EtherGeist

Improved remote station audio link

With the Flex 5000 up and running in my remote station over the phone patch, I wanted to be able to get stereo sound from the Flex to my operating position. The Flex uses both audio channels to provide binaural audio or to feed the audio from each receiver (I have the second receiver installed) to separate channels.

I started looking at a number of low delay solutions for digitizing audio channels and moving them over the Ethernet that provides remote control of the PC driving the Flex. There are commercial solutions that provide audio over Ethernet for recording studios but they are expensive! I started thinking about building a custom solution using a micro-controller when one of my friends suggested using a balanced audio link as a much lower tech solution.

The remote station is located about 500 feet from the house. The Ethernet run is all Cat 6 twisted pair cable in three spans – each span interconnects via an Ethernet switch running at 100 MBps. Fortunately, when the cable was installed I had dual cables pulled to each drop (cable is inexpensive) so I had a spare cable on each span. Each span connects to the next with a short Ethernet cable.

There is probably 700 feet of cable between the three spans – driving unbalanced audio over this length of cable was a sure recipe for hum pickup and disappointment. Balanced audio was a slam-dunk choice for the link.

Cat 6 has four twisted pairs so for my configuration I decided on the following:

• Two pairs for stereo audio from the Flex
• One pair for microphone audio to the Flex
• One pair for PTT control of the Flex

To make sure I had enough audio to drive the microphone pair back to the Flex, I added an audio preamplifier based on an LM386. Here is a link to the schematic – it's very simple!

Design Notes

1. The electrolytic capacitor on pin 7 of the LM386 provides decoupling of any low frequency noise on the power supply. The National Semiconductor data sheet does not show a rejection ratio for the 330uF capacitor used in this design. I tried a 50uF (the largest value for which the data sheet plots a rejection response) but it was not enough to completely nail the hum – the 330uF removed all hum.
2. The 0.1uF capacitor across the input provides a path to ground for strong AM signals picked up by the microphone cable. Some of the AM stations in the Bay Area put -15 dBm signals into my receivers – as a result, this capacitor is critical! Without it, there was low-level pickup of KGO AM 810. Note that even in the configuration shown; the LM386 has at least 26 dB gain – more than enough to render a good imitation of a crystal set!
3. The 16-ohm resistor (R3) in the schematic serves as the load for the LM386.
4. All the transformers are 1:1 ratio with 600-ohm impedance. Radio Shack carries a suitable transformer (PN: 273-1374 ). A typical Radio Shack store only carries a couple of these in inventory so you will likely have to visit a couple of stores – call ahead and ask them to check the inventory system to tell you which stores have the transformers in stock.
5. I built the link in two units – local and remote – separated by the twisted pairs of the CAT6 cable.
6. I couldn't find bulkhead mount RJ-45 connectors (if indeed they are made!). Instead, I used a wall plate with a modular jack insert. This dictated a larger box than I would have liked but it made for a much simpler mechanical construction.
7. Local and remote units have separate power supplies – this avoided having to deal with resistive losses in the cable – this is typically about 28-ohm per 1000 feet for CAT6 cable.
8. The PTT switch on the microphone provides the bias to the base of the transistor controlling the relay in the remote unit. Only 5 mA flows on the PTT line with little voltage drop due to IR losses in the cable.

The remote audio link works well and provides full stereo from the Flex to the operating position in the house. The stereo link goes into a pair of Logitech powered speakers and sounds great!

Remote Operation of a Flex 5000

I've made slow progress over the last few weeks on my Gen 2 WSPR system due to the (pleasant!) distraction of integrating a new Flex-Radio 5000 into my remote station.

I thought long and hard about the choices for a new primary transceiver to displace the Kenwood TS-B2000 I've used for the last 6 years. It came down to a choice between the Elecraft K3 and the Flex Radio 5000. Both are excellent radios with outstanding receiver performance. I opted for the Flex 5000 because of the attraction of being able to extend the radio's functionality through software and PC horsepower. The choice was easier for me as I've operated remote for so many years I don't miss the tuning knob and physical buttons of a radio control panel!

