LoRa Range on Indian Farms: A Link-Budget Field Guide for Ponds, Sheds and Fields
136 dB. That is the most path loss a LoRa link at spreading factor 9 can tolerate when a 14 dBm node talks to a gateway with a 6 dBi antenna and the receiver needs 10 dB of fade margin. It sounds like a generous allowance until a farm starts spending it. A row of coconut palms takes a few decibels. A tin roof between the node and the mast takes several more. A pond surface that reflects the signal back into itself turns a steady link into one that fades in and out through the day. By the time the budget has paid for all of that, the 20-kilometre range in the brochure has become a few kilometres on a good day and a few hundred metres behind a shed.
This is a field guide to spending that budget on purpose. It takes the link budget apart term by term, puts numbers on each term for Indian farm conditions, runs a calculator you can reproduce, and ends with a field-test procedure that replaces the guesses with measurements. The audience is a team about to place gateways and nodes across ponds, sheds and fields and wanting to know where the weak links will be before the first site visit.
Two honest notes on method. The calculator and its outputs were executed and the numbers below are quoted from the runs. Where a figure is a typical planning value rather than a measurement from this post (the path-loss exponents, the foliage loss per metre, the roof loss), it is labelled as an assumption, and the guide's central advice is to replace every assumption with a measurement from your own site.
A link budget is a plan, not a promise. The brochure range is the budget with nothing spent on the site; the farm spends most of it.
The Link Budget in One Line
Every radio link obeys the same accounting. The signal leaves the transmitter with some power, is boosted by the transmit antenna, loses energy across the path, is boosted again by the receive antenna, and must arrive above the level the receiver can decode, with a safety margin. In decibels the accounting is one line:
maximum allowed path loss = transmit power + transmit antenna gain + receive antenna gain − cable loss − receiver sensitivity − fade margin − extra losses
Term | What it is | Typical value for a farm node and gateway |
Transmit power | Output of the node's radio | 14 dBm is a common default; some radios go to 20 dBm or more |
Transmit antenna gain | Gain of the node's antenna | About 2 dBi for a small whip, less if the antenna is poorly matched or enclosed |
Receive antenna gain | Gain of the gateway antenna | 3 to 9 dBi for a collinear mast antenna |
Cable loss | Feeder and connector loss | 1 to 3 dB for a short run of good cable; more for long runs or cheap cable |
Receiver sensitivity | Weakest signal the receiver can decode | Depends on spreading factor; about minus 123 dBm at SF7 to minus 137 dBm at SF12 for a 125 kHz channel |
Fade margin | Reserve for fading, weather and ageing | 10 to 15 dB is a common planning value; more for critical links |
Extra losses | Foliage, buildings, body shadowing, water | Site-specific; the part you must estimate and then measure |
The first five terms are numbers from datasheets, and most farm deployments differ little in them. The last two are where the farm lives. A link budget exercise that spends an hour on antenna gains and thirty seconds on the extra losses has its priorities upside down.
India's Band and What the Rules Allow
LoRa links in India operate in the licence-exempt band of 865 to 867 MHz, so no operator and no licence is needed to run your own network. The LoRaWAN regional parameters define a plan for this band, IN865, and the specification text notes that the Indian band supports a maximum of 16 channels. The regional document also sets a default maximum effective isotropic radiated power for the band, which the reader should confirm in the current version rather than take from this post.
India's delicensing notification for the band is commonly summarised as allowing a transmitter power of up to 1 watt and an effective radiated power of up to 4 watts, with a limited channel bandwidth, but the exact conditions (power definitions, bandwidth, any duty-cycle or use restrictions) come from the current Department of Telecommunications and WPC notifications and should be read from the source before a product is certified or a network is planned. This guide's examples use a 14 dBm node, which sits comfortably under any plausible limit, and the link budget formula applies the same way at any legal power.
Check before design | Where to confirm |
Permitted frequency range and channel plan | Current DoT and WPC delicensing notification; LoRaWAN regional parameters for IN865 |
Maximum transmitter power and EIRP | The same notification; note whether limits are conducted or radiated |
Any duty-cycle, listen-before-talk or bandwidth conditions | The notification text; your module's certification documents |
Equipment authorisation for the node and gateway | WPC equipment type approval route and the test lab's advice |
Outdoor installation conditions | Your enclosure and antenna approvals, and any site rules |
For a product sold commercially, treat the regulatory check as a design input. For a farm network built from approved modules, confirm that the settings of the gateway and nodes match the plan.
