RF survey & site plan
Telstra vs Optus · 5G CPE on a 10–15 m tree mast · container-anchored build
Container anchor -41.372605, 146.378650 · PID 7673283 (8.26 ha) · Kentish, NW Tasmania
Rebuilt 22 October 2026 ACCC Mobile Infrastructure Report 2026 site data AWS Terrain Tiles z14 DEM ITU-R P.526 / P.833
Your container sits in a hollow at 245 m. To the north-east, east and west the ground rises 60–100 m within 500–900 m, and that rim is what decides which towers exist for you. The one direction that gets out is ESE — the Sheffield–Deloraine fall-away — and that is why a 27 km tower beats a 4 km one.
The charts below are real terrain profiles from the AWS Terrain Tiles 14-zoom DEM (about 7 m per pixel, cross-checked against SRTM 30 m at eight points, largest disagreement under 8 m). Dashed blue is the radio ray from a 15 m mast; the pale blue band is the first Fresnel zone. Clearance is what matters, not line of sight: at 850 MHz the zone is roughly 50 m wide at mid-path, so a path can look “open” and still be grazing.
Container → Telstra 7304001 (Deloraine, 27.4 km). The ridge at 700–900 m is the block. At 15 m the ray just grazes it — clearance -0.2 m — which is a working but marginal path.
Container → Telstra 7307001 (Sassafras, 12.5 km). Closer tower, but the near ridge is squarely in the way: no usable path from the container at any sensible mast height.
Container → Telstra 7306008 (Sheffield town, 4.2 km west). The closest site and completely buried — around 100 m of terrain in the way. This is why distance alone tells you nothing here.
101 candidate mast positions inside your boundary, each profiled against the eight towers that matter. Colour is modelled Telstra throughput at a 15 m mast. Hover any dot for the detail; the green ring is the recommended spot from the fine search.
The gradient is stark: the container end of the block is red/amber (8–13 Mbps), the SE corner is green (30–60 Mbps). The 225 m move changes which tower you are talking to, not just how well you hear the same one.
| From container | Ground m | Telstra tower | Mbps | Optus tower | Mbps |
|---|---|---|---|---|---|
| 225 m ★ | 249 | 7307001 LTE700 · 63° |
75 | 7307011 LTE700 · 63° |
75 |
| 200 m | 242 | 7307001 LTE700 · 64° |
75 | 7307011 LTE700 · 63° |
75 |
| 179 m | 242 | 7307001 LTE700 · 64° |
36 | 7305003 LTE700 · 53° |
23 |
| 165 m | 247 | 7307001 LTE700 · 64° |
22 | 7305003 LTE700 · 54° |
24 |
| 128 m | 258 | 7304001 NR850 · 124° |
13 | 7305003 LTE700 · 55° |
21 |
| 117 m | 257 | 7304001 NR850 · 124° |
13 | 7305003 LTE700 · 55° |
22 |
| 28 m | 246 | 7304001 NR850 · 125° |
13 | 7305003 LTE700 · 56° |
16 |
| 40 m | 246 | 7304001 NR850 · 125° |
13 | 7305003 LTE700 · 56° |
16 |
| 20 m | 246 | 7304001 NR850 · 125° |
13 | 7305003 LTE700 · 56° |
16 |
★ / the ring on the map is the recommended point (-41.37431, 146.38011, 225 m from the container, ground 249 m) — found by a second, finer pass: the coarse 101-cell grid saturates at the band ceiling, so the fine search re-scored 248 parcel points on Fresnel clearance alone and picked the highest, on a small rise 4 m above the pad. Ground heights come from the same DEM.
Your instinct to keep it near 10 m is right, and at the SE corner it is more than right. At the container the path to 7304001 improves smoothly with height — about half a dB per 2.5 m, roughly 0.4 Mbps per metre — with no cliff edge, so a 10 m mast is not a mistake, just a slower one. At the SE corner the curve is flat: the path is already clear at 5 m (11 m of Fresnel clearance at 10 m, 16 m at 15 m) and every height from 5 m up models at the ceiling. Go to 10 m for wind and tree-sway margin, not for signal.
