CHANGSTAR: Audiophile Headphone Reviews and Early 90s Style BBS
Lobby => Headphone Measurements => Topic started by: Marvey on April 22, 2015, 06:25:00 AM
-
MEASUREMENT FAQ
Q: Can't measurements bias one's perception of a headphone?
A:
Of course! However, lots of other things can influence how a headphone is going to sound like. Comfort, fit, construction, build quality, marketing, advertising, Jude's YouTube videos, hype (deserved or undeserved), price, et. al. will also influence how a headphone is going to sound like. The good thing about measurements is we don't have to look at them. Longtime readers of this site will know that I always listen to a headphone first to jot down subjective impressions before taking measurements.
We can also make a counter argument: measurements might help us become better (more critical) listeners.
Q: Can measurements "lie"?
A:
Absolutely. Measurements don't tell the whole story, the information they provide is too limited. This is why I like to post as many different kinds of measurements as easily obtainable: frequency response, cumulative spectral decay, and distortion plots. I like to think of measurements as a tool. Measurements can more easily tell us if something is wrong with a headphone than if a headphone is awesome.
The frequency response measurements have perhaps the strongest correlation to what we hear as tonal balance. CSDs seem to be most effective at discerning any issues (peaks) from the midrange on up; and have some limited correlation to articulation and resolution. Distortion measurements tend to relate to overall clarity, "blackness", "fidelity" of sound; and appear to be the most difficult for measurement newbies to correlate to subjective sound quality.
Q: Should I use measurements to determine whether I should buy a headphone or not.
A:
Best to get the headphone and try it out for yourself. Otherwise use numerous sources. Talk to friends who have heard it. Talk to people you trust. Use measurements. Consider all sources of information for proper due diligence. Use measurements here. Use Tyll's measurements. Most of all, consider negative information, whether they be measurements or subjective impressions. It's impossible to get transducers to do everything right.
Q: Your measurements say bad things about headphones like the Grado, therefore I don't agree with your measurements.
A:
That's fine. I respect John Grado for tuning his headphones according to a vision. He had a vision for a certain sound, and he's certainly hit the target right on the mark. I'd be more worried if Grado headphones sounded all over the place.
While you may not "agree" with the measurements, the measurements do tell us that Grados (RS-1 and below) have a consistent house sound that is very different from that of Audezes headphones. I'd argue that measurements tell us more about the differences between a Grado and an Audeze and than two fanboys arguing over which one is better.
Finally, measurements have nothing to do with personal preference.
Q: Your measurements look different from manufacturer provided measurements.
A:
Measurements can be used as marketing tools or inadvertently improperly presented. One "trick" I've seen recently with measurements is to present distortion graphs using a linear instead of logarithmic scale on the Y-axis. We hear on the logarithm scale. The use of linear scale can hide "problems". Also, SPL is a huge factor in distortion measurements. One way to lower distortion is to run at softer volume levels.
Other methods can involve setting the floor too high on CSD plots, using excessive smoothing, visually compressing the Y-axis while expanding the Y-axis range on frequency response graphs, etc.
I tend to measure at high SPL and present measurement visualizations in a manner which I feel slightly exaggerates issues so we can "see" what is going on. Changstar is not trying to sell headphones or get sponsorships. We want to know all possible issues with a headphone - with the understanding that nothing is perfect.
Q: Can I compare measurements using one method with measurements using another?
A:
In general no, unless you are really familiar with the methods and how they tend to deviate from each other. However, this does not mean you cannot look at a measurement here, and then check InnerFidelity to see if there are certain consistent patterns.
Q: I don't understand the measurements here.
A:
If you own several headphones and we have measurements of them here, a good way to start out is to listen to the headphones and study the measurements. The more we do this, the better we get at understanding measurements. There are various ear training resources on the Internet which might give you a head start in understanding the frequency response measurement. Finally, you can always ask. Asking how measurements work is not a stupid question. Please ask questions.
This site started a few years ago with only a few headphone measurements. It has definitely been a learning experience even for me to see how measurements correlated to qualities of sound.
Like all technical things, it does make a little bit of effort and intelligence to grasp.
Q: What is the ideal frequency response for your graphs?
A:
This is a tough question to answer. My current V2 measurement rig is set up to be most comparable to a measurement taken at the listening position with a wand microphone for a pair of speakers. The ideal response for me personally is a flat response up to several hundred Hz with a gentle downward slope where by 20kHz it's 6db down more or less. This is pretty standard for how things are setup in a quality mastering studio.
