SLIM-SFA-1013 rev 0
SAW Filter / Amplifier for 1013 MHz

Released 12-11-11.
Original release.
Updated 12-13-11.  Note regarding IMD
Updated 12-17-11.  Add section - Test, Evaluation, and Characterization
Updated 2-18-12.  Add sections h. and i.
Updated 5-2-12.  Add link for downloading schematic as Express file.
Updated 5-16-15.  Change capacitor C5. Parts list now Rev A.

SLIM-SFA-1013, SAW Filter / Amplifier, size-A
All needed documentation can be found within this web page.
a.  SKSLIM-SFA-1013 rev 0 Schematic.
b.  PWB-SFA rev 0, PWB artwork, in ExpressPCB software. Use for ordering from Express.
c.  Component Layout. Use for locating parts on board.
d.  PLSLIM-SFA-1013 rev A, Parts List.
e.  Construction
f.  Test Results
g.  Test, Evaluation, and Characterization
h.  Spurious Responses
i.  Conclusions

    The SLIM-SFA-1013 is a combination of three cascaded SAW Filters and an amplifier. The purpose of this module is to replace the Coaxial Cavity Filter used in the MSA. There are two expressly separate circuit sections on a single pwb. It is the intention to cut the pwb before assembly. Although both sections are intended to operate together, it is important to separate the sections so that full perimeter fencing can be installed on each section. This is very important to minimize section cross-talk and maximize ultimate filter response rejection.
    The basis of this design was to band pass the center frequency of 1013.3 MHz, which is the nominal First I.F. of the MSA. A more important design feature is to maximally attenuate frequencies near 1034.7 MHz, which is the image frequency of the MSA input. Test results indicate attenuation at this image frequency is about -91 dBc. This is not a good as the Cavity Filter (better than -100 dBc) but it is very good compared to the minimum attenuation required for Spectrum Analyzer operation, which is -70 dBc.
    The total gain for the design is approximately 0 dB. The combined loss of the attenuators and the filters is compensated by the amplifier's gain of approximately 18 dB. The maximum input power of any of the SAW filters is +10 dBm. Therefore, the maximum input to the Filter/Amplifier section is +13 dBm.
The maximum input to the dual filter section is +10 dBm. When the dual cascaded SAW Filter section follows the Filter/Amplifier, the maximum input to the Filter/Amplifier section must be kept below -2 dBm.
    When substituting this module for the MSA's Cavity Filter, connect Mixer 1 output to the
Filter/Amplifier section, followed by the Dual Filter section, followed by the input to Mixer 2.
    Although this web page is specific to the SLIM-SFA-1013, other filter frequencies could be used for other projects. The vendor specifies several SAW Filters with the identical footprint. They can be found at

12-13-11 Note: There is one area of concern that I have overlooked and should mention. IMD, Inter modulation Distortion.  Because the Bandwidth of this scheme is about 18 MHz it is possible that two incoming signals separated by 10.7 MHz will enter Mixer 2. The product of the two signals, 10.7 MHz will not be displayed as a spur. It will manifest itself as a noise floor increase. The noise floor might even "modulate" at a rate of the same modulation of either of the input signals. This interference will be most pronounced when using a very wide bandwidth Resolution Filter (the two signals don't have to be exactly 10.7 MHz apart). More information is presented later on this page. This interference is highly unlikely when using the Coaxial Cavity Filter, since its bandwidth is much narrower, about 2 MHz.

a. SKSLIM-SFA-1013 Schematic. Click to download Express file.
    The ERA-33SM (U1) can be replaced by a variety of MMIC amplifiers.  Also, the 10 volt supply voltage can be reduced to as little as 5 volts, along with R1 and R2 value reductions. This would decrease the amount of heat generated by R1 and R2. This would allow faster temperature stabilization of the SAW/Amp section.

b.  PWB-SFA, PWB Artwork. Click to download Express file. Use for ordering from ExpressPCB.

c. Component Layout, Top (Component) View
    This layout can be used for locating or placement of components. Connectors J1-J4 and P1 are mounted on the bottom side (ground plane) of the PWB. Size is 1.2 inch x 1.2 inch.

d.  PLSLIM-SFA-1013, Parts List Master, future changes will be made on this web page only.
Rev A, 5/16/2015
Des    Value            Mfg. Part Number          Digikey Number        Cost                      
C1    .01 uF            C0805C103K5RACTU    399-1158-1-ND        0.04   
C2    .1 uF              C0805C104K3RACTU    399-1168-1-ND        0.05   
C3    .1 uF              C0805C104K3RACTU    399-1168-1-ND        0.05   
C4    .1 uF              C0805C104K3RACTU    399-1168-1-ND        0.05   
C5    10uf/35v ceramic    GMK316F106ZL-T    587-1352-1-ND     0.35   Rev A, deleted
C5    10uf/16v     AVX, TPSB106K016R0500   478-5230-1-ND     0.65   Rev A, better esr cap
C6    100 pF            C0805C101K5GACTU    399-1121-1-ND       0.05   
C7    .01 uF            C0805C103K5RACTU    399-1158-1-ND        0.04   
C8    100 pF            C0805C101K5GACTU    399-1121-1-ND       0.05   
C9    100 pF            C0805C101K5GACTU    399-1121-1-ND       0.05   
FB1    Ferite Bead    BLM21PG221SN1D    490-1054-1-ND     0.06      
J1    SMA Female    Molex, 538-73391-0070    WM5544-ND    3.78   
J2    SMA Female    Molex, 538-73391-0070    WM5544-ND    3.78   
J3    SMA Female    Molex, 538-73391-0070    WM5544-ND    3.78   
J4    SMA Female    Molex, 538-73391-0070    WM5544-ND    3.78   
L1    68 nH            LQW18AN68NJ00D    490-1178-1-ND       0.39   
FL1    1013/15       SAW Filter, PN MA08712, Model TA0477A
FL2    1013/15       SAW Filter, PN MA08712, Model TA0477A
FL3    1013/15       SAW Filter, PN MA08712, Model TA0477A
P1    Conn, 2 pin     S1012-36-ND (row of 36 pins)    S1012-36-ND    0.06   
R1    75 1/4W        RHM75.0F             RHM75.0FCT-ND        0.08   
R2    75 1/4W        RHM75.0F             RHM75.0FCT-ND        0.08   
R3    294             MCR10EZHF2940    RHM294CRCT-ND       0.04   
R4    294             MCR10EZHF2940    RHM294CRCT-ND       0.04   
R5    17.8            MCR10EZHF17R8    RHM17.8CRCT-ND     0.04   
R8    0                 MCR10EZHJ000       RHM0.0ACT-ND         0.04   
R9    294              MCR10EZHF2940    RHM294CRCT-ND      0.04   
R10    294            MCR10EZHF2940    RHM294CRCT-ND      0.04   
R11    17.8           MCR10EZHF17R8    RHM17.8CRCT-ND     0.04   
R12    294            MCR10EZHF2940    RHM294CRCT-ND       0.04   
R13    294            MCR10EZHF2940    RHM294CRCT-ND       0.04   
R14    17.8           MCR10EZHF17R8    RHM17.8CRCT-ND     0.04   
U1    ERA-33SM    Minicircuits ERA-33SM                               1.72   
PWB    Circuit Board    PWB-SFA                                            3.28   
Note: The cost of the SAW filters is not included.