Sensitivity and Airtime: What Each Spreading Factor Buys
LoRa trades data rate for sensitivity through the spreading factor. Each step up from SF7 to SF12 improves sensitivity by roughly 2.5 to 3 dB and roughly doubles the time on air. Typical datasheet sensitivities for a 125 kHz channel on a common radio family are listed below, with the time on air of a 33-byte frame (a 20-byte application payload plus 13 bytes of LoRaWAN overhead) computed with the standard airtime formula.
Spreading factor | Typical sensitivity | Airtime for a 33-byte frame | Comment |
SF7 | −123 dBm | 72 ms | Fast and frugal; shortest range |
SF9 | −129 dBm | 247 ms | A good default balance |
SF12 | −137 dBm | 1,810 ms | Longest range; 25 times the airtime of SF7 |
The airtime numbers matter for three reasons. Energy: a long transmission costs proportionally more battery, as our earlier posts on power budgets and solar sizing show. Collisions: a long frame occupies the channel for longer and is more likely to overlap with another node. And regulations or fair-use rules may limit how much airtime a device may use. An adaptive data rate scheme lets the network move a node to the lowest spreading factor that still works, which saves energy and channel time, and a node whose link is marginal will drift to SF12 and pay for it.
Sensitivity figures are datasheet-typical values for a given module and a clean 125 kHz channel. A noisy site, a poorly matched antenna or a receiver desensitised by a neighbouring transmitter will deliver less. When planning, use the datasheet number for the radio you are buying, not a general table.
Path Loss: From Free Space to a Farm
The simplest model of path loss is free space, in which the signal spreads as a sphere and the loss grows with the logarithm of distance and frequency. At 866 MHz the free-space loss is 91.2 dB at 1 kilometre and 105.2 dB at 5 kilometres. Free space is the best case: a clear line of sight with nothing in the way. It sets the lower bound on the loss and the upper bound on the range.
Real sites lose more, and the usual way to model the extra is a log-distance exponent. Loss at distance d equals the free-space loss at a reference distance plus 10 times an exponent times the base-10 logarithm of the ratio of d to the reference. An exponent of 2 reproduces free space. Planning values for low antennas over ground are higher.
Environment (assumption) | Typical exponent | What it represents |
Free space | 2.0 | Clear line of sight, high antennas; rare near the ground |
Open farmland, low antennas | 3.0 to 3.5 | Ground reflection and low clearance |
Scattered trees, sheds | 3.5 to 4.0 | Intermittent obstruction and clutter |
Dense vegetation and buildings | 4.0 to 4.5 | Heavy obstruction |
These exponents are planning assumptions of the kind found in propagation textbooks, not measurements from this post. They matter enormously: moving the exponent from 3.0 to 4.0 cuts the predicted range by a factor of about four at the same path loss. That sensitivity is the reason the field-test section exists. Measure the exponent on your site instead of picking one.
Antenna Height, the Radio Horizon and the Fresnel Zone
Two geometric limits sit behind every path-loss number.
The radio horizon. With standard atmospheric refraction, the line-of-sight limit between two antennas of heights h1 and h2 metres is about 4.12 times the sum of the square roots of the heights, in kilometres. A node antenna 2 metres above the bund and a gateway antenna at 10 metres have a horizon of about 18.9 kilometres, and 3 metres against 15 metres gives about 23.1. Beyond the horizon the ground blocks the direct path and losses climb steeply. For farms, the horizon is rarely the limit; trees and buildings are.
The Fresnel zone. Radio energy travels in a volume around the line of sight, not a thin ray. The first Fresnel zone has a radius at the midpoint of the link that grows with distance and shrinks with frequency. Obstacles inside the zone, even ones that do not touch the line of sight, reduce the signal. A common planning rule is to keep at least 60 percent of the first zone clear.
Link length | First Fresnel radius at 866 MHz | 60 percent of the radius | Earth bulge at midpoint |
1 km | 9.3 m | 5.6 m | 0.01 m |
2 km | 13.2 m | 7.9 m | 0.06 m |
3 km | 16.1 m | 9.7 m | 0.13 m |
5 km | 20.8 m | 12.5 m | 0.37 m |
The numbers say something uncomfortable for ground-level farm links. At 2 kilometres, clearing 60 percent of the first Fresnel zone at the midpoint needs about 8 metres of clearance above any obstacle. A 3-metre node antenna looking across a field of 4-metre crops or a row of trees will sit inside the zone for much of the path. The earth's curvature is negligible at these distances, so it is the objects that cost you, and the cheapest remedy is height at one end, usually the gateway.