| Height | Clearance | RSRP | Mbps |
|---|---|---|---|
| 5 m | -6.4 | -87 | 8 |
| 7.5 m | -4.9 | -86 | 9 |
| 10 m | -3.3 | -86 | 10 |
| 12.5 m | -1.7 | -85 | 11 |
| 15 m | -0.2 | -85 | 12 |
| 17.5 m | +1.4 | -84 | 13 |
| 20 m | +2.9 | -84 | 15 |
| 25 m | +6.1 | -82 | 17 |
| 30 m | +9.2 | -82 | 19 |
| Height | Clearance | RSRP | Mbps |
|---|---|---|---|
| 5 m | +6.3 | -69 | 75 |
| 7.5 m | +8.8 | -69 | 75 |
| 10 m | +11.2 | -69 | 75 |
| 12.5 m | +13.7 | -69 | 75 |
| 15 m | +16.2 | -69 | 75 |
| 20 m | +21.2 | -69 | 75 |
| 25 m | +25.9 | -69 | 75 |
RSRP is the modelled reference-signal level at the CPE; −85 dBm is usable, −75 dBm is comfortable. Throughput figures are model output with a network ceiling applied — they are hypotheses to test, not promises.
Every site within 30 km that has a band your CPE can use, at a 15 m mast, from the container — 45 sites in the full dataset, ranked here by the best usable band. Bands are the 2026 ACCC Mobile Infrastructure Report site records, i.e. what each carrier actually has switched on rather than what a coverage map claims. “Usable” here means sub-1 GHz: 2.6 GHz and 3.5 GHz are modelled and reported in § 5G reality check, but they need clear line of sight this block does not have, so ranking on them would flatter the site.
| Site | Carrier | km | Bearing | Best usable band | Clr m | Diff dB | RSRP | Mbps | Verdict |
|---|---|---|---|---|---|---|---|---|---|
| 7305003 Sheffield north (Claude Rd) |
Optus | 4.4 | 56° | LTE Band 28 (700 MHz) | -98.8 | 35.9 | -90 | 10–16 | deep shadow — unlikely |
| 7304001 Deloraine |
Telstra | 27.4 | 125° | LTE Band 28 (700 MHz) | -0.1 | 7.5 | -85 | 7–15 | marginal |
| 7304001 Deloraine |
Optus | 27.4 | 125° | LTE Band 28 (700 MHz) | -0.1 | 7.5 | -85 | 7–15 | marginal |
| 7304004 Mole Creek Rd / Cethana |
Telstra | 22.8 | 145° | LTE Band 28 (700 MHz) | -2.6 | 10.9 | -86 | 5–14 | marginal |
| 7304007 Mole Creek Rd |
Optus | 22.9 | 145° | LTE Band 28 (700 MHz) | -3.1 | 11.1 | -86 | 5–14 | marginal |
| 7304002 Lower Barrington |
Telstra | 15.2 | 111° | LTE Band 28 (700 MHz) | -33.7 | 22.1 | -91 | 3–9 | speculative |
| 7304002 Lower Barrington |
Optus | 15.2 | 111° | LTE Band 28 (700 MHz) | -33.7 | 22.1 | -91 | 3–9 | speculative |
| 7306008 Sheffield town |
Telstra | 4.2 | 270° | LTE Band 28 (700 MHz) | -96.7 | 42.2 | -95 | 0–14 | deep shadow — unlikely |
| 7306008 Sheffield town |
Optus | 4.2 | 270° | LTE Band 28 (700 MHz) | -96.7 | 42.2 | -95 | 0–14 | deep shadow — unlikely |
| 7305001 | Telstra | 5.6 | 34° | LTE Band 28 (700 MHz) | -102.3 | 42.7 | -99 | 0–2 | deep shadow — unlikely |
| 7310014 | Telstra | 12.3 | 336° | LTE Band 28 (700 MHz) | -244.3 | 37.9 | -104 | 0–0 | deep shadow — unlikely |
| 7310014 | Optus | 12.3 | 336° | LTE Band 28 (700 MHz) | -244.3 | 37.9 | -104 | 0–0 | deep shadow — unlikely |
| 7307001 Sassafras / Harford |
Telstra | 12.5 | 64° | LTE Band 28 (700 MHz) | -60.1 | 30.8 | -97 | 0–0 | deep shadow — unlikely |
