Thursday, October 8, 2020

HP 85046B Conversion to 50-ohm Configuration using Low-cost Third-party Bridges (by WB0GAZ)

 A few months ago David Feldman, WB0GAZ, wrote to me regarding his project replacing the 75-ohm bridges in an HP 85046B S-Parameter Test Set with low-cost 50-ohm bridges purchased from eBay.  The topic intrigued me, and I offered to append his information to my own conversion article (http://k6jca.blogspot.com/2014/03/modifying-75-ohm-hp-85046b-s-parameter.html) when he had something available for me.

When I received his write-up, I realized it really deserved a blog post of its own (rather than appending it as an afterthought to mine).  And so...this is that post!

- Jeff, k6jca


****************************************

HP 85046B Conversion to 50-ohm Configuration using Low-cost Third-party Bridges 

By David Feldman, WB0GAZ 
wb0gaz@yahoo.com


The Opportunity:

A used HP 85046B 75-ohm test set was purchased this year, and upon arrival turned out to have been relieved of its transfer switch and step attenuator and had unrepairable damage to one of the two 75-ohm bridges.

Based on prior work by K6JCA (http://k6jca.blogspot.com/2014/03/modifying-75-ohm-hp-85046b-s-parameter.html), but without 50 ohm test set bridges in the junkbox, a decision was made to fit the test set with low-cost bridges sourced via eBay, replace the step attenuator with one from the junkbox, use a junkbox RF splitter/divider to provide a function normally supplied by one of the two HP bridges, replace the missing transfer switch with a low-cost mechanical version (HP 33311b), and accommodate the wiring changes with various surplus semi-rigid SMA jumpers from the junkbox.

Like most examples of this series test set, a solid state transfer switch permits the 8753-series analyzers (beginning with the 8753B with version 3.00 firmware) to display two different traces at (roughly) the same time by continuously switching the transfer switch to route the current signal of interest to the corresponding input port of the 8753 analyzer. The test set can (per description from K6JCA below) be configured for a (much lower-cost) mechanical relay switch by activating a firmware restriction on continuous switching. Near the end of this article is the result of an experiment to work around this firmware restriction.

No attempt was made to replicate the DC bias tee capability of the HP test set. This function could be implemented with external bias tee components if the need arises.

 

Phase 1: Replacement Splitter, Bridges and Mounting/Wiring:

HP 85046A/B test sets split the source signal into two paths, one sent to the analyzer R port (via 14 dB attenuation, a value that compensates for the nominal coupled-port loss of the type of asymmetric directional bridge used); the other sent through the test set’s step attenuator to the transfer relay and on to the two directional bridges. One of the HP bridges includes a co-located (but electrically isolated) 2-way resistive splitter (the 85047A test set uses a somewhat different arrangement with asymmetric resistive bridges that reduce source-to-DUT and DUT-to-receiver insertion losses by several dB.)

A pair of 3000 MHz bridges were sourced from eBay seller 60dbmco in Ukraine; these arrived in about three weeks. SMA port locations significantly differ from the stock HP bridges, necessitating significant changes to RF signal connections. These bridges were selected as they incorporate a metal housing.

Most of the existing SMA jumpers were removed from the test set and new ones formed from junkbox surplus cables. It appears the exact length of these jumpers is not very critical, however, one of the jumpers is quite long (providing time delay for the signal going back to the analyzer R port); this jumper was retained but somewhat rerouted to accommodate the replacement bridges.

From the junkbox, a 3-port resistive splitter provides the splitter function no longer available from the original HP bridges. Resistive splitters are simply three 16.7 ohm resistors in a WYE configuration. These have something over 6 dB insertion loss but work over very wide frequency ranges. Resistive splitters differ from other types of RF splitters in that they support wide frequency range (typically including DC) at the expense of higher insertion loss than other types of RF splitters.

A pair of surplus Nf-SMAf single-hole panel mount (bulkhead) adapters combined with  a pair of SMAm-SMAm adapters convert the two bridges’ test-port female SMA connectors to N connectors, and three SMAm-SMAf right-angle adapters provide access for cabling within the 85046B to three SMA test ports on the sides of the bridges.

