The information on this page is intended to support the assembled boards which came from the March 2004 group build.
In both the kit and assembled boxes, you will find a number of small plastic zip bags that are labeled with one or more 3-digit numbers. The number, the Inv. or inventory number, identifies the parts in the bag. The numbering scheme goes back to the first Group Build in January, and has not been changed, although a few part numbers have been added.
The part number is turned into a part specification through an Excel spreadsheet. With the first group build, I included a paper copy of the spreadsheet in the box. In order to reduce weight and shrink the box size, the spreadsheet is not included in this build. It can be downloaded from this link. If you cannot view the spreadsheet, the text is available in HTML by following this link.
This spreadsheet was created for the first Group Build. That group was 20 units. I was hoping that increasing the group size to 60 units would reduce the parts costs. Sadly, there was little change. This is because most of the expensive parts, such as the master oscillator and the DDS chips, did not have a price reduction in their relatively low quantities. The parts that did have a quantity price reduction were ones that were already inexpensive, such as some of resistors and capacitors that only cost a penny ($0.01 USD) or two each. So, I did not update the spreadsheet for this build, and the figures remain nearly unchanged. I did add parts, however.
I also attempted to add in some of the miscellaneous costs that are associated with kitting and shipping. I had to pay state sales tax for some of the parts, when the part was sold from a company which has an office in my state, Ohio. The other costs are for the boxes and bags, as well as the costs related to shipping parts to me in the first place. These are only estimates. I learned from the first group build that I tend to forget and underestimate these costs. So, with this total near $186, and then adding in a few more dollars for shipping and insurance, I believe that that $200 price for the kits is close, and should guarantee that I don't subsidize the group, which I'm trying hard to avoid.
When you open the box, you should find the following contents.
Cover letter and shipping documents, if international.
A static shielding bag containing the main VNA board.
A static shielding bag containing the T1-6T board.
The three female VNA board connectors, and the pins/contacts for the connectors. These are Inv. numbers 015, 017, 019, and 020.
Two, 100 Ohm 0.1 percent tolerance resistors, intended for a calibration standard load. The bag is labeled 039L.
The DB-25 male connector, with mounting hardware, part numbers 055 and 056.
One foot of ribbon cable, part number 057.
Two feet of RG-316 coaxial cable, representing part 058, located under the bottom foam.
Both boards, the T1-6T bridge, and the main VNA board, have been tested. The bridge boards were tested with a VNA manually assembled from this goup, and each VNA was tested with the gnat program.
The bridge boards were tested in the following way.
After the shunt DUT capacitors (Inv. 011) were soldered to the board, the signal line was checked for infinite DC resistance. Of the 50 bridge boards constructed, one capacitor did initially measure as a DC short. It was replaced.
After each pair of resistors (Inv. 039) were soldered to the board, the DC resistance of the pair was checked. Each pair is composed of two, 100 Ohm, 0.1% tolerance resistors, so their parallel combination should be very very very close to 50 Ohms. My 8000 count DVM read 50.1 Ohms for all pairs on all bridges, which is probably a sign of inaccuracy on it's part, since the resistors are very accurate.
After the transformer was soldered to the board, various DC resistance checks were made though the transformer, on both the input and output sides. The primary and secondary of the transformer are nearly shorts at DC, so once the transformer is on the board, several new DC test paths are available.
Each board was connected to a working VNA and a complex load, and the impedance of the load was measured from 1 to 50 MHz with 100 KHz steps. The test and results are described below.
The first three tests were made as the boards were assembled. The functional test was made in a large batch, after all of the boards were assembled.
The functional test complex load was a length of RG-174 coax terminated with a 150 Ohm resistor. The cable was approximately 25 feet long. This relatively simple setup creates a range of complex impedance values across a frequency sweep. RG-174 is a relatively lossy cable, and that loss will cause the resistance and reactance curves to taper towards the line impedance as the frequency rises. The resistance approaches 50 Ohms, and the reactance approaches 0 Ohms. This creates an easy to recognize graphic signature. Another way to say this is that the SWR on a perfect (no loss) 50 Ohm cable would be 3 (150/50). The impedance values on the input side of the line would take on values which lie along the SWR = 3 circle on a Smith Chart, as the line length was changed, or the frequency was changed. Due to loss, the values will not lie along the circle, but spiral towards the center (Z = 50 - j0 Ohms) as the line gets longer, or the frequency rises.
