Today, I will be comparing Sennheiser’s open-back flagship, the HD800 to its successor, the HD800S – a topic that always sparks discussion among those who have heard both. This article is partly in reply to a request made here.
The most notable difference is the reduction of the 6k resonance, which strongly improves on the treble „sharpness“ that the original 800 is so often associated with.
But there is also * difference in the low frequency region.
Tyll at Innerfidelity measured the HD800S to have slightly higher harmonic distortion at low frequencies.
He postulated that this was in fact done deliberately, because additional harmonics at low frequencies help to increase the "perceived" bass impact. Not a new concept, this is regularly used during music production.
A while ago I was told that Jude from Head-Fi had made similar measurements, and that he had in fact not found a difference in the bass region at all.
The company I work for has a measurement setup similar to Jude’s (Gras 43AG coupler and KB5000 artificial ear, Audio Precision signal generator and analyzer), so I figured why not try and reproduce those measurements and see if I can confirm either Tyll’s or Jude’s measurements.
Sennheiser HD800 vs HD800S | A Comparison Of The SPL Frequency Response Measurements
Fig. 1: A comparison of the SPL frequency response measurements. Or what people generally refer to as a "frequency response".
I measured this at 100 dB @ 1 kHz.
You can clearly see the influence of the HD800S-resonator around 6 kHz, how it reduces that peak by 4-5 dB. This is pretty much the same thing that the SDR-Mod attempts to do.
But you’ll also see how the HD800S has slightly but consistently more bass starting below 100 Hz. It’s less than 2 dB, but it’s there.
A Comparison Of The THD (Total Harmonic Distortion) Measurements
Fig. 2: A comparison of the THD measurements. THD here is expressed in %, this is the type of graph that you’ll also see at the Innerfidelity measurements.
No argument here, the HD800S clearly shows higher distortion at low frequencies (below 100 Hz).
Note that this is with 100 dB @ 1 kHz, so pretty loud. You normally wouldn’t listen to music that loud, unless you particularly hate your ears.
At more quiet listening levels, the distortion is much lower. I chose to measure it at such a high level because I wanted to clearly see the difference between the HD800 vs HD800S, and THD measurements are easier at high levels because as long as the noise floor remains the same, you have a higher signal-to-noise-ratio.
An interesting observation can also be made at around 3 kHz: This is half the frequency of the 6k resonance peak! And since that very resonance is reduced with the 800S, any 2nd order harmonic distortion produced at 3 kHz is also reduced (because the resonance does not amplify it anymore), leading to reduced THD compared with the HD800.
Sennheiser HD800 vs HD800S | A Comparison Of The Frequency Reponse and Harmonic Distortion Product (100dB @ 1kHz)
Fig. 3: This is where it gets a bit unusual - you may not be used to seeing THD displayed that way.
What you see here is the "normal" frequency response of the headphones (black). This is basically "when I play a specific frequency, and it corresponds to how loud the sound that the headphone produces at that frequency is".
The other two pairs of lines (yellow and green) are the sound pressure of the distortion products, but expressed in dB, not %. This is basically "when I play a specific frequency, how loud is the sound that the headphone produces at twice and three times that frequency? How loud are the distortion products?"
If I would add the green (H3) and yellow (H2) line and calculate the ratio between (H2 + H3) and the black line, I would end up with the "normal" THD graph expressed in %.
The yellow line is the sound pressure level of the 2nd order harmonic distortion (H2), the green line is the sound pressure level of the 3rd order harmonic distortion (H3).
How to read this graph:
- Choose a frequency (for example 20 Hz).
- The black line (at 20 Hz) shows you that sound pressure level of the headphones at that frequency (20 Hz) is about 95 dB.
- The yellow line (at 20 Hz) shows you that the sound pressure level which the 800S produces at 2 * 20 Hz = 40 Hz is 73 dB.
- The green line (at 20 Hz) shows you that the sound pressure level which the 800S produces at 3 * 30 Hz = 60 Hz is 54 dB.
So for every frequency you can see the sound pressure level of the desired frequency (the black line) and the sound pressure level of the 2nd (yellow) and 3rd (green) harmonic distortion product.
Now, looking at fig 3 we can see that for the most part of the spectrum, it’s only 2nd-order distortion that makes up the THD (remember: the sum of green and yellow line is the THD, *total* harmonic distortion) – only the frequency range between ~100 and 700 Hz has mostly third-order distortion.
We can also see that the 2nd order distortion is a lot higher on the HD800S, and the 3rd order distortion is a little lower.
Let’s have a closer look at the bass region in figure 4…
A Comparison Of The Frequency Reponse and Harmonic Distortion Product (100dB @ 100Hz)
Fig. 4: This is the same type of graph as fig 3, but now we’re only looking at the bass frequency region. I did measure it a little differently, (longer excitation times, DC coupling) to get a more accurate result.