e.  Construction
Cut the PWB-SFA into two sections with a hacksaw blade that has a cut less than .075 inch.
Sand the sides of each section to the perimeter ground planes. This will allow easy installation of the perimeter fencing.
See the Construction Tips web page for good advice on the following procedures.
Install top components using your favorite soldering method. I still use the old fashioned low wattage soldering iron and a hot plate from a coffee brewer.
Clean the board.
Clean it again, because I know you could do better!!
Modify the SMA connectors for a flush fit to the pwb.
Mount P1 and J1-J4 from the bottom side of the boards.
Clean the boards again.
Test. This is a "did I short anything out" type test. The test results will not be anywhere close to "good".
Install the three internal fences.
Clean the boards again.
Install the perimeter fencing.
Clean the boards again.
Test before installing the top covers. The results will be better than the first test, but still not "great".
Install the top covers.
Test. Results should be very good.

SawFilter/rawpwb.gif SawFilter/filtamp.gif
Raw PWB.                                                         Components installed.

Showing one of the internal fences.

Sections Integrated

f.  Test Results (Click on images for larger display)

SAW/Amplifier Section, Data taken by Neal Martini.

As a comparison, the following plot is one I (Scotty) took of a newly completed SAW/Amp Section.
    Marker 5 is not at -130 dBm. It cannot be measured since this MSA range is limited to 1036 MHz. This can be seen as the "fall-off" just above Marker 4. Marker 4 is the attenuation near the image frequency of the MSA (1034.7 MHz). Neal got a rejection ratio of -31.4 dBc and I got -32.3 dBc. Very close for two different sets of SAW filters.  This is a good indication that the SLIM is reproducible.

Dual Section, data taken by Neal Martini.
    The input section of the module is fully shielded. It shows excellent attenuation characteristics and minimal crosstalk.  Well within the desired design goal.

As a comparison, the following plot is one I (Scotty) took of a newly completed Dual SAW Section.
    Marker 1 is not really at -130 dBm, but it is below -120 dBm. Marker 4, the attenuation near the image frequency of the MSA (1034.7 MHz) is the important data point. Neal got a rejection ratio of -62.2 dBc and I got -65.5 dBc. Very close for two different sets of SAW filters.

Final Results of Cascaded Sections
, data taken by Neal Martini.
    Both sections cascaded together, the SAW/Amp section first.

As a comparison, the following plot is one I (Scotty) took of the cascaded sections.
    Markers 1, 2, and 6 are not really at -130 dBm, but they are below -120 dBm. Marker 4, the attenuation near the image frequency of the MSA (1034.7 MHz) is the important data point. Neal got a rejection ratio of -89.5 dBc, but his shielding was not complete. I got -99.1 dBc. Actually, Marker 6 (1034.875 MHz) is closer to the image frequency and should be used. Neal got -97.7 dBc and using -120 dBm for my Marker 6, I got -107 dBc. Anything better than -90 dBc is excellent

g. Test, Evaluation, and Characterization
    The purpose of this section is to offer my personal notes while building the SLIM-SFA-1013. These are the steps and notes as I progressed.

Dual Cascaded SAW Filter Section (no attenuators)
1. Acquired raw PWB and SAW Filters.
SawFilter/pic2893.gif  SawFilter/pic2896.gif
2. Cut sections of pwb and trim edges.

3. Install SMA connectors.
SawFilter/pic2900.gif  SawFilter/pic2901.gif
4. I swept the dual SAW section to determine amount of J3 to J4 crosstalk on this pwb. It is difficult (if not impossible) to predict what a pwb's crosstalk will be. A "perfect" pwb would have no input to output RF energy flow. In reality, the input to output crosstalk needs to be better than the expected input to output isolation of the SAW filters.

    This plot simply proves that this pwb does have some crosstalk. The white trace is used as a reference. It is a transmission response of two SMA connectors in open air, about 1 inch apart. The yellow trace is the actual J3 to J4 crosstalk response. The Verification MSA has an upper limit of 1035 MHz, as shown by the abrupt loss near the end of the sweep. The input signal level to J3 is -12.2 dBm at 1035 MHz, and indicating about -99 dBm at J4. Therefore, crosstalk is about -86.8 dBc.
 At this point in testing we don't know how much of the crosstalk is due to air transmission (J3 to J4 acting as antennas), surface effect (RF flowing across the surface of the pwb), or substrate conduction (RF flowing inside the FR 4 material). In any case, we want to minimize it so that the resulting crosstalk does not materially affect the SAW Filter responses. We predict that the two cascaded filters should have at least -62 dB of rejection ratio at the frequency of interest (1034.7 MHz). It is essential that pwb crosstalk be minimized at least 10 dB below this number. That is, pwb crosstalk must be better than -72 dBc at all frequencies near our SAW Filter. This pwb's crosstalk
of -86.8 dBc does meet the goal, but there are several thing we are going to do to this board which may make the crosstalk worse.

