Clifton Laboratories 7236 Clifton Road  Clifton VA 20124 tel: (703) 830 0368 fax: (703) 830 0711



To search within the Clifton Laboratories site, enter your search term below.


Software Updates
Softrock Lite 6.2
Adventures in Electronics and Radio
Elecraft K2 and K3 Transceivers


February 2007 Archives

27 February 2007

Yesterday, I received the prototype printed circuit boards for my new project, the Z100 CW tuning aid. I assembled one board today and tested it. I found a couple of small errors, but nothing that can't be fixed with an Xacto knife and a piece of wire for the test boards. I hope to send two prototype kits to a couple of volunteer builders by Monday.  The Z100 will be a two or three hour kit, with most of the time being installing the LEDs so that they align with the board edge and are equally spaced. (Spacing is set during assembly with temporary 0.009" paper spacers.)

The boards look good:

The front of the PCB has 24 LEDs. The frequency step for each LED is 25 or 50 Hz, depending on an option switch setting.
The bottom of the PCB has option switches and a center frequency tuning potentiometer. These are on the bottom as I plan access holes in the case bottom.

I've also worked today on Z90-Display version 2.0. I can now read parameters from the Z90 and set parameters on the Z90, as well as display a standard trace and a waterfall display. The standard trace has the same three line status/title/date-time display found in version 1.0.

The waterfall image below shows all the controls populated with data linked from the Z90. The signals displayed are from connecting an antenna to the Z90, set for 10 MHz. I hope to provide a selection of palettes, not just the gray scale shown.

There's still a lot of work to be done, but I hope to get an alpha test version out next week.

26 February 2007

I've started re-writing the Z90-Display program in Delphi 2006. The current version is written with Liberty Basic v 4.03, and although it works, some limitations have recently surfaced. At the request of a Z91 owner, I've decided to offer Z91 owners a new firmware load with considerably faster sweep speeds. The speed increase is a consequence of splitting the Z90 and Z91 firmware and deleting code associated with LCD control from the Z91 firmware. I do not plan to make a general release of the revised Z91 firmware until Z90 Control version 2.0 is ready, as the current release produces objectionable flicker when running with the increased sweep speed permitted by the new Z91 firmware.

At the moment, I have an engineering version of version 2.0 running but it's a long way from even a test release. Assuming other things do not get in the way, I hope to have a test release in a couple of weeks. Version 2.0 will add a waterfall display option and may add other features as well. Of course, the waterfall display will work with either a Z90 or Z91.

A preliminary view of the version 2.0 screen is shown below. The main differences from version 1.0 are:

  • Tab selection of amplitude / frequency display and waterfall display
  • Vertical position is set via a slider control.
  • Progress bar shows incoming data receipt.

The extra buttons and display window at the bottom of the display are for program development and will not be in the release software.


When is a 0.1uF capacitor not a 0.1uF capacitor? I've also added a new page Capacitor Voltage Change (click here to read it) presenting data I collected showing the variation in measured capacitance versus applied DC voltage for six capacitor types. Some high dielectric capacitors show major variation in capacitance with applied DC polarizing voltage. If these capacitors are used for coupling signals with large voltage excursions, harmonics and intermodulation distortion can be generated by these "passive" components.

I also note that Andy, G4OEP, has newly revised pages on the effect of turns spacing on toroidal inductors,  as well as his analysis of the G3UUP crystal oscillator method of measuring motional parameters.  Andy's home page is and you will find many interesting topics linked there.


21 February 2007

I've worked more with reading output from my Advantest signal generator, and have a waterfall display working. In testing the code this afternoon, I looked at the frequency range 144-154 MHz, using my 2 meter 6 dB gain omni-directional antenna, at 120 ft above ground. To improve the noise figure, I've added a Minicircuits ZFL-500LN preamplifier ahead of the spectrum analyzer.

There's an interesting occurrence in the data. Here's a small image:


Even in the small image, the diagonal line is quite visible. Here's the full scale image of that part of the data.

The center white line is 147 MHz, and the lines at either side are 146 and 148 MHz. The diagonal trace represents the sequential keying of repeaters, with their outputs in the 146.6 - 147.39 MHz range. I can identify repeaters at 146.79, 146.91, 146.97, 147.24 and a couple others.

If you look carefully, three "blips" on the repeater input range below 146.6 MHz can be seen, and a weaker, but similar sequence on the input side of the repeaters operating above 147.39 MHz is also visible. I've marked these signals in the next illustration.

The three input blips for repeaters operating in the range 146.6-147.0 MHz are inside the left ellipse, and the input signals for the repeaters operating in the range 147.0-147.39 MHz are parallel to the yellow line. The spacing matches the standard 600 KHz VHF repeater offset, -600 KHz for repeaters below 147 MHz and +600 KHz for repeaters above 147 MHz.