Your ideal frequency response may be different from mine.
Q: Wait, are you saying a flat line on your current measurement rig will sound a little bright?
A:
Yes.
Q: I don't hear this "high" distortion I see on your graphs.
A:
Distortion is probably the trickiest to hear or describe. Distortion in most cases here (which is usually dominated by 2nd order), does not sound like an amplifier clipping. Distorted bass tends to sound muddy and blurred. Distorted midrange tends to sound dirty or unclear. Distorted treble often coincides with peaks an octave higher.
Based on discussions with people here, there seems to be high variability on people's ability to discern certain kinds of distortion. The other factor is what headphones are you comparing against? If you don't hear distortion on a headphone with high levels of measured distortion, perhaps it's likely that the other headphones you've owned have had similar distortion characteristics.
Finally, it's possible that you are not sensitive to distortion. That could be a blessing.
-
Great post purr1n, may be it needs a sticky in Headphone and IEM measurements forum.
-
Please ask questions. This includes newcomers or lurkers.
-
Marv,
The following has always bothered me: You have a rig, you measure and come up with graphs and plots. Now you use the same HP and connect it to a source, dac and amp. Clearly, the afore-mentioned three will most likely vary from person to person. So what is then the correlation between the generated graphs and what you actually hear ? Granted, it should be very close but source "coloring" could present a different picture right?
-
Marv,
The following has always bothered me: You have a rig, you measure and come up with graphs and plots. Now you use the same HP and connect it to a source, dac and amp. Clearly, the afore-mentioned three will most likely vary from person to person. So what is then the correlation between the generated graphs and what you actually hear ? Granted, it should be very close but source "coloring" could present a different picture right?
Not according to the basic measurements or to those that live and die by them only. There's a reason he uses my 'Objective' 2 as the amp for these measurements and initial subjective impressions. Same rig. You can plug and play the upstream of more than 95% of the rigs people use out there and not measure an appreciable difference at the transducer unless something is very wrong or broken.
Besides, the transducer is where most of the measurable stuff is happening. Including that fabled awful tube distortion which is rather inconsequential to what the transducer is presenting anyway.
Remember, most of us here do NOT believe that the basic measurements often presented on many forums tell us the whole story of what we hear. That's not this place, no matter what anyone would have you believe elsewhere.
-
Marv,
The following has always bothered me: You have a rig, you measure and come up with graphs and plots. Now you use the same HP and connect it to a source, dac and amp. Clearly, the afore-mentioned three will most likely vary from person to person. So what is then the correlation between the generated graphs and what you actually hear ? Granted, it should be very close but source "coloring" could present a different picture right?
Frequency response, CSD, and distortion, etc. are crude measurements only capable of distinguishing gross, but not minute differences.
FR and CSD are based on impulse responses which in turn are derived from noise or sweeps. To get FR or CSD, we need to process the impulse response through a time window. FR is an averaged response for each excited frequency through such a time window. In essence, all information of how a sound wave changes in time is lost. That's the trade-off for getting a nice visualization / presentation that we humans can understand.
We can always look at the waveform through time, however humans cannot decipher waveforms in a visual sense. Waveforms are not a good visualization for helping us understand how something sounds like.
For distortion measurements, those reflect only one excited frequency in a steady-state. Unfortunately, music does not comprise of steady state single frequency signals.
This is why so few measurements on DACs and amps. Unless there is something horribly wrong, the measurement differences are minute and below the theoretical threshold of audibility. In fact, not just minute, but dominated by the transducer. Transducer non-linearities are magnitudes higher than those of DACs and amp.
It's very strange. Global negative feedback will do wonders for measurements like lower distortion, but too much feedback, we collapse the soundstage, kill PRAT, kill liveliness, make things sound dead. The challenge is balancing (clean sounding but dead) vs. (dirty but lively) with the right amount of feedback. The point is that there something missing from the current suite of typically used measurements. As far as I know, there isn't much that can "measure" soundstage or PRAT other than our ears.
-
Adding some external links such as Tyll's articles could be helpful and complementary of this FAQ too .
i plan to write to complete series of article in french to sum up the whole problematic :
1. read and understand Measurements
2. Correlate measurements and sound
3. Correlate measurements , sound and music enjoyment.
IMO very few people understand the whole mechanism and I often meet people who don't understand how frequency response can affect the perception of the music. Very few people know something abou harmonics for example.
