Published Nov 12, 2026 · Last updated Nov 12, 2026 · 4 min · IndieRF
IP3 Calculation: HMC8413 IM3 Meets the Floor at −27 dBm
Two tones at −27 dBm per tone (2.0 µW) at the HMC8413's input put its IM3 products at its own noise floor in 1 MHz, −112 dBm at the input. Above that drive, IM3 climbs 3 dB per dB.
At that drive the products leave the LNA 85 dB under the tones and level with its output noise, so −27 dBm per tone is the most this LNA takes before IM3 shows in a 1 MHz receiver. The SFDR row reads 85.1 dB at any source power, because IIP3 and the floor set it.
What the data sheet prints
Analog Devices' HMC8413 data sheet, Rev. 0, page 3, Table 1 gives the 0.01 GHz to 7 GHz range at 5 V and 95 mA: gain 19.5 dB, noise figure 1.9 dB, OIP3 35 dBm and OP1dB 21.5 dBm, all typical. OIP3 is an output number, so subtract the gain to move it to the input:
Each third-order product, at and , rises three times as fast as the tones that make it. Referred to the input, with per tone,
The input noise floor in 1 MHz is dBm. Set the product equal to it and solve: dBm. The gap from that level to the floor, 85.05 dB, is the spur-free dynamic range, printed as 85.1 in Cascade's SFDR row.
ADI measured the 35 dBm with 5 dBm out per tone, which is −14.5 dBm in. At that drive the products sit at dBm, 60 dB under each tone, and Cascade reads −55 dBm on the same block. That is the measurement behind the typical number, with each tone 16.5 dB under OP1dB.
Three dB of IM3 per dB of drive
At −27 dBm in, each tone leaves at −7.5 dBm and the products leave at −92.5 dBm, 85 dB under the tones and level with the output noise. Push both tones 10 dB harder, to −17 dBm, and the products rise 30 dB while the tones rise 10, so the gap closes by 20 dB. The products now sit 30 dB over the noise floor. A weak signal that lands on needs about 30 dB more level to keep the SNR it had, and in a receiver that is 30 dB of sensitivity lost on that frequency.
The 3:1 slope holds while the tones are well under compression. Cascade's IM3 row is one scalar for the pair, with no frequency of its own, so it assumes the products fall in your channel. Whether they do depends on the tone spacing: two tones 1 MHz apart put their products 1 MHz either side of the pair.
Both printed ranges
The data sheet prints two frequency ranges, each with its own typical column at the same bias. Every row below is the Cascade engine on the HMC8413 block with two −27 dBm tones and a 290 K source, checked against the formulas above. IM3 out is each product at the LNA output, floor in is the noise floor in 1 MHz referred to the input, and tone at floor is the per-tone input where IM3 meets it.
| Range | Gain (dB) | NF (dB) | OIP3 (dBm) | IIP3 (dBm) | IM3 out (dBm) | Floor in (dBm) | Tone at floor (dBm) |
|---|---|---|---|---|---|---|---|
| 0.01 GHz to 7 GHz | 19.5 | 1.9 | 35 | 15.5 | −92.5 | −112.08 | −27.03 |
| 7 GHz to 9 GHz | 19 | 2.8 | 33 | 14 | −90.0 | −111.18 | −27.73 |
Table 2, for 7 GHz to 9 GHz, loses 1.5 dB of IIP3 and gains 0.9 dB of noise figure. The floor rises with the NF and lifts the crossing back up, so the tone where IM3 meets the floor drops by only 0.7 dB. In a 1 MHz receiver the two ranges behave the same within a dB.
Typical, at one bias
Every number here is the typical column at one bias, 5 V and 95 mA, set by the 787 Ω RBIAS resistor. The data sheet prints a minimum for gain (17.5 dB) and OP1dB (19 dBm), and only a typical for OIP3.
The IP3 problem that cost me the most time was a cascode device that was very sensitive to its second-gate bias, VG2. Process variation meant one VG2 setting did not give the same IP3 from part to part.
The next stage behind the LNA counts at the input with its own IIP3 minus 19.5 dB. Duplicate the first case above, add the mixer or second amplifier you plan to use after the HMC8413, and Cascade's Cumulative IIP3 row at that stage is the receiver's number.
References
- Analog Devices, HMC8413 Low Noise Amplifier, 0.01 GHz to 9 GHz, data sheet Rev. 0, 10/2021, page 3, Tables 1 and 2.
- D. M. Pozar, Microwave Engineering, 4th ed., section 10.3: nonlinear distortion, intercept points and dynamic range.
- B. Razavi, RF Microelectronics, 2nd ed., section 2.2: the two-tone test and the input intercept.
Related
Questions? Contact
Discussion
No comments yet. Start the thread with a measurement, a correction, or a worked example.