to those who want to hear their music as closely to the master as possible.
-
So I've been tuning the frequency response of my modded LCD-2 slightly, and it seems like boosting 4-4.5KHz by a very small amount can help a lot with snap, speed, transparency, etc...
Given that the LCD-2 is a bit de-emphasized in this area, perhaps this is why it sounds so dark despite having a spike at 10KHz?
Anyway, will investigate more.
Edit: not quite so. Darkness is something else. But 4-4.5KHz is definitely the edge/snap. I think "sharpness" is higher up (5-6KHz?) since I'm not hearing sharpness, but definitely a lot more snap, more speedy after bumping 4-4.5KHz.
-
So... going with the discussion at hand, it really seems like sharpness is in the 5-6KHz region now. I think beyond 6KHz is where sibilance may occur.
CEE TEE recently introduced me to Adele's Chasing Pavements track, and I got the iTunes version following my meeting with him. Dear God, this recording is sibilant to no end. What I have found is that it seemingly has a boost to the entire upper region, and the boost is wide enough that it reaches past 13KHz. It gives some extra definition and clarity for something like... say, the Beats by Dre, but it is detrimental to a lot of high-end headphones with an emphasis at higher frequencies. Going by what I'm measuring from my phone, the 12KHz peak is unusually high compared to many other recordings, and then there's also a peak around 7KHz. CEE TEE also mentioned "squeakiness", which I think is around 1-2KHz. And so... when I looked up the Beats By Dre Pro's frequency response on Innerfidelity, I saw just that:
http://www.innerfidelity.com/images/MonsterBeatsPro.pdf
Hm... so to actually hear that one recording (or perhaps the entire album) the way it was intended to be heard, I think we all need to buy Beats By Dre Pro.
:P
-
CEE TEE recently introduced me to Adele's Chasing Pavements track, and I got the iTunes version following my meeting with him. Dear God, this recording is sibilant to no end.
Being a glutton for punishment, I had to check this out on Spotify. Ouch. Even with my HD600, which are pretty forgiving in this regard, there is painfully pronounced high frequency lift. I'm scared to listen to it through HD800's. :vomit:
-
Frequency peaks when I measured Adele's Chasing Pavements with my phone (notice gray line and ignore the rainbow stuffs, that's just current ambient noise levels - noise floor, also cursor is measuring lower peaks, so no idea what the higher peaks are, but you can count relative to 1KHz, which I used to volume match both headphones):
Modded LCD-2: (peak at 12.5KHz)
(http://s21.postimg.org/kdhx4khjr/image1.png)
Modded HE-560: (peak at 10KHz, and overall treble from 5-8KHz is more "lifted" than LCD-2, but not terrible)
(http://s10.postimg.org/9g5zerbsp/image2.png)
Those spikes at 10 and 12.5KHz... ow. Luckily, I did mod these headphones to be a bit on the "dark" side so they don't really spike up that crazy, and they are sort of bearable with this song, but I'd expect the HD600 and HD800 to spotlight that section pretty bad.
Just a thought: we can probably measure frequency response or CSD using songs instead of pink noise or sine sweeps, and maybe it'll give us some insights into how a headphone behaves with certain songs.
Oh, and recessed 500Hz? Very audible. Makes the song lack body/soul/fullness, etc...
Edit: and the LCD-2 sounds a bit more distant/far away compared to the HE-560, even though the HE-560 is a bit brighter, and yet fuller at the same time. I'd guess... decay characteristics or speed? But I wouldn't know.
-
Haha, imagine a few months later, I look back at this and feel like I was just talking out of my arse.
Since I have a much better measurement system now, I have been measuring and observing the effects. So far, I've found spectrogram graphs to give a more complete picture into what's happening at lower frequencies.
-----
For instance, let's inspect the spectrogram of Code-X:
(http://www.changstar.com/index.php?action=dlattach;topic=2092.0;attach=9251;image)
Notice how 80Hz has amazingly fast decay in Code-X? This is audible to me as amazingly fast and snappy drum hits that allows the drum in Eagles' Hotel California - Hell Freezes Over (K2HD version) to actually sound like "drum hit".