Kudos to Pyrates who take the time to share their knowledge. it"s a small light in an ocean of obscurantism .
-
Please ask questions. This includes newcomers or lurkers.
Can you recommend some common or popular headphones with high levels of distortion, and those with very low distortion, so that any of us who have heard or own those headphones can try to listen for the difference?
-
Great post, Marv. I think Tyll's talk at Axpona on "Finding Flat: How to Interpret Headphone Measurements" will be a pretty interesting addition to the information that you provided here.
-
This may not be the best place for these thoughts but I am reflecting upon measurements biasing impressions of a headphone.
RE: other influences that may bias one's impression of a headphone:
1. acute pain from injury or surgery
2. terrible mood from bad day (work or otherwise)
3. great mood from good day (work or otherwise)
4. libations
5. sleep deprivation
6. stress
7. new toy syndrome
you get the idea. this is not a comprehensive list.
In my opinion, it takes living with a headphone for a period of time, through many situations, to assess how it fits with one's personal preferences. As mentioned, the measurements are a tool, albeit one that is likely to be more consistent over time than we humans.
-
Thank you Marv. for an awesome thread.
I generally get the CSD and FR plots and how they're derived, but i'd like to ask, if I may, what the 'distortion' caluesa actually measure and also what is meant by 2nd, 3rd and 4th order distortion. Do you excite the diaphragm with one freq. and measure amplitudes of 2x , 3x and 4x freq.? Or am I totally wrong?
Thank you,
C
-
In my opinion, it takes living with a headphone for a period of time, through many situations, to assess how it fits with one's personal preferences.
This can't be overstated. Unless, you know, it's an Ultrasone.
-
Thank you Marv. for an awesome thread.
I generally get the CSD and FR plots and how they're derived, but i'd like to ask, if I may, what the 'distortion' caluesa actually measure and also what is meant by 2nd, 3rd and 4th order distortion. Do you excite the diaphragm with one freq. and measure amplitudes of 2x , 3x and 4x freq.? Or am I totally wrong?
Thank you,
C
Close. Let's take for example the 2nd harmonic of 1kHz. So when you see <x> amount of 2nd harmonic distortion at 1kHz, that means there is a signal at 2kHz that shouldn't be there when only playing 1kHz.
The limitation with harmonic distortion measurements is that they only show signals at shouldn't be there at integer multiples of the original signal.
This is a very interesting topic I will follow up on.
-
This can't be overstated. Unless, you know, it's an Ultrasone.
I can tell if a headphone is craptastic in 15 seconds if I am familiar with the recordings. There are many a time at headphone meets where I simply sit there bored staring at the ceiling because I don't want to be impolite and get up after 15 seconds.
I can tell if a headphone is fantastic within three recordings if I am familiar with the recordings and they are played back on good rigs I am familiar with. Headphones are easy to subjectively assess with experience and a consistent methodology. Amps a little bit harder. DACs most difficult
-
Can you recommend some common or popular headphones with high levels of distortion, and those with very low distortion, so that any of us who have heard or own those headphones can try to listen for the difference?
Headphones with highish harmonic distortion that intrudes into the mids:
SRH1840
SRH1440
HE-400
Alpha Dog Prime
Headphones with lowish overall distortion:
HE-560
SR-009
Dynamics with relatively good bass distortion: TH900, Abyss
Dynamics with relatively bad bass distortion: MDR-1R, SRH1840
One interesting thing is that distortion in the bass tends warm up the sound. If there are two headphones with the same frequency response, but one has more bass distortion than the other, the one with more bass distortion is likely to sound warmer. Makes sense since instead of a straight 50Hz, we would get a little bit of 100Hz, 150Hz, 200Hz, etc. This is not a necessarily a bad thing and I would argue that most people, including myself, are accustomed to some bit of bass distortion as normal. Good quality bass is really difficult to get right.
P.S. If you ever hear me say something is really clean sounding (subjectively), it probably (but not always) means means distortion is low.
-
I can tell if a headphone is fantastic within three recordings if I am familiar with the recordings and they are played back on good rigs I am familiar with. Headphones are easy to subjectively assess with experience and a consistent methodology. Amps a little bit harder. DACs most difficult
Yeah, I was really thinking along the lines of "you have to spend a bit of time with it in order to know if you can live with it".