Each bridge was connected to the VNA with a jig designed to keep the lead length as short as possible, but was also based on flat clips to make it fast and easy to change boards. The VNA was swept in 100 KHz steps from 1 to 50 MHz, 490 total points. The resulting data was graphed and compared against my initial bridge, and all of the other bridges.
Here is a picture of the test setup, taken in the middle of testing.
Obviously this is not going to be the most accurate setup, with the board hanging out in the breeze, supported by clips. But, it does lead to an easy to change fixture, and it should be stable and consistent. A single OSL (open, short, load) calibration was done at the beginning of the session, and was used for all boards. Again, the focus was on repeatable results, that all of the boards produce similar results, and that those results are similar to a bridge that is has been used for several months, and is considered to function correctly.
Here is a graph of a typical sweep.
|T1-6T Bridge Board Test Sweep|
All 50 boards have very similar sweeps, and looked nearly identical on the graphs. A high degree of sameness is reasonable since all of the resistors on the board are 0.1% tolerance.
The data values for all of the boards were saved in files. I decided to do a quick statistical analysis of the variation between boards. Since I had 50 boards, each with 490 data points, each with resistance and reactance data, there were a lot of numbers to consider. I decided to deal with only 20 boards, picking the first 10 and last 10 from the test session. For each frequency, and for resistance and reactance, I computed the minimum, maximum, and standard deviation across the 20 boards. I then took the 490 resistance standard deviations, and 490 reactance standard deviations, and computed the standard deviations of those two sample sets.
For resistance, the deviation of the deviation was 0.89 Ohms. For reactance, it was 1.10 Ohms.
Given that both resistance and reactance swing around 100 Ohms, this represented less than a 1 percent variation, which seemed like a good result. While looking at the minimum/maximum numbers, however, I would sometimes see a difference of several Ohms between the boards across some frequency. My concern became that while most of the boards clearly clustered together, perhaps a small number had substantially different values, which could indicate a flaw.
I went back to the data, and made charts for selected resistance and reactance values across the 20 boards at some frequency. Here are two sample charts, picked at random, one from resistance, and one from reactance.
|Example Resistance Values||Example Reactance Values|
It is immediately clear that boards 6 and 7 have a greater deviation from the average than the others. The fact that these two boards were next to each other on the graph suggested that the error was not in the boards, but in the test fixture. While it is possible that some of the boards could have flaws, what are the chances that I would put a few flawed boards into the bag of 50, then as testing them, happen to pick out the two flawed boards in a row? That is not very likely. More likely is that something was going on while I was testing boards 6 and 7, and that was the source of the variation.
I believe that the change is due to how I clipped the boards to the fixture. The clips were copper spring clips with flat jaws (not alligator clips). Half of the clips (3) were ground clips, and were clipped to the board at a point where both sides were ground. The signal clips had a signal trace on one side of the board, and ground on the other side. So, I could not simply clip to the board, since that would short out the signal.
My solution was to put heat shrink tubing on the ground side of the clip jaw. I used two layers, shrinking the first, then adding a second.
I initially wanted to put as much of the clip as I could on the board in order to get the best connection. I found, however, that if I put too much on the board, I could go beyond the heat shrink tubing on the back, and short out the board, which was obvious on the graph. From playing around with this, I ended up going the other way, and put the least amount of clip on the board. I suspect that boards 6 and 7 were tested at the time when I was putting a lot of clip over the board, but then stopped. That jaw with insulator would act as a capacitor, coupling the signal to ground.
In any case, it was my conclusion that all 50 boards were functional and correct, and each was then placed into a static shielding bag. Since there are no active devices on the board, however, this is probably not necessary. If you are interested in the raw data, please email me, and I will forward it. It is too large to put in the web site.
Since the VNA boards were reflow soldered, incremental testing is not possible. The power/data connectors, master oscillator, stacked parts, and transformers were added by hand as a second step. The following steps/tests were then performed on each board.
Connect the board to the test box with the two power connectors and the one data connector. Apply power, check for smoke.
Connect a frequency counter to J150, and measure the buffered master oscillator at 148.344 MHz.
Using the gnat program, set the RF and LO DDS outputs to 10 MHz.
Connect a standard high impedance scope probe to J170, and measure 2 V peak to peak on the open RF output.
