Mixer LO to IF Isolation vs. 1*LO - 0RF Spur

Question

I am working on a simple digital radio receiver and am confused by the difference between the LO to IF leakage vs. the 1LO - 0RF spur. These two tables in the datasheet of the Analog Devices HMC218B passive mixer are the source of my confusion.

The "Spurious Outputs" table lists the 1LO-0RF as being 3.4 dBc. Working from the test conditions at the bottom of the table, this means that the mixer would have an output spur at a frequency of 5.25 GHz 3.4 dBc below the power of the desired IF output, in this case 100 MHz. If the conversion loss of the mixer is 7 db (as is specified in the 2nd "Electrical Specifications" table), then for a RF power of -10 dBm the IF power at 100 MHz should be -17 dBm, so the power of the 1LO-0RF spur at 5.25 GHz should be -20.4 dBm.

If the LO to IF isolation given in the "Electrical Specifications" table is used to find the power of the LO signal at the IF port, one finds that the power of this unwanted LO at the IF port is -19 dBm. This calculation uses the specified LO to IF isolation from 4-6 GHz to be consistent with the frequencies used in the "Spurious Outputs" table.

The 1.4 dBm difference in 5.25GHz power at the IF port given by the "Spurious Outputs" table and the "Electrical Specifications" table is minor, but in my case the input RF power is around -50 dBm (@ 5.8GHz), not the -10 dBm that is used in these tables. This decrease in RF power significantly changes the calculated LO power at IF depending on which table is used to find the LO at IF power. For example, if the "Spurious Outputs" table is used in conjunction with the conversion loss from 4-6 GHz in the "Electrical Characteristics table" I find that the 1LO-0RF spur should be -60.4 dBm, 3.4 dBc below the -57 dBm IF output at the desired IF. However, if I use an LO power of 13 dBm (the LO power used throughout the datasheet), and I calculate the LO power at IF using the LO to IF isolation given in the "Electrical Specifications" Table, I find that the LO power at IF would be -19 dBm. Which method should I used to find the LO power at IF? Is there a fundamental difference between LO to IF isolation and the 1LO-0RF spur? Perhaps my understanding of dBc is wrong and is throwing off my calculations.

Thanks for your help.

Answer 1

The first table is only usable at an RF power of -10dBm, and only at an LO frequency of 5.25 GHz.

You would expect the power of a 0RF spur to stay constant, regardless of the input RF power, so it's not correct to try to use the first table to take 3.4dB with respect to a different RF input power.

Given that the LO to IF isolation is given as typical 32dB, minimum 15dB, quite a range, I'd say 1.4dB between the two tables is as near as you can expect to spot-on. Note that the 3.4dB is quoted at 5.25 GHz LO, the 32dB isolation is typical over the range 4.5 GHz to 6 GHz.

I see. Even though the values given in the table aren't valid for different RF powers, is it acceptable to use them as general guidelines while frequency planning? For example, some of the spurs are as weak as 90 dBc for the given test conditions. Is it ok to ignore those spurs even though the spur table will change with varying RF power.

The table 1 guidelines are to allow you to quickly eliminate mixers that are 'obviously' wrong for your application, they are not a substitute for proper receiver design. Once you have a shortlist of those that may do, you must then get either more comprehensive figures from the manufacturer, a physical model and test it at your power levels, or ideally do both.

The purpose of table 1 is to give you guidelines as to the levels of spurs to use when doing frequency planning. However, as it's given at one power level only, you will have to use your skill and judgement to estimate what it will be at other power levels.

In general, the nRF product will change at ndB per dB of signal level change, for typical diode mixers, when the signal level is small. When I've used RF mixers, they often have a table 1 taken at two different power levels, which allows you to determine whether signals are still 'small' at that level. For a badly behaved mixer, the spurs may become better or worse as you get to the 'big signal' regime, so extrapolation down to your low signal level is very uncertain.

Whether -90dBc is 'OK to ignore' is down to your application. If you're building an amateur data receiver, almost certainly. If you're building a commercial synthesiser, probably not. If you're building a commercial data receiver, then it's not going to affect your in-band demodulator performance, but what are your blocking specifications? You mention signal levels of -50dBm, which in my experience is low, even for a spectrum analyser, so you probably are pushing the spurious performance.