When I first setup the Kenwood TS-B2000, I worked very hard to avoid having a computer at the remote location (in my case, about 500 feet from the house – close to the antennas). I opted to control the Kenwood via an Ethernet connected terminal server (4 RS-232 ports capable of operating over Ethernet at up to 115.2 Kbps) and transfer bi-directional audio over an extension of the small PBX in the house. An auto-answer telephone coupler and a modified MFJ auto-patch interface the phone to the remote radio and my desktop computer.

I use three ports on the terminal server:

• CAT control of the TS-B2000 at 57,600 Bps
• PTT control & audio interface via a RigBlaster Nomic
• Serial connected relay controller for switching

I modified the MFJ auto-patch to operate in reverse as audio coming from the telephone line goes to the audio input on the PC sound card; audio from the microphone needs to drive the telephone line. I added a small audio amplifier (LM386) to provide more drive between the microphone and the line. A 1K resistor over the connector to the handset port simulates "off-hook" to keep the telephone line open. The PBX port auto-dials the remote auto coupler when the phone patch goes "off-hook".

Initially I used an automatic hang-up detector at the remote end that detected the line disconnect tone from the PBX. This was fine for voice or operation but would sometimes false detect when operating CW or digital modes and hang up in mid-stream! I solved this problem by using the relay controller to drive the hang up control line on the auto coupler.

A side-note – my first idea for the audio transport was to use Voice over IP to support bi-directional audio. I tested this out but found there was too much delay introduced both on receive and transmit. Cracking a pile-up was almost impossible. Since then, a lot of progress has been made on codec performance and PC horse power – the Skype SILK codec works pretty well – still not as good as a analog solution but acceptable for rag chewing and casual DX operation.

I ordered the Flex and set about building a simple audio switching unit so that I could select either the Flex 5000 or the Kenwood to source and sink audio via the auto-coupler. A relay on the serial relay controller determines the audio source – Flex or Kenwood.

The Flex requires a PC at the remote end to run PowerSDR – the software "guts" of the radio that provides the DSP capabilities. Since I was going to control the remote PC using UltraVNC (a great open source solution for remote desktop control), run PowerSDR plus a raft of digital mode software (Spectrum Lab, WSPR, MixW…), I wanted to make sure I had plenty of PC horsepower to spare. I purchased an HP Pavilion desktop machine with an Intel Quad Core (2.33 GHz/core) which unfortunately came with Vista X64.

Vista doesn't provide anything I need and between the 64 bit OS and driver signing issues, is frankly a pain. It works fine and its stable but the "enhanced security" with its endless "do you really want to do this" dialog boxes is a hassle. This has NOTHING to do with Flex – indeed, once configured Vista X64 runs the Flex software perfectly.

At some point in the next couple of weeks, I'm going to add a Windows XP dual boot capability to the system and run XP as the default system.

With UltraVNC, PowerSDR, DXWin logbook, MixW or WSPR running, typical CPU utilization is less than 15%!!! I have had NO issues at all with the PowerSDR software disconnecting from the Flex Firewire driver due to excessive latency. Screen updates and mouse control is very snappy – UltraVNC runs about 20 MBps over the Ethernet (100BaseT in my case) and causes very little load on either my desktop or the remote machine running the Flex.

The Flex 5000 is a spectacular performer on all modes. Having brick wall filters with full control over bandwidth and center frequency makes operation on all modes a piece of cake.

 

Feeding a DDS with a GPS disciplined oscillator – part 2

With the frequency multiplier completed I had a nice 30 MHz sine wave to be able to feed the DDS-60 with an external reference clock.

