MSA Temperature Testing
Released 7-01-10.
Updateded 7-03-10. Added section on Thermometer used for testing.
Updateded 7-22-10. Add more temperature data.
Updateded 8-09-10. Add "Comment on Compensation" at end of page.

    The MSA can be used in ambient temperatures that vary from below freezing (<+32 F.) to desert conditions (>+100 F.).  Ambient temperatures in normal home usage would typically range from 65 F. to 85 F.  The MSA's frequency accuracy and both Magnitude and Phase Measurements are affected by temperature changes to components within the MSA.  If the temperature of each sensitive component could be regulated to a fixed temperature, the accuracy of the MSA would rival the most expensive commercial Analyzers.  This is not practical for the MSA.  The MSA's components that are the most sensitive to temperature variations are:
Master Oscillator: It has a positive frequency vs. temperature coefficient of approximately .15 ppm (parts per million) per degree F. This means if the temperature of the Master Oscillator changes 1 degree F, its frequency will change .000015 percent.  For a 64 MHz oscillator, this is 9.6 Hz.  All frequencies generated within the MSA are dependent on the stability of the Master Oscillator, since it is the only frequency reference in the MSA.
Final Xtal Filter (Resolution Filter): Temperature variations of this filter will change Insertion Loss (Magnitude variation) and Center Frequency (Phase variation). Crystal filters (narrow BW) are usually much more susceptable to temperature variations than lumped element filters (wide BW).
    I characterized the Final Crystal Filter (SLIM-MCF-FL096) used as the primary path (Path 1) in the Verification MSA. 
The Magnitude change vs. Temperature change is negligible. It is less than .01 dB / 1 deg F.  The resonant center frequency will decrease at a rate of -4.51 Hz / 1 deg F, which is also a phase vs. temperature coefficient of -.777 deg / 1 deg F (a negative temperature coefficient).
Updateded 7-22-10  I tested a dual pole, 10.7 MHz ceramic filter (Murata), with a 320 KHz BW.  It has a negative temperature coefficient: -500 Hz / 1 deg F. The phase change is -.57 deg / 1 deg F.
Coaxial Cavity Filter: Temperature variations of this filter will change Insertion Loss (Magnitude variation) and Center Frequency (Phase variation). The characteristics of this filter depend solely on its physical dimensions. Since copper and brass expand (and contract) over temperature, it is natural for this filter to be affected by temperature.  The diameter of the inner center stubs and the diameter of the outer cavity will increase with temperature (although I don't know how much). These dimensions will not alter the resonant frequency. The length (height) of the center stubs will increase with temperature, causing the resonant frequency to decrease. The length (height) of the outer cavity will increase with temperature, but should have no affect on resonant frequency. However, since its height increases with temperature, the tuning screws will move further away from the top of the stubs. This will cause the resonant frequency to increase. The combination of these factors create a temperature coefficient that is unpredictable.
The coaxial cavity filter in the Verification MSA has a positive Phase vs. Temperature coefficient of 1.0 deg/1 deg F (5.56 KHz / 1 deg F). The coaxial cavity filter in the Original MSA has a negative Phase vs. Temperature coefficient of -.20 deg/1 deg F (-1.11 KHz / 1 deg F). This proves "unpredictability".

Temperature Characterization of the MSA
    Even though the individual componentss of the MSA can characterized for temperature variations, the integrated MSA can only be characterized as a single unit.  This is because the temperatures of the individual SLIMs and components will vary widely. "Hot spots" are common in integrated equipment. When characterizing a temperature coefficient for an integrated system, the question is then "what temperature is used as a reference".  I suggest that the exhaust air of the MSA be used as the Reference Temperature. This is equivalent to the average temperature of the internal components and SLIMs.
If the MSA is not well ventillated, the internal air temperature will stagnate, allowing the temperature of the individual SLIMs to equalize to each other. However, internal temperature stabilization will take a very long time and the difference between outside ambient temperature and internal stabilization temperature will be quite large.
    If the MSA is well ventillated, the
internal temperature will stabilize in a shorter period of time and be only a few degrees higher than the ambient temperature.  The Verification MSA has a small muffin fan, and the inside air temperature (exhaust temperature) is about 6 degrees (F) higher than the outside ambient temperature.
    I characterized the Verification MSA by testing Stabilization Time, Frequency variation, and Magnitude/Phase variation (VNA) vs. temperature.