Height at the gateway is the cheapest decibel on the farm. Lifting the antenna clears the Fresnel zone, shortens the shadow of every shed and tree, and does it for every node at once.
Palms, Tin Roofs, Ponds and Cattle: Where the Extra Losses Come From
The extra-loss term is the one that varies from farm to farm. The values below are illustrative planning figures to be replaced with measurements.
Obstacle | Typical effect (assumption) | Notes |
Dense foliage | About 0.3 dB per metre of depth through the canopy, so 20 m is about 6 dB | Wet foliage after rain is worse; coconut and mango canopies differ |
Metal roof or wall between node and gateway | 8 dB or more | A shed with a metal roof can shadow or reflect strongly |
Concrete or brick wall | 6 to 12 dB per wall | Depends on thickness and reinforcement |
Pond surface | Reflection that causes fading rather than steady loss | A low antenna over water can see deep fades as the reflected and direct paths interfere |
Animals and people | A few dB of body shadowing | Matters for tags on cattle, where the animal's body blocks the antenna |
Metal tanks, vehicles, pipework | Local shadows and reflections | Position the antenna clear of large metal objects |
Rain and humidity | Small at these frequencies, but wet foliage adds loss | The fade margin should cover it |
Water and metal deserve special care. A pond is a flat reflector, and a node on the bund and a gateway at a similar height can see a direct path and a reflected path that interfere constructively and destructively as temperature, wind and water level change the geometry. The result is a link that works at noon and fails at dusk without anything on the site changing. Fade margin is the answer, along with diversity in the form of a second gateway or a slightly raised antenna. Metal roofs are strong shadows and strong mirrors, and the best gateway position is usually clear of them, not tucked beside one.
The Calculator
The code below puts the pieces together: free-space loss, a maximum allowed path loss from the link budget, a range estimate from a log-distance model, the radio horizon, and the Fresnel radius. It uses the sensitivities from the table above.
# lora_link.py -- link budget, range from a log-distance model, horizon and Fresnel zone
import math
SENS = {7: -123.0, 9: -129.0, 12: -137.0} # dBm, 125 kHz, datasheet-typical
def fspl_db(d_km, f_mhz):
return 32.44 + 20*math.log10(d_km) + 20*math.log10(f_mhz)
def max_path_loss(ptx_dbm, gt_dbi, gr_dbi, sf, cable_db=1.0, fade_db=10.0, extra_db=0.0):
return ptx_dbm + gt_dbi + gr_dbi - cable_db - SENS[sf] - fade_db - extra_db
def range_km(pl_db, f_mhz, n, d0_km=1.0):
return d0_km * 10 ** ((pl_db - fspl_db(d0_km, f_mhz)) / (10 * n))
def horizon_km(h1_m, h2_m):
return 4.12 * (math.sqrt(h1_m) + math.sqrt(h2_m))
def fresnel_r1_m(d_km, f_mhz): # radius at the midpoint of the link
return 17.32 * math.sqrt((d_km/2) * (d_km/2) / ((f_mhz/1000.0) * d_km))
f = 866.0
for extra, label in ((0.0, "clear path"), (14.0, "20 m foliage plus roof or wall (assumption)")):
for sf in (7, 9, 12):
pl = max_path_loss(14, 2.15, 6.0, sf, cable_db=2.0, extra_db=extra)
print(label, f"SF{sf}", f"{pl:.1f} dB", {n: round(range_km(pl, f, n), 1) for n in (3.0, 3.5, 4.0)})
With a 14 dBm node, a 2.15 dBi node antenna, a 6 dBi gateway antenna, 2 dB of cable loss and 10 dB of fade margin, the run gives these upper-bound ranges in kilometres for three exponents:
Case | Max path loss | Range at n = 3.0 | n = 3.5 | n = 4.0 |
Clear path, SF7 | 133.2 dB | 25.0 km | 15.8 km | 11.2 km |
Clear path, SF9 | 139.2 dB | 39.7 km | 23.5 km | 15.8 km |
Clear path, SF12 | 147.2 dB | 73.3 km | 39.7 km | 25.1 km |
14 dB of clutter, SF7 | 119.2 dB | 8.6 km | 6.3 km | 5.0 km |
14 dB of clutter, SF9 | 125.2 dB | 13.6 km | 9.3 km | 7.1 km |
14 dB of clutter, SF12 | 133.2 dB | 25.0 km | 15.8 km | 11.2 km |
Read these numbers as ceilings, not forecasts. The clear-path ranges exceed the radio horizon for farm antenna heights (about 19 to 23 kilometres), so the horizon, not the radio, caps them. The model also ignores the extra loss that low antennas suffer over real ground, which is one reason field ranges on farmland are often a fraction of the model's. The use of the table is comparison: it shows what 14 dB of clutter costs (a factor of two to three in range), what a spreading-factor step buys, and how much the exponent matters.