After some hunting on the Analog Devices web site, I found Application Note 419 – "A discrete, low phase noise, 125 MHz crystal oscillator for the AD9850 complete Direct Digital Synthesizer". In some ways, I wish I'd found this article earlier as it shows an elegant solution to generate a CMOS compatible clock using an ECL to TTL level converter (MC10ELT21) – this part needs only about 200 mV peak-to-peak input to generate rail to rail outputs even though its almost $6 in single quantity from some of the parts suppliers. Using this part could have simplified the parts count for the frequency multiplier.

This AppNote also shows a very simple way to condition the clock for input to the DDS-60 – all that is needed is a capacitor and a pair of resistors to provide a potential divider and bias the clock at 2.5v DC level.

The AppNote is also helpful because it provides a simple explanation of the minimum clock required to meet the DDS chip spec – 3v peak to peak.

My frequency multiplier design provides about 4.5 V peak-to-peak when terminated in a 100 ohm resistor so I added that at the input before the capacitor for this simple circuit:

The DDS-60 works fine with the GPSDO and frequency multiplier as the source of the reference clock.

One thing I have noticed with the DDS-60 even with the canned SMT oscillator on the PC board is a roughly -45 dBc spur of the 30 MHz reference clock. This is present at all output frequencies of the DDS. It's possible that this is stray pickup from my test bench set up – ground continuity and screening is definitely a compromise – but it is strange. At any operating frequency below 10m, this isn't a problem as the transmitter low pass filter will attenuate the 30 MHz signal without any problems.

I'll add an update here if (once!) I track down the source of the problem. That said, snooping around the DDS-60 board, the 30 MHz reference clock seems to be everywhere!

Feeding a DDS with a GPS disciplined oscillator – part 1

One of the steps towards building my next-gen WSPR system was figuring out how to generate a very accurate and stable reference clock for the DDS oscillator that serves as the transmit and receive LO. I acquired a 10 MHz GPS disciplined oscillator (GPSDO) on eBay and set about developing a 10 to 30 MHz frequency multiplier to generate the reference clock for the DDS-60 that I had built.

This is a two part post that describes how I use the 10 MHz GPSDO to clock the DDS-60:

• Part 1 – the 10 – 30 MHz frequency multiplier
• Part 2 – feeding the DDS-60 with an external clock

 

The Frequency Multiplier

There are many ways of building a frequency multiplier – after some research via Google and the excellent book published by the ARRL "Experimental Methods in RF Design", I chose a diode based frequency multiplier based on work by Wenzel. This design is simple and effective. I built a prototype using both regular diodes (1N4004) as well as a surface mount packaged pair of Schottky diodes connected with common cathodes (BAS40-5). Both worked as multipliers with the BAS40 based prototype generating slightly richer harmonics.

I opted to follow the basic frequency multiplier with a double tuned LC band pass filter driving a transistor based amplifier and a second filter to clean up the resulting signal.

As a learning experience, rather than using a canned design for the amplifier I decided to design one from scratch. After removing much mental dust (I learnt about transistor amplifier design in University a while ago!) and experimentation I arrived at the following design.

You can click on the diagram for a larger view of the schematic.

Astute viewers may recognize that this diagram was extracted from a SPICE circuit simulator – in my case, LTSpice from Linear Technologies. This excellent (and FREE) tool is a fully featured SPICE simulator maintained by Linear Technologies – to whom, tremendous thanks! It includes full library support for LT products as well as some basic devices like diodes, transistors etc but you can use additional libraries or component descriptions you generate yourself. The data sheets of a lot of modern semiconductor parts include the SPICE model for the part – I generated the SPICE model for the BAS40 using this strategy.

LTSpice was invaluable in refining this design – for example, I optimized component values in the frequency multiplier to improve the return loss (SWR) presented to the GPSDO as well as maximizing the 10 MHz voltage presented to the first diode.

Although the diagram shows an amplifier using a pair of 2N3904 transistors, I opted to use a 2N3553 for the second (output) device. I wanted a little more available power than the 2N3904 would generate without getting excessively hot – also, in experimenting with different amplifier designs, the 2N3553 generated a cleaner output spectrum when run with a higher standing current.