Stabilization vs. temperature (and time)
The purpose of this test is to determine how long it takes for the MSA to "warm up" before measurements will be valid.  It also characterizes the Verification MSA for changes in Magnitude and Phase vs. Temperature (both Ambient and internal) from turn-on to stabilization. The following sweep took 66.7 minutes.  The Tracking Generator output is connected to the MSA input through two 10 dB attenuators with 6 inches of RG-188 coaxial cable.  The MSA is commanded to 12 MHz (in the VNA-Transmission Mode), but any frequency or power level could be used. "Ambient" is the air temperature entering the MSA (room air). "Exhaust" is the air temperature exiting the MSA.
It takes about 40 minutes from turn-on to stabilization (from first data point to Marker 5)
    Time   Ambient   Exhaust   Notes    See Graph for actual Magnitude and Phase.
L            0 min     81.5 F      81.5 F
R            9            81.6         86.9
1            11           81.6         87.2
2            18           81.8         87.6
3            25           82.0         87.9
4            28           82.2         88.1
5            40           82.2         87.9       Stabilization
6            51           82.2         87.9
The ambient temperature is rising because I turned off the home air conditioner to prevent sudden temperature changes in the ambient surroundings.  After stabilization, the Ambient-Exhaust temperature differential is 5.7 degrees F.

The Magnitude shows very little change from start to stabilization, Marker 2, 18 minutes. Phase has stabilized by Marker 5, 40 minutes. If the user is satisfied with a Phase error of 1 degree (
Marker 2, 18 minutes) he could begin taking data after 18 minutes of warm up.

The following is an example of the Verification MSA with poor or no ventillation.  It is a continuation of the previous sweep, except I partially blocked the entrance and exhaust holes from time 0 to Marker 1.  If allowed to continue, I believe the Phase would stabilize at about 22.5 degrees in another 30 minutes. This was enough test time to tell me that poor ventillation is unacceptable.
From Marker 1, the holes were totally blocked, to simulate an MSA with no ventillation. In both sections of the sweep, the internal muffin fan is circulating the internal air.  T
he ambient temperature was rising since the home air conditioner was off, but then it started raining outside and the ambient temperature began to decrease.
Time   Ambient   Exhaust    Phase    Marker   Notes
0 min    83.1F       88.7 F   17.68
3           83.8         89.0      17.27
8           84.3         93.0      17.91
      L        Magnitude is stable
10         84.5         90.6      18.50
15         84.9         91.2      19.84
22         85.2         91.2      20.87
27         85.2         91.2      21.34
29         85.1         92.1      21.60
32         85.2         92.8      21.76       1         Cover vents, the "exhaust" temp is internal stale air temp
35         84.9         93.5      22.65        
         Started raining outside, ambient temp is decreasing
40         84.3         94.4      23.90       R
44         84.2         94.8      24.76
53         83.8         95.1      26.19
62         83.8         95.3      27.09
64         83.8         95.7      27.28
67         83.8         95.7      27.46
120       84.0         96.9      28.94        NA    Final stabilization, not shown on this graph. See next graph.
  Notice that the ending Ambient-Internal temperature differential is 12.9 degrees F. This is over twice the differential of normal operation.  Also notice, the Magnitude has changed only +.05 dB.  This is excellent Magnitude stability for such a large internal temperature change.

The following will test to see how long it takes for the MSA to stabilize from a very high internal temperature. After the previous test, I removed the blockages from the vents and allowed the MSA to stabilize with normal air flow (muffin fan).
Time   Ambient   Exhaust    Phase    Marker   Notes
0 min    84.0 F     96.9 F    28.94                  This is the final stabilization from the previous test.
1          84.0        96.9        28.91        L       
Removed the blockages from the vents
17        84.0        90.1        21.02        R
27                                     19.73        1         Phase is well within 1 degree of final stabilization
40                                     19.43        2         Stabilization
67         84.2       89.7        19.42        3
Notice that the final data point (Marker 3) temperatures are about 2 degrees higher than the first graph. This is due to the room temperature changing over a period of 2 hours. However, the Ambient-Exhaust temperature differential has not changed much (5.5 degrees F).

Magnitude/Phase variation (VNA) vs. Temperature
    We can use the previously taken data to characterize the Magnitude and Phase variations vs. Temperature.  Here is data taken two hours apart, with a small temperature change:
Exhaust    Phase        Magnitude    Notes
87.9 F     16.63 deg   -30.59 dBm
89.7 F     19.42 deg   -30.60 dBm    Data after 2 hours
+1.8        +2.79          -   .01 dB       Delta changes
Exhaust temperature change = 1.8 deg F; average phase change = 2.79 degrees
Phase vs. Temp Coefficient = 2.79/1.8 = 1.55 deg / 1 deg F
This would suggest that the Verification VNA phase could be expected to drift +1.55 degrees for every +1 degree F temperature change.  This would seem to be a significant drift, but any phase or magnitude drift can be "calibrated out" when the VNA is "line calibrated" before a critical data measurement.  Measurements usually take less than a minute, and significant temperature changes inside the MSA are unlikely to occur in such a short period of time (unless, of course, your wife turns up the furnace 10 degrees and the outlet is blowing directly on your MSA).
The Magnitude vs. Temperature coefficient is negligible. It is less than .01 dB / 1 deg F.

