HIFIMAN Susvara

Four rigs, two classes, and the class predicts the verdict. Computed from sai's own raw exports, level-matched to the Harman over-ear 2018 target across 200 Hz–1 kHz, the Susvara sits about 6.1 dB below target from 20–40 Hz, 3.9 dB below through 1.2–2.2 kHz, 0.6 dB below at 2.5–4 kHz, level at 4–6 kHz and 1.9 dB above from 10–16 kHz. Measured instead against its own 200 Hz–1 kHz level, the bass is only 1.1 dB down at 20 Hz — flat, not thin. Audio Science Review's GRAS 45C reads the same deficits and calls them faults. Both MiniDSP EARS rigs read it as exemplary: Soundnews calls it 'the most linear and extended frequency response I've measured around here', and headphones.com finds 'The bass extends like a flat line all the way into the sub-bass on the HPN compensation.'

⚠ vs. listeners — Nobody is misreading a graph; they are choosing different rulers, and two of them are different instruments. Against a bass-boosted Harman target the Susvara is measurably short of bass and short through the presence region. Against its own midrange — or against a diffuse-field-style compensation, which is what both in-house rigs used — it is close to a straight line. Both MiniDSP EARS users say plainly that their rig is not industry-standard and should not be compared with ones that are, so the disagreement is partly instrumental and partly a question about whether a bass shelf belongs in the target at all.

The two rigs that published distortion data disagree flatly. Audio Science Review finds 'Even at 94 dBSPL we are exceeding our target between 4 and 5 kHz', reports that one dB past 114 dBSPL 'the drivers bottom out and just buzz', and calls the group delay the messiest it has seen across the audible band. Soundnews, on a MiniDSP EARS rig, reports 'Their overall THD sits at 0.27% which is world class by my standards' with the highest distortion in the treble rather than the bass, plus a fast decay and clean waterfall. Amir's own hedge matters: he notes at least one other published measurement shows a similar distortion issue, so the finding is not a broken sample.

⚠ vs. listeners — This is the deepest conflict in the record and it does not resolve. One GRAS-class rig measures serious distortion and resonance where hearing is most sensitive and a driver that audibly gives up at high level; one in-house rig measures the lowest distortion its owner has recorded. Both then listened, and heard opposite things. What the objective camp adds is a mechanism rather than a taste — that resonances and a lifted top octave manufacture the sensation of resolution — and the subjective camp has no equivalent rebuttal beyond the listening itself. A buyer should know that the one thing everyone praises is the one thing the measurements most seriously question.

Computed from sai's raw exports against the Harman over-ear 2018 target, 10–16 kHz averages 1.9 dB above target and peaks about 6.0 dB above at ~12.8 kHz, with 4–6 kHz sitting level. Audio Science Review, on a different GRAS-class rig, arrived at the same place by ear: after EQ he writes 'I broke my usual rule of not messing with > 8 kHz and dialed in a negative notch filter at 12.7 kHz.' The MiniDSP EARS rigs disagree — headphones.com reports 'it doesn't have too much energy above 12khz like many other flagship planars tend to go for', with its only caveat 'a bit more energy at 7khz than some may prefer'. sai's own fix is a 3 dB cut at 11,500 Hz, Q 2.5. The deep notch near 9.5 kHz on GRAS traces is a pinna artifact, not an audible dip.

⚠ vs. listeners — The two independent GRAS-class readings converge on a lift centred near 12.5–12.8 kHz, one measured and one found blind by a listener reaching for EQ. The two MiniDSP EARS readings converge on flat, and both of their owners state that their rig is not IEC-standard and should not be treated as reference. Above 8 kHz, ear-canal geometry also varies enormously between people, which sai flags himself. So 'the smoothest treble HIFIMAN has made' and 'it can get a tad hot and zingy' are both honest reports, and which one a given listener recognises depends on their ears as much as on the headphone.

The dip is measured and agreed. Computed from sai's exports, 1.2–2.2 kHz runs 3.9 dB under the Harman over-ear 2018 target, deepest at about 4.9 dB down near 1.75 kHz, while 2.5–4 kHz sits within a dB of target. The EQ prescriptions converge on filling exactly that: sai uses a +3.0 dB peak at 1,700 Hz with Q 2.0, and Audio Science Review's filters at 1,900 and 4,500 Hz 'filled in the treble area'. headphones.com measures the same feature and reads it as benign — 'The mids have a slight dip near 2khz' — crediting 'sufficient energy at 3-5khz to provide clarity for female vocals and piano tones'.

⚠ vs. listeners — One curve, two experiences, because the dip sits below the presence peak rather than on top of it: vocal body and chest weight step back while articulation and clarity stay forward. Two physical variables the sources name themselves widen the gap further — sai's positional measurements show that where the headphone sits on the head shifts the balance, and Currawong reports the same thing by ear; and pad rolling measurably lifts the region, with sai noting that velour and heavily perforated pads bring up 1–3 kHz.