Connect a standard high impedance scope probe to J180, and measure 1 V peak to peak on the open LO output.
Connect a frequency counter to J120, and measure the unfiltered LO output at 10 MHz.
Using #30 wire wrap wire, jumper the LO output to the detector LO input, and the RF output to the detector RF input. These jumpers will be left on the board when shipped.
Using the gnat program's overlap test, verify that the test output is a sine wave of full amplitude on the computer control area. The voltage peaks occur at approximately 1.0 volts (plus and minus 1 volt).
Using the gnat program's overlap test, verify that the ADC conversion rate is approximately 6.5 conversions per second. This implies 154 mS per conversion. With 10 mS of setup time, the ADC conversion time is 144 mS, which matches the 6.875/second internal clock conversion rate. Times can vary a little because the computer timing is not precise.
Check the power consumption of the VNA. With my test setup, using a single power supply (see below), the typical draw is 255 mA at 9.6 volts. This is with both DDS chips running (not power down mode), and a JITO-2 oscillator. With the Valpey-Fisher oscillator, the power consumption rises to 290 mA.
There is no end to the number of enclosure possibilities.. The N2PK web site has several pictures of some of the alternatives built to date. He also comments on enclosures in the second article. In selecting an enclosure, there are several issues to consider. These issues include whether the power supply will be included in the enclosure, the number of external RF connectors, and, the choice of RF connector. Speaking for myself, I prefer the die-cast aluminum cases because they have good RF shielding, and are nearly indestructible. I also like to keep the power supply (AC transformer, fuses, etc) away from the main board (in a separate box). I've also chosen N connectors up to this point in time, and their size will influence of the size and dimensions of the enclosure.
A version of the T1-6T bridge boards was laid out so that it would fit snugly within the Hammond 1590A die-cast box. This is one of the few boxes that I am aware of that have the right size for the bridge board. My construction experiences with that box and board are described on another page.
The Hammond boxes can be expensive, and hard to find. One source is Digi-Key. Some of the popular sizes have been duplicated by other makers, and can be a little less expensive (but still hard to find).
I've only seen one alternative to a Hammond 1590A for the bridge. This is the KAB3321, available from Alltronics.
I put my first VNA into a larger enclosure, with six external connectors. This enclosure came from Circuit Specialists, and is their number 03-125C. This box is about the same size as a Hammond 1590R1, but much less expensive.
My second VNA, built from parts in the second group build, was put into a smaller enclosure, since I only wanted to bring out four N connectors. This box was a Hammond 1590C, which I purchased from Digi-Key. In my opinion this box is about as small as you can go for the VNA. I did have my to tuck my simple power regulator away on one side, and that did take up a small volume.
Here is a picture of the enclosure, with a bridge built into a Hammond 1590A on top. You can click on the picture for a larger view.
The signal jumper between the LO output and detector LO input is made with a male to male elbow directly attached to a male to female elbow. This results in a U shaped jumper with two male ends, ready to screw onto the female N connectors on the case. Finding a male to male 90 degree N adapter can be a little tough. Some suggested sources can be found a little further down this page.
There are several power supply alternatives. N2PK has designed and documented a high quality supply with a number of protection features. This is described in his second article. The VNA can also be powered from the popular wall wart, and even batteries. The quality of the supply can impact the quality of the data obtained from the VNA.
The VNA has a 24-bit analog to digital converter (ADC), which means it has 16 million counts spread over a 2.5 volt range. Each count should represent a change of 0.156 micro volts. It's very easy for all sorts of noise to creep into measurements unless you are very careful. One of the sources of noise can be the power supply. In addition to a being a source of noise, the power supply can be a source of destruction if in puts out a nasty voltage spike. The N2PK design is highly regulated and loaded with protection features.
Although the VNA schematic in the second article shows a +5 VDC and +12 VDC power requirement, the +12 VDC turns out to be quite flexible, and can drop down to around 7.5 VDC. This means that a single supply around 8 to 12 VDC can be used to power the VNA, assuming that a 5 VDC regulator is used to produce the required +5 VDC level. That's what I've been doing here. I bring in around 9 VDC from a wall wart, and use that directly for the +12 VDC input. A 7805 regulator provides the +5 VDC supply. Capacitors are used around both sides of the regulator, and a diode provides reverse voltage protection at the input. Due to the diode drop. my power supply really should stay above 9 volts. Nothing fancy.
