October 2020

Headphone Shake-Up
Keith Howard supplements his headphone experience with a kick in the posterior, and find himself shaken by the results.
Review By Keith Howard Of HIFICRITIC

  It's a well-worn clich้ that headphone listening can never match the sheer impact of using speakers because only our ears are engaged, not our whole body. In a soundfield we are subject to music-related vibrations that reach us via the air itself and via structure-borne sound, conveyed through the floor to where we are standing or sitting. We may not always be conscious of these vibrations but, the trope goes, they have a significant effect on our perception of live or reproduced music – an effect that headphone listening necessarily lacks.

Actually, this is more than mere clich้ or trope: there's hard evidence to back it up – although, strangely, you're unlikely to have read about it in any audio magazine. In fact research evidence gathered over the last decade or so shows that even loudspeaker listening is improved, and by significant margins, if additional physical vibration is delivered to the listener's body. So if you've previously thought that seat shakers are just for home theatre enthusiasts looking to recreate cinema LF effects, think again. While it turns out that there's more to achieving the optimum outcome than simply connecting a power amplifier to the LFE channel and using it to drive a seat shaker, in essence that's what we're talking about here. At the right level, with the right spectrum and optimum signal conditioning, listening seat vibrations enhance the audio experience.

If this is true of loudspeaker listening – I'll describe some of the evidence for that in a moment – imagine how much more significant might it be for headphone listening, where there is negligible body vibration to begin with? On the face of it, seat vibration could be a game-changer for headphone users – and yet I've not seen any published research in this area. What seems like an obvious extension of the loudspeaker findings appears not yet to have been explored. This article is intended to plug that gap, at least a little, and address the question every headphone enthusiast would surely want answered: does seat vibration elevate the headphone listening experience to a new level?

Loudspeaker Research
References 1 and 2 provide details of some of the most important research that has been performed on using seat vibrations to enhance loudspeaker listening. Key facts from these papers are that equalization of the vibration signal is necessary to ensure that the acceleration versus frequency response at the seat, with a given human in place – what Altinsoy and Merchel refer to as the Body-Related Transfer function (BRTF) – is flat, and that other signal conditioning measures such as downwards pitch-shifting or dynamic range compression can be applied to the vibration signal to enhance its effect.

When the signal processing is optimized, the addition of body vibration resulted in listeners scoring sound quality significantly higher than with loudspeakers alone.

Seat And Shaker
Experiments like those I've just described are made using large, floor-mounted electrodynamic shakers constructed rather like a moving coil drive unit on its back. The subject sits on the vibrating platform and is shaken by signal current passing through a large 'voice coil' beneath, the assembly being compliantly suspended from the body of the shaker.

Figure 1. The Red Wood seat shaker used for the listening, bolted to the underside of the office chair.

A more unlikely object to have in a domestic living space, in place of the more traditional armchair or sofa, is hard to imagine. Any practical solution for the audiophile headphone listener is likely to use instead one or more of the many commercially available seat shakers, which again are like moving coil drive units but apply shaking force inertially rather than directly. The 'chassis' of the shaker is bolted to the chair or sofa, and conveys magnet reaction force generated by a suspended voice coil and attached mass within. I chose to use a Rockwood BS301-L 4-Ohm body shaker for my experiments, rated at 100W and available for about ฃ30 from Amazon.

Any thoughts of attaching this to my listening room sofa were banished by it actually being a sofa bed, which not only eliminates some obvious points of attachment but also makes the seating inherently resonant. So I decided instead to mount the shaker to the underside of an office chair which normally serves my electronics workbench. I reasoned that, because it is small and light, it would obviate the shaker being unable to deliver sufficient acceleration. But in the event that proved not to be an issue. Figure 1 shows a photo of the shaker attached to the underside of the seat. To ensure as rigid an attachment as possible, the plywood was drilled to accept threaded nut inserts, and the shaker bolted hard down on to them.

Measurements
Figure 2 shows a photo of the equipment I used to measure BRTF on my chosen seat. It comprises an aluminum sheet slightly smaller than the seat area, a Measurement Specialties ACH-01 piezoelectric accelerometer attached to the sheet using double-sided Tesa tape, and IB-ACH-01 accelerometer amplifier. An earth wire between the amplifier case and the aluminum sheet quelled what was otherwise excessive mains-induced hum pickup.

Attaching an accelerometer using adhesive tape is normally deprecated because it introduces some compliance which can affect the results. Mechanical fastening or gluing with epoxy is normally recommended. But thin double-sided tape is fine when you're only concerned with low frequencies, as here.

Figure 2. Equipment used to measure the BRTF (body-related transfer function) for me sitting on the listening chair. The piezoelectric accelerometer, attached to the aluminum sheet by double-sided tape, is the small block in the foreground, with its amplifier box behind. The crocodile clip, connecting the metal sheet to the case of the accelerometer amplifier, is an old trick to reduce hum induction.