Two more sensitivities from the same calculator, for SF9 at an exponent of 3.5 on a clear path:
Change | Result |
Fade margin 6 / 10 / 15 / 20 dB | 30.5 / 23.5 / 16.9 / 12.2 km |
Gateway antenna 2.15 / 6 / 9 dBi | 18.2 / 23.5 / 28.6 km |
Each 5 dB of fade margin you give up costs a quarter to a third of the range. Each 3 to 4 dB of antenna gain adds about a fifth. Neither substitutes for clearance, but both are worth specifying deliberately.
The Field Test That Replaces the Assumptions
A day of measurement beats a week of modelling. The procedure below produces the two numbers the model needs, a path-loss exponent and an extra-loss figure for each environment, and it tests the fade margin at the same time.
Set up a reference gateway at its intended height and position, with the final antenna and cable, powered from its final supply.
Use a test node with a fixed transmit power, the same antenna type as the product, and a logger that records the received signal strength and signal-to-noise ratio at the gateway for each message.
Walk or drive a set of distances along representative directions: 100 m, 250 m, 500 m, 1 km, 2 km and further if the site allows. At each point send at least 20 messages at the same spreading factor and record the average and the worst.
Place the node at its real height. A test node held at chest height tells you nothing about a node on a 1.5 metre post at the pond edge.
Cover the awkward places. Behind the shed, under the trees, on the far bund, next to the aerator panel, beside the metal tank.
Repeat at a different time and after rain. Fading shows up as the same spot giving different readings at dawn, noon and dusk.
Compute the path loss at each point: transmit power plus antenna gains minus cable loss minus the received signal strength.
Fit the exponent: plot path loss against the logarithm of distance and fit a line. The slope divided by 10 is the exponent, and the intercept at 1 km is the reference loss.
Record the worst case at each location and compare it with the sensitivity of your chosen spreading factor.
Set the fade margin from the spread you saw, not from a rule of thumb.
A short script makes the fit routine, and the same data feeds the decision about gateway position and the choice of spreading factor.
Measure the exponent on your own farm. It moves the predicted range more than any antenna you can buy, and it takes one day to find.
Fitting the Exponent and the Intercept From a Walk Test
The field test produces a table of distances and path losses. Two numbers come out of it, and the second is the one that surprises people. A straight-line fit of path loss against ten times the logarithm of distance returns the exponent as the slope and the loss at one kilometre as the intercept. The model in the calculator above anchored the loss at one kilometre to free space, 91.2 dB at 866 MHz. A farm with low antennas and clutter often has an intercept tens of decibels higher, and that gap moves the range more than the exponent does.
# fit_pathloss.py -- fit PL(d) = PL1km + 10 n log10(d_km) to walk-test points, and size the fade margin
import numpy as np
def fit_pathloss(d_km, pl_db):
x = 10 * np.log10(np.asarray(d_km))
A = np.column_stack([np.ones_like(x), x])
(pl1, n), *_ = np.linalg.lstsq(A, np.asarray(pl_db), rcond=None)
sigma = (np.asarray(pl_db) - A @ [pl1, n]).std(ddof=2) # scatter around the line (shadowing)
return pl1, n, sigma
def fade_margin_db(sigma_db, coverage=0.95):
z = {0.90: 1.28, 0.95: 1.64, 0.99: 2.33}[coverage]
return z * sigma_db
def range_km(ptx_dbm, gain_dbi, cable_db, sens_dbm, fade_db, pl1, n):
max_pl = ptx_dbm + gain_dbi - cable_db - sens_dbm - fade_db
return 10 ** ((max_pl - pl1) / (10 * n))
Run on a synthetic walk test of twelve points between 100 m and 2.5 km, generated with an exponent of 3.9, a loss of 124 dB at one kilometre and 6 dB of random shadowing, the fit returned a loss of 124.6 dB at one kilometre, an exponent of 3.66 and a scatter of 6.0 dB. The fade margin for covering 90, 95 and 99 percent of locations came out at 7.6, 9.8 and 13.9 dB. Feeding the fit into the range function for SF9, a 14 dBm node, 8.15 dBi of total antenna gain, 2 dB of cable loss and the 9.8 dB margin gave a range of 2.5 kilometres, and 2.9 kilometres with the 90 percent margin.