Flat extension, no shelf. Computed from sai's raw exports, the Susvara sits only 1.1 dB below its own 200 Hz–1 kHz level at 20 Hz and 0.7 dB below through 50–120 Hz — essentially a straight line into the subsonic — while measuring about 6.1 dB under a Harman-style target from 20–40 Hz. sai reports it 'extends impressively down to 20Hz and below human audibility with minimal roll-off' and that 'Throw a 10dB bass shelf on it and the Susvara will do just fine', which is the excursion headroom the flat trace implies. headphones.com measured the same shape on a different rig and drew the same distinction against the Harman bass elevation.

⚠ vs. listeners — The 'plenty of bass' versus 'lacks bass' argument is entirely about the reference point, and the graph resolves it. There is no roll-off: the headphone plays 20 Hz at essentially the same level as its midrange, which is why listeners describe real sub-bass rumble and why it EQs up so cleanly. It is also several dB under a target that deliberately boosts bass, which is why one measurement source records a deficit. Where sources genuinely differ is on tactility rather than level — and even the enthusiasts concede a good dynamic driver hits harder.

The power requirement is arithmetic, not opinion: 60 Ω at a rated 83 dB/mW, which Soundnews turns into 'for a loudness level of 110 dB (fairly loud), HE1000SE would need only 25 mW and Susvara would need 502 mW – that’s almost 20 times more power'; sai measures 60 Ω and reads sensitivity lower still, at 79 dB/mW. The ceiling is measured too — Audio Science Review found that one dB past 114 dBSPL 'the drivers bottom out and just buzz', and that a Benchmark AHB2 power amplifier ran out of voltage before a Topping A90 did, so the limit there was the headphone, not the amp.

⚠ vs. listeners — Two things are true at once and the sources talk past each other. The Susvara needs genuinely unusual power to reach a given loudness — that is spec-sheet arithmetic nobody disputes. Whether it needs a speaker amplifier to sound RIGHT is the actual argument, and it is not settled: Soundnews reports a night-and-day transformation on speaker taps, Currawong ran it on capable headphone amps and does not feel a speaker amp is necessary, and an ASR poster who owned one and drove it from a Topping A90 calls the amp issue immensely overblown. Amir's finding cuts across both camps: past a certain level no amplifier helps, because the driver itself gives out.

No rig measures imaging, but channel matching is measurable and it is excellent: Soundnews reports matching 'at less than 0.5 dB' after repeated remounts, and sai publishes a dedicated flat-plate channel-matching graph alongside his frequency-response set. Sources also offer a mechanism rather than a verdict — the open Window Shade grille removes reflective structure behind the driver, and Soundnews notes the sound changes drastically for the worse when the grilles are covered by hand.

About 450 g without cable (448 g measured by sai) on a wide perforated leather suspension strap, with clipped pads roughly 110 × 105 × 20–25 mm, medium clamp and around 240° of headband swivel; sai's own note reads 'Comfort: 7/10 (slightly weighty)', the mildest reservation anywhere. Soundnews weighs the comfort case plainly: 'At 450 grams (exactly a pound) I find them lightweight and super comfortable long term.' For scale, the LCD-4 it is usually compared against passes 700 g and the AB-1266 TC reaches 630 g.

No rig measures stage, so sources offer mechanisms instead. sai frames the character rather than the size — the soundstage is 'exemplary of what might be called ‘realistic.’' rather than the grandest, and he explicitly contrasts it with headphones criticised for sounding artificially large. Soundnews credits the fully open Window Shade grille and reports the presentation collapsing when the grilles are covered. Both note that amplification changes the perceived scale substantially.

An aluminium frame with the Window Shade grille and a wood-veneer ring, a full-metal headband and yoke, clipped pads and dual 3.5 mm TRS sockets, with a 6.35 mm and a 4-pin XLR cable supplied and no 3.5 mm cable — a deliberate omission, since the headphone cannot be driven from a portable. Soundnews confirms the construction: 'Their metal structure was created via CNC milling and hand polishing'. The cable criticism is unanimous and specific: 'the cables are horrendous' (an owner of two-plus years), 'The cables that are in the box are gummy feeling and not very professionally finished', 'Those default cables need an upgrade'. Packaging did improve — the 2017 unit shipped a novelty velour bag and no case, later units a leather carrying case.

Fully open by construction: the Window Shade grille is designed to protect the diaphragm while leaving the rear of the driver acoustically clear, and the stealth magnets are shaped to let waves pass between them, so leakage runs both directions by design. sai publishes a leakage-tolerance measurement showing how compromising the seal at the front changes the frequency response — which is also why glasses and headphone position matter more here than on most designs.

The price history is the argument, and it is unusual: HIFIMAN lists and sells the Susvara at $6,000 as of July 2026, in stock, unchanged from 2017. The new unit is never discounted; the cheaper routes are HIFIMAN's own refurbished stock at $3,299 (2.5 mm) and $3,599 (3.5 mm), open-box at $4,499, or a $4,000 trade-up into the $8,000 Susvara Unveiled. The comparison the sceptics keep making has a number too — the same week, HIFIMAN's HE6se V2 sold for $812.77 and the Arya Stealth for $616.19 on Amazon US, and sai's verdict is that 'For a fraction of the cost, the HE6se series offers a similar sound profile, though with slightly less tonal accuracy and technical prowess.' Total cost of ownership is the other hidden figure: nearly every source says a serious amplifier is part of the purchase.