| 7307011 Sassafras / Harford |
Optus | 12.7 | 64° | LTE Band 28 (700 MHz) | -61.3 | 31.0 | -98 | 0–0 | deep shadow — unlikely |
| 7310021 West Kentish |
Optus | 15.4 | 270° | LTE Band 28 (700 MHz) | -109.3 | 38.6 | -108 | 0–0 | deep shadow — unlikely |
| 7307008 | Telstra | 15.6 | 13° | LTE Band 28 (700 MHz) | -93.3 | 39.7 | -109 | 0–0 | deep shadow — unlikely |
| 7307002 | Telstra | 15.6 | 28° | LTE Band 28 (700 MHz) | -92.9 | 34.6 | -104 | 0–0 | deep shadow — unlikely |
| 7307012 | Optus | 16.2 | 10° | LTE Band 28 (700 MHz) | -91.8 | 37.2 | -107 | 0–0 | deep shadow — unlikely |
| 7307009 | Optus | 17.0 | 21° | LTE Band 28 (700 MHz) | -91.7 | 34.4 | -105 | 0–0 | deep shadow — unlikely |
| 7306015 | Telstra | 17.4 | 251° | LTE Band 28 (700 MHz) | -305.0 | 50.8 | -122 | 0–0 | deep shadow — unlikely |
| 7310001 | Telstra | 18.5 | 347° | LTE Band 28 (700 MHz) | -229.5 | 44.0 | -116 | 0–0 | deep shadow — unlikely |
| 7310001 | Optus | 18.5 | 347° | LTE Band 28 (700 MHz) | -229.5 | 44.0 | -116 | 0–0 | deep shadow — unlikely |
| 7310016 | Telstra | 18.9 | 352° | LTE Band 28 (700 MHz) | -242.6 | 46.6 | -119 | 0–0 | deep shadow — unlikely |
| 7310009 | Telstra | 19.2 | 346° | LTE Band 28 (700 MHz) | -224.8 | 44.0 | -116 | 0–0 | deep shadow — unlikely |
| 7310023 | Optus | 19.2 | 359° | LTE Band 28 (700 MHz) | -124.8 | 34.0 | -106 | 0–0 | deep shadow — unlikely |
| 7304012 | Telstra | 19.5 | 172° | LTE Band 28 (700 MHz) | -275.6 | 48.3 | -121 | 0–0 | deep shadow — unlikely |
| 7304013 | Optus | 20.3 | 171° | LTE Band 28 (700 MHz) | -381.1 | 50.0 | -123 | 0–0 | deep shadow — unlikely |
| 7310019 | Telstra | 20.3 | 350° | LTE Band 28 (700 MHz) | -236.4 | 45.4 | -118 | 0–0 | deep shadow — unlikely |
| 7310022 | Optus | 20.7 | 252° | LTE Band 28 (700 MHz) | -58.0 | 29.4 | -103 | 0–0 | deep shadow — unlikely |
| 7306002 | Telstra | 20.9 | 231° | LTE Band 28 (700 MHz) | -48.4 | 23.9 | -97 | 0–0 | speculative |
| 7310004 | Telstra | 21.2 | 356° | LTE Band 28 (700 MHz) | -129.9 | 36.7 | -110 | 0–0 | deep shadow — unlikely |
| 7310005 | Telstra | 21.5 | 349° | LTE Band 28 (700 MHz) | -234.4 | 44.5 | -118 | 0–0 | deep shadow — unlikely |
| 7310005 | Optus | 21.5 | 349° | LTE Band 28 (700 MHz) | -234.4 | 44.5 | -118 | 0–0 | deep shadow — unlikely |
| 7310025 | Optus | 21.7 | 356° | LTE Band 28 (700 MHz) | -130.6 | 39.0 | -113 | 0–0 | deep shadow — unlikely |
| 7310018 | Telstra | 21.9 | 355° | LTE Band 28 (700 MHz) | -149.8 | 42.7 | -117 | 0–0 | deep shadow — unlikely |
| 7307003 | Telstra | 22.0 | 16° | LTE Band 28 (700 MHz) | -88.7 | 33.9 | -108 | 0–0 | deep shadow — unlikely |
| 7310020 | Telstra | 22.8 | 353° | LTE Band 28 (700 MHz) | -192.7 | 44.0 | -119 | 0–0 | deep shadow — unlikely |
| 7310012 | Optus | 22.8 | 356° | LTE Band 28 (700 MHz) | -130.3 | 38.5 | -113 | 0–0 | deep shadow — unlikely |
| 7310006 | Telstra | 23.0 | 356° | LTE Band 28 (700 MHz) | -130.6 | 38.9 | -114 | 0–0 | deep shadow — unlikely |