The single-hole N female bulkhead adapters require a 5/8” ID mounting hole; the outer layer (the panel is 2 layer) of the test port openings on the HP 85046B is 7/8” ID; the inner layer of these openings is about 1-3/8” ID. Low-cost thin “aircraft” washers (5/8” ID, 1-1/8” OD, type AN960-160L) were sourced from eBay seller.

Owing to the 85046B's internal mechanical layout near the test port connectors, it is easier to pre-assemble the replacement bridge (with the test port adapters in place) as seen in Photo 2 (below) prior to installation in the test set. Carefully tighten the adapter connectors, and pay particular attention to tightening the 5/8" nut used to hold the bridge and adapters to the test set front panel. Two wrenches were used for this step - one holding the "flats" of the N female connector (visible in Photo 1 right hand connector) in place, while another was used to tighten the nut; this prevented inadvertently loosening the adapter interconnections leading to the bridge. No attempt was made to obscure the "75 ohm" nomenclature still visible in Photo 1 (below).

 

Step Attenuator

The step attenuator from the junkbox had different steps (1, 2, 4, 8, 16, 24, 24, 24 dB) compared with the removed HP attenuator (10, 20, 40 dB), and it was controlled with 5V TTL logic signals vs. 24V coils for the HP attenuator.    

The three original attenuator control signals (10, 20, and 40 dB) from the 85046B logic board were converted to appropriate logic levels and mapped to provide the desired attenuation of 0, 10, 20, 30, 40 and 50 dB using the junk-box step attenuator. The 60 and 70 dB attenuation settings are rendered as 68 and 78 dB respectively (the analyzer still displays 60 and 70 dB in the corresponding attenuator settings); a somewhat more involved logic translation would remove this limitation.

 

Phase 2: Bridge Replacement:

After the test set was in service for about a month, a decision was made to search for a bridge with improved directivity over the 3 GHz passband of the analyzer. One earlier-generation “2.5 GHz, 40 dB” bridge sourced a few years ago from a Chinese eBay seller was tested and found to have improved directivity above about 500 MHz vs. the Ukrainian bridge.

A question about the new bridge (vs. the 60dbmco bridges which were selected in part because of their metal housing, a trait they share with the original HP bridges, which internally have milled area separations and other high-cost manufacturing properties) was whether the new bridges, having a plastic housing, would be adversely affected when enclosed in the 85046B metal enclosure.

 The result is that there is some influence, however, once in the enclosure their directivity still appears to be an improvement over the 60dbmco bridges. This influence is likely to appear in terms of directivity degradation (if not severe, this should “calibrate out”), and crosstalk, leakage or “isolation,” which would impact dynamic range as an unwanted signal from the first coupler makes its way into the second coupler during a S12 or S21 measurement.

Isolation is also a concern in the transfer switch (the point where the single source signal is steered between the two bridges.) Most (but not all) of the HP 8753 calibration procedures will serve to compensate for isolation effects when two ports are in use. The method used to informally assess sensitivity of the plastic enclosure was to first calibrate the analyzer with the bridge external to the 85046B, then move the 85046B into position and measure behavior (with open, short and load) while retaining the baseline calibration state. This method was then repeated with a new pair of measurements to assess impact of the 85046B top and bottom cover (calibrate for baseline, add covers, measure behavior.) These tests can be seen in plots 4, 5, 6 and 7 (at the end of this post).

A second example of this type bridge was procured and, upon arrival, the two generations of the Chinese “2.5 GHz 40 dB” bridges were compared. The older generation bridge met its 40 dB dB directivity claim except for an area around 1.1 GHz (a peak around 32 dB directivity) and (as expected) in the 2.5-3.0 GHz range.  The newer generation bridge met its 40 dB directivity claim up to 2.5 GHz without the 1.1 GHz “bump”. The newer bridge was designated for the left 85046B port; the older bridge for the right port.

The Chinese “2.5 GHz 40 dB” bridges are slightly wider (in the long dimension) when compared with the packaged Ukraine bridge, however the difference is slight and as shown in the attached photos, the bridges and right-angle adapters fit OK.

  

The photos that follow show the reworked internals of the test set using the plastic-housed Chinese “2.5 GHz 40 dB” bridges that replaced the Ukrainian metal-housed 3 GHz bridges previously installed. An RF comparison of three bridge alternatives tried for this project is described later in this article.


Photo 1 -- 50 ohm N female test port connectors in place

 


Photo 2  -- The Chinese “2.5 GHz, 40 dB” bridge sourced from eBay, along with junkbox adapters and 5/8 x 1-1/8” thin aircraft washers. 