5. I placed a small brass box over the J3 area. The purpose was to eliminate all air transmission from J3 to J4. I did not plot the response, but the total crosstalk dropped below the MSA noise floor (the -130 dBm line). This would conclude that all of the crosstalk measured in step 4. was attributed to air transmission. This is good news. It means that the design using plated-thru holes to minimize
substrate conduction was successful. Air transmission crosstalk is far easier to minimize than surface effect or substrate conduction. The same brass box over the J4 area resulted in the same response (as predicted).

6. I installed three temporary 100 pfd capacitors in place of R8, R11, and R14. A used a fourth 100 pfd to bridge over the uninstalled FL 3, This is to provide a low impedance path from J3 to J4 so that I can test the true substrate conduction crosstalk under FL2. I added the total perimeter fence using .2 inch height coffee can material (tin plated steel). This picture looks inverted, but connected to the MSA (under the blue cloth) the signal direction is from left to right (Tracking Generator Output to J3 and J4 to MSA Input).

The following plot shows the crosstalk with only the capacitors and perimeter fence in place. The top white trace is J3 to J4 crosstalk. The yellow trace is the same, but with a temporary lid on the module.
    This crosstalk looks extremely high, but it is due to the 4 coupling capacitors that lowers the impedance from J3 to J4. We need to shield the circuit internally to minimize the air transmission effect before obtaining viable data for the
substrate conduction crosstalk under FL2.

7. I added an internal fence over the ground strip that separates the input and output of FL2.
    The white trace is a reference, as used in the previous plot in step 6. The yellow traces are my attempt at holding a temporary top cover over the J3 section. This shows the crosstalk passing through the FR4 material from the input pad of FL2 to its output pad. With an input at -12.3 dBm and output at -100 dBm, this crosstalk is about -88 dBc. Although this crosstalk seems high, it is actually pretty darn good. Consider this: these two input/output pads are separated by only .050 inch. If the pads were not separated with a ground strip and ground vias, the internal FR4 crosstalk would be horrible. As it is, we only need crosstalk to be about 10 dB better than the expected SAW filter response at 1034.7 MHz, which is predicted (by specification) to be -31 dBc. We beat this by 57 dB. Woo-Hoo!
    Now let's see how a single SAW performs with this really good pwb.

8. Install FL2. (Holy Flyspeck Batman, this thing is small!).
    If you construct this board according to the Construction procedure, the SAW will be installed before the SMA connectors and the perimeter fencing. This allows solder paste and reflow technique. This would be the easiest way to install this SAW. However, I could not do this since I wanted to characterize the pwb before installing the SAW. Here is how I installed the SAW using a standard fine tip soldering iron. I flipped the SAW on its back and applied a tiny solder bead on the input lead. I then positioned the SAW in its correct position on the pwb. I then applied the soldering iron tip on the pwb trace as close to the input lead as possible. The heat transfer caused the solder bead on the SAW to flow onto the pwb's trace. I then soldered the output of the SAW in the same manner, but adding solder this time. I then soldered the 4 ground corners using the same method. I cleaned with 99% isopropyl. ( I let dry and then I cleaned it again).

9. I added the internal fence over FL2.
SawFilter/pic2913.gif  SawFilter/pic2914.gif  Dual Filter Assembled
    I cut the fence to size (roughly) using coffee can lid and metal shears. I used a Dremel Tool with a carbide cut-off wheel to trim for final shape and added the notch for the SAW. I soldered the internal fence sides to the perimeter fence first and then to the ground strip than extends under the SAW. While still hot, I added solder to the fence directly above the SAW and moved the soldering iron tip down to the top of the SAW and let the solder flow. The SAW accepted the solder quite nicely. The dual section, completed. Notice that there are three seperate lids.

10. Sweep for results. There is a temporary lid tacked over the J3 area only.
    Results are extremely close to both Neal's data and the Vendor's specs. Insertion loss is (-12.27 dB/-15.9 dB) = -3.63 dB. Attenuation at the image frequency (1034.5 MHz) is
(-15.9 dB/-47.86 dB) = -31.96 dB.
    The following wide band sweep shows the out-of-band attenuation at lower frequencies. With my set-up I cannot sweep above 1035 MHz.

11. I removed the temporary 100 pfd at FL3 and installed the SAW at FL3. I fashioned a second internal fence and installed it over FL3 in the same manner as done for FL2.
Insertion loss is (-12.27 dB/-18.21 dB) = -5.94 dB. Attenuation at the image frequency (1034.5 MHz) is (-18.21 dB/-88.54 dB) = -70.33 dB. I do not show a wide band sweep below 988 MHz because it would be a waste of web page space. The isolation is well below the noise floor of the MSA.
  You might notice that I showed the insertion loss of the single FL2 SAW to be -3.63 dB. Here, I show the combined insertion loss of FL2 and FL3 to be -5.94 dB. This would indicate that the insertion loss of the FL3 SAW is (-5.94/-3.63) = -2.31 dB. However, after installing FL3 I noticed that the Tracking Generator SMA connector was loose. This might account for the descripancy. I am inclined to believe that each SAW has an insertion loss of -5.94 / 2 = -2.97 dB.

    This shows the 3 dB bandwidth (2 cascaded SAW's) to be 16.4 MHz. It also shows that the in-band response has over 1 dB of "slope". Not really a concern and we can't do anything about it, anyway.

    I added both internal 3 dB attenuators and added the lids permanently. This is the sweep of the finalized section:
    The insertion loss is now
(-12.27 dB/-24.18 dB) = -11.91 dB.
I added a 2.8 dB pad on each port (center trace). Then added another 2.8 dB pad on each port (bottom trace).
    The top trace is a carry-over from the previous sweep. Adding the external attenuators causes distortion. Weird!

12. I did a quick test of phase change versus temperature change. I will do more, but preliminary testing indicates that the phase change is inversely proportional to temperature change. That is, as temperature increases, the phase decreases. It is about 1 degree of phase for each 1 degree F. Center frequency insertion loss does not change at all over a 20 degree temperature change. I will do further testing to see if the image attenuation changes vs. temperature.