An interesting set of data was taken with
the Final Xtal Filter (Resolution Filter) removed and replaced with a short coax.  Using the MSA's exhaust air temperature as a reference, the phase is measured at two different temperatures:
At 89.9 deg F, the phase measured 163.91 degrees;
 at 92
.9 deg F, the phase measured 167.78 degrees
This is
Phase vs. Temp Coefficient = 3.87 deg/3.0 deg F = 1.29 deg / 1 deg F
Since the Final Xtal Filter is removed and replaced with a short piece of coax cable, I expected the coaxial filter phase change would have been the only contributor to the MSA phase change (any phase change caused by a frequency change is insignificant to total phase). Subtracting the
1.0 deg / 1 deg F of the cavity filter from the total MSA of 1.29 deg / 1 deg F would be 0.29 deg / 1 deg F.  This has to be the contribution of components other than the coaxial cavity filter and the final crystal filter. It is probably the phase contributions of the Log Detector, and the I.F. Amplifier.
Updateded 7-22-10  I performed a test on the Original MSA, since removing the Coaxial Cavity Filter in the Verification MSA is inconvenient. I tested the Phase change vs. Temperature of the Original MSA with both the Cavity Filter and Final Crystal Filter removed (and replaced with short coaxes). It showed a positive Phase vs. Temp Coefficient: .2 deg / 1 deg F.  This is quite close to the calculated drift of the Verification MSA (.29 deg / 1 deg F), with its cavity filter coefficient removed.

Frequency Variation vs. Temperature
    The most obvious way to characterize the MSA's Frequency/Temperature coefficient is to measure the frequency of the Master Oscillator at two different temperatures and calculate it. However, with the MSA completed and covers in place it is not possible to access these frequency points inside the MSA.  Since the spare output of DDS 1 is accessable from the outside of the MSA, I measured it with a frequency counter that is accurate to .1 Hz.  It is extremely stable with ambient temperature changes.  The Verification MSA has been previously calibrated to WWV. The Master Oscillator was exactly 63.999560 MHz at an ambient temperature I don't remember. In this test, I am not interested in the exact frequency of the Master Oscillator, I am only interested in how much it changes over temperature.
Using the Special Tests Menu, I commanded DDS 1 for a frequency of 16.0 MHz, with DDS Clock at
63.999560. The following are the data:
At 88.7 deg F the DDS 1 output measured 16,000,004.1 Hz. This is .256 parts per million high.
At 83.0 deg F the DDS 1 output measured 15,999,986.4 Hz. This is .850 parts per million high.
The Frequency vs. Temperature coefficient is = 1.106 ppm/5.7 deg F = .194 ppm/1 deg F. This states that the integrated Master Oscillator changes .194 ppm for every degree F of exhaust temperature (average inside temperature).  This seems to conflict with data we already have for the Master Oscillator Slim (.15 ppm/1 deg F). Actually, it just means that the Master Oscillator changes temperature more than the average temperature change of the MSA.
    Even though there is no temperature probe directly on the Master Oscillator, we can make a calculation as to what the temperature changed at the Master Oscillator SLIM.
As an individual module, the Master Oscillator has a Freq/Temp coefficient of .15 ppm/1 deg F.
The test showed that the MSA temperature variation of 1 degree F results in a Master Oscillator deviation of .194 ppm.  Comparing .194 ppm/.15 ppm = 1.293. Therefore the Master Oscillator temperature changed 1.293 times the average change of the MSA temperature (exhaust temp).
The Master Oscillator temperature changed:  1.293 x 5.7 deg F = 7.37 deg F