-
Most of us divorce our headphones within a year or two anyways. :)p1 The other solution is to practice headphone polygamy.
-
Yeah, I was really thinking along the lines of "you have to spend a bit of time with it in order to know if you can live with it".
That was my intended message as well, fitting one's personal preference. Especially if one has not spent time developing and refining critical assessment skills, which requires practice. I will add that I would not trust my own judgement if I was having a bad day and give the headphone being evaluated another listen on a day when I'm in a better mood.
-
I haven't heard a headphone where my impression genuinely took a 180 degree turn after the first 10-20 seconds tbh.
Having said that, I thought the Beats (no idea which one) sounded ok at a meet, maybe I did need to give it more time :D
-
I have never noticed my hearing abilities to be mood dependent. Though it is very much time dependent for me. Best critical listening time is in the morning, daytime is ok too just have to concentrate a bit harder. Evening and night time is the worst by far, everything starts to sound more rough/edgy and harsh. Never noticed this with loudspeakers though.
Before every critical listen I normally do a simple massage which helps with both loudspeaker and headphone evaluation equally well (has something to do with equalizing the air pressure between the ears, stretches and warms up the eardrum and exercises some parts of the inner ear and bone conductivity.
-
Thank you for the information, purrin.
And have some questions about your measurement; what is compensation target of your measures? And want to know about dBSPL(or dBA) of you harmonic distortion analysis. Thanks!
-
I can tell if a headphone is craptastic in 15 seconds if I am familiar with the recordings. There are many a time at headphone meets where I simply sit there bored staring at the ceiling because I don't want to be impolite and get up after 15 seconds.
I can tell if a headphone is fantastic within three recordings if I am familiar with the recordings and they are played back on good rigs I am familiar with. Headphones are easy to subjectively assess with experience and a consistent methodology. Amps a little bit harder. DACs most difficult
^This. Marv is honestly being generous with that 15 seconds for some phones. I think for us there are certain sonic characteristics that are just deal breakers immediately.
-
I think for us there are certain sonic characteristics that are just deal breakers immediately.
Definitely : LCD-XC : 3s > poo , Chord Hugo > 2s > poo
-
Thank you for the information, purrin.
And have some questions about your measurement; what is compensation target of your measures? And want to know about dBSPL(or dBA) of you harmonic distortion analysis. Thanks!
The V2 measurements have absolutely no compensation. The rig was designed to not require any. See this: http://www.changstar.com/index.php/topic,2305.msg63409.html#msg63409 (http://www.changstar.com/index.php/topic,2305.msg63409.html#msg63409)
For the most recent measurements, the 0dbFS line = 100db SPL. -10db = 90db SPL. This is very very loud and tends to exaggerate distortion numbers. But I like it this way because it tends to separate the really good stuff from the mediocre stuff. It's more difficult to see differences if we don't push headphones hard.
-
Can anyone explain what I'm looking at and/or what I should be looking for when I look at a CSD graph?
For example I'm looking at the graph for the PM-2 linked at the bottom of the home page.
Thanks for taking the time to do all this! It's very much appreciated.
-
Can anyone explain what I'm looking at and/or what I should be looking for when I look at a CSD graph?
For example I'm looking at the graph for the PM-2 linked at the bottom of the home page.
Thanks for taking the time to do all this! It's very much appreciated.
Don't take my word for it, but the way I understand it, the csd plots show how fast a transducer settles. In other words, how much residual energy or resonance is left after a tone has been produced.
The interest is that a cleaner plot is the result of a transducer which can reproduce a tone more precisely, avoiding smearing between tones. Timing will also be better, avoiding delay errors.
-
Can anyone explain what I'm looking at and/or what I should be looking for when I look at a CSD graph?
http://www.changstar.com/index.php/topic,120.0.html Maybe this will help.
-
CSD description: http://www.changstar.com/index.php/topic,120.msg971.html#msg971 (http://www.changstar.com/index.php/topic,120.msg971.html#msg971Types)
Types of driver ringing: http://www.changstar.com/index.php/topic,71.msg219.html#msg219 (http://www.changstar.com/index.php/topic,71.msg219.html#msg219)
In terms of detail retrieval: http://www.changstar.com/index.php/topic,593.msg12256.html#msg12256 (http://www.changstar.com/index.php/topic,593.msg12256.html#msg12256)
-
^This. Marv is honestly being generous with that 15 seconds for some phones. I think for us there are certain sonic characteristics that are just deal breakers immediately.
