Cloning the HP35676A Reflection Transmission Test Set for the 3577A/B VNA

[Jay_Diddy_B]

Hi,

In the testing of the a CM/DM Noise Separator I designed, I needed to test a frequencies lower than the HP34677A S-parameter Test Set. The S-parameter is designed to work down 100kHz.

I found that HP also produced the 35676A Reflection Transmission Test Set. The R/T Test Set will work down to 5Hz.

You can buy the R/T Test Set on the usual places for about $300-500.

Note:

The 'B' Version is not an improvement on the A version, the A is 50 and the B is 75 .

Information Search

I found a nice summary of the 35676A:

Link: http://www.hparchive.com/Manuals/HP-35676-SCHEMATIC.pdf

I will attach a copy of the document to the end of this message.

I looks to be a pretty simple circuit, just 7 resistors and four connectors. Most of the challenges are in mechanical construction.

Schematic

I entered the schematic into KICAD. The test points are connected to the chassis mounted connectors.







PCB Design




The coplanar Wave Guides are 50 .

Complete Project



Here is a picture showing the circuit board mounted inside a Hammond 1550P Diecast box.

Mechanical Design


The mechanical design was done using Fusion 360. I started with a simplified drawing of the HP3577A front panel:



I downloaded the STEP file for the Hammond 1550P box and added the holes for the connectors and the PCB:



Here is a picture showing the assembly:




and the project  installed on the HP3577A:



Testing

After installation I did the one port full calibration with Open, Short and Load.
At high frequencies the results agreed with my S-parameter Test Set. At lower frequencies the R/T Test Set produced the expected results.

I have attached the pdf file.

Regards,
Jay_Diddy_B

 kicad schematic.JPG (29.48 kB. 687x379 - viewed 727 times.)

[Jay_Diddy_B]

Hi,

This is a picture of an original one. They were supplied in a case with cables, a load and a short for calibration:



The screws that hold it together are under the top label.

The kit was about $1800 USD in 1989 about $3600.00 in today's money. This probably why they didn't want you to look inside 

Regards,
Jay_Diddy_B

[brob]

Hi Jay

Would you be able to post any DXF, or PCB layout?

I am looking at making the same test set for my 4395A.  Looks like you milled your PCB, I would also be looking at doing the same thing. I have not worked with controlled impedance PCBs before.

Any help would be appreciated

Thanks

[bson]

Very cool, I'd like one of these.  My 3577A is one of my most used instruments! What connectors did you use, and care to share the KiCad project?

[KE5FX]

I'll confess I don't understand how this circuit achieves any sort of directivity, much less 40 dB.  All I see are a couple of T attenuators at the reference and receiver ports, with a bypassed resistor between them.  If the test port is shorted, the A receiver channel gets nothing from the source.  If the test port is open, the A receiver gets everything   Where's the balanced-bridge element?

[Jay_Diddy_B]

Very cool, I'd like one of these.  My 3577A is one of my most used instruments! What connectors did you use, and care to share the KiCad project?

Hi,

I can't be 100% sure what the connectors are. The word you are looking for is 'extended' teflon. I bought them from eBay a few years ago.

These connectors from Amphenol should work. You may have to trim the extended dielectric.

Regards,
Jay_Diddy_B


 72224_24_customer_drawing-1279450.pdf (55.1 kB - downloaded 149 times.)
 72260_10_customer_drawing-1279511.pdf (62.92 kB - downloaded 111 times.)

[Jay_Diddy_B]

I'll confess I don't understand how this circuit achieves any sort of directivity, much less 40 dB.  All I see are a couple of T attenuators at the reference and receiver ports, with a bypassed resistor between them.  If the test port is shorted, the A receiver channel gets nothing from the source.  If the test port is open, the A receiver gets everything   Where's the balanced-bridge element?

Yes, it is pretty hard to understand how this works so well for a simple circuit.
The HP 87512A/B are similar devices. The 'A' is 50 and the 'B' is 75 .

Link: http://www.miedema.dyndns.org/co/2016/4395a/trans-reflect/HP-87512B-tear-down-schema-en-manual.pdf

A lot of the 'magic' happens in the VNA calibration routine.

I can testify that they work relatively well, although not as good as the s-parameter test set. They work at lower frequencies than the s-parameter test sets.

Regards,
Jay_Diddy_B

[KE5FX]

It must be a variation on what Joel Dunsmore's book calls a 1+gamma reflectometer, except that the signal at the b1 receiver (labeled 'A' here) actually is allowed to go to zero when the DUT is shorted. 

This sort of topology requires near-perfect stability since directivity errors are magnified immensely.  Probably a safe-enough assumption at VHF and below, but I'm still surprised they didn't go with an active bridge of some sort (to incorporate another thread by reference).

[Jay_Diddy_B]

Hi,

These measurement will give you an idea of how well this works. I set the VNA to measure S11 (reflection) and then I performed a full port calibration with open, short and load.

I then measured some attenuators.

First a 2x (6dB) attenuator. The reflection should be 12dB. 6dB attenuation, reflection and then 6dB attenuation.



Now a 5x (14dB) attenuator. The answer should be 28dB.



a 10x (20db) attenuator. The answer should be 40dB.



and finally a 50  a different one than the one I used for the calibration.



These measurements were take with the clone of the HP35676A R/T test set.

Regards,
Jay_Diddy_B

[Jay_Diddy_B]

Hi Jay

Would you be able to post any DXF, or PCB layout?

I am looking at making the same test set for my 4395A.  Looks like you milled your PCB, I would also be looking at doing the same thing. I have not worked with controlled impedance PCBs before.

Any help would be appreciated

Thanks

Hi group,

I have been working with forum member brob on this issue.

The R/T test set for the HP 4395A was HP 87512A or 87512B. The 'B' versions are 75 .

There is is a subtle difference between the 4395A and the 3577A Network analyzer.

On the 3577A the spacing between the R input and the A input is 1.700 inches.

On the 4395A the spacing between the R input and the A input is 1.600 inches.

I have made a second artwork that is for the 1.600 inch spacing.

Regards,
Jay_Diddy_B

[bson]

I think it's just a 2-way coupling bridge.  When the 2-port input impedance is 50Ω the receiver sees the calibrated 1x signal level.  When it's less, the coupled signal drops.  When it's more, it increases.  So it operates on measuring loss rather than reflection per se.  Since the main use of S11/S22 is to measure input and output impedances (and, yes, there are impedances at low frequencies as well, and being able to sweep and measure them both by frequency and amplitude is useful) it doesn't really matter how they're derived.

[RomDump]

I have made a second artwork that is for the 1.600 inch spacing.

Regards,
Jay_Diddy_B

Wondering if you can post the PCB Artwork?

Thanks.