| 7304021 | Telstra | 23.3 | 211° | LTE Band 28 (700 MHz) | -675.1 | 57.5 | -132 | 0–0 | deep shadow — unlikely |
| 7315006 | Optus | 23.5 | 336° | LTE Band 28 (700 MHz) | -257.7 | 38.2 | -113 | 0–0 | deep shadow — unlikely |
| 7310007 | Telstra | 23.6 | 335° | LTE Band 28 (700 MHz) | -255.0 | 39.0 | -114 | 0–0 | deep shadow — unlikely |
| 7310028 | Optus | 24.0 | 346° | LTE Band 28 (700 MHz) | -292.7 | 45.1 | -121 | 0–0 | deep shadow — unlikely |
| 7307005 | Telstra | 25.1 | 30° | LTE Band 28 (700 MHz) | -90.1 | 34.2 | -110 | 0–0 | deep shadow — unlikely |
| 7307005 | Optus | 25.1 | 30° | LTE Band 28 (700 MHz) | -90.1 | 34.2 | -110 | 0–0 | deep shadow — unlikely |
| 7315004 | Telstra | 26.0 | 317° | LTE Band 28 (700 MHz) | -248.3 | 46.9 | -124 | 0–0 | deep shadow — unlikely |
| 7315002 | Telstra | 27.1 | 328° | LTE Band 28 (700 MHz) | -162.7 | 42.4 | -120 | 0–0 | deep shadow — unlikely |
| 7315002 | Optus | 27.1 | 328° | LTE Band 28 (700 MHz) | -162.7 | 42.4 | -120 | 0–0 | deep shadow — unlikely |
| 7315009 | Telstra | 28.3 | 290° | LTE Band 28 (700 MHz) | -213.9 | 48.0 | -126 | 0–0 | deep shadow — unlikely |
| 7306005 | Telstra | 28.4 | 235° | LTE Band 28 (700 MHz) | -190.2 | 42.6 | -121 | 0–0 | deep shadow — unlikely |
| 7304016 | Telstra | 28.7 | 126° | LTE Band 28 (700 MHz) | -63.4 | 27.1 | -105 | 0–0 | speculative |
| 7315001 | Telstra | 29.4 | 324° | LTE Band 28 (700 MHz) | -210.2 | 44.3 | -123 | 0–0 | deep shadow — unlikely |
| 7315001 | Optus | 29.4 | 324° | LTE Band 28 (700 MHz) | -210.2 | 44.3 | -123 | 0–0 | deep shadow — unlikely |
Clr = worst-case terrain clearance on the path (negative = terrain intrudes into the ray). Diff = diffraction loss, central estimate between a principal-edge model and a multi-edge (Deygout) model. Range is the span of the two models scaled for uncertainty — read the upper end as “best case with a good install”.
| Site | Carrier | km | 700 | 850 | 1800 | 2100 | 2600 | n28 700 | n5 850 | n7 2600 | n78 3.6G |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 7304001 Deloraine | Telstra | 27.4 | ● | ● | ● | ● | ● | · | ● | ● | ● |
| 7304001 Deloraine | Optus | 27.4 | ● | · | ● | ● | · | ● | · | · | · |
| 7307001 Sassafras / Harford | Telstra | 12.5 | ● | · | · | · | · | · | · | · | · |
| 7307011 Sassafras / Harford | Optus | 12.7 | ● | · | · | · | · | · | · | · | · |
| 7304004 Mole Creek Rd / Cethana | Telstra | 22.8 | ● | ● | ● | ● | · | · | ● | · | · |
| 7304007 Mole Creek Rd | Optus | 22.8 | ● | · | · | · | · | · | · | · | · |
| 7305003 Sheffield north (Claude Rd) | Optus | 4.4 | ● | · | · | · | · | · | · | · | · |
| 7306008 Sheffield town | Telstra | 4.2 | ● | ● | ● | ● | ● | · | ● | ● | · |
Filled dot = band present at that site in the 2026 data, for the eight towers that matter. Telstra runs 5G on the Deloraine site (n5, n7, n78) and is 4G-only at the two nearer ones; Optus has 5G (n28) at Deloraine and at the western site 7310021, and is LTE-only at Sassafras.