The BRTF I obtained – by placing the accelerometer plate on the chair, sitting on it and using ARTA (www.artalabs.hr) to perform the measurement – is shown in Figure 3, where the major vertical divisions are 10dB. As you can see, the response is far from flat (and quite different from that typically obtained using large floor-mounted shakers), with what are almost certainly seat structural resonances apparent at about 46Hz and 83Hz, with approximate Qs of 11.5 and 5.2.

Figure 3. The BRTF measured using the equipment in Figure 2. Major vertical divisions here are 10dB with resonances obvious at 46Hz and 82.5Hz .

The Process
Once the BRTF was measured, an inverse filter had to be designed to flatten the acceleration versus frequency response at the seat. In a commercial realization you'd probably want to perform this task using an IIR digital filter as, particularly at the low frequencies inherent in seat vibration, it will be much more computationally efficient than an FIR alternative and not introduce significant latency. But a few initial trials persuaded me that I didn't have time to design an effective IIR filter for these experiments.

Figure 4.Acceleration versus frequency response at the seat once inverse filtering was applied to the shaker signal (calculated, not measured). Here the major vertical divisions are just 1dB.

Figure 5. Uncorrected frequency responses for the left (blue trace) and right (red trace) capsules of the Audio-Technica ATH-A2000Z closed-back headphone used for the listening .

So I took a brute force, FIR filter approach instead, making use of Angelo Farina's invert.xfm and convolve.xfm Aurora plugins for version 3 (or earlier) of Adobe's Audition audio editor. The first was used to create the impulse response of the inverse filter, and the second to apply it to the shaker signal. Applying the inverse filter to the BRTF impulse response showed the corrected response to be that shown in Figure 4. Note that the vertical divisions here are just 1dB, and that the response is effectively flat down to below 30Hz, and within +/-0.3dB limits down to 20Hz.

The precise processing steps were:

1) Combine the left and right channels into a mono WAV file.

2) Pitch-shift the mono file downwards by an octave, using Audition 3.0's own pitch-shifting tool.

3) Apply the inverse filter to both the mono and pitch-shifted mono files.

4) Apply a 10-order Butterworth low-pass filter to both mono files using Audition's Scientific Filters tool.

5) Assemble two three-channel WAV files using the left, right and shaker signals, with and without pitch-shifting.

I wrote a simple software utility to perform the last step, taking into account that the inverse filter added 15052-sample latency to the shaker signal and that it is of slightly different length to the L/R signals. Only 44.1kHz/16-bit music files were used but the shaker channel processing was performed in 32-bit floating point format, with the processed mono files converted back to 16-bit as the final step before assembly of the three-channel output file.

System Hardware
To replay the left and right channels of the original music file over headphones simultaneously with the processed shaker signal, of course you need a multichannel reproduction system. I used an RME Fireface 800 FireWire audio interface to replay the three-channel WAV files from computer, with channel 3 carrying the shaker signal. The outputs of channels 1 and 2 were fed to the left and right inputs respectively of my resident Teac HA-501 headphone amplifier, while the channel 3 output was routed, via a multi-turn potentiometer acting as a shaker level control, to an old Exposure XVIII Mono power amplifier and thence the shaker itself. For some reason the Fireface 800 reproduced this signal on its channel 4 output, but no matter. The Fireface channel matrix – which normally routes channels 3, 5 and 7 to channel 1 and channels 4, 6 and 8 to channel 2 – was reconfigured to ensure that the shaker signal was not passed to the right headphone channel.

Because the shaker proved not to be silent in operation, it was logical to use a closed-back headphone for the listening comparisons to suppress its audible contribution. Audio-Technica's ATH-A2000Z filled the role admirably, although – as Figure 5 shows – it doesn't have the extended bass response typically provided by planar magnetic alternatives. This may or may not be a factor in optimizing results with a seat shaker, but that's a detail to explore on another occasion. There is evidence that LF vibration does indeed influence preferred bass level [3].

Listening
I selected seven disparate WAV files, representing various musical genres, to process for the listening. In no particular order they were:

1) 'The Saga Of Harrison Crabfeathers' (from Wood, Brian Bromberg, A440 Records 4001 ) – solo double bass introduction [excerpt duration 0:69].

2) 'The Turtle Dove' (from Scarborough Fair, James Griffett, Regis RRC1112) – solo tenor recorded in a large natural acoustic [duration 3:01].

3) 'Private Investigations' (from Love Over Gold, Dire Straits, Vertigo 800 088-2) – wide dynamic range rock/pop [duration 6:46].

4) 'Be My Number Two' (from Body And Soul, Joe Jackson, A&M CDA65000) – large-scale pop/soul [duration 4:21].

