How to Read a Frequency Graph

Hi, I'm new to the forum.

I read many good reviews about the Sennheiser PX 200 II headphones.
I found the frequency response graph here (under the tab "Specs"):
Sennheiser PX200II

I know some dips in the highs are normal, but to be sure I would like to know how to interpret the deep dip at 7 kHz.
Do I notice any sort of lacking sound at 7 kHz?

Thanks.

 

I have several of this headphone, and the highs are soft and rolled off some, but I have not observed a large dip in the area of 7 khz with one channel or the other. But there is some small variation between samples. The overall sound of the PX200II is greatly improved with EQ, i.e. the 'rock' EQ setting on Apple players.

 

I'm not a scientist so you will have to excuse the lack of supporting information to this but this is how I understand the reason headphones (and speakers) do not have a flat output for maximum accuracy.

As human beings our ears and brains are tuned for our environment (presumably by thousands of years of evolution) so that some frequencies stick out and other frequencies (past 20Khz) are inaudiable. If you were making a record and were therefore EQing the music and removing pops, clicks and other unwanted sounds you want to hear exactly what is there, good or bad, so you can change it. Once this work has been done, record released and now someone is listening to it with headphones there is no need to accentuate all parts of the frequency range equally to the detriment of the all over sound.

If you download a tone generator you can have a listen to what note is present at any point in a frequency graph. Above 7Khz you are out of the range of the notes you will find on a piano (as a pure note) but the sound at the top (and the very bottom) of this scale is still part of the harmonic effect of the music and effects the interplay of the audiable sounds too. The tones around 7Khz are pretty tough on the ear, at any volume these tones should make you cringe a bit. Real instruments do not produce pure tones, their character comes from their sound being unevenly spread across a large frequency range.

As I said before, some frequencies appear louder than others, almost certainly this is to help with speech recognition. I do not know whether the ear has developed physically for this or if it is a function of the brain. The better quality the headphone the less it is necessary to prune what is there - by producing a lot of detail a well balanced headphone should appear natural. A cheaper headphone will produce a sound with more distortion - not necessarily horrible noises, just more difference in the output sound compared to the signal you put into it. The easiest way to make an average driver sound better is to tune down frequencies which stick out or frequencies where "mud" gathers thereby making it sound more natural and less confused or inaccurate.

What I am trying to say is - the less accurate the headphone driver the more shaping that will be needed to make it sound good or natural or "right".

That might be more than you wanted to know.

 

Just realised I didn't even answer the question!

Forget frequency graphs and read well written reviews from people like Dale Thorn, Marcusd or Quadpatch - they are great at describing the vibe of a headphone which is virtually impossible to accurately interpret from a graph. The Graphs work well if you are comparing headphones, they often help me work out why two headphones have a very different feel or where one is missing out over another. In other words, they are probably more help after you already own the headphone!


Sennheiser PX200-II Foldable Headphones in Black (PX 200-II)
£56.95


Headphone reviews

 

I just remembered something that explains the treble dips, partially at least. Back a long time ago, Beyerdynamic and Sennheiser published FR (freq. response) graphs that looked like their headphones actually sounded. But later on when I bought a Stax headphone, I got a FR graph in the box that was all choppy on the high end, and there was an explanation written there that said words to the effect "This is how the headphones measure and the way our ears hear" etc., and following that somewhat rude awakening to choppy FR graphs, I've wished ever since that the industry would get together on a standard for normalizing the upper end of those graphs. Normalizing is simple - you take perhaps the 10 best headphones for treble in the world as agreed by respected audiophiles, then plot out the treble measurements for all 10 and average the output into one curve. Once you have that, you make an inverse curve and apply that to each new headphone you test, so the FR graph now shows the *deviation* from the erstwhile flat, neutral, or ideal curve. Then anyone who publishes measurements could publish both the raw and normalized graph together. Or, if the industry agreement results in 3 standards instead of one (for example), the computer can apply all 3 curves and print them as easily as one.

The point of all this (very important) is to make it easy for someone like myself to scan across a series of FR graphs and quickly identify how the headphones sound, because the choppy stuff will be replaced by a line that represents the actual peaks and dips. Even if the normalization curves aren't perfect, anyone can quickly adapt to their favorite and apply the compensation mentally. However, there is going to be a lot of manufacturer resistance to this, for obvious commercial reasons. i.e. not everyone wants you to know just how choppy that high end really is.

 

That is exactly right! Ultimately most FR Graphs are not really that comparable - there being no standard of how to display the information - as they all look slightly different. Who knows how they measured the response in the first place and how much the graph you are studying is a nice graphic made by a packaging designer.

 

Has anyone else seen this website? I'm sure Dale has at least.

Effin' Ringin' - Headphone Measurements: Frequency Response and CSD Waterfall Plots

It's one guy testing a bunch of headphones and making the FR charts but also other charts like waterfall graphs and driver matching. I only found this one recently and find it interesting but still hard to interpret.