The output waveform is nice and clean – here's a couple of shots taken from the oscilloscope of the wave form and output spectrum when fed into a 50 ohm load.

Output Waveform into 50 ohms

Output spectrum

For these measurements, the GPSDO was running on the bench without a connected GPS antenna so ignore the frequency displayed on the 'scope!

The two hottest spurs on the output are almost -44dBc down on the 30 MHz signal – this was more than clean enough for my requirements to feed the DDS-60. If you want a cleaner output, better filtering is required (triple LC tuned filter for example) at the amplifier input and perhaps an additional 10 MHz trap between the first and second amplifier stages.

I built the completed design on a hand scribed PC board using the technique I described in my last post.

Here's a shot of both sides of the completed board.

Top Side

Bottom side (see note below)

Note the isolated ground area in the bottom right of the board. I isolated the ground plane under the frequency multiplier and 10 MHz input in an effort to avoid ground plane coupling of the 10 MHz signal into the amplifier. It gave a slight improvement over the roughed up prototype that I'd built on scraps of PCB before committing to the final design.

Last but not least, here's the completed unit boxed without the cover.

The completed 10-30 MHz frequency multiplier

The Trimble Thunderbolt 10 MHz output is a nominal 10 dBm into 50 ohms – mine is a little hotter at 12 dBm. The output of the completed frequency multiplier is almost 16 dBm. Generating less power resulted in hotter spurs in the output spectrum – most of the power will be dissipated in a pad as part of the interfacing circuitry to the input of the DDS-60.

 

Quick prototype printed circuit boards

I've made simple PC boards for analog designs for many years. Rather than etching these simple designs, I've used a sharp knife to cut the copper foil and create isolated islands to which I could solder parts.

On a recent visit to the local hardware store, I found a tool that really speeds up the process – a tungsten carbide tipped scribe. The scribe is usually used to mark metal prior to cutting and the tungsten carbide tip is intended to leave a mark in hard metal like stainless steel. On softer copper, it makes short work of cutting through the foil down to the fiberglass.

The copper is soft enough that it doesn't put much (if any) wear on the tip of the scribe and its much easier to use than wielding a knife and having to put significant pressure on the blade to cut the copper. A couple of passes with the scribe against a ruler is sufficient to cut through the foil.

The scribe is fine enough that you could (with patience!) create boards that would accept integrated circuits but at some point its clearly quicker to etch!

Here's an image of the scribe and a prototype board I put together for testing an RF amplifier design for my 30 MHz tripler project (more on this in the next post).

Once the board is scribed, I check that each island is isolated by using a DVM to check for shorts. Shorts tend to be at the corners of an island – a couple more passes with the scribe clean them up.

Here's a shot of the completed board – note I didn't go for "small" here – I wanted easy access for changing components.

This is a basic 2 transistor broadband RF amplifier using a 2N3904 and a 2N3553. The first stage is transformer coupled into the second while the output stage uses an RF choke and capacitor to take RF from the collector.

For cutting PC board, I use a cutting wheel in a Dremel hand drill – gloves, safety glasses and a face mask are required for this operation!

Next-Gen WSPR System

Good hardware design, like good software design, starts with a set of requirements. After completing the first generation QRPP WSPR transmitter I've been giving a lot of thought to its successor and the system requirements. This is a much more ambitious system than the first one!

Requirements

• Frequency agile covering 160-10m
• High level of frequency stability
• Variable power under program control (0 to 5 watts)
• WSPR or Graphical modes (QRSS CW, FSK CW, others perhaps)
• Initially transmitter only but with capability to add a receiver at a later date
• PIC controller
• Ethernet interface for command and control

Initial Design thoughts & consequences

• Use a DDS for frequency control
  
  Implemented using a DDS-60 from AMQRP – already built and tested. To get maximum frequency stability will require a "crystal oven" approach to minimize drift of the reference oscillator with temperature. The DDS will also require some frequency "trimming" due to the reference oscillator not being exactly 30 MHz.
  
