Published Nov 20, 2026 · Last updated Nov 20, 2026 · 6 min · IndieRF

Mixer Spur Chart: Half-IF at −55 dBc, Image at −28 dBc

HMC213B, 500 MHz IF: the preselector misses the half-IF.

HMC213B at 3.5 GHz, 500 MHz IF, Johanson 3350BP39A0500001E in front: image −28 dBc, half-IF −55 dBc. The half-IF at 3250 MHz sits in the 3100 to 3600 MHz passband for any low-side IF up to 800 MHz.

Where the half-IF lands

An input at f converts into the IF when m·f − n·fLO comes out at ±fIF, so the spurs sit at f = (n·fLO ± fIF)/m. The image is the other 1x1: fLO − fIF, 3000 − 500 = 2500 MHz. The half-IF is the 2x2 on the wanted signal's side of the LO:

2f−2fLO=fIF⇒f=fLO+fIF/2=3250 MHz2f - 2f_{LO} = f_{IF} \Rightarrow f = f_{LO} + f_{IF}/2 = 3250\ \text{MHz}

That is the channel minus half the IF. A filter sized for the band has to fall from passband to stopband in IF/2, 250 MHz here, and the 3350BP's passband alone is 500 MHz wide.

The level comes from the HMC213BMS8E data sheet, Rev. 0, page 18, Downconverter M × N Spurious Outputs, measured at 3.5 GHz and −10 dBm RF with a 3.6 GHz LO at 13 dBm. The 2x2 entry is 55, printed as dB below the IF output; Cascade and the spur chart take levels with their sign, so it goes in as −55 dBc. The 0x1 entry, LO leakage, is printed −6: 6 dB above the IF.

What the spur table in a sheet leaves out

A spur table copied into a spreadsheet gives each row at the table's drive with no filter in front: half-IF −55 dBc, image 0 dBc. Cascade takes the same table and reads two more things from the lineup.

The first is the preselector at each spur's own input frequency. Johanson prints guaranteed minimum attenuation at a few points on page 1 of the 3350BP39A0500001E sheet, Revision 3.0: 35 dB over 500 to 2000 MHz, 30 dB at 2500 MHz, 5 dB at 2950 MHz, and 2.0 dB maximum loss over the passband. Cascade draws the curve through those points and reads it at every spur. For the image that is the printed 30 dB less the 2.0 dB the wanted signal also loses at 3500 MHz, 28 dB, so the image goes from 0 to −28 dBc. For the half-IF it is 2.0 dB less 2.0 dB, nothing.

The second is the drive. The table is measured at −10 dBm; here the source is −47.5 dBm, the filter takes 2.0 dB and the HMC8413 adds 19.5 dB, so the mixer sees −30 dBm. A product of RF order m moves m − 1 dB for each dB of RF drive against the wanted IF, so the 2x2 output product at |2 × 3500 − 2 × 3000| = 1000 MHz reads −55 dBc on the sheet and −75 dBc in the receiver. LO leakage goes the other way: the 0x1 is +6 dBc in the table and +26 dBc at −30 dBm, since the leak stays put while the IF falls 20 dB. Both are far from the IF and an IF filter takes them; the rule holds below compression and stops near the HMC213B's 11 dBm input P1dB.

Input spurs stay at the table's own drive in the planner, since it cannot know an interferer's level. A half-IF interferer weaker than −10 dBm lands lower by the same m − 1 dB per dB.

Picking the IF against this filter

Every number below is Cascade's spur planner on the linked lineup with the LO 3500 MHz minus the IF, checked against the equations and the printed points worked by hand. Each spur frequency lands on a printed point or inside a printed band. The spreadsheet column would read 0 dBc for every image and −55 dBc for every half-IF.

IF (MHz)LO (MHz)Image (MHz)Image after filter (dBc)Half-IF (MHz)Half-IF after filter (dBc)
1003400330003450−55.0
27532252950−3.03362.5−55.0
50030002500−28.03250−55.0
75027502000−33.03125−55.0
110024001300−33.02950−58.0

The 100 MHz IF is the one the data sheet measures at, and it puts the image inside the passband too: both spurs reach the mixer at full table level, and only a narrower filter or an image-reject mixer helps. The image also carries the LNA's noise; the image noise note works that side, and 30 dB at the image is enough there.

The half-IF clears the 3100 MHz passband edge once the IF passes 2 × (3500 − 3100) = 800 MHz. At 1100 MHz it lands on the printed 5 dB point at 2950 MHz and gains 3 dB, which is all the sheet promises that close to the band.

I'd build the 750 MHz IF on this filter. The image sits on the 35 dB floor and comes out at −33 dBc, the half-IF is −55 dBc at any IF below 800 MHz anyway, and the HMC213B's IF port runs to 1.5 GHz. Getting the half-IF lower takes a second filter with a steeper low skirt; pushing the IF past 800 MHz buys only the 3 dB the 2950 MHz point guarantees.

Limits of these numbers

The HMC213B table is typical and printed for one plan: RF 3.5 GHz, LO 3.6 GHz on the high side, 100 MHz IF. Cascade applies the same entries at every LO in the table above; the real levels move with LO frequency, LO drive and injection side. The filter numbers are guaranteed minimums, and the typical curve is better. Between printed points Cascade's curve is a straight line, which moves no number above by more than 0.03 dB across the 1 MHz channel. The curve runs from 500 MHz to 10 GHz, the range the sheet prints, and its 8 to 10 GHz segment is the 20 dB the sheet marks "Ref."; planner rows outside that range read "curve ends".

References

  • Analog Devices, HMC213BMS8E, 1.5 GHz to 4.5 GHz, GaAs, MMIC, Double Balanced Mixer, data sheet Rev. 0, 2/2018, page 3, Table 1, and page 18, Downconverter M × N Spurious Outputs.
  • Analog Devices, HMC8413 Low Noise Amplifier, 0.01 GHz to 9 GHz, data sheet Rev. 0, 10/2021, page 3, Table 1.
  • Johanson Technology, 3350BP39A0500001E 3.35 GHz Band Pass Filter, Revision 3.0, page 1, General Specifications (attenuation minimums from 500 MHz to 8 GHz; 8 to 10 GHz printed as 20 dB "Ref.").

Questions? Contact

Discussion

No comments yet. Start the thread with a measurement, a correction, or a worked example.