Let's inspect the Abyss now:
(http://www.changstar.com/index.php?action=dlattach;topic=2092.0;attach=9253;image)
Bigger red region is due to an FR notch at upper sub bass and lower mid bass that gives a sense of authority and body/thud. In practice, Abyss also has a small region right above 100Hz that has amazingly fast decay. Any song that has bass in this region will sound insanely "fast" with the Abyss here, but overall, Abyss just has a greater sense of body/impact compared to many other headphones.
Now let's take a look at my LCD-2: (haha, ignore the settings window! Unless you wanna reproduce my graphs)
(http://www.changstar.com/index.php?action=dlattach;topic=2092.0;attach=9257;image)
It has good bass definition (yellow region is fairly small), but I think the green region trails off a bit too far, almost twice as long as Code-X and Abyss comparatively. Subjectively, this means the LCD-2 does sound very very slow, bloated, full down there. It's much blurrier compared to the other 2 headphones.
Then... I did some acoustic mods to it and achieved this:
(http://www.changstar.com/index.php?action=dlattach;topic=2092.0;attach=9255)
It's much better than how it was before. Mid bass is still not as clean/clear as either Abyss or Code-X, and that contributes to a slightly blurrier bass impact, but it does have fairly decent body and thud in comparison to the Code-X, which can sound... kinda bass-light in comparison.
-
EDIT: For 80 Hz 100 ms might be OK.
-
I'm using 50ms -> 20Hz. You can see my window parameters in the spectrogram for the LCD-2 there. WIP14 is just a work-in-progress mod of the LCD-2.
Also I edited the previous post with some comments. :)
-
This is really cool stuff Bill. I'm really interested in seeing how some dynamic headphones behave using the same system.
-
Well, glad ya asked. :)
Attached Spectrograms of the ATH-M50 and Beyer T5P.
See... I kept wondering how you'll ever be able to see bass "speed" with CSD. CSD will only show ringing and how clean/clear treble is, but it won't show this stuff. Now that I know, I'll include a spectrogram in every single measurement I'll make moving forward. It'll at least help give a more complete picture of how a headphone should sound like.
Anyway, a few notes on the dynamic headphones measured:
M50 shows quite a bit of decay in the bass, and the yellow region from 20-40Hz is like... very blurry. Subjectively, I think that's why it sounds very loose in the sub bass region. Impact has good definition (clear yellow line), but also excess decay blurs and muds up this bass. Orthos do much better here IMO (except for LCD-2). I think stock LCD-2 tuning is warm in order to mask the muddy bass... But that may just be me.
T5P barely has any bass here to begin with, so there is almost no decay to speak of. Some say this makes it sound "tight", and I guess... that's to be expected.
Oh, and bonus... attached modded HE-560 graph. Amazingly close to Code-X. :)p1 But notice the yellow region from 20-40Hz is again quite blurry, so definition is somewhat lost, but still, impact and everything else is very impressive.
Edit: to do -> inspect spectrogram at different windows. I think I'm seeing other behaviors like "warmth" and "smooth" here, too. Oh man, this is like a friggin' treasure chest! :)p1
-
Bill, I could be wrong but a CSD and a spectrogram are roughly equivalent, and kind of different visualizations of similar things...
Utrasone 880 CSD
(http://www.changstar.com/index.php?action=dlattach;topic=2092.0;attach=9265;image)
Ultrasone 880 Spectrogram
(http://www.changstar.com/index.php?action=dlattach;topic=2092.0;attach=9267;image)
As you can see the "ringing" at 4 kHz and 5 kHz are visible either way. In particular there seems to be a small notch (discontinuity) around 5 kHz which in time will generate some ringing (due to a mechanism similar to the Gibbs phenomenon). For the 4 kHz case, the effects may be due to a narrow peak. Phase may get wacky around there as shown here (phase is the plot below the magnitude):
(http://www.changstar.com/index.php?action=dlattach;topic=2092.0;attach=9269;image)
When phase gets wacky, frequencies are not delayed by the same amount. Frequencies 5 kHz and up may lag a little the rest of the spectrum immediately to the left. The inflections may correspond to the "ringing". How does that sound? Proly weird. There are some descriptions about it here:
http://www.changstar.com/index.php/topic,71.0.html
Similar stuff may happen in the bass area for this particular phone.
I'm not sure if these will tell you much about "speed" or how "clean" things are. Distortion plots might help with the clean aspect of things. Speed can mean many things depending on context I guess.