Additional information on the power requirements are on my first VNA page. One of the nice aspects of moving to a single supply around 9 volts is that you can power the VNA off of batteries. I use an 8-cell AA NiMH pack and it does the job. I can take it outside with my laptop computer for antenna measurements.
When I first put the supply together I spent some time going through my bag of salvaged wall warts, finding ones with an acceptable output voltage. After connecting the supply, I observed the ADC output voltage (open detector), reported by a monitoring program. The quality of the supply did show up in the stability of the readings. A poor supply would cause the readings to jump around more. I ended up with a fancy wall wart that came from an expensive HP scanner. The scanner long ago died, but the wall wart lives on.
Inside the box you will find one foot of 16 conductor, 28 gauge, ribbon cable. In is in bag 057. I've been using this cable to wire up the three power and data connectors in the VNA. Of course you are free to use whatever cable you like, this is just something that I know works. There is a 2-pin power connector (+12 VDC), a 4-pin power connector (+5 VDC), and a 10-pin data connector which connects to the 25-pin DB-25 printer cable. The three board-side connectors are contained in a single bag, with Inv. numbers 015, 017, and 019. These plastic AMP connectors work with individual metal female contacts that are first soldered to the end of the wire, then inserted into the plastic shell. There is a correct orientation to the contact so that a tab fits into the shell in a way that makes it hard to pull out the contact. The contacts are Inv. number 020, also in the same plastic bag as the connectors.
The remaining connector in the box is marked with Inv. 055 and 056. This is a 25-pin male DB-25 D connector, with mounting hardware. The schematic in the second N2PK article shows the connections between this connector, which is usually mounted to the VNA enclosure, and the 10-pin connector which plugs into the VNA PCB. Assuming you have a standard printer or parallel port on your computer, then all you need is a standard printer cable to connect the VNA to the computer. It's male on one side, and female on the other, with straight-through wiring.
At least two, and as many as six RF connections must be made to the VNA board. I suspect that the number four is common, which brings the RF output, LO output, detector RF input and detector LO input off board. You can go down to two connectors if you want to strap the LO output to the detector LO input, and, in fact, the board is shipped that way because it was tested that way. The other two RF connections are the buffered master oscillator output and the unfiltered LO output. Another variation to think about is having two detectors in the same box, which would add two more connectors (up to eight). This would imply an additional board, however.
I have found that an appropriate coaxial cable to use for these signals is RG-316. This cable is the same physical size as RG-174. The main reason for using it is that it has a Teflon dielectric, which withstands abuse and heating. RG-174 seems to melt very easily and is not very rugged. In Europe, there is a cable named RG-188 which is about the same as RG-316. Another possible alternative is LMR-100. Another cable which sometimes shows up is RG-178. While this is a Teflon cable, it is much smaller in diameter than the RG-174/RG-188/RG-316 cables. It will not work with some connectors, although it is otherwise a fine, albeit small, cable.
In the box your should find 2 feet of RG-316 cable. In the case of the assembled boards boxes, the cable is put under the bottom foam. This cable is part Inv. 058.
I have been directly soldering my cables to the VNA board, coming up from the bottom side. I use two small lengths of brass tubing which act as a support guide and ferrule.
The first brass tube is approximately 1/4" long, and has an outside diameter of 1/8". This tube is pushed into a connection hole from the bottom, and soldered to the board so that it is flush with the top of the board. This tube is the guide sleeve. The center conductor of the coax is threaded into the tube, and it emerges on the top of the board. The outer insulation is removed from the end of the coax where the coax covers the inner tube. In order to hold the braid in place, a second tube of a larger diameter, 3/16", is forced down over the outer jacket, and it traps the braid between the outside of the inner tube, and the inside of the outer tube. This is a nice tight fit, and does not need to be soldered, if you are are experimenting with the board. Once I know I have the right length cable in place, I do solder the bottom of the outer tube to the board. The result is a good mechanical and electrical connection, and with very little cost.
The left picture shows the smaller 1/8" diameter tube in a hole in a blank board. Of course this step is usually done after assembly of the board, this is just an example. I solder the smaller tube in place around it's perimeter, where it meets the board. The larger tube section to the right is the collar that is brought down over the braid after the coax center conductor is inserted into the smaller tube. When finally soldering the outer tube down onto the board, the braid will probably have to be trimmed, and certainly keep any excess braid from touching a surrounding component or trace.