Photo 3
showing the Chinese “2.5 GHz, 40 dB” bridges in place (the 3 GHz metal-housed bridges were slightly smaller but mounted similarly) as well as the junkbox step attenuator a small logic board used to adapt the 3 TTL attenuator selection signals from the 85046B logic board to the steps available on this step attenuator (1, 2, 4, 8, 16, 24, 24, and 24 dB).



Photo 4
showing the resistive splitter used to split the source signal between the analyzer R port path and the switched test port path.


Continuous Switching with a Mechanical Switch: 

One limitation of the mechanical (vs. more common, but very costly and electrically fragile solid-state) switch in the 85046 series is a signal that is sent from the test set to the analyzer informing the analyzer that the switch is mechanical and that dual display configurations necessitating continuous operation of the switch are to be precluded by analyzer firmware, owing to concerns about mechanical switch lifetime. 

As K6JCA described in his 2014 blog-post on an 85046B retrofit, the logic board inside the 85046 (A or B) can be altered to send the “no continuous switching” indication by opening two easily accessed through-hole resistor/jumpers on a small PC board.  This was verified experimentally only with HP 8753B analyzer running 3.00 firmware (the first type and version of the HP 8753 series that enabled continuous updating of display for two different measurements.) 

During the Phase 2 bridge retrofit, the now-open resistor/jumper circuits cited in K6JCA’s original work were brought to the rear panel of the test set and terminated in a DPST switch (a push-button switch is shown in Photo 5 as that was the only type in the junkbox.) An unknown was how the HP analyzer (8753B with version 3.00 firmware in this case; other type analyzer was not available to repeat this experiment and may behave differently) would behave if presented with a false indication of switch type. 

A brief experiment confirmed that the analyzer will work with a continuously switching mechanical switch, leaving it up to the user to decide how long to let such a condition persist during a test session. 

  • Power-up or PRESET with switch in “no continuous” setting (both paths open) – analyzer behaves as if mechanical switch is in place and will not continuously switch if the test configuration needs this. Moving the switch to the “yes continuous” setting (both paths closed) after the analyzer has been powered up or PRESET does not change this result. 
  • Power-up or PRESET with switch in “yes continuous” setting (both paths closed) – analyzer behaves as if solid state switch is in place and will continuously switch if the test configuration needs this. The rate of switching appears to be about 2 cycles per second at the preset sweep setting of 201 points. 
  • Once powered up or PRESET with switch in “yes continuous” setting, then switch moved to “no continuous” setting allows continuous switching to proceed, however, the cycle rate is now about 1 cycle per second (this can be switched in this fashion during a test session.)  Moving the switch (after power-up or PRESET, provided it was at “yes continuous” at power-up or PRESET” can be used to set the cycle rate to 1 (no continuous) or 2 (yes continuous) cycles per second, again at the preset sweep setting of 201 points.

The switching cadence of 2 cycles per second is influenced by the number of sweep points and other settings that impact per-point measurement rate (preset points value is 201 on the HP 8753B with version 3.00 firmware; other settings related to filtering were in their respective preset values).  The switching cadence may differ on other models of the 8753 even with the same settings - this was not verified. 

Changing the sweep points to 3 (the minimum possible on the 8753B) resulted in a faster cadence (about 5 cycles per second with the modification switch in "yes continuous" setting). Changing the sweep points to 1601 (the maximum possible on the 8753B) resulted in a slower sweep cadence (less than one cycle per second). 

At all settings tried, there was no sign of impairment in the measurements due to the sweep starting before the relay transfer switch settled, however, this test was not exhaustive. 

A photo of the (currently push-button) switch added to the rear panel of the 85046B is shown in Photo 5.

Photo 5 – addition of switch allowing control over the firmware prohibition against continuous operation of a mechanical switch (operation verified in HP 8753B with version 3.00 firmware only.)


RF Comparison of Three Bridge Types: 

Three bridge types were compared (the original Ukraine packaged bridges, the Chinese “2.5 GHz 40 dB” bridges, and a simple resistive RF splitter (also sourced from China); the latter experiment was performed to see if a resistive RF splitter, which has no directional properties at all, could be imbued with some RF bridge properties when used alongside 1-port VNA calibration and error correction. 

