N2PK VNA - Some initial plots

Quick sweep (3 to 29 MHz) of my G5RV clearly shows the low SWR points. It also reminds me that I need to do some minor tuneup of the antenna. The red vertical cursor is at the CW portion of the 20m band - the other SWR dips are in agreement with what should one expect from the classic G5RV. A VNA makes the ultimate antenna analyzer! One extremely valuable feature is the ability to perform calibration at the end of the coax / antenna feed point thus removing the impedance transformation effect of the coax from the measurement! It is like having the VNA connected right at the antenna feed point. (A good overview of various antenna analyzers and designs (incl. N2PK) can be found here )

The noise floor of Detector 2 (the one I am planning to use for transmission measurements) is below -125dB (at the slowest ADC speed) and flat. The plot for Detector 1 looks the same! This is a pretty good range for home-brewed equipment!

N2PK - Final Assmebly

Finally, about two months after I started working on it (with the "partial support" of my patient wife :) - the finished VNA is waiting for its aluminum cover. Really glad, that the build is over - it was a fun project but I am not sure if I am going to build a second one.
I am very satisfied with the results - a very precise, lab grade instrument in a professionally looking package - one can hardly tell that it is not a "commercial product".
The initial testing went just fine as expected. I spent a lot of time, making sure that everything is done right, inspecting every single solder joint and checking every component so I was not expecting any major problems. Everything worked the first time I plugged it in!

A note to myself: Next time I decide to build something with 7 bulkhead BNC connectors on the front panel - get a Greenlee D-hole punch. Shaping that many D-shaped holes with a regular drill and a file was tedious work.

My method of using Clear ink-jet mailing labels and 3M laminating sheets resulted in a superb quality panel finish. The colors on the panels are just a hint how much fun is to use the instrument :-)

The last touch before closing the unit was to attach a spare fuse to the PS board :) (hopefully I won't need it)
It is a pretty tight install - once everything is in place and connected, there is not much space wasted in the Hammond enclosure.

2010 Update: This picture shows the final version, using semi-rigid coax (RG-405) for Detector inputs and RF DDS out. The coax is in black heat-shrink tubing - just did not like the idea of exposed conductors crossing above the VNA PCB. The control/power cables were re-organized too and an accessory connector was installed.

The Hammond enclosure comes with two elegant bezels for the front and rear panels. When the bezels are installed, they stand between the enclosure and the panels. Under "normal" circumstances (DC/low frequency applications) this is OK. Since these bezels are plastic, the face-plates are electrically connected to the enclosure only with the 4 corner screws. In the RF world things are a little different. The electrical gap which occurs because of the plastic trim can cause signal leakage to and from the VNA. It degrades the overall RF shielding of the aluminum enclosure and affects the impedance-to-ground of the front panel connectors. I solved this issue with 6mm strips of self-adhesive copper foil - I wraped the 3 surfaces of the inside plastic edge with copper foil, covering all 4 sides of the bezel. When the plastic bezel is sandwiched between the panels and the enclosure, the copper foil serves as electrical bridge between the aluminum face-plates and the enclosure, while creating continuous RF shield on the inside.

This is the complete set - VNA, Reflection bridge and OSL calibration standards.
Since the calibration standards determine the accuracy of the measurements, I paid extra attention while building them. For the 50 ohm load I used small (0603) thin-film resistor - optimized for high frequency use (up to 20GHz) (Digikey P/N FC0603-50BWCT-ND or FC0603-50BFCT-ND). The construction of the OPEN and SHORT standards was also done very carefully. I am planing to make 2 more sets of calibration loads - N-type, SMA and UHF loads. This way I can use adaptors BNC-to-"X" and calibrate after the adaptor or at the end of the jumper cables regardless of their type and without using more adaptors.
The last thing left to do is to organize everything in a Pelican 1450 protective case.
Hmm...now thinking about this RF-IV sensor ... it doesn't look like the build is really over :-))

Feb 2011 Update - My complete VNA kit fits nicely in a Pelican 1500. This case is water-tight, very durable and makes it convenient to store and transport the VNA with all of it's accessories.

N2PK VNA Main board RF shielding

I fabricated the custom DDS ICs heatsink out of a used old Pentium CPU heatsink (the green heatsink on the picture). Using a Dremel tool, I shaped the heatsink to fit in between the nearby taller components (ICs, inductors and a transformer). It was a tricky job! The two holes on the PCB provided for heatsink mounting are hardly in their optimal location. Because the holes are located only to one side of the DDS chips, it was very tricky to mount it in a way where the heatsink is pressing on both chips with equal force without tilting to the side of the screws. I used some washers as spacers to achieve this. A second heatsink (the black square heatsink) was attached to the side of the main DDS heatsink for cooling the Valpey-Fisher Master Oscillator. Two smaller heatsinks were attached to the ADC chips using self-adhesive heatsink interface.

I made the RF shield cans out of tin-plated brass sheet. This material is very easy to work with and solder. Due to the high density of components and the really narrow solder pads for the shield, a great level of precision is needed while fabricating the RF shields. There is no need to solder completely the shield along its whole length - just a few solder points per side is sufficient. The extra heat used during a complete soldering is unnecessary and dangerous to the components. In addition, if the shield ever needs to come off it will be much easier that way.