^This again. If it's craptastic enough 15 seconds is plenty of time.
*Stares in direction of Ultrasone HFI-780*
OTOH I find that the better (or just closer to my tastes) something is the more time I need to spend with before I know if I want to keep it or not.
-
Marv, Can I ask what is good scale for FR and also for CSD
Do you have recommendations for ARTA and REW?
-
Marv, Can I ask what is good scale for FR and also for CSD
Do you have recommendations for ARTA and REW?
Floor from -30db to -35db depending upon how you normalize the signal up top. Going even lower actually reveals more though. Think I already posted this link: http://www.changstar.com/index.php/topic,593.10.html
I use ARTA. UB uses REW. REW use to have some screwy settings that wouldn't let me do CSDs properly for the timescales I wanted. I believe this has been fixed.
-
^This again. If it's craptastic enough 15 seconds is plenty of time.
*Stares in direction of Ultrasone HFI-780*
Oops there it is... the one you were staring at that is... it paid me a visit today (sadly enough for my ears)
Don't know if the rough treble it has is that obvious in the FR plot already ::) but the THD plots also showed > 1% for several frequencies and the CSD also showed resonances (though short lived).
(https://diyaudioheaven.files.wordpress.com/2015/05/hfi780-fr-up.png?w=921&h=582)
couldn't agree more on the sound.
This is how Ultrasone measured it:
(http://i837.photobucket.com/albums/zz296/solderdude/HFI780frequencyplot_zpse9fdc166.jpg)
-
30 seconds with the Beats by Dre Solo HD mark I taught me one thing: bass farts exist.
It was a demo. If it was my property I would have used that headphone for target practice in "play darts with garbage bins."
Some headphone sound so bad that the ears and brains say no. Never ignore that signal...
-
How to measure distortion the way innerfidelity does?
-
you need special software that has a sliding notch filter in the measurent path that runs along with the sweep frequency.
Latency is of importance here for accuracy.
I don't see the need though and REW and ARTA can also give you THD plots.
-
How to read REW and ARTA distortion measurements?
(http://s3.postimg.org/q136xb5rn/2015_05_15_11_01_05.jpg)
-
Well first you need to determine the THD.
Usually this is slightly above the highest harmonic contents you see in the plot.
Then you have to look at each individual frequency (say 100Hz for instance) and determine the distance between the fundamental (upper trace) and the highest harmonic peak and calculate the percentage from the dB scale: http://www.sengpielaudio.com/calculator-thd.htm (http://www.sengpielaudio.com/calculator-thd.htm)
the distance is 55dB in this case which works out at 0.18% for 100Hz.
AFAIK Marv measures at around 90dB SPL so this should be close the blue/red traces.
Tyll's plot shows the THD somewhere between 0.1% and 0.2%
Ignore Tyll's peaks at 20Hz, 200Hz and 2kHz.
Personally I think the plots showing 2nd, 3rd and 4th harmonics to be more telling/informative than the THD.
-
A few quick very noob questions
1) In ARTA, when calibrating interface/soundcard's full scale input, I've found that every time you estimate the max input mV for the soundcard's full scale input, the estimated Max Input mV changes after consecutive runs. Is that normal?
2) I've noticed that the distortion levels measured in ARTA can vary drastically with different usb interface gain settings and amplifier gain settings/volume levels. how would would do I calculate the ideal settings for my specific equipment for each pair of headphones? any other factors that I should be aware of that impacts those squiggly lines?
3) what amplifier specifications should I be looking or measuring to test how well my amplifier will work for measuring applications?
4) the distortion squiggly lines... how much lower should it be in a good measurement? if I am understanding the explanation above correctly, it is really most important to look at the highest peaks of the highest squiggly line? does the elevation in the lower frequency on the squiggly line count as a peak?
sorry for the newbie questions, but I appreciate any help I can get in further my understanding on this subject. Thank you so much for the help!
-
1) I've given up on calibrating with ARTA. One day I will figure it out.
2) I take distortion measurements where the 0dbFS line in ARTA = 100dbSPL. You'll have to get a SPL calibration with your mic. I did my calibration via a second system / CLIO. In general, I try to line up the lower mids/mids to about ~90dbSPL, which is extremely loud, but I think it gives a clearer measurement result that more easily correlates to what we hear.