  An alternative would be to a GPS disciplined 10 MHz oscillator and triple it up in place of the 30 MHz canned oscillator on the DDS-60 board. In my case, I'd have to build the 10 MHz reference oscillator around one of the GPS units I already have. Having a 10 MHz reference would be a "good thing™" but is another development step. Have to mull this one over… On the plus side, it would eliminate the need for the "oven" on the DDS-60 and having to build a frequency calibration table.

  
  Update: I did some digging into the design of a GPS disciplined oscillator and found several excellent designs using a temperature controlled precision reference oscillator – while I was looking at buying one of these via EBay, I found that there were two surplus GPSDO systems available – the HP Z3801A and the Trimble Thunderbolt. I successfully bid on a Thunderbolt and now have that up and running – next step is to build a tripler to feed the DDS-60 with its reference clock.
  

• Variable output power
  
  Building an RF deck with constant gain and output power across a range of 2-30 MHz is a complex affair! To simplify design, I've decided to use a variable attenuator (PIN diode assembly with voltage control to select the attenuation) and to measure the actual output power of the transmitter. Software will control the level of attenuation while monitoring output power for each band in operation.
  
  I'll use the Analog Devices AD8307 to provide the power measurement and feed the output of this device into an A/D channel on the PIC.
  
  After some digging, I found that Avago has a packaged pin-diode product (HSMP-3814) for use in a voltage-controlled attenuator – there is a nice write up in the Avago Application Note 1048. This will need the PIC to generate a control voltage to drive the attenuator – from the app note, a control voltage of 15v fully saturates the PIN diodes –there is a 2 dB minimum loss through the attenuator.
  
  The control voltage will need a D/A out of the PIC (in this case probably a R/2R resistor ladder) and an external voltage amplifier (op-amp) to generate the 15v from the PIC output which will provide a maximum of 3.3v. A more PIC-pin-economical method would be to use an external D/A converter connected via the PIC I2C bus.
  
• Ethernet command & control
  
  Having Ethernet access to the transmitter will allow some interesting opportunities for control. As a minimum, I want to be able to set a schedule for the system to transmit different modes at different combinations of time, frequency and power level. This way the schedule can accommodate different band and propagation conditions. A side effect of this is that the PIC controller will have to be Ethernet capable and maintain a time of day clock.
  
  Using the PIC Ethernet controller adds a requirement – the PIC must have a 25 MHz clock when the embedded Ethernet controller is used.
  
• PIC Controller
  

  I'm torn between two different approaches – use a PIC with an Ethernet interface (like the 18F97J60) or add an Ethernet interface using the I2C interface The 18F97J60 is a very capable part with more than enough IO and communication ports to do the job. I haven't used this part before and it's a 100 pin TQFP surface mount package. I can prototype the PIC controller using an adaptor board that mounts the surface mount part and provides a through hole breakout. Ugly solution but practical!

  
  I have a couple of other projects in mind that could benefit from the 'J60 part so on balance, I'll bite the bullet and get up the learning curve.

 

• Implications of frequency agile operation
  

  The system will need a broadband RF deck to deliver 0 to 5 watts across the complete frequency range. The output of the DDS-60 will run through the attenuator first then into a driver stage capable of generating at least 0.5 watts at 30 MHz. This is turn will be fed into a power amplifier to add the remaining 10 dB of system gain.
  
  The PA output will have to be filtered via a band specific low pass filter to remove harmonic content. This PIC will select the filter per band using relays to control RF flow to the antenna. There is an I2C compatible relay driver chip that I've seen – need to hunt down the reference.

   

  With this broad a frequency coverage range, one antenna isn't going to do the job. Best to design the system to be capable of driving up to 4 different antennas – selectable by PIC control and relays.