This concludes the characterization of the Dual Cascaded SAW section.

SAW/Amplifier Section (no attenuators)
1. Install SMA connectors
2. Install .25 inch perimeter fence
3. Install C8 and C9 as a perminant installation
4. Temporarily install two 100 pfd capacitors. One across the pads that R5 will occupy and one across the pads that U1 will occupy. This is to create a low impedance path from J1 to J2 so that I can measure the isolation between the two pads that FL1 will occupy.
5. Temporarily install the FL1 internal fence. It has no cut-out for the SAW. The SAW is not installed yet.
7. Tack a temporary lid over the J1 area.
8. Measure the transmission from J1 to J2 to characterize
the crosstalk passing through the FR4 material from the input pad of FL1 to its output pad. I would expect it to be the same as the crosstalk number taken during the Dual Filter Board testing, which was -88 dBc.
    The input power is -12.3 dBm and the output is -100 8 dBm. Therefore the FL1 gap crosstalk leakage is -88.5 dBc. We needed at least -41 dBc and we beat that by 47.5 dB.
By golly, we have another winner.

9. Install FL1 and its
internal fence. (no amplifier or internal pad)
Tack a temporary lid over the J1 and J2 areas. Sweep for characterization.
    Insertion loss is
(-12.27 dB/-15.74 dB) = -3.47 dB. Image point (R) is 32.3 dBc. Quite similar to the results of FL2 taken earlier. However, the out-of-band response is about 10 dB poorer than FL2. I'm not going to spend much time trying to determine why the difference. Since the out-of band levels are at least 10 dB below the image response it won't make a difference.
    Here is a wide band sweep of the response.
(no amplifier or internal pad)

    Here is a narrow sweep to show the 3 dB Bandwidth.
(still no amplifier or internal pad)

    I added the internal 3 dB attenuator and the MMIC amplifier. Permanently installed lids. The SAW/Amp section is now complete.  The following plots are sweeps with external attenuators. The top trace is with two external 3 dB attenuators, one on each port (J1 and J2). It is somewhat distorted due to it overloading the MSA input. The maximum input without distortion is -11 dBm. The center trace is with 6 dB attenuation on each port. The MSA is within its linear dynamic range. The bottom trace is with 9 dB attenuation on each port.
    According to the top trace, the insertion loss (gain) is
(-12.27 dB/-5.14 dB) = +7.13 dB. This is taken with 6 dB of external attenuation, so the total insertion gain of the SAW/Amp section is +13.13 dB.

Integrating the Two SAW Sections (no amplifier or attenuators)
    The previous paragraphs show data of the two sections during construction and after completion. During construction, I tested the uncompleted sections cascaded together. The purpose was to get total insertion loss and total bandwidth data for 3 series SAW filters with no amplification or internal attenuators. These are the results.

1. I connected the MSA's Tracking Generator to J3 of Dual SAW. Its J4 is connected to J1 of SAW/Amp section. Its J2 is connected to the MSA input. The Dual SAW section has no internal attenuator resistors. The SAW/Amp section does not have its amplifier or internal 3 dB attenuator resistors installed. A coupling capacitor is bridged over each of the missing components. This simulates a test of 3 series SAW filters.
    The heading of this graph is confusing. It really should read "Data of 3 series SAW Filters".
    The total insertion loss is (-12.3 dBm - 20.94 dBm) = -8.64 dB. The image point (R) is displayed as being below -120 dBm, which is the lowest level this MSA can measure effectively. There is no calibration below -110 dBm. Even if the image point was at -110 dBm, the image rejection would be -89.06 dBc. Obviously it is better than this, but at this point in test, I don't know precisely what it is. I will determine this later.
    This plot is impressive. The design stipulates 3 dB attenuators between SAWs to minimize interaction. However, I'm beginning to think the SAWs can tolerate feeding each other without padding. I will know more once the attenuators are in place.

2. Plot the 3 dB points of the 3 SAWs in cascade.
(no MMIC amplifier or internal pads)
    Yuck. We may need those pads, after all. The peaking may be caused by the interaction the SAWs are having with each other. It is also possible that the MSA is causing interference, since I am sweeping through the 1013.3 MHz of its First I.F. Anyway, the relevant information is the 3 dB bandwidth, which is about 14.9 MHz.

Integrating the Two completed Sections (full shielding, internal pads, and amplifier)
    The following plots are the responses to the SLIM-SFA-1013 in its final configuration. Tracking generator to J1, J2 to J3, and J3 to MSA input. The top trace is with no external attenuators on the ports (J1 and J4). The center plot is with a 3 dB pad on each port. The bottom trace is with 6 dB of attenuation on each port.
    The Tracking Generator power to J1 is -12.3 dBm. Therefore the total gain of the two cascaded sections is +1.1 dB. Yes, those internal attenuators are needed to help isolate the SAWs from each other.
I did try adding a 3 dB attenuator in front of FL2 to smooth things, but it made no difference.

SAW Input Impedance Test
    According to the data sheet, the SAW is close to 50 ohms but not exactly. I ran a Reflection measurement (S11) of J3 (SAW/Amp section) to see what the impedance really is.
  This is what I got:
  As suspected, the port is not 50 ohms at 1013 MHz, it is closer to 75 ohms.

    Here is an S22 Reflection measurement of the SAW/Amp Section (J2). Basically, it is an output impedance measurement of the ERA-33 amplifier.
    One would expect to see a flat 50 ohm response. However, this takes place only if the ERA-33 amplifier's input impedance remains constant. Here, the SAW filter is absorbing and reflecting
the amplifier's reverse coupling (reverse isolation) energy, and that changes the source impedance of the amplifier's input. Here is the Smith chart for the above data.

    The following plot is a comparison of the 
SLIM-SFA-1013 to the Coaxial Cavity Filter. The Cavity Filter is tuned to 1014 MHz, which is the First IF of the Original MSA (from which the filter was removed). I made no attempt to re-tune the Cavity just for the purpose of this comparison. It has an insertion loss of -8 dB. The SLIM-SFA-1013 has an insertion loss of +1.2 dB.