Quantity | Free-space ceiling model | Field-fitted model (synthetic walk test) |
Loss at 1 km | 91.2 dB | 124.6 dB |
Exponent | 3.0 to 4.0 (assumed) | 3.66 (fitted) |
Fade margin | 10 dB (assumed) | 7.6 to 13.9 dB (from scatter) |
SF9 range on a clear path | 16 to 40 km (horizon-limited) | 2.5 to 2.9 km |
The synthetic data is built to resemble a low-antenna farm site and is not a measurement from any real one, but the lesson transfers. The brochure and the ceiling model answer the question "what is the most this radio can do", and the fitted model answers the question "what does this farm give me". They differ by an order of magnitude for ground-level nodes, which is why the field-test section sits at the heart of this guide. A site that fits at 2.5 kilometres can be a very good site; the mistake is to plan for 20.
The intercept is as important as the exponent. If your fitted loss at one kilometre is thirty decibels above free space, you have found the reason your range is a tenth of the brochure's.
Antenna Choice and Mounting
Antenna and mounting decisions are cheap and permanent, and several recur on farms.
Choice | Effect | Practical rule |
Omnidirectional collinear on the gateway | Even coverage in azimuth, higher gain with a narrower vertical beam | Good for a central gateway; the gain comes from squeezing the beam toward the horizon, so keep the mast vertical |
Directional antenna at the gateway | Much more gain in one direction, nothing behind | Use for a long strip of ponds from one end, or to reach a distant cluster |
Small whip on the node | Low gain, tolerant of orientation | Keep it vertical, away from metal and the body of the enclosure |
Antenna inside a sealed enclosure | Loses several decibels to the box and the electronics | Use an enclosure designed for the antenna, or an external one with a short pigtail |
Height above ground | Clears ground reflection and Fresnel obstruction | A metre or two on a post can be worth several decibels at the node |
Separation from metal | Metal near the antenna detunes and shadows it | Keep the antenna well clear of tanks, roofs and the aerator panel |
Cable and connectors | Loss and failure points | Short runs, quality cable, weather-sealed joints |
Orientation on cattle tags | The animal's body blocks one side | Mount on the upper side of a collar, and test on a live animal |
Two details trip teams up. First, the ground plane: many small whip antennas assume a metal surface to work against, and a plastic enclosure on a post does not give them one, so their real gain is lower than the datasheet. Second, polarisation: a vertical antenna at one end and a tilted or horizontal one at the other loses several decibels. Fix the orientation in the installation drawing, and check it on commissioning day.
Watching the Margin After Deployment
A network that worked on the day of the test can slide. Trees grow, a new shed goes up, a connector corrodes, water gets into a feeder cable. The cheapest defence is to keep the signal numbers from every uplink and watch their trend.
Measure | What it tells you | Alert rule (starting point) |
Received signal strength, per node | Gradual path-loss change | A fall of 6 dB or more over a week, or margin to sensitivity under 10 dB |
Signal-to-noise ratio, per node | Interference or a failing receive path | A sustained drop, or a spreading factor that keeps stepping up |
Spreading factor chosen by the network | How hard the link is working | A node stuck at the highest factor |
Message loss per node | Collisions and weak links | Loss above a few percent over a day |
Gateway-wide noise floor | Nearby interferers or a gateway fault | A step change in the floor |
Time since last message | A dead node or a dead link | Missing two reporting periods |
Treat these as the link's equivalent of the battery trend in our post on solar sizing: a slow decline is an early warning that a person should look at an antenna, a cable or a shadow, long before the node goes silent. Store the numbers with the node's position and installation notes, so that a technician with a laptop can see what changed since the last field test.