Plan pricing is deliberately not quoted here: fixed-wireless plan structures and introductory pricing change monthly, and a BYO-CPE data plan is priced differently from a supplied-router 5G home plan. What matters for your build is that you want a data-only SIM with no CPE lock-in — the ATL 5G is not on either carrier's supported-device list, so tell the sales process you are putting the SIM in your own router and check the plan does not require their hardware.
You bought a 5G CPE. Here is the honest position on whether 5G helps at this address.
| Band | Where it is | What it would give you | Verdict |
|---|---|---|---|
| n5 — 850 MHz | Telstra 7304001, 27.4 km | 10–13 Mbps at 10–15 m from the container | Your realistic 5G, and it is a 4G-class speed |
| n28 — 700 MHz | Optus 7304001; Optus 7310021 (west, blocked) | Same path as Telstra n5, similar result | Works, no advantage |
| n78 — 3.5 GHz | Telstra 7304001 | Modelled 90–110 Mbps, but the path grazes the Fresnel zone by −0.1 m at 27 km | Do not design for it. A single tree or a wet day kills it |
| 2.6 GHz (n7 / LTE2600) | Telstra 7304001, 7306008 | Needs the Fresnel zone to itself at 15–27 km | Not available to you |
The useful consequence: the fast link on this block is 4G, not 5G. At the SE corner the best tower is LTE700 only and it still beats every 5G option by a factor of three. Your CPE is a Cat20 LTE modem plus 5G, so it handles either — you are not wasting the purchase, you are just not buying speed with the 5G label.
Short version: the ATL 5G cannot take an external antenna. It is a sealed outdoor unit with four mid-band and four low-band elements moulded in, one PoE port and no connectors for anything else. So "which antenna has the most gain" is really "what would I gain, and what would it cost to be able to use one" — both answered below.
The bands that are actually usable from the container are 700, 850 and 900 MHz, all on Telstra 7304001 (Deloraine) at 27.4 km. That narrows the market sharply: an antenna has to do its work at the bottom of the spectrum, where gain only comes from physical size. Wideband "all-band" panels quote 10–11 dBi at 2.7 GHz; at 700 MHz the same box gives about half that, and it is the 700 MHz end that decides whether this path works.
| Antenna | Gain | Band | MIMO / pol | Beam | Size & weight | Price | Verdict for the container |
|---|---|---|---|---|---|---|---|
| Telco XPOL700-900PL | 16 dBi | 698–960 MHz | 2×2, ±45° X-pol | 30° × 30° | 600×560×156 mm, 12 kg | $445 | The pick: full 700–900 coverage, MIMO in one box, 2× N female, 198 km/h rating |
| Telco TEL-Y720 Yagi ×2 | 16 dBi each | 700–890 MHz | Single pol — two needed | 27–28° | 2,360 mm, 0.75 kg each | $792 | Same gain, but two 2.4 m booms, twice the wind load and a harder aim |
| RFI YB815-82 | 14 dBi | 850–930 MHz | Single pol | 30° | 1,600 mm, 1 kg | $249 | Good for n5/B5 — but it misses Band 28 entirely |
| Dual 11 dBi MIMO LPDA | 10–11 dBi peak | 698–2700 MHz | 2×2 | 50° × 40° | 445×265×65 mm, 1.2 kg | $178 | Cheap MIMO, wide beam — but about 5–6 dBi where you need it at 700 MHz |
| ZCG 22 dBi grid | 22 dBi | ≈700 MHz | Single pol | Narrow | Large dish | $5,500 | The most gain money can buy here, and the least sensible: heavy, single band, huge |
AUD inc GST, October 2026, Telco Antennas (Hemmant, QLD) and Remote One. "16 dBi on 700" is the manufacturer's number for the band, not a peak spread across the range.