    The following plot is a narrower sweep, showing the S21 phase of both filters. The SAW traces are yellow and the Cavity traces are white. Since the sweep is only 5 MHz wide, the S21 Magnitude of the SAW is a straight line at the top of the graph.
    The phase delay through the Coaxial Cavity Filter is a total of 360 degrees, 90 degrees for each cavity. The phase delay for the SAW Filters seem to be less, but this is deceiving, as this is plot is for a (relatively) narrow frequency sweep.

    The following plot shows the S21 for the SAW Filters for its entire 3 dB bandwidth.
Here, you can see that the phase delay is 3 complete rotations (1080 degrees) within its 3 dB bandwidth. This is 360 degrees for each SAW Device.

h. Spurious Responses
    For some MSA builders, the SAW Filter approach is easier to implement than constructing the Coaxial Cavity Filter. However, this approach does have its limitations. Replacing the Coaxial Cavity Filter with the wider band SAW Filters may result in more spurious responses while in the Spectrum Analyzer Mode. These spurious responses are somewhat predictable, and easily differentiated from real input signals. But still, they can be a nuisance, especially for novice spectrum analyzer users. The following paragraphs will explain how the "predictable" spurious responses are created. I should point out that spurious responses are created in an MSA with either the Coaxial Cavity Filter or the SAW Filters. I will address only those responses that will be higher using the SAW Filters, as opposed to the Cavity Filter.
    I define a spurious response as "the Magnitude indication of an input signal that does not exist". As an example, when the Spectrum Analyzer has a measurable Magnitude response when it is commanded to 100 MHz, yet there is no input signal at 100 MHz. The measurement is called a "Spurious Response". The MSA's commanded frequency (100 MHz) is called the "Spur Frequency", even though it does not exist. The actual signal that creates the spurious response can enter the MSA via its Input, be self-generated internally by the MSA, or it can be a combination of both. These signals are called "Spur Generators".
    In almost all cases, a spurious response is created by the product of two or more frequencies within a mixer. In the MSA, there are two mixers, Mixer 1 and Mixer 2.
Ideally, there would be only two RF frequencies that enter Mixer 2 and produce the proper IF of 10.7 MHz. These two RF frequencies are IF1 and PLO2. IF1 is the product of Mixer 1 and enters Mixer 2 at its R port. PLO2 is the 2nd local oscillator and is fixed at 1024 MHz, entering Mixer 2 at its L port.
the band pass (Cavity or SAW) filter between Mixer 1 and Mixer 2 would allow only an IF1 frequency of 1013.3 MHz to enter Mixer 2. The product of Mixer 2 would then be PLO2 - IF1 = 10.7 MHz. However, nothing is "ideal" and we must deal with "real" components.
    Mixers are not "ideal" and they will produce output frequencies with an action called harmonic mixing. "Mixer-Generated Inter modulation Distortion", "Intermods", and "Single-Tone IMD" are some of the terms used for this mixing action. I will generalize this form of mixing action as "IMD" or Inter Modulation Distortion. IMD is not very efficient, but none the less, it is how spurious responses are created in the MSA. IMD occurs in both Mixer 1 and Mixer 2. For either mixer, this is the mathematical formula:
±(M*LO) ±(N*RF) = Mixer Output  : where M and N are harmonic multipliers (0 through infinity), and LO and RF are the two frequencies to be mixed.

    For Mixer 2, the formula can be re-written as:
±(M*PLO2) ±(N*IF1) = Final IF

The frequencies of PLO2 and Final IF are fixed (1024 MHz and 10.7 MHz). Therefore, the formula can be re-arranged to find what IF1 frequencies will create a spurious response:
IF1 = (
±Final IF ±M*PLO2)/N, or
IF1 = (
±10.7 ±M*1024)/N
    By substituting M and N with multipliers,we can calculate all of the (single) IF1 frequencies at which Mixer 2 can create a spurious response at 10.7 MHz.
Usually, responses that are created by harmonics greater than about 9 are insignificant. Therefore, I will show only those IF1 frequencies using harmonics of 1 through 10. Also, I will show only those IF1 frequencies that can pass through the SAW Filters with less than 90 dB attenuation (990 MHz - 1035 MHz).
    The following table shows the IF1 frequencies that will potentially produce a spurious response in Mixer 2.
Harmonic Multiples                IF1 Frequency

using M*PLO2 - N*IF1 = IF2
for M=1    and    N=1,  IF1 = 1013.3 MHz    (The "normal" or Fundamental IF1)
for M=2    and    N=2,  IF1 = 1018.65 MHz
for M=3    and    N=3,  IF1 = 1020.43333 MHz
for M=4    and    N=4,  IF1 = 1021.325 MHz
for M=5    and    N=5,  IF1 = 1021.86 MHz
for M=6    and    N=6,  IF1 = 1022.21667 MHz
for M=7    and    N=7,  IF1 = 1022.47143 MHz
for M=8    and    N=8,  IF1 = 1022.6625 MHz
for M=9    and    N=9,  IF1 = 1022.81111 MHz
for M=10  and    N=10, IF1 = 1022.93 MHz

using N*IF1 - M*PLO2 = IF2
for M=10  and    N=10, IF1 = 1025.07 MHz
for M=9    and    N=9,  IF1 = 1025.18889 MHz
for M=8    and    N=8,  IF1 = 1025.3375 MHz
for M=7    and    N=7,  IF1 = 1025.52857 MHz
for M=6    and    N=6,  IF1 = 1025.78333 MHz
for M=5    and    N=5,  IF1 = 1026.14 MHz
for M=4    and    N=4,  IF1 = 1026.675 MHz
for M=3    and    N=3,  IF1 = 1027.56667 MHz
for M=2    and    N=2,  IF1 = 1029.35 MHz
for M=1    and    N=1,  IF1 = 1034.7 MHz
   (The Image Frequency)

    There are three mechanisms for producing these
IF1 Frequencies:
Case 1. Directly, by PLO1. The MSA can be commanded to a frequency at which PLO 1 will generate the IF1 frequency.
The frequency of PLO1 will always be 1013.3 MHz above the command frequency. Example: Command MSA to 5.35 MHz.  (PLO1 = 5.35 + 1013.3 = 1018.65 MHz). The power level of PLO1 at the input to the IF1 Filter depends on Mixer 1, L to R port isolation, but will be in the range of -20 dBm to -35 dBm. In my Original MSA, it is approximately -38 dBm.
Case 2. As a mixing product of Mixer 1. When an input frequency combines with PLO 1 (in Mixer 1) to create one of the IF1 Frequencies. This is called Single-Tone Inter Modulation Distortion.
Case 3. Two or more input frequencies to the MSA will combine in Mixer 1 to produce the IF1 Frequencies. This is a form of Two-Tone Inter Modulation Distortion.