Gateway Placement: A Decision, Not a Default
The gateway position decides how many nodes can be served and how many need a second gateway. Four questions settle most placements.
Question | Why it matters | Rule of thumb |
How high can the antenna go? | Height clears the Fresnel zone and shortens shadows | Mast or building top, clear of metal roofs; more height beats more gain |
Where are the nodes' clusters? | The gateway should sit near the middle of the heaviest cluster | Centre of mass of the node positions, adjusted for obstacles |
What power and backhaul are there? | A gateway without power or uplink is a mast | Mains with backup, or solar sized as in our earlier post; a cellular or wired backhaul with a fall-back |
Is there a second gateway for overlap? | A second gateway covers shadows and gives redundancy | Overlap the cells so that each node is heard by two gateways where possible |
A gateway needs mechanical and electrical care. Lightning is a hazard on a tall mast during the same weeks of the year when solar input is low, so use proper earthing and surge protection. The coaxial feeder is a loss and a failure point: keep it short, use good-quality cable and weatherproof every joint, because a water-logged connector costs more decibels than any amount of antenna gain recovers.
An Eighteen-Pond Farm Across Two Kilometres
A representative composite, genericised and not a named client, shows how the guide changes a plan. A farm with eighteen ponds and three sheds spread along a 2.2 kilometre strip of coastal land planned one gateway at a shed roof near the middle, with nodes on posts at each pond. The model, using an exponent of 3.5 and SF9, predicted comfortable coverage with margin to spare.
The field test told a different story. At the far ponds the received signal sat 12 to 15 dB below the model's prediction, and at three ponds behind a line of casuarina trees the signal faded by 20 dB at dusk. The fitted exponent for the strip came out near 4.0, not 3.5, because of the low antennas and the vegetation. The farm's response used the guide's order of operations. It raised the gateway antenna by 4 metres on a mast clear of the metal roof, which recovered about 8 dB on the far ponds by clearing more of the Fresnel zone. It moved the nodes behind the trees to 2.5 metre posts, bought another 5 dB, and kept a few on SF10 under adaptive data rate. It added a second small gateway at the north end, both for the leftover shadows and for redundancy. After the changes every node was heard with margin above 10 dB in the worst measurement, and the farm re-ran the field test after the first monsoon to confirm. The figures in this account are illustrative of the pattern. What generalises is the sequence: model, measure, discover the exponent was worse, spend the next decibels on height and placement before gain.
When LoRa Is the Wrong Link
Data-heavy applications. LoRa moves small payloads at low rates. Video, images and large firmware images belong on other links, or on a gateway that talks to the cloud over cellular while the nodes stay small.
Dense, chatty networks. Many nodes sending often at long spreading factors will collide. Budget airtime for the whole network, not one node.
Critical real-time control. A link with seconds of latency and a retry is the wrong tool for a safety interlock.
Sites where a licensed or wired link already exists. A cable in the ground and a switch may be cheaper and more robust than a mast.
Places with no unobstructed gateway position. If the only mast sits behind a hill, a repeater or a different technology may be cheaper.
Starting on Monday
Write down the regulatory limits for your band, power and equipment from the current notification, and confirm your modules comply.
Choose the radio and spreading-factor range, and note the datasheet sensitivity for each.
Run the calculator with your real antennas, cable and fade margin, and look at the sensitivity to the exponent.
Survey the site for obstacles: trees, roofs, tanks, ponds, and mark the Fresnel clearances at your longest links.
Pick a first gateway position with height, power and backhaul in mind.
Do the field test with a node at its real height, at several distances, times and weather.
Fit the exponent and the extra losses, and update the plan.
Add a second gateway or raise the first where the worst-case margin is under 10 dB.
Log signal strength and signal-to-noise from every node after deployment, and alarm on a falling trend.
Spending the Budget on Purpose
A LoRa link has a fixed number of decibels to spend, and the farm will spend them whether or not you plan it. Plan it: put height in first, antenna gain second, spreading factor third, and keep 10 to 15 dB in reserve for the days the pond and the weather disagree with the model. Then go and measure, because the exponent in your head is almost never the exponent on your farm.
EurthTech is an engineering-first product design and embedded systems company with a development center in Hyderabad. To discuss a project, write to connect@eurthtech.com or see https://www.eurthtech.com/case-studies.
Comments