This is the finished board ready to be installed in the enclosure. The lids of the two RF screening cans are attached with self-adhesive copper tape (with conductive adhesive). Again, this method provides sufficient electrical connection and allows for an easy access to the the detector circuits if ever needed.

N2PK VNA Power supply module

I mounted the power supply module onto a piece of double-sided FR4 board. The board is exactly the same width as the main VNA PCB and just slides in the lowest channel of the Hammond enclosure - the same way the main VNA PCB is mounted.

This makes up for a very nice mounting solution with no drilling and no screws on the bottom of the enclosure - I really wanted to keep the enclosure clean and free of unsightly screws on the outside. The actual PS PCB is screened with a tin-plated brass RF shield (soldered to the top copper layer of the FR4 board). Copper tape strips on each side of the module are used to improve the electrical connection between the aluminum enclosure and the two copper planes of the FR4 board, once it is inserted in its channel. The lid of the RF shield is attached with self-adhesive copper tape (with conductive adhesive). This allows for an easy removal of the lid should a fuse change is necessary. If needed, the whole power module can be completely removed for servicing just by detaching the connectors and sliding it out of the channel.

The PCB is bolted to the FR4 board inside the shield using small brass stand-offs/spacers (there are components on the bottom side of the PS PCB and some clearance is needed between the solder joints/components and the top copper plane).

The bottom side of the FR4 plate. I decided to move the +12V linear voltage regulator from the bottom side of the PS PCB to the bottom side of the FR4 mounting plate. When the plate is inserted in the very bottom channel of the enclosure, there is just enough space (aprox. 5mm) for the linear regulator to fit in. This, actually turned out to be a pretty good cooling solution. On the right side of the board are visible some current limiting resistors for the two LEDs and the VNA Detect circuit as well as by-pass and filter caps. There is also a second LDO voltage regulator for +9V line (along with some filter caps). The +9V line is wired to the Accessory connector on the back, powering Transverters, RF-IV sensor or S-parameter test set.

The linear regulators are using the bottom side of the aluminum enclosure as a giant heatsink. Some thermal grease and a small copper insert (shim) ensures the good mechanical contact between the aluminium wall and the IC. The bottom copper layer of the FR4 board serves as a secondary heatsink - the regulator is mounted with its metal tab facing down and it is "sandwiched" between the FR4 mounting plate and the bottom wall of the enclosure. I used thermal double-sided self-adhesive tape to attach the regulator to the FR4 board (to the copper shim actually, the shim is soldered to the FR4 along one of its edges so it can flex). Soldered to the board is a little brass stub that goes into the regulator's mounting hole. This provides extra mechanical stability when the board is installed/removed.

N2PK VNA Reflection brdge

The reflection bridge was fairly easy to build. Most of the work was mechanical. The two male bulkhead BNC were very difficult thing to find. I could not find a local source so I had to buy them from Hong Kong (tnx eBay!) . All RF connectors are attached to the aluminum enclosure using stainless steel hardware.

Great deal of precision is needed when drilling the holes for the male bulkhead BNC connectors. There is not much mechanical "play" in these connectors and they need to match perfectly the female BNCs on the front panel of the VNA. Both connectors are exactly 2 inches apart. This picture also shows the two L-shape brackets connecting the bolts of the female BNC to the ground plane of the PCB. The brackets are soldered to the ground plane on both sides of the center pin of the female connector.

Same L-shaped brackets for the male BNCs. I had to remove some of the solder mask in order to make better electrical connection. Good ground connection is essential! The male BNCs have their own ground pins on the back side (a nice feature) - also soldered to the ground plane (visible at the very left and right of the board).

The finished reflection bridge. This side is "up" when the bridge is fitted onto the Detector 1 port - the normal position for the bridge during reflection measurements.

The reverse side of the bridge. This side is "up" when the bridge is fitted onto the Detector 2 port (optional).

N2PK VNA Rear panel

Cutting out the rectangular openings for the switch and the DB25 connector was the real challenge here. The left side of the panel is dedicated to graphics - there is not much room behind it because of the power supply board. I left some empty space on the panel for the future installation of a small Accessory connector so I can connect different VNA accessories (like the RF-IV test head or Transverters). The power switch has a built-in LED connected to the regulated +12V line. The front panel's LED is connected to the +5V line (DDS power).

The rear panel is completely detachable. This makes life easier if I need to open and disassemble the unit. Ferrite chokes are installed on the control / data lines and power supply input. The little 4 pin connector supplies power to the +12V LED, VNA Detect line and +9V to the Accessory connector.