The right picture shows the top side of the board. I simply bend the center conductor down onto the board, and solder the end to the appropriate pad. Here's where the Teflon coax works well, since the center conductor can tolerate a lot of heat, and it does not melt. In the N2PK version of this technique, he uses a length of #30 wire to span the distance between the board trace and the center conductor of the coax, which is trimmed back to not even emerge onto the top side of the board.
The brass tubing concept is initially described by N2PK on page 29 of his second article. My addition is the second tube over the first, which I believe creates a neat and clean result, and can be used temporarily without soldering
I've been satisfied with purchasing N connector hardware from RF Parts, when I can't find what I need at a local hamfest. In the configuration shown above, the LO output and LO detector input connectors are very close together. They are jumpered with a male to male elbow connected to a male to female elbow. This creates the smallest radius all-metal jumper with two male N connectors on the same plane. The male to male 90 degree N adapters can be hard to find. One source is Fry's Outpost, and another is Transel.
If you really wanted the RF VNA board connections to be made through a connector, then it seems as if the best idea is to solder a vertical female SMA PCB mount connector to the bottom of the board, and then use a jumper wire between the center pin and the board. You will have to remove the four corner mounting pins prior to soldering, but the result should be quite nice.
In order to use the VNA you will need calibration standards. The ultimate accuracy of the VNA is largely derived from the standards, and the quality of their characterization.
The N2PK articles cover standards in some detail. To be precise, the standards are used with reflection or impedance measurements. If you only wanted to make transmission measurements, then they are not needed.
All of the VNAs, whether assembled boards or parts kits, include two, 100 Ohm, 0.1% tolerance resistors. When used in parallel, they create a 50 Ohm load.
There can be times when it seems as if you are doing nothing other than putting on and taking off standards. With the N connectors, I was screwing the outer shell on and off with very little enjoyment. What I did here was to use a Dremel tool with a cutoff wheel to remove the outer shell completely from the N connector. This left an N connector that is mated with a push-on friction fit.
This group of VNAs uses the LT2440 ADC (analog to digital converter), as opposed to the LT2410. The 2440 is a more flexible part, offering a range of data conversion rates. As the rate increases, the ENOB (effective number of bits) decreases. So, there is no free lunch. If you go faster, you get less accurate data.
As shipped from here, the board is configured for the slowest data rate, which makes the 2440 compatible with the 2410. All existing software designed for the original N2PK design should work as expected with the VNA. If you would like to experiment with faster conversion rates, you have several options.
On the hardware side, you have two choices. The simplest is to jumper pin 7 of U260, the ADC, to ground. This changes the sampling rate to 880 samples per second. Of course this implies that you should be using software which expects that rate, and can take advantage of it. If desired, the control pin could be connected to a mechanical switch which could select the rate.
The second choice connects the same pin via a jumper wire to the RF DDS data in pin, or some other appropriate output pin coming from the computer (on the parallel printer port). This enables the conversion rate to be set via software. While this is the most flexible solution, and allows all conversion rate alternatives, it also requires the cooperation of the software, since the software must actively set the rate for each ADC conversion. The software that has been developed here, the VNAgra, VNAccess, gnat, and cialog programs, have been modified to control the ADC in software. Of course this is not yet a complete suite of programs.
My own preference is to connect the ADC control pin to the RF DDS data in pin. This allows for complete control of the ADC in software. The reason to reuse the RF DDS data in pin is that it is available on the board, relatively close to the ADC chip, and it has been buffered and conditioned. It is certainly possible to use another appropriate signal, including otherwise unused signals on the parallel cable. It really is nothing more than a question of the configuration options in the software that you would like to use. The reason to avoid the RF DDS data in pin is that a software writer wants to load the DDS chips at the same time that the ADC is being read.
Let me get back to the hardware for a moment, and make sure that the 2440 changes are clear.
Pin 7 on the 2410 was grounded. With the 2440, that pin has been turned into a serial data in control pin. The conversion rate is set for every conversion. When you read out data from the chip, by definition you clock in the rate for the next conversion. The data sheet defines meaningful conversion values for all high data, and all low data. This means that it is not required to control the pin in software, it can be strapped either high or low, if those two data rates happen to be the desired data rates. Sadly, the rate of the original 2410, approximately 7 conversions per second, is the all high value. The all low value is 880 conversion per second.