Case 1. Direct Generation

    The following table shows the MSA Frequency Commands that will directly create the IF1 frequencies. This test is performed with the Original MSA, using the SAW Filters. The external input to the MSA is terminated into 50 
Ω (no signal input). IF1 (PLO 1) is approximately -38 dBm on the input of the SAW Filters. The power level of IF1 at the input to Mixer 2 is dependent on the frequency response of the SAW Filters. The Indicated Magnitude response level is a relative measurement. That is, if a "real" 5.35 MHz signal at -96 dBm were input to the MSA, the Magnitude measurement would be "-96 dBm". Here, the MSA is commanded to measure a 5.35 MHz input signal, but is actually measuring the internal PLO1 frequency power and indicating a spurious response measurement of -96 dBm.
IF1 Frequency                                                                Indicated Magnitude Response Level
IF1 = 1013.3 MHz,              MSA Commanded to 0.0 MHz                  -31 dBm

IF1 = 1018.65 MHz,            MSA Commanded to 5.35 MHz                -96 dBm
IF1 = 1020.43333 MHz,      MSA Commanded to 7.13333333 MHz     <-120 dBm
IF1 = 1021.325 MHz,          MSA Commanded to 8.025 MHz             
<-120 dBm
IF1 = 1021.86 MHz,            MSA Commanded to 8.56 MHz               <-120 dBm
IF1 = 1022.21667 MHz,      MSA Commanded to 8.91666667 MHz       -110 dBm
IF1 = 1022.47143 MHz,      MSA Commanded to 9.17142857 MHz     <-120 dBm
IF1 = 1022.6625 MHz,        MSA Commanded to 9.3625 MHz            <-120 dBm
IF1 = 1022.81111 MHz,      MSA Commanded to 9.51111111 MHz     <-120 dBm
IF1 = 1022.93 MHz,           MSA Commanded to 9.63 MHz                <-120 dBm

IF1 = 1025.07 MHz,           MSA Commanded to 11.77 MHz
               <-120 dBm
IF1 = 1025.18889 MHz,     MSA Commanded to 11.8888889 MHz       <-120 dBm
IF1 = 1025.3375 MHz,       MSA Commanded to 12.0375 MHz            <-120 dBm
IF1 = 1025.52857 MHz,     MSA Commanded to 12.2285714 MHz       <-120 dBm
IF1 = 1025.78333 MHz,     MSA Commanded to 12.4833333 MHz       <-120 dBm
IF1 = 1026.14 MHz,           MSA Commanded to 12.84 MHz               <-120 dBm
IF1 = 1026.675 MHz,         MSA Commanded to 13.375 MHz              <-120 dBm
IF1 = 1027.56667 MHz,     MSA Commanded to 14.2666667 MHz       <-120 dBm
IF1 = 1029.35 MHz,           MSA Commanded to 16.05 MHz               <-120 dBm
IF1 = 1034.7 MHz,             MSA Commanded to 21.4 MHz                 <-120 dBm
    A measurement of  <-120 dBm does not necessarily mean that there is no spurious response. -120 dBm is the noise floor of this MSA. (The responses shown will not be accurate for every MSA using the SAW Filter topology. However, the testing does indicate that the spurious responses will likely exist.)  This test was repeated with the MSA using the Coaxial Cavity Filter. None of the IF1 frequencies produced a measurable response. This is due to the much narrower bandwidth of the Cavity Filter.

Case 2. Single-Tone IMD
    The following table shows the MSA's Magnitude response levels when an input signal is combined with PLO 1 to create the
IF1 Frequency. The external input to the MSA is a fixed frequency at five different power levels: 0 dBm, -10 dBm, -20 dBm, -30 dBm, and -40 dBm. The MSA is commanded to the Input Frequency plus a Frequency Offset.  The IF1  Frequency is produced by the mixing action of the input signal and PLO 1. Mixer 1 output power level is approximately -7 dB below the input power level. The power level of IF1 at the input to Mixer 2 is dependent on the frequency response of the SAW Filters. For testing, I used an input frequency of 30 MHz (at the 5 power levels). However, any input frequency could be used as long as the same Frequency Offset were added to the MSA Command Frequency. For example, in the first step, the MSA is commanded to 30 MHz. In the second step, the MSA is commanded to 35.35 MHz, etc.

IF1 Frequency                            Frequency Offset            Indicated Magnitude Response Level (dBm)
IF1 = 1013.3 MHz                        0.0 MHz                            0      -10     -20      -30     -40
IF1 = 1018.65 MHz                     5.35 MHz                         
-38      -55     -75      -96     -117
IF1 = 1020.43333 MHz                7.13333333 MHz              -47     -69      -99   <-120  <-120
IF1 = 1021.325 MHz                    8.025 MHz                      -68      -95   <-120   <-120  <-120
IF1 = 1021.86 MHz                      8.56 MHz                        -84     -118  <-120  <-120  <-120
IF1 = 1026.22392 MHz                8.91666667 MHz              -93   <-120  <-120  <-120  <-120
IF1 = 1026.47856 MHz                9.17142857 MHz              -102 <-120  <-120  <-120  <-120

  The other 13 IF1 Frequencies were tested, but none created a measurable Magnitude level. The responses shown will not be accurate for every MSA using the SAW Filter topology. However, the testing does indicate that the spurious responses will likely exist. This test was repeated with the MSA using the Coaxial Cavity Filter. None of the IF1 frequencies produced a measurable response. This is due to the much narrower bandwidth of the Cavity Filter.

