The instruments. Six configurations, three targets.

A 60 × 60 cm cylindrical parabolic trough, hand-built from cardboard formers, a nichrome-wire skeleton, a tensioned plastic-film skin and an aluminium-foil reflector, with a linear half-wave dipole at the focal line. Aimed at GK-2A. Full-disk images dated 23 May to 1 June 2026 were captured with this instrument, before the 3D-printed dish at Instrument VI superseded it on 2 June.

Antenna

60 × 60 cm cylindrical parabolic trough, hand-built (23 May 2026)

Reflector

Cardboard formers · nichrome wire skeleton · plastic-film skin · aluminium-foil surface

Feed

Linear half-wave dipole at the focal line (8.9 cm @ 1.69 GHz), parallel to the spine

Pointing

az 103° true · el 25° (→ 128.2°E geostationary)

Front-end

Nooelec SAWbird+ GOES (LNA + SAW, externally powered; SDR bias-tee OFF)

Frequency

1692.140 MHz (LRIT, BPSK 128 ksymb/s post-Viterbi 64 kbps)

Sample rate

2.4 Msps cu8, --dc_block=true

Post

Sanchez geostationary projection · blue underlay · haze 0.6

Operative note GK-2A LRIT is right-hand circularly polarised; a linear dipole takes a fixed 3 dB cross-polarisation loss regardless of orientation. The trough's larger aperture partly compensates: the 24 May timelapse held lock for nine consecutive slots above the 1.5 dB Peak-SNR threshold this rig needs to decode. The cleanest recovery path is to swap the dipole for an RHCP helix feed at the same focal line; the trough geometry is unchanged.

View captures from this instrument →

A 12-turn axial-mode helix wound for RHCP, on a 25 × 25 cm aluminium ground plane, hand-pointed. It produced the first decoded GK-2A frames on 22 May 2026. Replaced the next day by the parabolic trough at Instrument I; kept here for provenance.

Antenna

12-turn axial-mode helix, RHCP, hand-wound

Reflector

25 × 25 cm aluminium ground plane

Pointing

az 103° true · el 25° (→ 128.2°E geostationary), hand-aimed

Front-end

Nooelec SAWbird+ GOES (LNA + SAW, ext. PSU; bias-tee OFF)

Frequency

1692.140 MHz (LRIT, BPSK 128 ksymb/s)

Sample rate

2.4 Msps cu8, --dc_block=true

Status

Retired 23 May 2026; replaced by the parabolic trough at Instrument I

Operative note The helix matched GK-2A's RHCP downlink, so unlike the trough's linear dipole there was no 3 dB cross-polarisation loss in the link budget. The trade-off was capture area: a 25 × 25 cm helix collects less signal than a 60 × 60 cm parabolic surface, and the larger aperture was preferred. Result on this chain: 22 May 2026, 3 IR105 full-disk images at ~2.2 dB Peak SNR.

View captures from this instrument (22 May 2026 plate) →

The same parabolic trough and SAWbird+ front-end as Instrument I, repointed west towards the 76°E geostationary slot to receive Elektro-L5's LRIT broadcasts. L5 is in commissioning and transmits only sporadically.

Antenna

60 × 60 cm parabolic trough (shared with Instrument I)

Feed

Linear half-wave dipole at the focal line

Pointing

Repointed west of GK-2A by ~52° in azimuth (target 76°E equator)

Front-end

Nooelec SAWbird+ GOES (LNA + SAW, ext. PSU)

Frequency

1691.000 MHz (LRIT)

Schedule

Sporadic: slots at :12/:27/:42/:57 UTC, mostly silent

Operative note L5 has been in commissioning since 2026-02-12, with broadcasts clustered around 08:00 UTC and rare outside that window. The first attempt on 2026-05-17 produced zero CADU: Viterbi never synced, and an FFT confirmed no signal in band. It is not yet known whether the antenna was off the bird, the polarisation sense was wrong, or L5 was not transmitting during that slot.

View captures (commissioning, no images yet) →

A pair of crossed half-wave elements on a rooftop pole, receiving the LRPT downlink during METEOR overpasses. The simplest chain on the rig, and the source of the only polar-orbit imagery in the register so far.