RF measurement observations: 

  1. The Chinese “2.5 GHz 40 dB” bridges had improved directivity beyond 500 MHz vs. the Ukraine bridges (plot 1 vs. plot 3.)

  2. The conversion to N connector involves two adapters (SMAm-SMAm followed by Nf-SMAf bulkhead.) The various bridges tested were compared without and with this conversion from SMA to N connector.  Plots 1 and 2 compare the Ukrainian bridge without and with the N-connector conversion.  Plots 3 and 4 compare the Chinese “2.5 GHz 40 dB” without and with this conversion.  And plots 8 and 9 repeat this comparison using the resistive splitter as a bridge. Note that plot 2 has a different reference value (10 dB) vs. all of the other plots (0 dB).

  3. A few Nf-SMAf bulkhead adapters were obtained from an eBay seller in China; these were compared with the junkbox parts on hand, and while outwardly similar in appearance and fit, the return loss of the Chinese adapters was substantially inferior to those in the junkbox. The Chinese adapters were not put into service.

  4. The resistive splitter can be used as a bridge and does exhibit limited directional behavior when used with calibration and error correction enabled. This experiment suggests that six 16.7-ohm resistors (two RF dividers) could suffice as a T/R test set (such as HP 85044A) for the 8753 series (D and E, only where Option 011 is configured) with only the cost of a few RF adapters, connectors and cabling. Return loss measurements up to about 30 dB were performed with this configuration, although extensive evaluation was not completed as of this writing. See Plot 8 (without N adapters) and 9 (with).

  5. Prior to testing a resistive splitter in the 85046B experiment, a trivial test set was experimentally created from just two of the $3 eBay RF splitter modules and a few interface cables; with calibration it provided some usability for reflection and transmission measurements.

  6. The resistive splitter appears to have higher uncorrected S11 (open and terminated port cases) in plots 8 and 9 than the other two types of couplers; this is due to no attempt being made to balance the source-to-reference path insertion loss with the source-to-test path insertion loss.  Note that the normal HP couplers as well as the Ukraine and Chinese couplers have asymmetric insertion loss necessitating an internal 14 dB attenuator in the path from the test set to the analyzer R port. The resistive splitters are symmetric and would require different compensation. These effects do not impact the accuracy of measurements made with this experimental configuration.

  7. The plastic housing on the Chinese “2.5 GHz 40 dB” bridges was a cause of concern regarding RF leakage, owing to the close proximity to metal surfaces in the 85046B analyzer. To estimate this effect, the analyzer was calibrated (S11) with one of the bridges mounted near (but not in) the analyzer, then the bridge was moved into position in the analyzer, and S11 was measured with corrections retained from the previous calibration. This showed a loss of directivity (about 7 dB) between 1.5 and 1.8 GHz. (see plot 5.) 

  8. Once the Chinese “2.5 GHz 40 dB” bridges were in place, a similar test was run to see the effect of putting the top and bottom covers on the 85046B. The difference (seen in plot 7) is less than 1 dB (at around 2.7 GHz.)

  9. The RF leakage testing involving the plastic-housed bridges was cursory and could be explored further.

  10. Without RF absorber material placed around the bridges (which would also potentially impact directivity), there may be some cavity and other internal propagation (internal RF emission) effects that would degrade dynamic range for S21 and S12 measurements; these effects would contribute to isolation degradation (along with coupling in the transfer switch, solid state or mechanical) which is compensated for during calibration at the expense of some marginal impact on dynamic range. 


For these tests, high-quality loads (HP 909 series) were used for N and SMA terminated and uncorrected plots, as well as for N and SMA 1-port S11 calibration (along with suitable short and open devices) for corrected plots. 

Note:  The plots cited above follow the conclusion section below.

 

Conclusion: 

The modified 85046B remains in service and is now in use for RF learning, experimentation and similar hobby-level pursuits.


Plots:











End of WB0GAZ's Article

****************************************


If you have any questions or comments for David Feldman, WB0GAZ, please feel free to contact him at:

wb0gaz@yahoo.com


Caveat:

Although I edited WB0GAZ's text, I cannot guarantee that there are no errors by WB0GAZ or errors introduced by me.  If anything looks confusing or wrong to you, please feel free to comment below or send me an email.

Also, I will note:

This design and any associated information is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

No comments: