HD650 technical measurements

HD650 technical measurements

And how single number metrics provide extremely limited information

Measurement setup

Klippel QC analyzer
Furman SP-20A headphone amp
ACO Pacific 7052SYS microphone
Brüel & Kjaer 4231 mic calibrator
Mr BlockHead™ flat plate coupler
density similar to average human head 1.08 g/cc
width consistent with average human head 15.25 cm



Mr BlockHead is mounted on a high mass lighting stand to minimize externally induced vibrations.



View of ACO Pacific 7052SYS in Mr BlockHead flat plat coupler
Alert assistant is nearby ready to help.



HD650 is specified to have 0.05% THD+N at 1 KHz with 1Vrms stimulus

Historically a low distortion 1 KHz sine generator would have stimulated the electro-acoustic transducer using a low distortion amplifier. Then a steep 1 KHz notch filter removes the stimulus and the measured level of what remains is compared to the stimulus level. That ratio results in THD+N.

Now consider the capability of a modern acoustic analyzer.

Klippel generator is adjusted to 1Vrms (circled in red.)
Low distortion amplifier is set for 0 dB gain to follow the generator settings displayed.
Headphones carefully positioned on flat plate coupler.

Sound level observed is 100.6 dB SPL which is consistent with published sensitivity.
THD+N at 1 KHz is about 0.05% THD+N (the red dot on the graph.)

Looking at this graphic limited to the single number data point offers minimal information.
SINAD at 1 KHz provides similarly very limited information about the characteristics of an amplifier. A single dot on a graph.



Expanding the view to include the entire analyzer 1 Vrms sweep from 20 Hz to 20 KHz vastly more information is observed

Upper left graph: Frequency response, rub+buzz
Upper right graph: THD+N, H2 (offset by -4% for visual clarity) and H3 (offset by -2% for visual clarity)
Lower graph: Impedance

Why would we want to limit our view of an electro-acoustic transducer or amplifier to a single number representing an infinitesimal aspect of performance?

Single number metrics used for audio component performance rankings are absurd when one realizes how much information is missing.



Notes about the analyzer graphs

1) frequency response is not compensated in any way. Data can be exported such that others may use their favorite compensation methods.
2) rub+buzz is a measure of all that remains when fundamental and harmonics are removed. It is sort of an acoustic crap factor.
3) impedance measurements are often noisy and unreliable below 0 dBu in normal environments outside an anechoic chamber or very well controlled sound space. Every electro-acoustic transducer can transform electricity to sound (speaker) and sound to electricity (microphone.) They are usually optimized for one or the other but still can function in reverse. To avoid outside influences impedance is typically measured with a sine sweep of between 0 and +10 dBu. It is a balancing act between outside influence and analyzer input level of which there are more considerations than will be discussed here.
4) rub+buzz below 40 dB SPL is inconsequential

Note the rub+buzz bump at 6 KHz corresponding to a small THD+N and H2 bump at 6 KHz. Watch these points in the following graphs.

 

Klippel rub+buzz orientation

With generator off the room residual noise is measured in the upper left graph.
Room ambient SPL, blue trace, is 40 dB SPL below 80 Hz.
rub+buzz noise floor is the black trace.
Keep these measurement noise floors in mind when viewing the following data.


-10 dBu (0.245 Vrms) observe 88.5 dB SPL
Impedance measurement should be ignored
Observe the changes in rub+buzz and distortion as stimulus levels are increased
THD+N is approximately 0.5% at 60 Hz and doesn’t rise above 5% even down to 20 Hz
All H2. H3 is very low.
Listening at low levels has many benefits including lower overall distortion.


0 dBu, 0.775 Vrms, 98.4 dB SPL
SPL and Frequency Response plot rise 10 dB as expected
THD+N is rising, especially at low frequencies.


Changing generator reference system from dBu to Vrms
1 Vrms, 100.6 dB SPL
1 Vrms is 2.2 dB higher than 0.775 Vrms
SPL and Frequency Response plot rise by 2.2 dB as expected


2Vrms, 106.5 dB SPL
SPL and Frequency Response curve rise 6 dB as expected
Observe THD+N now above 0.5% below 200 Hz and exceeds 10% at 40 Hz


Animation of previous four slides for easy visualization of distortion rise from -10 dBu (0.245Vrms) to 2Vrms


Expanded y-axis view of THD+N
Even at -10 dBu (0.245 Vrms) distortion is approximately 0.05% above 150 Hz, listening about 89 dB SPL


At 2Vrms distortion below 300 Hz is above 0.15% , and typically above 0.075% at higher frequencies, listening at approximately 107 dB SPL

As most of the distortion is H2 and is often above 0.1% there will be masking of upstream component low level distortions.

It is unwise to infer distortion performance of amplifiers and DACs or other source sound sources when an electro-acoustic transducer is producing this much distortion.

Keep in mind Sennheiser HD6x0 and HD800 headphones are already very low distortion headphones; these measurements demonstrate very good performance.