3) I use an Objective 2. It measures very well and has good power and low enough output Z. It is also cheap. Just don't overload the input because the pot after the voltage gain stage is a retarded design.
4) All the lines matter more or less. Overall high distortion is never good, so looking at the peaks doesn't necessarily make sense. The different lines represent different orders of harmonic distortion from 2nd to 4th order. How these kinds of distortions sound like is beyond the scope of this thread or website. We are still learning. Listening and looking at measurements to see how they correlate. What also matters is how headphone A compares to headphone B. Measuring without listening is stupid. Some measurements correlate to experience more than others. That being said, some people say even order or 2nd order distortion is more euphonic than odd or 3rd order. Personally, I think that is bullshit. 2nd order distortion sounds just as shitty as 3rd order, just a different kind of shitty.
-
Hah ... was typing at the same moment Purrin was.
1) In ARTA, when calibrating interface/soundcard's full scale input, I've found that every time you estimate the max input mV for the soundcard's full scale input, the estimated Max Input mV changes after consecutive runs. Is that normal?
I don't think so, it defies logic...
If anything a set reference level would be just that.
As I don't use ARTA I can't say what's happening there.
2) I've noticed that the distortion levels measured in ARTA can vary drastically with different usb interface gain settings and amplifier gain settings/volume levels. how would would do I calculate the ideal settings for my specific equipment for each pair of headphones? any other factors that I should be aware of that impacts those squiggly lines?
Distortion and noise levels all are level dependent so you must choose some reference you want to use.
For headphones usually a certain SPL at 1kHz (or using noise) would be used as a reference.
3) what amplifier specifications should I be looking or measuring to test how well my amplifier will work for measuring applications?
It should have the obvious properties, enough output voltage and current. Wide bandwidth with no phase shifts in the audible band and distortion that is very low.
Most likely the mic+pre-amp will be the limiting factors.
4) the distortion squiggly lines... how much lower should it be in a good measurement? if I am understanding the explanation above correctly, it is really most important to look at the highest peaks of the highest squiggly line? does the elevation in the lower frequency on the squiggly line count as a peak?
It's the distance between the highest peak(s) and the accompanying measured level at each frequency that says something about distortion levels at those frequencies.
Determining the THD should be done by the program.
The lower the better. Distortion between 0.1% and 1% is decent for headphones.
-
Thanks all for the insight. A question for measurement takers: If my understanding is right, when you record the input signal at higher sample rates it allows you to use longer windows for analysis, enabling the display of CSD data below certain threshold, say 200 Hz, why not do this?
-
Not higher sample rates, but longer window sizes. Bass waves are long and take a long time to propagate. Bass tends to be sensitive. A truck 1/2 mile down the street results in a lot of subsonics. I've wanted to look at bass decay for headphones on a longer scale, to the 10s of mSec, but haven't gotten to it. I still have a lot of raw data. Don't know if there would be anything enlightening or not.
-
Yup. Longer windows would be required, but sampling rate could actually be reduced as it mostly affect resolution much needed for high frequency characterization.
In fact, say you are constrained by the number of data samples you can take (limited buffer size). Then when doing measurements, going for 96 kHz would yield information just below 48 kHz in your measurements. But since you keep the number of data samples fixed, reducing the sampling rate in effect allow you capture a larger time frame which in turn permits much better resolution of lower frequencies. This is readily apparent in REW when one selects 44.1 kHz sampling vs 96 kHz.
As Marv said, large frequencies have larger periods. Therefore a few cycles of decay (or even fractions) in low frequency may correspond to a crap load of cycles on high frequency decay. IMO we would have to have two or more sets of CSDs for certain frequency ranges. Resolution may be frequency dependent. One could get an idea of decay though by looking at how smooth the low frequency responce is. Likely the smoother and flatter, the fastest the decay.
-
Understood, thank you both.
-
@purrin & @solderdude,
Thank you both so much for responding in detail to my query and I also appreciate the more layman usage of language!!!
@purrin, I will look into attempting a spl calibration on my mic. No idea how to do that, but will research the pertinent background information prior to immediately shooting out tons of other questions. I will also look into acquiring a O2 on the cheap just to ease my concerns about the amplifiers I am currently using. thanks again!!!
@solderdude, if you don't mind me further pestering you, I have a few more questions related to your direct responses (which were extremely helpful, though quite revealing of my still very incomplete grasp on everything). if you prefer to discuss this via PM, please let me know.