Thermometer used for temperature testing
    I don't have any thermocouples or thermal probes so I fashoned a system to make these Temperature Tests. I thought it was worth publishing, as it was a very easy and can be copied by anyone.
    I modified a commercial Indoor/Outdoor Digital Wireless Thermometer (La Crosse). Unknown price since it was a Christmas gift, but it can't be too expensive. Check Wal-Mart for similar items. This one will also read in degrees C. It seems to be accurate and reads to the tenth of a degree.
msapictures/therm1.gif msapictures/therm2.gif
    The (In)side "Master" unit is simple to disassemble (4 small screws on back). It has its thermister mounted on the side of its main pwb. I just clipped the leads and added two twisted wires to exit the side of the unit's vent holes. The wire is 6 inches, but can be any length. I installed some heat shrink tubing at the thermister to prevent stress on the leads. This is what I used to measure Exhaust Temperature.
    The (Out)side "Remote" unit is a transmitter (915 MHz) that squirts out its data every 10 seconds. It has very slow thermal response, so I cut the plastic area covering its thermister (the little do-dad below the marked surface mount resistor).
  This improved
thermal response time but since the themister is mounted directly on a pwb, the whole pwb must stabilize in temperature before the reading is accurate (about 5 minutes). At a future date, I will remove the surface mounted thermister and use extension wires similar to the Master unit modification. I used this "Remote" unit for the Ambient Temperature measurements.

Comments on Compensation. Added 8-09-10.
Let me add a few cents on the subject of temperature compensation for Frequency and one cent for Phase.
There are three components in the MSA/VNA that are susceptable to temperature changes:
Master Oscillator - Frequency Change vs. Temp
Coaxial Cavity Filter - Resonant Frequency Change vs. Temp (phase)
Final Resolution Filter - Resonant Frequency Change vs. Temp (phase)
All three have such a minor change in Magnitude vs. Temp, it is not even worth discussing (unless the Resolution Filter is extremely narrow).

The Master Oscillator
 My original MSA concept did not have the VNA capability in mind. For Spectrum Analyzer and Tracking Generator operation, a minor amount of the Master Oscillator's Frequency drift is noticeable only if the Final Resolution Filter is extremely narrow (less than a few hundred Hz.). Of course, if the Tracking Generator is used as a Frequency Source, its frequency is dependent on the accuracy of the Master Oscillator. With a Master Oscillator stability of .15 ppm/1 deg F (9.6 Hz/1 deg F, as measured in the Verification MSA), this equates to a TG error of about 150 Hz / deg F.
 The Master Oscillator frequency of 64 MHz was chosen because this is a "standard" that is available from multiple sources. The present oscillator (Cash chose this one for price, availability, and stability) is a good one. With a little redesign of the SLIM-MO-64, it could be temperature compensated to much better than its .15 ppm/1 deg F.
 For those who wish to use a precision frequency source as a replacement to this oscillator, these are the requirements:
A. Minimal requirement - 6 MHz to 100 MHz in 2 MHz increments (6,8,32,34,etc)
B. Nominal requirement - 12 MHz to 100 MHz in 4 MHz increments
1. For oscillator frequencies less than 22 MHz, the DDS's X4 multiplier can be utilized. The maximum division ratio in the DDS must be less than 2, due to Nyquist. Since the output of the DDS is appx. 10.7 MHz, its minimum clock must be greater than 21.4 MHz. Utilizing its X4 Multiplier, the minimum clock input must be greater than 21.4/4 or 5.35 MHz.
2. PLO 2 uses the LMX 2326 pll and its minimum clock divider ratio is 3. Its phase/frequency detector should operate at a minimum of 2 MHz (4 MHz is optimal). Therefore, its minimum clock input (Master Oscillator) must be 6 MHz (12 MHz is preferred to obtain the 4 MHz PDF for PLO 2).
3. The maximum frequency for the DDS is 128 MHz, but the maximum for the LMX 2326 is 100 MHz. Therefore, the Master Oscillator can not be greater than 100 MHz unless it is divided down.
C. Optional Requirement. Requirements A and B assume the MSA's 2nd LO is 1024 MHz. However, this is not absolutely necessary. Any frequency between 1010 MHz and 1030 MHz could be used. It is possible to use almost any oscillator frequency between 6 and 100 MHz, but changes to the MSA software would need to be performed.

The Filters
For VNA operation, the Cavity Filter and Final Resolution Filters become the dominant phase drift factors for the MSA. If I had designed an MSA for utilization as a VNA only, I would not have used a Cavity Filter or "Resolution Filter". They are not necessary for VNA operation. In fact, the final filter would not be called a "Resolution Filter". It would just be called a noise limiting filter. A very simple single pole filter could replace the Cavity Filter, so could a resistive attenuator. Either would exhibit little or no phase drift over temperature. Same goes for a wide band final noise filter.

To make a long story short, if you are extremely critical to Spectrum Analyzer frequency measurements, use a precision Master Oscillator. If you are critical for phase measurements in VNA mode, use a precision MO and modify the filter requirements for the First IF and Final IF.