   

• Implications of a receiver
  
  For now this will serve as a list of issues to think through…
  

1. The system will need additional antenna and band switching capabilities to switch the antenna(s) between the receiver and transmitter and select appropriate receiver filtering/amplifiers. I can simplify the process of adding the receiver later by incorporating a TX/RX relay in the system before the antenna relay control.
   
2. Band switching likely needs another relay controller – yet another relay controller on the I2C bus!
   
3. The DDS-60 output will serve as the LO for both the transmitter and receiver. Implies switching the LO output between the receiver and transmitter – this can be handled by one of the relay control chips and is best done post the attenuator.
   

With these requirements in mind, here is a functional block diagram for the system.

WSPR Spots & Propagation

I've spent some time experimenting with VOAAREA to generate prediction maps for my 200mw 30m WSPR transmissions. VOAAREA is one of the tools in the NTIA/ITS HF Propagation Prediction suite originally developed for Voice of America. VOAAREA generates maps for any of the prediction parameters generated from the underlying VOACAP prediction engine.

The tool suite is a professional HF propagation prediction package and you can download it from the VOACAP web site.

This web site also provides links to how to use the different programs in the package as well as some interesting articles about HF propagation prediction.

If you have EZNEC+ or EZNEC Pro you can model your antenna in EZNEC and then save it in a format that is readable by the VOACAP suite.

I started using the VOAAREA tools to try and explain why Larry WB3ANQ gets such great signal reports (even at 5mw power levels) from David VK6DI near Perth in Western Australia. I'll write up those results in the next post.

Using VOAAREA I generated a full days worth of propagation predictions for my 200 mW WSPR transmissions on 30m using a 160m inverted L antenna loaded up with a remote antenna tuner. I modeled the antenna in EZNEC+, saved as a model for the VOACAP engine and then did a batch run for each hour of the day (in UTC of course) for April 2009 with a predicted smoothed sunspot number (SSN) of 8 (the current forecast from NOAA).

Using Photoshop I took the map for each hour and generated a short video of the propagation progression through the day. You can watch the video by clicking on the video below (its hosted on YouTube).

The video views best in HD format by clicking through to the YouTube web page.

VOACAP uses a signal to noise ratio (SNR) normalized to dB/Hz. The WSPR spot database at WSPRnet.org records the SNR levels from WSPR which are in dB in a 2500 Hz bandwidth. To convert from the VOACAP SNR levels to WSPR you have to add approximately 33 dB. So a WSPR spot of -27 dB corresponds to a VOACAP SNR of -60dB/Hz. The map is color coded with a spot of -27 dB as the lowest level.

You can see spots in the WSPR database down to -32dB wsprnet.org – but in practice most stations seem to bottom out around -29 dB on receive so I modeled -27 dB as the lowest level of interest.

The model runs assume a "residential" noise level of -145 dBW/Hz – I suspect that most urban noise levels are higher than this by perhaps 10 dB (-135 dBW/Hz) so its worth taking that into consideration when comparing the predictions against spots in the database.

I haven't worked out a way to automate taking spots from the database and placing them on the map – I'll add that if I ever get around to it – however, comparing spots to predictions looks pretty accurate.

 

 

 

Using WSPR to compare antennas – part 3

Thanks to Bruce W1BW and WSPRnet.org, the entire database of WSPR spots can be downloaded in CSV format by the month. You can access the download page on the WSPRnet web site.

I wrote a simple PERL script to parse the data and convert it into a format suitable for plotting in EXCEL.

The first analysis was to take the spot data and aggregate it by antenna across several days. My rationale was that the propagation varied over the course of the measurement period on each antenna but would be fairly consistent because of the solar minimum.

There are some errors introduced as a result of this assumption that I'll discuss at the end of this post.

I wanted to plot the maximum distance of an observed "spot" by time of day and reported received SNR for a given power level. To do this, I built up a dynamic 2-D array of time slots (00:00, 00:02… 23:58) and ranges of received SNR. I chose to split the SNR range from -32dB (the lowest reported by WSPR) and everything greater than 10 dB) into 15 "buckets".