1) what measurement software do you personally use? I will like to give it a try :)
2) mmm... I guess I am not sure if I fully understand. This is how I am currently interpreting your statements. Please let me know if I am misunderstanding. Further explanation would be much appreciated!
So I am assuming that I can find from the manufacturer the reference they used and this spec will be listed as a certain SPL at 1kHz (or using noise). And then I will run a measurement trial, I will attempt to match the threshold of the measured distortion to what they say the spec should be using the gain settings?
if further explanation is too complex, I don't mind just resource referrals where I can slowly learn about it on my own free time.
3) that makes a lot of sense. sounds like you are saying I probably shouldn't worry about my amplifier as my mic and usb interface is the limiting factor? thank you
4) what program would I use to determine the THD? any recommendations?
I am very appreciative of the support shown to my probably very newbie-ish questions. Thanks again to both of you for taking the time to respond!! :)
edit: lol I just realized that the "@" tag does not operate the same here as on head-fi. just kind of a habit now to respond that way =S. hope that is okay!
-
1) what measurement software do you personally use? I will like to give it a try :)
REW 5.11 (http://www.roomeqwizard.com/ (http://www.roomeqwizard.com/))on Win XP, soundcard = EMU0204 on 192/24 samplerate, rig = WM61A on homemade rig + dedicated and compensated pre-amp.
The WM61A has a peak at 16kHz which I the pre-amp removes.
Furthermore a mic that is designed to be used in free air (a measurement mic as well) has a different LF response when mounted on a(n infinite) baffle so must compensate for this as well.
These compensations can also be applied in REW but haven't experimented with this.
The output of my mic amp is thus already compensated 'flat' (at least as far as I know) and outputs line level.
I use a homemade amp for driving the headphones.
It's irrelevant what you use as long as it has enough output power to reach 90dB SPL in each headphone (+ headroom) and doesn't alter phase and FR and has low distortion.
All amps that meet those minimum requirements ALL will give the exact same readings on the same headphone.
As Purrin also explained, REW needs also needs a separate dB meter for calibration (only for the SPL part).
Bought a cheap one from fleabay and already had a (modified) Realistic meter which I calibrated a while back when I had an opportunity to use it alongside expensive mics that were being calibrated.
The cheap fleabay meter is about 0.5 to 1 dB 'off' so cheap meters MAY be accurate enough (you never can know for sure though).
It really doesn't really matter if the actual SPL is 88, 90 or 92dB when it displays 90dB.
You set the SPL meter in the 90dB range and crank up the amp/vol control sliders in the calibration mode till it reads 90dB (using noise generated by REW)
Make sure you seal the headphone (using a CD with some felt on it ?) and poke the mic though the hole in the CD.
ENSURE you have a good seal around the mic and headphone to get a proper SPL reading.
Put that headphone on the rig.
Then you adjust the input of the soundcard so it is around -10dBFS (leaving headroom while maximizing S/N ratio)
Then you tell REW that its input level is 90dB SPL and never touch any of the controls again or note the settings so you can set them in the same position later on.
So I am assuming that I can find from the manufacturer the reference they used and this spec will be listed as a certain SPL at 1kHz (or using noise). And then I will run a measurement trial, I will attempt to match the threshold of the measured distortion to what they say the spec should be using the gain settings?
the dB spec the manufacturer gives is irrelevant, it only says something about its sensitivity at a certain frequency.
4) what program would I use to determine the THD? any recommendations?
I use REW (it can also give distortion in % and or dB) but ARTA (properly set up also gives good results.
Measuring headphones is not as easy as it looks and every one that does is produces different plots, some are somewhat comparable, others are not.
Some use smoothing others do not or use less.
A measurement is just an 'aid' and lies to you.
It's up to you to know if the rig is telling lies or not.
That takes knowledge and experience ... and thus takes time and lots of practice as well as understanding what the plots mean and learnig to correlate aspects to sound.
There will NEVER be a 100% correlation to the sound for many obvious reasons but such a rig can be helpful when modifying headphones or when looking for major flaws.
-
Marvin, I saw a video of you on youtube talking about the problems with professional rigs, like the not accurate artificial ear canals which might cause resonances.
Instead of trying to build the perfect-according to your standards-measurement rig wouldn't it be better to buy a pair of relatively cheap binaural microphones and perform the measurements on yourself?