Antenna

Crossed V-dipole, linear polarisation

Mount

Rooftop pole, ~3.6 m, no rotator

Frequency

137.900 MHz (LRPT primary)

Sample rate

1.024 Msps (Nyquist for 1024 ksymbol/s QPSK)

Gain

RTL gain 25 dB (no physical LNA in chain)

Decode

SatDump on NAS, --dc_block=true mandatory for cu8

Operative note The V-dipole is linear and LRPT is circular, so there is an unavoidable ~3 dB cross-polarisation loss. The dominant limit, though, is tuner gain rather than polarisation: pushing the R828D above 30 dB compresses the front end and raises the noise floor by 20+ dB. At 25 dB the rig is the quietest it has been measured.

View captures from this instrument →

The Lucknow L-band chain: a biquad feed in a tin-box reflector at the focus of a 1 m parabolic dish, pointed at GK-2A. This is the permanent-site build the 1 m dish was planned around; the trough and helix were the temporary Mumbai rigs that proved the link first. Build photographs are in the field record.

Antenna

1 m prime-focus parabolic dish (100 cm aperture), Lucknow site

Feed

Biquad (double-quad) at the dish focus, in a biscuit-tin reflector cavity

Polarisation

Linear, a fixed 3 dB cross-polarisation loss against GK-2A's RHCP downlink

Pointing

az 112.6° true · el 29.6° (Lucknow 26.85°N 80.95°E → 128.2°E geostationary)

Front-end

Nooelec SAWbird+ GOES (LNA + SAW, ext. PSU; bias-tee OFF)

Frequency

1692.140 MHz (LRIT, BPSK 128 ksymb/s)

Sample rate

2.4 Msps cu8, --dc_block=true

Also

The same dish repoints ~52° west to Elektro-L5 at 76°E

Operative note A biquad is linearly polarised, so like the trough's dipole it takes a fixed 3 dB cross-polarisation loss against GK-2A's right-hand circular downlink. The 1 m dish has roughly 2.5× the collecting area of the 60 cm Mumbai trough, which more than offsets that loss; the biscuit-tin box behind the biquad acts as a sub-reflector, tightening the feed pattern onto the dish. At a 1 m aperture the L-band beamwidth is only ~12°, so the dish has to be aimed far more precisely than the hand-pointed Mumbai rigs.

See the build photographs →

A 600 mm prime-focus parabolic dish printed as a flat-pack skeleton: eight PLA ribs slip-fit into a hub, with bamboo rings backing eight foil-on-paper gores at f/D 0.50. A coax-fed linear dipole sits at the focus with the LNA mounted right at the feed. It replaced the cardboard trough and decoded GK-2A at 4.66 dB Peak SNR from a Mumbai window. Full-disk images dated 2 June 2026 or later were captured with this instrument.

Antenna

600 mm prime-focus 3D-printed paraboloid · f = 300 mm (f/D 0.50, 75 mm deep)

Structure

Eight slip-fit PLA ribs, a hub and a feed boom; bamboo-skewer rim and purlin rings, zip-tied. Flat-packs.

Surface

Eight gores of two-ply chart paper laminated with aluminium foil (the foil is the reflector; ~0.5 dB faceting loss across 8 facets)

Feed

Coax-fed linear half-wave dipole (~42 mm arms) on a slide/rotate collar, with a Ø150 mm foil reflector at λ/4 behind it

Polarisation

Linear, matched to GK-2A's linear LRIT downlink: no cross-polarisation loss, where a helix would take 3 dB on a linear signal

Pointing

az 103° true · el 25° (Mumbai → 128.2°E geostationary); ~20° beam, hand-aimed

Front-end

Nooelec SAWbird+ GOES (NF ~0.8 dB) at the feed, then a long RG6 run to the SDR; externally powered, bias-tee off

Frequency

1692.140 MHz (LRIT, BPSK 128 ksymb/s)

Sample rate

1.024 Msps cu8 (pipeline-native; ~10× lighter than the trough's 2.4 Msps)

Gain

≈17.5 dBi theoretical at 50% efficiency

Operative note The LNA sits at the feed and the long coax comes after it by design: by Friis, with ~30 dB of gain up front, the cable loss and the RTL's own noise figure add only hundredths of a dB to the system noise, so a long RG6 run costs almost nothing in SNR. First light, reflectorless, returned 4.66 dB Peak SNR at Viterbi BER ~0.08, roughly double the cardboard trough and a clean margin over the ~1.5 dB this rig needs to decode. A Ø150 mm feed reflector has since been fitted to recover a further 3 to 4 dB. The chain live-decodes straight from the SDR with no IQ written to disk.

View captures from this instrument →