Case 3. Two-Tone IMD
    Multiple input signals to the MSA will combine with PLO1 in Mixer 1 to create an IF1 that can produce a spurious response in Mixer 2. There are literally millions of different frequencies that can produce this effect with both the SAW Filter and Cavity Filter. I will concentrate on only one circumstance that will affect the MSA with a SAW Filter much more than an MSA with a Cavity Filter. This is when the difference between two input frequencies, F1 and F2, is within the 90 dB bandwidth of the MSA's Final Resolution Filter. Also, the product of each input frequency and the frequency of PLO 1 must be within the 90 dB bandwidth of the First IF Filter. This may seem like a rare event, but it is not. Input signals with wide band modulation occur quite frequently. For example, two Digital TV channels are shown in the following Spectral Scan.
The MSA is tuned to 113.25 MHz, the guard band between the two TV channels. Marker 1 is indicating RF power at 107.91 MHz.
Marker 2 is indicating RF power at 118.614 MHz. The two signals meet the first criteria that their difference is within the 90 dB bandwidth of the Resolution Filter (118.614-107.91 = 10.704 MHz).
Since the MSA is tuned to 113.25 MHz, PLO 1 is at (1013.3 MHz + 113.25 =) 1125.55 MHz.
The product of F1 and PLO1 is 1125.55 - 107.91 = 1017.64 MHz.
The product of F2 and PLO1 is 1125.55 - 118.614 = 1006.936 MHz.
The 90 dB bandwidth of the SAW Filter is from 990 MHz to 1035 MHz. Therefore, both the F1 and F2 products of Mixer 1 meet the second criteria. Since both signals will pass through the SAW Filter, they will mix together in Mixer 2. (1017.64 MHz - 1006.936 MHz = 10.704 MHz). In the above scan, there is no indication that a spurious problem exists in this situation. Here is why:
The power level of each signal entering Mixer 2 is about -98 dBm, well below a power level required to create an IMD Magnitude response. If the input signal levels were high enough, we would see that the signal level at 113.25 MHz would be greater than the noise floor of the MSA (-120 dBm). We will perform that test in the next scan.
    This form of Two-Tone IMD is not likely in an MSA using the Coaxial Cavity Filter. This is because the second criteria is difficult to meet with the narrower bandwidth of the Cavity Filter.

    In the following scan, two input signals have amplitude great enough to create severe Two-Tone IMD.
The odd behavior of the IMD is the combination of Two-Tone IMD and other IF1 spurious combinations. We are dealing with four fundamental signals, which create a large combination of distortion products.

The following scan is a continuation of the previous scan. In each successive sweep, the power of both input signals are attenuated 10 dB. The power level of the IMD drops dB for dB as the input powers are attenuated.
The bottom trace is when both input powers are -43.9 dB and -43.6 dB.  It can be safely assumed that if both input signals are less than -50 dBm, the IMD will be too low to be measured. The following scan shows those results.

    The only way to decrease this type of Two-Tone IMD is to change the Final IF from 10.7 MHz to a frequency greater than about 20 MHz. A Final IF of 21.4 MHz is an option, and will decrease this IMD. However, to totally eliminate this potential IMD, the Final IF must be greater than about 28 MHz. 28 MHz is the -50 dBc attenuation points on the SAW Filter skirts.

1. Increasing the First IF to 1017.3 MHz will reduce the amplitude of the spurious responses, but not totally eliminate them. This is a simple change to the Hardware Configuration Manager to "tell" PLO 2 to operate at 1028 MHz instead of 1024 MHz. Changing
the First IF to 1021.3 MHz is a better option, which will be detailed in the next section.
2. Changing the Final IF to a frequency greater than 20 MHz would have a better affect of reducing all spurious responses. However, narrow Resolution Filters at higher frequencies are difficult to build or acquire. Also, a Final IF greater than 30 MHz will require an MSA hardware change. Both Mixer 2 and the SLIM-IFA-33 have cut-off frequencies of 30 MHz. Also, the MSA/VNA has not been tested with a Final IF greater than 21.4 MHz.
3. Changing the MSA topology from Dual Conversion to Triple Conversion would eliminate the spurious responses described in the previous paragraphs. However, new spurious frequencies may be introduced, depending on what frequency schemes are used. This would be a big change to both MSA Hardware and software. The software would have to include both topologies of an MSA (Dual and Triple Conversion). It would be more prudent to establish a Hardware/Software scheme totally separate from the present MSA.

Exploring Option 1
    The IMD levels created in Cases 1 and 2 can be somewhat reduced by operating the MSA's First IF at the higher frequency of 1017.3 MHz (or 1021.3 MHz) rather than 1013.3 MHz. This is because the IF1 frequencies are higher and are positioned on the upper slope of the SAW Filter. This is a simple change to the Hardware Configuration Manager, to tell the software that PLO2 is to operate at 1028 MHz (or 1030 or 1032 MHz) instead of 1024 MHz.
For a study of Option 1, spurious response data was taken with the First IF at four different frequencies. The sweep labeled "PLO2 at 1024" shows the SAW Filter Band-Pass Response with the First IF at 1013.3 MHz, the MSA's normal frequency configuration. Markers are placed on the response trace to indicate how effective the IF1 spurious frequencies are attenuated through the SAW Filter. The actual frequency at each marker is the Center Frequency (1013.3 MHz) plus the marker frequency.

PLO2 at 1024
Marker 1 is 1013.3 + 5.35 = 1018.65 MHz. This is the frequency that creates the 2*M - 2*N spurious response. As we can see, this MSA frequency configuration is not very effective in reducing the spurious frequencies below Marker 3.

The MSA frequency configuration is changed to take advantage of the upper skirt (slope) of the SAW Filter. Here, the markers still indicate where the IF1 spurious frequencies will reside.
PLO2 at 1028
    The Marker 1 frequency at 1017.3 + 5.35 = 1022.65 is the 
2*M - 2*N spurious response. It is attenuated by 5.4 dB. However, all higher spurious response frequencies are much more attenuated. Changing the MSA to this frequency configuration does not result in fewer spur frequencies, but it does attenuate their final spurious responses.