The PERL script processes the CSV formatted data and build the array. As each spot is processed, it captures the maximum distance of the spot for that combination of time slot and SNR.

There are some anomalies in the data – despite the time synchronization required for WSPR operation, some spots are reported on odd minute periods or as part of the next period. Odd minute spots are considered to be part of the previous period – there's no way to tell if the spot was reported on the next period so this issue (small in number in the database) was punted.

The processed data was dumped as a CSV file and imported into EXCEL.

 

Results

Here are the plots for two 10 days periods – March 1^st 00:00Z to 11^th 23:50Z on the inverted L and March 11^th 23:52Z to March 22^nd 19:26Z on the vertical dipole. Both plots are for my WSPR transmissions at 23dBm.

 

Inverted L

Vertical Dipole

Click on either image to download a PDF file of the image.

 

Comparing the two plots doesn't give a conclusive answer to which antenna is best...

My early conclusions…

• The inverted L generates far more spots from US based stations (out to about 4000 Km) than the vertical dipole
• SNR reports for spots in the same range are significantly stronger on the inverted L than the vertical dipole – further backing up the observation above
• The vertical dipole generates longer distance spots in more time periods than the inverted L
• Longer distance spots are generally weaker on the vertical dipole than comparable distances on the inverted L

When I first thought about this comparison method, I assumed that with the solar cycle at its current minima, there would be little variation in propagation during this 3 week period. After checking the archive of solar data (Solar flux level and A-Index) for the period, there were some significant perturbations:

So while the above results are interesting, they are not definitive. A better comparison would be to do a true A/B comparison against the two antennas – ideally with two identical transmitters driving each antenna simultaneously on slightly different frequencies.

I don't have the capabilities to do this (yet!) so I may opt for a simpler comparison – using a single transmitter alternately feeding each antenna during back to back WSPR time periods. It's a simple matter to have the PIC use one of the attenuator relay drivers to instead switch a changeover relay between the two antennas.

Qualitatively, the inverted L seems to be a much better antenna despite its modeled radiation pattern. There was a significant increase in spots (both near and far) when I switched back to the inverted L.

I'm still no closer to understanding how WB3ANQ's performance is so stellar! However, in the intervening weeks since I started this "experiment", I've seen spots at the 50mw level from both Australia and Japan while my 10mw signal has made it to both Japan and Wake Island – 8,323 and 7,084 kilometers respectively.

Not bad for flea power!

Using WSPR to compare antennas – part 2

Having settled on a 30m vertical dipole as the "experimental" antenna, I decided to build a wire version of the Force12 Sigma 40 but with a vertical section long enough to not require the center loading coil. I started the construction process by looking for a convenient tree (highest and closest to the radio). Of course "The Tree" turned out to be close to others and made a straight wire copy of the Force12 impractical – there wasn't enough space for the end loading wires.

After a modeling session with EZNEC, I settled on a slightly different design using X shaped end wires – the modeled antenna looks like this:

The dimensions are as follows:

• Each leg of the X is 4 feet long
• The vertical wire is 26 feet long – two 13 feet lengths from the feed point to the X end load

The antenna is suspended with the bottom X 10 feet above a "normal" ground (EZNEC – Average, Pastoral, Heavy Clay).

Here are the field and SWR plots for the modeled antenna.

 

Construction

In a word – simple!

The X end loads were made with 4 foot bamboo plant stakes pushed into a 4" section of 2x4 lumber with eye hooks top and bottom. The wires were measured, soldered together and then spread via the bamboo stakes – secured with PVC tape and tie wraps.

The eye hooks provide strain relief for the vertical section wires – each 13 feet long before pruning and tie off the support rope (at the top) and the stabilizing rope anchored by a large rock (at the bottom) to minimize movement in the wind.