I think that you would discover relative differences between cans that no mannequin can show and you could understand better why headphones sound different when they measure about the same(on various rigs).
-
There's already measurements like that on this site, Hans' specifically.
BMF on Head-Fi also used to measure using in-ears. The downside is the same, if my pinna is bigger than your pinna, then my curves won't match yours to a certain extent.
-
There's already measurements like that on this site, Hans' specifically.
BMF on Head-Fi also used to measure using in-ears. The downside is the same, if my pinna is bigger than your pinna, then my curves won't match yours to a certain extent.
Correct. But the same holds true for dummy heads. Only if you measure the response of top speakers in an accoustically treated room with the use of a dummy/human head will you have a good reference. Otherwise you can only make comparisons and even then you are not 100% sure.
I downloaded some time ago an extract from a german audio magazine(Profesional audio Musik and Equipment-Juni 2009), pertaining measurements of many models and the DT 990 pro measures ruler flat on Neumann KU 100).
Here is a picture from the pdf.
http://s13.postimg.org/yvx9xwsaf/Capture.jpg
http://s8.postimg.org/lotgax9wl/Capture1.jpg
Sorry for the quality. It's not my fault.
The caption reads "The third Beyerdynamic, DT 990 pro shines with an impressively linear frequency response". Even better than the HD800. And not radically different.
Would you trust this?
Would you dare saying that FR tells almost nothing about subjective experience when you don't know if the FR you have in front of you is correct in the first place?
EDIT: Because the scale on the left is not visible here is the result of a more colored can for comparison.
http://s14.postimg.org/ojqjymwch/Capture2.jpg
-
Would you trust this?
Would you dare saying that FR tells almost nothing about subjective experience when you don't know if the FR you have in front of you is correct in the first place?
As long as you have other measurements to compare to using the same measurement method, should be OK. Also should correlate to what experienced listeners hear.
-
Correct. But the same holds true for dummy heads. Only if you measure the response of top speakers in an accoustically treated room with the use of a dummy/human head will you have a good reference. Otherwise you can only make comparisons and even then you are not 100% sure.
I think what Marv and others are doing with their rigs is try to remove the ear from the equation and use the results as a basis for comparison. I like this 'big picture' approach more; I have many issues correlating what I hear with what Tyll's and other HATS-based plots show beyond 2 KHz for example. Maybe C.U.N.T + In-Ear/HATS measurements combined can show a more detailed picture, but the latter will always have a big YMMV factor associated.
-
https://www.youtube.com/watch?v=Yd5i7TlpzCk
Should this help with measurement?
(https://cdn3.vox-cdn.com/thumbor/b8B3AV1F_yuovShNiZencsPgnUE=/422x0:1502x1080/700x700/cdn0.vox-cdn.com/uploads/chorus_image/image/45673902/vup_111_binaural_ooyala.0.0.jpg)
Here is product description
http://3diosound.com/index.php?main_page=product_info&cPath=33&products_id=45
-
Maybe IEMs, or if you want to have to take into consideration the frequency dependent gain effects of the ear only to remove them later via approximate compensation.
As far as their offerings, you could buy their entry level ear models:
http://3diosound.com/index.php?main_page=product_info&cPath=34&products_id=44
I bought these:
http://www.amazon.com/Thumbs-Up-UK-Headphone-Cable/dp/B0082C4GOE
Just cut here and there for your DIY and your good.
EDIT: Yup, it's also pretty good for getting a taste of sound localization given cross channel information.
-
As far as their offerings, you could buy their entry level ear models:
As far as their offerings, you could buy their entry level ear models:
http://3diosound.com/index.php?main_page=product_info&cPath=34&products_id=44
That definitely looks interesting. Would be really easy to take a standard mic measurement in front of speakers, and then a measurement with this in front of speakers... voila, we get a nice compensation curve. C.U.N.T.2.
-
Maybe IEMs, or if you want to have to take into consideration the frequency dependent gain effects of the ear only to remove them later via approximate compensation.
As far as their offerings, you could buy their entry level ear models:
http://3diosound.com/index.php?main_page=product_info&cPath=34&products_id=44
Hmm... They look flexible. I might get one of those.
Er, for measurements, and such.
Yeah...
-
I could take some measurements of the Mackies with the Tidy ears I use for IEM measurements and see whazzup :)p7