What we see in this graph is that the HD800S does clearly show increased 2nd harmonic below about 100 Hz, and a slightly decreased 3rd harmonic.
Another way of looking at this is to just watch the spectrum when the headphone plays a 20 Hz sine signal.
Fig 5 and 6 show the spectrum of a 20 Hz signal, where you can see the distortion such a signal causes.
The signal was played at 0.591 Volt RMS, which equals 100 dB at 1 kHz
Sennheiser HD800 vs HD800S | A Comparison Of The Frequency Spectrum Of A 20 Hz Sine Signal 0.591 Vrms (100dB @ 100Hz)
Fig. 5: The frequency spectrum of a 20 Hz sine signal played over the HD800.
Fig. 6: The frequency spectrum of a 20 Hz sine signal played over the HD800S.
Again we can see that at this low frequency the HD800S produces higher H2 (2nd order distortion) and lower H3 (3rd order distortion) than the HD800. In fact in this case the H2 of the HD800S is a little over 14 dB louder, while the H3 is about 8 dB more quiet.
And since it may interest some of you: How do you specifically change the harmonic distortion in a headphone?
Harmonic distortion is created by nonlinearity of the forces that move the diaphragm. A nonlinearity can either be symmetrical (be the same whether the diaphragm moves forward or backward) or nonsymmetrical (affecting forward motion in a different way than it affects backward motion of the diaphragm).
I’ll spare you the mathematics and just say this: symmetrical nonlinearity introduces odd-order harmonics (H3, H5, H7, … but mostly H3), nonsymmetrical nonlinearity introduces even-order harmonics (H2, H4, H6, … but mostly H2).
Now with the HD800S we see an increase of 2nd order Harmonics. How was this achieved, when the driver looks identical from the outside?
In this case my best guess is that the magnet assembly of the 800S was tweaked slightly. Either the geometry of the magnetic field was changed or there are physical tweaks to how the bottom of the magnetic gap is vented. This affects the „springiness“ of the entrapped volume of air in the magnetic gap, which acts as a restoring force on the diaphragm, and together with streaming through venting holes can impose nonlinearity, in turn introducing distortion.
A Comparison Of Treble Resonance
Fig. 7: This is the same type of graph as fig 3 and fig 4, but this time we’re only looking at the treble range, 1 kHz upwards.
See how the dashed black line of the HD800 shows a big peak at about 5.8 kHz? This is what people talk about when they say "treble peak of the HD800".
You can also see that this peak is surgically reduced on the solid back line of the HD800S.
There’s still a peak left (shifted a bit down to 5.5 kHz), but it’s about 4 dB less than on the HD800.
Now for the interesting part – This actually lowers distortion as well.
For example: When the headphone is playing a 3 kHz sound, it will be slightly distorted. There’s always some level of distortion.
We already know that the HD800 produces mostly 2nd order harmonics (H2), so a 3 kHz signal will also show a small peak at 2 * 3 = 6 kHz.
And since the HD800 has an earcup resonance at 6 kHz, this H2 will be amplified by this resonance. The result is higher distortion at 3 kHz!
Now we know that the HD800S reduces this 6k resonance, and we can also see that the harmonic distortion at 3 kHz is reduced – because the distortion is not amplified by the resonance anymore.
Isn’t it nice when theory and practice check out?
So, to answer the question – could I confirm Jude’s or Tyll’s measurements?
Well my results do show different distortion behavior with the HD800S, very similar to what Tyll’s measurements showed.
Does that mean Jude made a mistake? No, I don’t think so.
I did measure one HD800S (my own) and one HD800 (belonging to a friend of mine). There could very well be deviations between individual specimen. Sennheiser's Quality Control is usually excellent and tolerances for the HD800-series are incredibly tight (as can be expected for that price from a German company), but tolerances still exist.
Also I did measure at higher levels than Jude, which causes higher distortion in general. I also employed a different measurement technique – whereas he fed the headphone with a single test signal and looked at the resulting spectrum (averaged over a certain amount of time to reduce noise) I used a measurement technique that employs self-correlation with the original test signal – in theory this is more accurate and more precise as it can filter out almost any amount of noise.
I assume Jude is using the same measurement program that I’m using (Audio Precision’s APx500 suite), so if he wanted (and if he had time to do so) he could replicate my results using a “Stepped Frequency Sweep” measurement that is implemented in the Audio Precision measurement suite. That way he should get the same results in the low frequency range.
Don‘t forget that while it is possible to measure a lot of things, in the end headphones are not made to be measured, they are made to listen to music. Enjoy the music!
Acoustic engineer by day, recording and mixing engineer by night.