In Amplifiers, Mixers, Transmit/Receive/Synthesizer Architectures

Modern battlefields are awash in electronic signals from radars, jammers, and radio communications. Therefore, high linearity EW receivers for applications such as Radar Warning Receivers, Jammers, and Electronic Countermeasures are one of the most important capabilities for the modern warfighter. Any electronic warfare system, regardless of its ultimate goal, must first detect the electromagnetic signals around it using a receiver. One of the most common subcomponents in an electronic warfare receiver is a microwave tuner, which scans a given bandwidth (frequently from 2-18 GHz) and converts these signals to a baseband receiver.

Theoretically this could easily be accomplished with a direct downconversion using a mixer such as the M1-0220, ML1-0220, or MM1-0224 and either a high or lowside LO:

The major issue with this scheme is that unwanted signals will create significant spurs at the IF unless a strong (and tunable) channel selection filter is used. For example, using the Marki Spur Calculator we can see that if the user tunes to detect 5.5-6.5 GHz using a 7 GHz LO, the following spurs will appear in band:

Wow that’s a lot of spurs! With the image plus almost every spur you’ve ever heard of, this is clearly a problematic frequency plan. From here you have three main options to improve the situation.

  1. You can use a tunable filter or a switched filter bank at the front end before the mixer (good luck finding that tunable filter!)
  2. You can break the band into multiple channels using fixed filters, and downconvert them all at once using multiple downconversion channels,
  3. You can upconvert to a higher frequency, filter there, and then downconvert to the IF. This heterodyne tuner architecture using multiple conversions is extremely common.

Depending on system requirements (and in particular whether the entire spectrum needs to be visible at once, or whether the receiver is trying to focus on a single threat at a time) all of these techniques are used. In this tech note we will focus on the double conversion architecture from option 3, as shown below:

At first this architecture may seem overly complex, especially as it uses higher frequency components that are more difficult to realize than in a direct conversion scheme. However, this architecture dramatically reduces the in-band spurs. If we imagine the same scenario where the user wants to identify a 6 GHz threat, the tuner would use a 28 GHz LO and see the following spurs in band:

As you can see there is a drastic reduction in both the number and intensity of the spurs that are present in the converted band (in particular the elimination of the image frequency). In the first scheme the isolation from RF to IF is a concern at low frequencies, and isolation from the LO back to the RF path is a concern at all frequencies. In this double conversion scheme none of the isolations are critical (other than that they determine spur suppression). A much more detailed spur analysis is given below, but the spur problem is greatly reduced compared to the direct conversion scenario.

Component Requirements

The change from a direct conversion receiver to a high frequency double conversion architecture dramatically reduces the upfront filtering requirements, but requires a mixer with an ultra-broadband IF port and LO frequency coverage to much higher frequencies. It also requires a much higher LO to be generated and delivered with sufficient power to drive the mixer.

Working from right to left, the second mixer in the architecture has an easy enough job, but it should still have sufficient dynamic range to prevent distortions in the downconversion. Marki recommends the MM1-0626S for this slot. While the bandpass filter does not have stringent requirements, it is narrower than the filters that Marki typically offers. Numerous filter specialists can provide this filter without too much difficulty. The first section is the tricky one, with the high LO frequency and broad IF.

Mixer Selection

The first mixer slot has been a source of continuous innovation for Marki Microwave for over 20 years. For many of those years the M9-0942 in the ES package was the only mixer with the IF bandwidth to accommodate this frequency plan in a planar structure. Later the Microlithic line of mixers was developed with several models targeted towards this frequency plan. Currently Marki recommends the MM1 series of MMIC mixers for new designs. As of 2019 the following models are available:

Model RF/LO Low RF/LO


IF Low IF High LO Power Low LO Power High Conversion Loss [dB] Isolations L-R [dB] IIP3 [dBm] Surface Mount Bare Die Module RoHS
MM1-1467L 14 67 DC 21 +9 +13 7 53 +12 N Y Y Y
MM1-1467H 14 67 DC 21 +11 +20 7 53 +17.5 N Y Y Y
MM1-1850H 18 50 DC 20 +12 +20 8 43 +21 Y Y Y Y
MM1-1850S 18 50 DC 20 +17 +23 8.5 43 +25 Y Y Y Y
MM1-1857L 18 57 DC 21 +7 +15 8 35 +13 N Y Y Y
MM1-1857H 18 57 DC 21 +13 +21 7.5 45 +20 N Y Y Y
T3-1040 10 40 1 18 +13 +25 8 28 +25 N N Y N

Of the above, the MM1 mixers are all available as bare die or modules (and some as surface mount), while the T3-1040 is only available as a module. Also, the MM1 mixers are all RoHS compliant, while the T3 is built with lead. Based on the table above, the MM1-1850S has the best IP3, the MM1-1467 has the best conversion loss and isolations, and the MM1-1857L has the lowest LO power requirement. However, these numbers are all based on full band performance, and not performance specifically as a tuner mixer. This is very important, since things like the RF balun loss are included in the above data, but don’t matter for the tuner application. Below we perform a much more rigorous analysis of these potential candidates for the mixer 1 slot.