For example, here are two charts for a topic that Robin and I are discussing at the moment. One if for the £600 Denon (although no longer sold) and the other is the £1500 Fostex TH900. Both headphones are made by Foster Electronics and look nearly identical. They sound similar but the Fostex is better in most ways, some by quite a bit:

Denon D7000 - http://www.changstar.com/index.php?action=dlattach;topic=29.0;attach=25;image
Fostex TH900 - http://www.changstar.com/index.php?action=dlattach;topic=351.0;attach=1373;image

Now I can't look at these and say ah yes there it is, that's why the Fostex is so much better. It looks like the Fostex should have way more bass - not really. The Denon looks terrible on the treble - not really. They both have big dips around 500Hz which look terrible, but believe me they do not sound terrible there. Maybe that's to curb the powerful upper bass... Then I look at something like the Beyerdynamic DT880 - very flat by comparison and no dip at all at 500Hz... 0.o

I think it can be hard to get much out of FR graphs unless something is drastically wrong.

Check links on the lower right-hand side of the main page for links to charts for each headphone that they've tested.

 

Nice! I have seen these graphs in posts on other websites and wondered where they came from.

It isn't surprising that most headphones are not that flat. When mixing a musical recording there is so much useless information to get rid of. by the time a mix is nearly there you should be rid of nearly everything below 30hz, have a decent dip at 240-250hz probably a little one 450-500hz and everything should begin to roll off after 5Khz. The sounds below 30hz are virtually inaudiable and simply interfere with other sounds, 240-250hz and 450-500hz are where all the mud and hiss are and as you disappear over 5Khz you get to the harsh part of a childs scream, the really piercing bit which you can't ignore. You get there by cutting the bits of the instruments that will interfere with each other (or that just sound bad) but when you watch the metering it should be where you get to. The process of voicing headphones (for music listening) must be similar, there would be no point in accentuating any of the frequencies which will irritate.

 

Changstar are the so-called "pirates" that Innerfidelity featured recently. Besides being a mean bunch (as you would expect with pirates) and having certain other issues, their stated purpose as I remember is to reveal what's wrong with headphones through graphs and measurements. My view of this is that the typical audiophile who really loves music and wants to get some good headphones with the least possible research effort, will look for reviews that describe the sound and the anomalies in plain English, rather than lead them into graphs and charts that require a techie perspective that works well only for electronic hobbyists. When I was growing up I knew several people who were electronics hobbyists, and I greatly admired their goodies and some of the results they got, but what I didn't go for is descending into the hobby, which in my view was a distraction from the end goal. As it turned out, whenever I had a question for my hobbyist friends, I could never get an answer because they were more tuned into the hobby (the gear) than into using the gear to achieve the ends I was seeking.

So in summary, I think Changstar is going to serve people much like themselves, and have little or no usefulness outside of gear hobbyists.

BTW, on a related note, my Youtube stats show that the viewers are 94 percent male, which is hugely disappointing to me. I don't know if that's entirely because of the hobby aspect of headphone etc. gear itself, or whether Youtube's image plays a role in that. But I see as many women wearing earphones as men in public, and I'd like to make a contribution to bringing them to headphone sites, where they wouldn't feel isolated by the likes of pirates and so forth.

 

I know around 85% of headphone purchases from our shop are to men. The percentage of women buying goes up hugely at Christmas. From my days chained to the phone (I loved it really) I can say that women ask excellent questions and are much less suspicious of the answers. They are much better at purchasing things for other people (by describing what those people like) but much less into researching things for themselves - therefore they do not tend to spend as much on themselves. I will say that we do have some hardcore Gadget Girls who buck that trend completely

 

I thought that dip could mean lack of sound at 7 kHz... but as you say probably it is better to read a good review than trying to figure out a sound from a graph.

Any suggestion for neutral sounding and affordable headphones?

 

Yeah, the deeper the dip the less sound at 7 khz. But some dip is normal, so the trick is, how much? The only real problem I see is if the peaks and dips are exaggerated with respect to each other, so that the normalized curve (if you could see it) would also be really jagged, meaning the sound would be harsh. Until someone comes up with a scheme to make these readable it's less useful the way it is.

The best and most neutral sound I've heard for $200 USD or less is the ATH ESW9A. But that's a small on-ear believe it or not, so for more size and a larger sound that's excellent, here: GermanMAESTRO GMP 8.35 D Monitor Closed Back Studio Headphones

Here's a great one for less: Beyerdynamic DT770 PRO Closed Back Headphones (DT 770) - Professional Studio

Getting into the $100 and under category, the Beyerdynamic DTX501p I had was very neutral. I've heard it's a OEM'd Soundmagic P30, but I can't confirm they sound the same.

 

These are surprisingly good for their price but the pads will not suit everyone (I think they are comfy but I know not everyone agrees)


AiAiAi Capital Closed Back Headphones with Mic in Midnight Black
£89.95

 

I've been meaning to ask for ages, how do you pronounce that manufacturer's name? I wonder how many sales they loose because people are too afraid to mention it

 

I think you are supposed to pronounce it as if it is written eye, eye, eye. Some people just say A I. There is a definite danger of this, it doesn't seem to effect Sennheiser - for whatever reason people say this Schnizer, Sinizer, Shennhouser and my personal favourite Schizer (which I'm sure they are not pleased about!).