The board change that was made in this batch removes pin 7 of the 2440 from ground, and connects it to +5 VDC through a 100 K Ohm pull-up resistor. This resistor, Inv. 059, is shown in the middle of the above left picture. If you are building a kit, you should add the resistor. This resistor is located on the bottom of the board, under the ADC chip.
Without the resistor, the data in control pin will float, and operation will probably be erratic at best. With the resistor, and no other changes, the control pin is pulled high, and the conversion rate will match the 2410, meaning although a 2440 chip is in use, the VNA will act as if it is using the 2410 chip.
If you are happy with that behavior, then you are done. If you want access to different conversion rates, then you have some more work to do. This is true whether you have a kit, or assembled boards.
If you ground the pin side of the resistor (not the supply side!), you will select the all low control rate, which is 880 samples per second. If you want to do this, a small opening in the solder mask has been provided to access ground. In the above left picture, you would simply run a jumper from the through hole which connects to pin 7 (the one that has a red wire in the picture), and connect that hole to the ground immediately to its upper right. Now, you are fixed at 880 conversion per second.
If you want access to all of the conversion rate choices on the chip, then the pin must be connected back to the computer so that it can set the rate for every conversion. My own solution has been to take a #30 wire wrap wire, and starting at the hole, coming up from the bottom of the board, solder the wire to the hole on the top side. This is pin 7 of the chip. I ran the wire on the bottom of the board, towards the bus interface chip, the SN74ACT1284 (U160). The insulated wire transitions from the bottom to the top of the board through larger ground hole. This can be seen in the above right picture. The wire then is routed to pin 19 of the bus interface chip, U160. I soldered the wire directly to the pin, above the level of the board. This is shown in the above right picture. Pin 19 is the buffered RF DDS data in signal.
Whatever you decide to do with your hardware, it should be a change which is acceptable to the software that you want to use. The only 100% safe choice is to leave the board in the 7 conversion per second mode, since that is the historic operating mode of the VNA.
Since this was first written, N2PK has expressed his plans for extending his VNA architecture. This includes parallel port pin assignments. His use of the existing RF DDS data in pin to drive both detectors is consistent with my above suggestion to connect the ADC serial data in control pin to the RF DDS data in pint.
As of August 2004, I am aware of four different software efforts for the N2PK VNA.
G3SEK (now GM3SEK)
It seems to me to be a good thing that there are multiple efforts, since each will probably provide different capabilities based upon the interests and needs of the authors. At the same time, there are variations in hardware. The 2440 ADC on this board is a hardware change that requires software changes in order to allow it to reach its full potential. This is not the only hardware change that I am aware of. The group working with OZ1DUG has added an auxiliary board which supplies an external clock to the 2410 chip. This provides three additional conversion rates, and requires two control lines.
Perhaps over time, other hardware enhancements will emerge.
The question becomes, will some software not work with some hardware due to an interface incompatibility?
I would hate to see that situation emerge, but it can easily happen, purely by accident.
My only answer to this problem at this point is to offer the following philosophy.
All software authors should make a reasonable attempt to accept reasonable hardware variations while maintaining the software functionality. People making hardware variations should openly publish the specifications of their changes so that software writers don't have to guess at what needs to be changed.
I don't think it's necessary for all software to embrace all enhancements, but the enhancement should not cause software to fail in its basic function. For example, a hardware variation might add a remote controlled switch in front of the VNA to pick between several bridges, Software should not be forced to support the switch selection at a high level, but it should know how to select a default bridge and then carry out its function as if there were a single bridge. Software need not support multiple ADC conversion rates; but, the software should know how to set the ADC to the default conversion rate. It a hardware variation reassigns any of the historic VNA control signals, I do not believe that needs to be supported by software, since it is not a reasonable modification.
This is my opinion on the best approach to the situation.
On page 29, step 18, of the second N2PK article, N2PK describes nine ground to ground jumper wires added to improve ground plane integrity. Those wires were not part of the automatic assembly process. If you want to add them, then you will need to use a hobby knife or other sharp object to pick away the solder mask at the points shown on the pictures in the article and build info files, and then add the wires.
Back to my March 2004 Group Build Page
Back to my Experimentation Page
Back to my Amateur Radio Page