LO Amplifier Selection

Before we begin the analysis of mixer options, we need to discuss the importance and strategy to generate a proper LO drive signal. Recall the first mixer commandment: Thou shalt not starve the mixer of LO drive. Traditionally generating sufficient drive in the mm-wave frequencies has been difficult. In order of importance, the LO drive amplifier should have the following qualities:

  • Sufficient power to drive the mixer across the band without suckouts (across temperature and packaging losses, plus LO attenuation if required)
  • Sufficient gain to reach required power levels from LO signal source power (across temperature)
  • Sufficient output return loss to absorb return signals from the LO port of the mixer*
  • Sufficient reverse isolation to prevent return signals from the LO port of the mixer from contaminating preceding LO generation stages
  • Low enough DC power consumption and heat dissipation to meet system requirements

*Recall that the LO port of a mixer not only reflects the input signal, but acts as a source of many different harmonic signals

Note that harmonic generation is critically important in lower frequency LO amplifier applications (such as for the T3 high linearity mixer), but since the harmonics will be so high in frequency they are not a critical factor for this application.

It is with these targets in mind that Marki developed the AMM-6702 LO driver amplifier. With ~20 dBm of saturated output power (across temperature) and over 20 dB of gain (across temperature) the AMM-6702 is capable of driving an H or S diode mixer from a 0 dBm input source from -40C to above 85C. It can be biased to consume less than a watt of power even in heavy saturation. For this reason the AMM-6702 is recommended for use with all mm-wave Marki mixers.

An additional benefit of the AMM-6702 is that it operates above the traditional 45 GHz cutoff of GaAs pHEMT amplifiers providing enough power to drive an H diode mixer as high as 55 GHz. This adds an extra degree of freedom for the microwave tuner designer. Going back to our original frequency plan, if we shift the intermediate stage frequency from 22 GHz to 29 GHz the spurs become much simpler. Take for example the 6 GHz tuned frequency, which we now lock in with a 35 GHz LO instead of 28 GHz:

The number of crossing spurs is reduced from 10 to 7 using the higher frequency LO.

mm-wave LO Generation

Generating a low phase noise LO in the 20-50 GHz range can be significantly more challenging than developing an LO in the 2-18 GHz range. Since the LO range spans nearly an octave, high harmonic suppression is required. This means that multiplying a low frequency source with an NLTL is not a good option.  Most likely multiple stages of doublers with filtering in between is required. For the final stage doubler Marki offers the MMD-1648 (with outstanding isolation) and the extremely efficient MMD-2060. Marki offers several MMIC multiplier chain clean-up filters, including the MFB-3300, the MFB-3450, and the MFB-3475. These extremely compact filters can be inserted directly into the LO chain to improve subharmonic isolations. If additional multiplication is required, Marki recommends the MMD-1030.

Packaging and Layout Considerations

For the uninitiated, mm-wave packaging can be daunting. Although research in surface mount designs is rampant, the typical design still uses bare die components. Of particular importance is the radiation effects of higher frequencies degrading the isolation of mixers and doublers and causing potential instabilities in amplifiers. New users of these technologies are encouraged to contact [email protected] for guidance on how to properly design and install packaging for these chips.

Detailed Mixer Comparison and Analysis

The first consideration for a mixer is the conversion loss, which also determines the noise figure of the converter (assuming an insignificant noise contribution from the LO). Below is the conversion loss of each of the candidate mixers under the following conditions:

Input Signal: 10 MHz to 20 GHz at 0 dBm

Output Signal: 22 GHz

LO Drive: 22.010 GHz to 42 GHz generated by the AMM-6702UC with 5 dBm input power (and an additional 6 dB attenuator between the amplifier and mixer on the MM1-1467LUB, MM1-1857LS, and MM1-1850HS)

The lowest conversion loss is seen with the MM1-1467HUB in the B configuration, but all of the MM1 options have similar conversion losses between 6 and 8 dB at low frequencies and between 8 and 10 dB at the high end. The T3-1040LN has a higher conversion loss, especially at very low and high frequencies.

IP3 Analysis

The dynamic range of a system is limited by noise (for low power signals) and distortion (for high power signals). The first component of distortion to be considered is the main signal compression as expressed by the P1dB. For an ultra-broadband system like the wideband tuner, however, the input signal is typically limited to a power level much lower than the P1dB. Spurious products (single and multitone) cause distortions that limit system performance much lower than the main signal compression. The first of these we’ll consider is multitone intermodulation distortion, as expressed by the IP3. Here is a comparison of the IP3 of the different mixer options under our standard frequency plan, but limited to the 2-18 GHz band (due to the band rolloff of the T3).