N2PK VNA - Power supply PCB and front panel

I received the PCBs for the PS board from the board fab house (Sunstone circuits) and I must say that the fabrication quality is pretty good. The schematics are based on the N2PK/OM3LZ power supply board with very minimal changes on my part. Basicly, there are two voltage regulators: +12V Low-Dropout linear type and +5V switching regulator, LC noise filters for each regulator and two "crowbar" type over-voltage protection circuits combined with over-current and reverse-polarity protection. In other words - the power conditioning is very solid and provides a good protection for the VNA board from a number of potentially dangerous events.
There is an additional +9V regulator circuitry (installed on the bottom of the PS support plate) for the Accessory connector - it is used to power transverters, S-parameter test fixture or the RF-IV sensor.
The PCB layout design was done with Cadsoft Eagle. I started with the OM3LZ board as a reference but at the end I changed the layout a bit. These pictures are of my ver. 1.0 board. The final version of the PS PCB is ver. 2.0 and that one is even more compact and with a smaller footprint. Not sure if the 2.0 board will ever see daylight since I am all set with my PS needs for now.

The PS is done with mixed thru-hole/SMD technology - first the SMDs are installed and then the rest of the components. I added pads for an SMD LED on the +5V line and a current limiting resistor for it. This LED serves as a reminder that the unit is powered when operating with the covers off and it is totally optional to install.

On the bottom side I have a space for an optional LDO 12V linear regulator (Digikey p/n 576-2206-ND (Mircrel MIC29150-12WT)) . This allows for an external power supply with a wide range of voltages - typically filtered DC 13V to 24V. Because of the extremly low drop-out voltage ( 0.35V @ 1.5A) of this chip, the VNA can be powered by a standard 13.8V PS and still be in regulation. For portable use I'll probably power the unit from a 2 x 7.2V Li-Ion battery packs. This regulator can be omitted and bypassed with a jumper but then extra attention needs to be paid to the input voltage in order to prevent the over-voltage protection from triggering (at the price of a blown fuse). There is a couple of extra bypass capacitors associated with this regulator.

Unfortunately, it wasn't cost-effective to put a silk-screen for the SMD components on the bottom but I don't think it is a big issue.

The front panel of the VNA. First step is to drill all holes in the panel. The graphics for the panel are actually a "sandwich" of two layers - printed layer and protective layer. I used Corel Draw to design the graphics layout (any other vector graphics software like Adobe Illustrator will work too). The layout is then printed with a Color Ink Jet printer (printer driver: best quality, transparency) on a sheet of "Avery Clear Full-Sheet Labels" (Avery 8665 or 18665 (better) from OfficeMax/Staples). The aluminum panels must be cleaned, de-greased with alcohol and dry. During application of the printed layer, one should be careful not to smudge the printing or leave fingerprints and must try to prevent any air bubbles from forming at the same time. (I used a piece of the base paper from the label stock (the glossy, waxy side as an "applicator", rubbing the sheet while applying it to the surface). The graphics must be carefully aligned with the holes on the panel during application. Then, the print layer is protected with a second layer of a durable clear self-adhesive plastic sheet - 3M Scotch Laminating Sheets (LS854- 10M or -10G (the last number shows how many in a package, M for Matte, G - for Gloss). After applying the first (printed) layer, compressed air and soft brush were used to remove any dust particles, then applying the protective layer is done the same way - applied slowly while watching out for air pockets . (Remember - it is a "one shot" deal - if something goes wrong during the application of the laminating layer there is no going back - you have to start over with new print layer).

The resulting surface is smooth, dirt and scratch-resistant and because the printed layer is transparent, the front panel has almost the same brushed aluminum/metallic look as the enclosure. Instead of using transparent print layer, a solid-color stock can be used too, but IMHO it looks "flat" and not as attractive as the natural metallic look. Once the two layers were "sandwiched" and pressed well together, I use scalpel blade to carefully cut out the openings and the excess around the edges.

The finished panels came out very nice and professionally looking - practically "commercial product" grade. I am really happy with the results - I think this will be my method for printing front panels from now on.

While looking for a front panel layout solution, I came across an interesting product - Ink Jet printable laptop skin (sold in Office Max). It is a white, self-adhesive vinyl sheet and I think it could be used for front panel labeling as well but I really wanted to preserve the aluminum finish look so I opted for the see-through label sheets. Another possibility is to use one of the products by http://www.texascraft.com/

The front panel is installed on the enclosure along with the BNC connectors and LO jumpers. I decided to color code the connectors because of their number on the front panel. It will be easier to work with the VNA and keep track of all connections.

All of the RF interconnects are done! Initially, I was going to use microwave semi-rigid coax as it provides the best shielding and phase stability but this stuff is too exotic (read: difficult to find/install/work with/using specialized connectors) so I opted for a special mil/aerospace version of the RG-316 by Semflex called SI316. The regular RG-316 is double-shielded with two silver-plated round braids. The SI316 is the same silver/teflon coax but it is triple shielded - it has metalized kapton foil layer between the outer round braid and the inner flat braid. This results in lower attenuation and much better shielding (aprox. 35 db better or >90 dB) than the regular "plain-vanilla" RG316 - the only thing better then this cable would be to use semi-rigid coax (shielding >110 dB).

Next item on the list is the wiring harness and the power supply board mount.

Update: I made a set of bulkhead (f) BNC to (m) SMA internal RF interconnects, using semi-rigid hand-formable RG-405 coax. I left the old SI316 cables for the LO DDS and replaced only the ones connecting both detector inputs and the DDS RF OUT to the front panel. I did not observe any better detector noise figures.