 

BK891 LCR measurements

Cal check with SR1 1K laboratory standard reference resistor


HD650 measurements


SR1 1K impedance sweep (calibration check) 0.5Vrms


HD650 impedance sweep 0.5Vrms


HD650 capacitance sweep 0.5 Vrms


HD650 inductance sweep 0.5Vrms


HD650 impedance sweep 1.0Vrms


HD650 capacitance sweep 1.0Vrms


HD650 inductance sweep 1.0Vrms


Notes:
Where capacitance goes negative the load is more inductive
Where inductance goes negative the load is more capacitive

All data for above measurements in the attached zip file
Added a csv file with all data converted to a more easily imported format

 

reserved for additional data and corrections if necessary.

 

Stellar job, Bob! My experience is that systems with multiple significant distortion sources subjectively start to sound noisy. It's like signal correlated noise. I've often encountered cases like these in clubs and concerts where it's impossible to converse, despite there not being much info in the vocal range. With clean systems one could easily hear and understand speech despite ridiculously loud music.

 

Epic X2 for animation.

 

Wow I haven't seen one of those in decades... did you specifically pull that one out to make a point of not using a modern ultra low sinad amp?

 

I get that we can't judge the performance of an amp measured with real-world loads in isolation. But what about in comparison to other amps? I found the Topping vs Jotunheim measurements interesting. What's wrong with finding an amp that is known to be well built with low distortion and using that as a baseline of comparison with tests like these? I'm sure I am missing something.

EDIT: I re-read your posts again. I think I get it but I'll wait for your follow-ups.

 

The two major points of my post are:

1) in evaluating measurements for an audio component would you rather have

THD+N at 1 KHz = 0.05%

or



Ranking components with highly reduced information is a terrible idea.

2) attempting to hear amplifier distortion of less than 0.01% while listening at 107 dBSPL where the headphone has between 0.075% ~ 0.25% distortion across the frequency range, greater than 1% below 150Hz and rising exponentially at lower frequencies, is a categorically Bad Idea™

AB and ABX comparison of level matched amps is useful and discrimination achievable at 75 to 85 dBSPL where the headphone distortion will be significantly lower than HD650 at 2Vrms, 107 dBSPL.

 

Good eye. I acquired the Klippel QC in 2012 when I had an idea to create an improved driver that would be part of a system to improve speech intelligibility for medical devices used in noisy environments. See US Patent US9713728B2, which is one of many of my ideas for which I obtained patents. The driver constraints included mylar diaphragm due to IP65 rating for the device including sound system. The driver was designed to have a specific distortion profile which changed in a desired way at specific amplifier drive levels.

Also about this time I needed to develop a mobile system which could demonstrate sound lab psycho-acoustic experiences. It is expensive to fly people to my campus acoustic lab. Enter the need to measure headphones and know how well they could perform when driven to 115 dBC SPL. The user was required to wear hearing protection if wishing to have more than a minute of psycho-acoustic video demonstration on the mobile system. HD650 was one of very few that made the cut.

The SP20A was also used for cue mix during recording sessions. I have three of them and with the companion distribution boxes could handle cans for approximately 20 musicians in the studio.

 

Added LCR measurements in post 3

 

Great job! I wonder if, using such measurements, it is possible to detect the difference between different interchangeable cables, to see how the cable actually changes the sound and at what frequencies?

 

pardon my ignorance . Do those show any time domain performance. Like if there is ringing or how fast the drivers are?
I got the HD600 once awhile ago and while they were pretty nice, they weren’t impactful enough for me.
Does the HD650 measure and perform better?

thanks

 

While I definitely think that time domain elements are really important in the personal audio space and severely underappreciated in the hobby at large, at least insofar as at least the difference in perceived impact between the HD600 and HD650 go I do think that the different distortion profiles of the two Sennheisers and even just the greater low end elevation of the HD650 can go a long ways towards explaining its greater "impact".

There was a comparative thing showing decay between the JAR600 and an HD650 here though: https://www.superbestaudiofriends.o...e-hd800-sr-207-hd650.3688/page-12#post-296621

Now that I think about it was there ever a comprehensive head to head between the two insofar as burst response and decay measurements go? FR and distortion are long established but I can't remember if there was more than that.

 

No, these are not time domain measurements. Look for those using an oscilloscope and square wave or burst stimulus.

 

thanks for the link.
I’m trying to decipher the literature and measurements.

so yeah I’m trying to find a headphone that can give me the quickness of my Beyer 1990 ( I EQ them of course ) and the HD600 didn’t quite do it.
But I liked the overal tone quite a bit.
Maybe the 490 pro will do it for me.

I listen to mainly Classical, so that’s a heavy burden to any headphone and/or speaker.
Contrary to what some fools think, classical is hardly relaxing music. Well depends on the piece of course. But you need something with very good dynamics to capture the cadence of let’s say a Beethoven Symphony which combines many many swings in volume and intensity.
Not to mention Classical also deals with so many instruments that Timbre needs to be well done, which many think is a characteristic of good transient response as well to recreate. So you don’t mix up an oboe with a Flute for example.
If transient response is slow, strings will also sound grainy and bow dynamics won’t be so apparent and will be smeared. You wont hear a clear strike of the bow, which is percussive and then draw of the bow which should be more textured and larger sounding. It will sound more like just one sound and not distinct. Why recording strings using difficult and some classical recordings render the bow and string sound well and others just sound smeared and screechy or occupy a very small freq range. If you have heard different classical recordings. The older ones especially didn’t record strings well at all. Newer recordings have fixed this a lot.
Hope that’s helps and makes sense.

thanks