Modelled at 15 m, Telstra 7304001, planning loss 7.3–7.5 dB, against the 3.5 dBi the current model assumes for the integrated array:
| Antenna gain at 700–850 MHz | RSRP | SINR | Modelled | What it means |
|---|---|---|---|---|
| 3.5 dBi — ATL 5G as shipped (model assumption) | −84.7 | +0.5 | 12 Mbps | Usable, but a graze that rain and leaves erode |
| 11 dBi — cheap MIMO LPDA | −79.7 | +5.5 | 25–27 Mbps | Doubles the link; enough for two 4K streams |
| 13 dBi | −77.7 | +7.5 | 31–35 Mbps | Comfortably above the SE corner's floor |
| 16 dBi — XPOL700-900PL / Yagi pair | −74.7 | +10.5 | ceiling | The path stops being the limit; the cell and its 20 MHz do. Expect 35–55 Mbps real |
The dB also buys uplink, which is what breaks first on a 27 km path: at SINR +0.5 dB your phone call and your camera uploads are the things that stutter, and antenna gain is the one fix that helps both directions at once.
If the storage is fixed at the container, the antenna route is the right answer and the XPOL700-900PL is the right antenna — 16 dBi across the whole 700–960 MHz range in a 12 kg panel is about as good as a single-box solution gets before you are buying a 22 dBi grid for $5,500. If the storage can move, spend the money on cable instead.
880 W of panel (2 × Powerbay PB-DH96NE-440, series: ~70 V Voc, ~16 A Isc), a 12 V 200 Ah LiFePO4 (2.56 kWh, 2.3 kWh usable) and an MPPT into a 12 V bus. Solar resource from NASA POWER 2001–2020 climatology at your coordinates, transposed to a 45° north-facing array with a numerical solar-geometry model — i.e. one that knows the sun is 72° high in December and 25° in June, and that a 45° tilt is worth roughly twice a flat panel in midwinter.
| Device | Typ | Peak | Source |
|---|---|---|---|
| ATL 5G CPE (PoE) | 8 W | 10 W | 802.3af/at, 12-56 V, 10 W max |
| WAVLINK AX3000 AP | 10 W | 14 W | assumed 802.3af / 10 W — check your unit |
| Reolink TrackMix WiFi | 6 W | 12 W | DC 12 V 2 A, 12 W max with IR |
| 12 V industrial PoE sw | 5 W | 8 W | assumed incl. PoE conversion loss |
| Victron MPPT self-use | 0.6 W | 1 W | standby draw |
Typical 29.6 W continuous, 45 W peak, 0.85 duty → 0.60 kWh/day.
| Month | GHI | POA | °C | kWh/d | Balance |
|---|---|---|---|---|---|
| Jan | 6.30 | 5.45 | 16 | 3.59 | +2.98 |
| Feb | 5.42 | 4.81 | 16 | 3.20 | +2.60 |
| Mar | 3.98 | 3.58 | 14 | 2.44 | +1.84 |
| Apr | 2.56 | 2.32 | 12 | 1.63 | +1.02 |
| May | 1.85 | 1.74 | 10 | 1.24 | +0.64 |
| Jun | 1.51 | 1.46 | 8 | 1.05 | +0.44 |
| Jul | 1.66 | 1.63 | 8 | 1.17 | +0.56 |
| Aug | 2.40 | 2.26 | 8 | 1.60 | +1.00 |
| Sep | 3.32 | 3.01 | 10 | 2.10 | +1.50 |
| Oct | 4.49 | 3.97 | 11 | 2.73 | +2.13 |
| Nov | 5.43 | 4.72 | 12 | 3.19 | +2.59 |
| Dec | 6.20 | 5.34 | 14 | 3.55 | +2.94 |
Result: annual average 2.29 kWh/day generated against 0.60 kWh/day consumed. The worst month is June at 1.05 kWh/day — still a +0.44 kWh/day surplus. Battery autonomy at that load is 3.8 days. The system is correctly sized with a modest winter margin, which is what you want: it should be able to carry two or three dark days without touching a generator.
The model brackets the answer but cannot settle it: terrain diffraction at 27 km has ± 6 dB of legitimate model spread, and cell loading is invisible from here. Two prepaid data SIMs settle it in an afternoon.
/interface/lte/monitor in RouterOS). Repeat with the Optus SIM.Conventions, as in the earlier Sheffield reports: clearances are signed metres (positive = ray clears terrain);
bearings are true north; distances are geodesic km; coordinates are WGS84; prices are omitted where they would go stale.
Everything here is reproducible from the project files in scratch/sheffield5g/ — profiles.jsonl,
grid_results.jsonl, power_budget.json.