The antenna is fed by 50 ohm coax and a 1:1 balun. The balun is housed in a PVC "T" piece with end caps on the straight through portion of the T. Bolts are drilled through the end caps and secure the vertical wires to the balanced side wires of the balun.

The side connection of the "T" houses the balun and supports a three foot length of PVC pipe that helps take the coax away at right angles to the dipole wires.

For wire, grab what's at hand! I used PVC insulated 18 AWG wire – cheap by the 500' from your friendly hardware store…

 

Adjustment

Some pruning was required to bring the antenna to resonance – about a foot off each vertical section of the dipole. I used an antenna analyzer to make exercise fast! The end result delivered an SWR response that was almost identical to the EZNEC model.

 

Comparison

Given that we're at the low of the solar cycle, I thought I could get away with running each antenna (the inverted L and the vertical dipole) for a 7 to 10 days and comparing the data. This was the easiest and quickest approach as I already had a ton of data for the inverted L.

 

Part 3 will summarize the results…

 

Using WSPR to compare antennas – part 1

As you can imagine, having built my QRPp WSPR transmitter and getting everything on the air, I started watching the spots database on WSPRnet with even more interest! I was curious to see how well the different output levels were heard and moreover, to see just how far I could get at these lower power levels.

When I first started using WSPR on 30m I had taken a simple approach to a 30m antenna – load up one of the existing antennas and see how it gets out. The choice of antennas was simple – take the 160m inverted L (with a 65 foot vertical section) and use a remote antenna tuner at its base to deal with the impedance match required.

The antenna worked great at the 5w level from my TS-B2000 and WSPR with spots all over the WSPR world – the bottom of the solar cycle taken into account!

Hams being a competitive crowd, I started looking at the different WSPR stations across the US and the quality (distance & SNR) of their spots compared to mine. Larry WB3ANQ came out hands down ahead of my results – stunningly so!

Larry's <1mw WSPR signal had made it to VK6DI in Western Australia while the best I'd been spotted by VK6DI was at 200mw even though the path from Larry to VK6 is almost 4000 km longer than my path!!!

This warranted some investigation!!!!

Digging into Larry's web site at www.wb3anq.com I found that Larry was using a Force12 Sigma-40XK for his 30m antenna.

The Force12 Sigma-40 is a vertical dipole that looks like an "H" on its side. The antenna uses a combination of end loading (the flat ends of the H) plus a center loading coil.

I decided to compare a simple vertical dipole with end loading mounted 10 feet above ground versus my 160m inverted L – both operating on 30m – using EZNEC. The results were very interesting:

30m Vertical Dipole

As expected the vertical dipole is omni-directional with a peak radiation angle of 21 degrees. I modeled a vertical dipole with 14 foot end elements and adjusted the length to get resonance at 10 MHz – this was a rough model for the Force12 Sigma-40.

 

N6TTO Inverted L

 

My inverted L is mounted roughly NE-SW with the horizontal flat section running to the SW – in this diagram that corresponds from left (NE) to right (SW). The pattern is much more symmetrical on 160m than the above plot – not really a surprise because the antenna is many wavelengths long at 30m. The antenna is still mostly omni-directional but with higher gain to the NE. The radiation angle on 160m is about 30 degrees but rises to almost 40 degrees on 30m according to EZNEC.

 

Could the antenna make THAT much difference?

I thought about this for a while and decided to try another modeling step – using VOACAP and W6ELProp to model the effects of the different antenna radiation angles on projected receive signal levels over the path from the West Coast of the US to the West Coast of Australia when using 200 mw output. This is likely model misuse since the VOACAP model was developed for modeling kilowatts not milliwatts and W6ELProp was designed around 100w typical ham transmitter.

Despite the likely misuse, the comparison against radiation angles on each of these models was the same – some projected paths using the vertical dipole but none on the inverted L. Of course, I'd had a number of spots using the inverted L but the model comparison made me suspect that signal strengths would be better using a vertical dipole.

 

The next step was obvious – build a 30m vertical dipole and see how it worked!