5) Vivaldi: Four Seasons, 'Spring' (from Seasons, Trafalgar Sinfonia, Signum Classics SIGCD 437) – small-scale orchestral work [excerpt duration 1:40].

6) 'Well WellWell' (from Hurricane, Grace Jones, Wall of Sound WOS050CD) – compressed pop with a hallmark Sly/Robbie groove [duration 3:50].

7) Boulez: Notations, '1 - Fantasque Mod้r้' (from Explosant-fixe…, Pierre-Laurent Aimard) – high dynamic range solo piano [duration 1:00].

Of these seven tracks only one – track 2, the solo tenor item – didn't really benefit from having the shaker operating. It was possible to achieve some extra sense of vocal warmth and acoustic space by cranking up the shaker level, but to such an extent that I suspected it was an artifact rather than a genuine contribution to fidelity.

All of the other tracks benefited from the shaker, although to different degrees and there was some variation in optimum level setting. In all cases what I chose as the optimum level caused the seat to vibrate quite obviously on loud passages, which I'd rather avoid if possible – but maybe that isn't possible while still enjoying the shaker's benefit. If I had to summaries in a word what that benefit is I'd say 'solidity'. Solidity in terms of more impressive large-scale dynamics, and solidity in respect of firmed-up stereo imaging.

Simply More Real
The two tracks which benefitted most – to a point where I feel it's appropriate to talk of a transformation – were the solo piano item (track 7) and Private Investigations (track 3). Pierre-Laurent Aimard's piano took on a weight and power that came a lot closer than usual via headphones to conveying the percussive impact of a concert grand. All told, the result simply sounded more real, and because of that more arresting: it commanded attention in a way that the shaker-less replay fell well short of matching.

That Private Investigations benefited so clearly from the shaker was, I admit, a relief. It's such a pot-boiler of a demo track that I was loath to use it, but had an inkling it would be a good choice – and that proved to be correct. As well as adding dynamic heft, the shaker did extraordinary things to the stereo image – which deepened, broadened and solidified – and to the various elements within it. All the little 'noises off' had texture and detail I've normally not experienced, and the overall result was at least as attention-grabbing and impactful as I've ever heard from loudspeakers.

Conclusion
Exploratory and incomplete as it was, this trial has convinced me that the addition of seat vibration improves headphone listening in a fundamental way that no amount of, for instance, DSP-metered inter-channel crosstalk ever could. Maybe seat vibration allied to cross-feed processing might be better still, but there's little doubt in my mind that seat vibration has the more significant effect on improving headphone sonics in a way that enhances musical enjoyment.

What will I do differently next time (there will be a next time: this is too important to let lie)? Well, I'd choose a more comfortable – but thinly padded – chair allowing the shaker to be placed immediately below the torso, rather than the upper thighs as dictated by the pillar of the office chair. I'd also add a second shaker on the back of the chair, perhaps with two possible locations: lumbar and thoracic. I have no idea whether mixing vertical and horizontal vibrations will improve matters or, if so, how the two might be optimally balanced – but it 's worth a try.

Nothing in my trial run suggested that more than one, inexpensive shaker is necessary (per plane of vibration) nor that a particularly powerful drive amplifier is needed. But a shaker volume control is definitely required, and I'm pretty certain that results can be improved by applying some dynamic range compression to the shaker signal, to increase vibration at low signal levels and trim it back on high signal levels. There may be other useful signal conditioning tricks, but my initial experience of pitch-shifting the shaker signal downwards by an octave did not convince me that this is the optimal approach, despite it working well in the speaker research. Still, there's no reason not to try it a second time, along with obvious alternatives. The only good reason for using a more costly shaker, I think, would be if it had reduced radiated sound level.

Of course, I'm fortunate enough to have the necessary hardware and software to perform the BRTF measurement, to design and apply the inverse filter, to perform the signal conditioning, and to assemble and replay a three-channel WAV file. Few audiophiles can say the same nor, I imagine, would they wish to spend time pre-processing stereo source files. A turnkey hardware solution – a seat with shakers, signal processing and amplification built-in – is obviously what's needed to make seat vibration a practical accompaniment to headphone listening for most. I do hope somebody is working on the concept.

References
1) Altinsoy, M E and Merchel, S, 'BRTF (Body Related Transfer Function) and Whole-Body Vibration Reproduction Systems', Audio Eng. Soc. 130th Convention (May 2011).

2) Altinsoy, M E and Merchel, S, 'The Influence of Vibrations on Musical Experience', J. Audio Eng. Soc., vol 62, no 4 (April 2014).

3) Simon, G; Olive, S E and Welti, T, 'The Effect of Whole-body Vibrations on Preferred Bass Equalization of Automotive Audio Systems', Audio Eng. Soc. 127th Convention (October 2009).

Subscribe!

Click here to subscribe to HIFICRITIC