Conversion Loss (dB)
Mixer Configuration Median Worst Case Median Worst Case Median Worst Case
MM1-1467HUB A -6.7 -7.9 26 20 19 14
MM1-1467HUB B -6.3 -7.0 27 20 21 15
MM1-1467LUB A -7.8 -9.3 15 10 8 1
MM1-1467LUB B -7.2 -7.7 14 11 7 4
MM1-1850HS A -7.4 -9.2 20 16 13 8
MM1-1850HS B -7.3 -8.7 20 16 14 8
MM1-1850SS A -7.0 -9.3 28 23 21 15
MM1-1850SS B -7.3 -8.6 28 24 21 16
MM1-1857HS A -7.8 -10.6 21 19 14 9
MM1-1857HS B -7.6 -10.8 23 18 16 10
MM1-1857LS A -9.7 -12.6 13 10 4 2
MM1-1857LS B -9.2 -12.4 13 11 4 0
T3-1040LN A -9.8 -12.5 24 21 15 8
T3-1040LN B -9.7 -11.9 25 20 15 12


The MM1-1850SS in Config B shows the best typical and worst case IP3 on both an input and output referred basis. This is not surprising since it is the only S diode tuner mixer available. Surprisingly, however, the MM1-1467H is close behind in both Config A and Config B (within 1 dB). The T3, which has by far the best IP3 of any mixer topology at low frequencies, cannot provide higher IP3 at these frequencies due to the lack of higher frequency harmonics (therefore the switching time of the T3 is the same as the MM1).  The L diode mixers (MM1-1467L and MM1-1857L perform a little worse than expected (10 dB lower rather than the expected 6 dB), but this is most likely related to specifics of the LO drive conditions. Here is the IP3 performance across frequency of the MM1-1467 (L and H) and MM1-1850S in Config A and B:

The IP3 performance varies significantly across frequency. The exact IP3 performance of the converter will depend on the specifics of LO drive, packaging, and port matching. The MM1-1850SS is slightly better in Config B than the other similar mixer, but all of them show significant improvement over the L diode mixers. A worst case OIP3 of 16 dBm (for the MM1-1850SS) indicates a worst case dynamic range of 48 dBc at a 0 dBm input (or 68 dBc at a -10 dBm input).




Single Tone Spurious Analysis

Previously we showed the spur web for a given conversion. A more detailed analysis shows that the following single tone spurs will be present with a complete 2-18 GHz input bandwidth:

IF Order LO Order IF Start Frequency (GHz) IF Stop Frequency (GHz) LO Start Frequency (GHz) LO Stop Frequency (GHz)
2 0 10.75 11.25 23.5 40.5
3 0 7.16 7.5 23.5 40.5
-2 1 2 9.25 26 40.5
-3 1 2 6.17 28 40.5
-4 1 2 4.625 30 40.5
-2 2 12.5 18 23.5 29
-3 2 8.3 18 23.5 38
-4 2 6.25 14.75 23.5 40.5
-3 3 16.17 18 23.5 25.3
-4 3 12.125 18 23.5 31.33
-4 4 18 18 23.5 23.5
-5 4 14.4 18 23.5 28
3 -1 15.17 18 23.5 32
4 -1 11.375 15.625 23.5 40.5
4 -2 17.25 18 23.5 25


All spurs were measured at all frequencies for all mixers under study by creating a large spur study template on our PNA network analyzer. Note that special care is required to ensure that the harmonics seen are generated by the mixer, and not converted harmonics of the network analyzer source. This can be accomplished either by filtering the source or by limiting the output power while making the receiver as sensitive as possible by reducing the IF bandwidth (at the expense of sweep time).

Here is the summary table of spur performance, showing the typical (median) and worst case for each spur for each mixer in each configuration with an input power of -10 dBm. The best performance for each category is highlighted in green, and the overall worst typical spur and worst overall spur is highlighted in the last column:

Overall the MM1-1660HUB in either configuration shows the best typical performance, while the MM1-1850SS in the A configuration shows the best worst case performance. With a 0 dBm input, the MM1-1467H shows better performance than the MM1-1850S, since the 3LO – 3IF limits the MM1-1850S. Here is the 2LO -2IF spur plot, which is the worst case spur for both mixers:

Again the exact values of the spurs depends on the specifics of LO drive, port termination, frequency plan, and packaging. The MM1-1467LUB performs surprisingly well due to the fact that multitone spurs can be cancelled with well balanced baluns.


A detailed performance comparison was performed on all of the tuner mixer configurations recommended by Marki Microwave as of May of 2019. Of the candidates, the MM1-1467H and MM1-1850S offered the highest dynamic range, limited by IP3 and spurious products. While exact details of LO drive, port termination, frequency plan, and packaging will determine ultimate system performance, these two mixers are a good starting point for any tuner application.


Recommended Posts

Leave a Comment