Field note · Selection

The Twelve Numbers That Decide a Radar Project

A typical mmWave radar datasheet lists forty to sixty parameters across six pages. Most of them are either derived from a handful of primitives or exist to fill a comparison table. Twelve decide whether the project works.

Why the headline number misleads

Almost every datasheet puts "detection range" in the first table, in large type. It is the least reliable number on the page, for three reasons.

• The conditions are missing. Range depends on target size, aspect angle, weather, mounting height and the detection probability the vendor assumed. A single figure without those four is not a specification, it is a best case.
• The tests are idealised. Free space, no clutter, room temperature, target directly on boresight. Real installations have a fence, a wall, rain, and a target that is not cooperating.
• Definitions differ. One vendor means 50 % probability of detection. Another means 90 %. Same hardware, two very different numbers.

So instead of reading the range, read the four parameters that produce it. Then calculate the range yourself for your own conditions — that is what the range calculator is for.

Group one: the four that set your range

1. Transmit power (or EIRP)

Usually given in dBm, sometimes as EIRP which already includes antenna gain. Watch which one you are reading: 10 dBm conducted with a 12 dBi antenna is 22 dBm EIRP, a twelve-fold difference that people mix up constantly.

2. Antenna gain

In dBi. Higher gain means a narrower beam for the same aperture, so it is not free — you trade coverage angle for reach. A 20 dBi antenna on a presence sensor is a warning sign, not a feature.

3. Noise figure

In dB, typically 6–12 for integrated mmWave front ends. This is how much noise the receiver adds to the thermal floor. Every 3 dB here costs you about 19 % of range, and it is the parameter most often omitted entirely.

4. Noise bandwidth

This is where datasheets become actively misleading. Range scales with the fourth root of bandwidth, so a factor of 100 in bandwidth changes range by a factor of 3.16. Vendors rarely state which bandwidth they used.

For an FMCW radar, the honest figure is the bandwidth of one range bin after the FFT — roughly the ADC sample rate divided by the FFT length. A sensor sweeping 1 GHz with a 10 MSPS ADC and a 1024-point FFT has a noise bandwidth near 10 kHz, not 1 GHz. If a datasheet quotes range without saying which, assume the most favourable interpretation.

Group two: the three that set your angular performance

5. Aperture size

Physical antenna dimensions. Angular resolution is set by aperture in wavelengths, not by software. A 40 mm wide array at 60 GHz has about eight wavelengths of aperture; no algorithm turns that into a 1° beam. If the datasheet claims fine angular resolution on a small board, check the aperture before believing it.

6. Channel count

Number of transmit and receive chains, written as 2T4R or 3T4R. More channels mean more virtual array elements and better angular resolution — but also more power, more processing and more cost. For presence detection, 1T2R is often enough. For tracking multiple people in a room, it is not.

7. Beamwidth, in both planes

Datasheets often quote only azimuth. Elevation matters just as much: a sensor with 12° azimuth and 60° elevation detects the floor as readily as the intruder, and clutter processing has to deal with it. Always ask for both.

Group three: the five that decide whether it survives deployment

8. Update rate

In Hz. A 10 Hz sensor cannot track a running person reliably; a 20 Hz sensor may be enough for perimeter work. This number is rarely the constraint until it is.

9. Power consumption

Average and peak. Peak matters if you are on battery or PoE with a limited budget. Many datasheets quote only average, which hides a 3× peak during chirp transmission.

10. Operating temperature

Range quoted at 25 °C does not survive a sealed enclosure at 60 °C. Ask specifically what happens to the noise figure and the oscillator stability at the top of the range, not just whether the part "operates".

11. Ingress protection

IP rating. IP65 is a starting point for outdoor perimeter work. IP67 if it will be washed down or submerged. A radar module with no enclosure rating is a component, not a product.

12. Band approval for your market

The one item that can make everything else irrelevant. A module legal in Shenzhen can be non-compliant in Frankfurt. Check the band compliance lookup before you fall in love with the specification, and ask for the actual certificate, not a promise.

The reading order

Three questions that expose a weak datasheet

These work on any vendor, in any language. Ask them in writing and watch how long the answer takes.

1. "The detection range figure — what RCS, what weather, what probability of detection, and what mounting height does it assume?"
   A vendor who knows their product answers in one message. A vendor who does not will send you the datasheet again.
2. "The noise bandwidth — is that measured after the range FFT or at the ADC?"
   This separates engineers from salespeople faster than any other question.
3. "Can you send the FCC or CE certificate for this exact variant, and which bands does it cover?"
   The exact variant matters. A certificate for the 60 GHz version does not cover the 24 GHz one, even if the enclosure is identical.

What to do with this

Run the numbers yourself before you trust a number someone else ran. The range calculator takes fifteen seconds and tells you whether the claimed range is even physically plausible for the stated power and antenna. If it is not, you have saved yourself a project.

And if you want a second opinion on a specific part, send the datasheet — that is what the form below is for.