    The PLO2 is operating at 1030 MHz, causing the First IF to be at 1019.3 MHz. PLL 2 must use a Phase Detector Frequency of 2 MHz instead of the normal 4 MHz.
PLO2 at 1030
    The Marker 1 frequency at 1019.3 + 5.35 = 1024.65 is the  2*M - 2*N spurious response. Marker 5 indicates where the Image Frequency will fall.

    Finally, the MSA frequency configuration is changed to take maximum advantage of the upper skirt (slope) of the SAW Filter. Here, the markers still indicate where the IF1 spurious frequencies will reside.
PLO2 at 1032
    The Marker 1 frequency at 1021.3 + 5.35 = 1026.65 is the  2*M - 2*N spurious response. It is now greatly attenuated. And, all spurious response frequencies higher than that are even more attenuated.

    The following table shows the Case 2 spurious responses with the four different MSA frequency configurations. The responses shown are not absolute, but can be used as a comparison to each other. It should be noted that the input signal creating the IMD response has a power level of 0 dBm. This is a much higher signal power level that would normally be input to the MSA. Therefore, this data is a worse case situation. As a rule of thumb, when the input is lowered by 10 dB, the spurious response will decrease about 20 dB.
Numeric Spur     Freq Offset                  1013.3/1024    1017.3/1028    1019.3/1030    1021.3/1032

1M-1N               0.0 MHz                        0   (ref)            0   (ref)          0   (ref)          0   (ref)
2M-2N               5.35 MHz                     -43 dBc            -51 dBc         -75 dBc         -101 dBc
3M-3N               7.13333333 MHz          -53 dBc            -93 dBc
4M-4N               8.025 MHz                   -75 dBc
5M-5N               8.56 MHz                     -93 dBc
6M-6N               8.91666667 MHz         -102 dBc
7M-7N and greater spurious responses are below the measurable noise floor of the MSA.
A response that has its measurement blank is
below the measurable noise floor of the MSA.
    The following is a comparison of the Case 1 spurious response with the different PLO2 frequencies:
Numeric Spur                  1013.3/1024    1017.3/1028    1019.3/1030    1021.3/1032
2M-2N                             -96 dBm          -107 dBm       -118 dBm       <-120 dBm
Others are too low to be measurable.

    It would seem that the best option is to change the MSA's frequency configuration to use 1021.3 MHz as its First IF frequency. However, there is a concern. 1021.3 MHz is on the upper skirt of the SAW Filters. This can be seen as Marker 1 on the following "
Temperature Shift" graph. SAW Filter characteristics will change with temperature. It is likely that the SAW frequency response will shift enough over temperature to cause the absolute Magnitude measurements to change. The two traces in the following graph are the Magnitude responses of the SAW Filters. The white is taken during a cold start. The yellow is after a warm-up of 10 Minutes.
Temperature Shift
    The markers shown on the graph are for cold. They are not displayed for warm. However, they are listed below:
Marker     Cold               Warm            Change vs Temperature
L            -30.21 dBm    -30.18 dBm        .03 dB
1            -34.67 dBm    -35.26 dBm        .59 dB
The change at the center frequency of the filter (Marker L, 1013 MHz) is minor, but the change at the operating point (Marker 1 at 1021.3 MHz) is significant. The main contributor is the SAW Filter (FL1) in the SAW/Amp section, which is getting extremely warm from the ERA-33 circuitry. This heating effect could be minimized with one or both of the following modifications:
1. The ERA-33 is normally running at 38 ma of current.
As designed, this section consumes about .4 watts. This allows full compliance to the Minicircuits specifications of gain and compression. However, we don't need +13 dBm at compression. We could get by with a much lower compression point, about +2 dBm. By lowering the current to 15 ma (from its nominal 38 ma) the compression point would be about +6 dBm. The total power consumption would be reduced to .15 watts. The gain will be reduced a little, but we don't need much gain with this SAW Filter replacement for the Cavity Filter. This is easily accomplished by changing the values of R1 and R2 to 200 ohms. They can be 1/10 watt, 0805 size.
2. The power consumption of the ERA-33 at 15 ma is about .06 watts. The rest of the power (.09 watts) is consumed by the bias resistors, R1 and R2. These could be positioned outside of the module where their heating would not affect the SAW filter. At 15 ma, R1+R2 should be about 400 ohms. A 390 ohm axial lead resistor could be placed in the 10v feed line, external to the module. The internal R1 and R2 would be replaced with 4.7 ohm resistors.
3. The ERA-33 could be replaced with a lower compression, lower current MMIC device. The MAR-6 would do this. (R1 and R2 would need re-calculating.)
    I performed an experiment where I lowered the input voltage to the SAW/Amp section from 10 volts to 6.8 volts. This lowered the MMIC current from 38 ma to 17.5 ma, a total power dissipation of 118 within the module. R1 and R2 were left in place, dissipating 23 mw each. The MMIC dissipates 72 mw. This decrease in power consumption (from 400 mw to 118 mw) made a significant difference in the temperature rise of the module. The total gain only decreased 1.7 dB. The gain vs temperature change at 1021.3 MHz (Marker 1 in the previous sweep) decreased only .2 dB (was .59 dB).
i. Conclusions
    When substituting the SLIM-SFA-1013 in place of the Coaxial Cavity Filter, the best MSA results will be obtained if these suggestions are followed:
1. Use 1021.3 MHz instead of 1013.3 MHz as the 1st IF to minimize CASE I spurious responses (< –101 dBc)
2. Use 1021.3 MHz instead of 1013.3 MHz as the 1st IF to minimize CASE II spurious responses (< –120 dBc)
3. To avoid Two Tone IMD problems (CASE III), reduce MSA input levels for strong signals using external attenuation.
4. Lower the MMIC current in the SAW/Amp section to reduce temperature effects.
5. Allow the MSA a minimum of 10 minute "warm up" stabilization before any calibrations or important data measurements.