My first thought was this is the result of a sweeping spurious signal hitting repeater inputs. However, if it is a true spurious, however, there should be a dead band in time, where the spur sweeps the repeater output frequencies, but I don't see that, which says  this is most likely the result of someone going through the various repeater input frequencies and keying them up in sequence.

When I have the waterfall display software completed, I'll probably do a simple version for the Z90/91. The main problem with general release of a Z90/91 version of the program is that writing individual pixels to a computer screen is very CPU-intensive in Liberty Basic and the program does not coexist well with other programs. My development computer is a Gateway duo-core machine, and the program grabs 100% of one core, or 50% of total CPU resources. On a single core computer, almost all resources will be grabbed, unfortunately. I can throttle back the resource grabbing, but the paint speed becomes painfully slow.


19 February 2007

Emachineshop says that the Z91 front panel job will be re-made again, starting from scratch. Based on past history, this process will run about 75 days, which places delivery around the end of April.

I'll contact the customers awaiting Z91 deliveries and advise them of the delay and again offer to deliver the kits with interim front panels and final rear panels.

I've also added a new page with measurements of various bypassing strategies. It can be found via the index at the top left of the page or by clicking here.

I've also added a power rating update to the dummy load page based on Celwave catalog information provided by Tom, N0SS.


18 February 2007

I've cleaned up the quick and dirty program I wrote several days ago to capture data from my Advantest R3463 spectrum analyzer, and looked at an hour's signal strength data from CHU at 7335 KHz. The data was taken about 30 minutes either side of noon here in Northern Virginia. The R3463 was connected to an M2 log periodic antenna at 100 ft above ground. The spectrum analyzer was set for 200 Hz resolution bandwidth and 10 Hz video bandwidth, with a 20 second sweep, 1001 points per sweep, corresponding to 50 ms per data point.

The raw data is shown below. Of course, all 180,000 data samples can't be shown at this resolution. The red line is a 60-second moving average to show the longer term trends. The data shows at least two superimposed fading patterns. One of relatively short duration, with deep fades of 20 to 40 dB and the second a slower trend, over the space of tens of minutes, although rather sharp excursions in the moving average are also visible

Plotting the relative distribution signal strength probabilities shows, as expected, a non-Gaussian distribution.

The two references I rely upon in HF progagation matters are both by Kenneth Davies. They are Ionospheric Radio Propagation (National Bureau of Standards Monograph 80, April 1, 1965) and Ionospheric Radio Waves (Blaisdell Publishing, Waltham MA, 1969) If you are seriously interested in HF propagation, these books should be in your library.

Before discussing fading statistics, it's important to know that we are not considering slow variations in signal strength, e.g., measured over the space of hours as the ionosphere changes state from night to sunlight or vice versa. Rather, by limiting our data to the space of an hour, during a time when both ends of the radio path (Ottawa Ontario and Northern Virginia) have been in sunlight for several hours, we examine the relatively fine variation in signal strength, fades measured in terms of a few milliseconds to a few seconds. Mixed in with this fast fading data is, of course, the longer term changes in signal level, as evidenced by the red moving average line in the first graph. A better analysis would remove these longer term trends, but we will content ourselves with noting that the total slow variation in signal strength is approximately 10 dB, and for the most part the the variation is less.

Prof. Davies states that HF signals, looked at over a short time frame exhibit generally fading that can be described by one of two statistical distributions:

  • Rayleigh fading�the signal that arrives at the destination is the sum of a large number signals, each with a time changing path so that their vector sum increases and decreases as the individual components phase and amplitude relationship changes.
  • Ricean fading�the signal is dominated by a single path, with several other signals also present, but with the other signals being weaker. Again, the receiver detects the vector sum of these multiple signals.

The names Rayleigh and Rician derive from the names of the scientists that developed the statistics, Lord Rayleigh and Stephen Rice. Rice worked at Bell Laboratories and studied random noise processes and Lord Rayleigh, of course, was one of the foremost 20th century physicists.

A later extension by Weibull generalized the statistics developed by Rayleigh.

I've plotted a typical Rayleigh distribution using Excel. The blue probability curve shows a definite resemblance to the fading data. (The curves need to be flipped left-to-right to match the scale).  Altering the parameters used to define the Rayleigh distribution will tighten or loosen the curve and alter the tail.

The Excel spreadsheet actually uses Weibull statistics, but with the parameters set to reduce to a Rayleigh distribution.


The last plot shows a cumulative distribution of the measured data. As with the earlier data taken, the fades provide an almost perfectly linear plot over a 40 dB range.

16 February 2007

I finished the last fixes in the Z100 tuning aid PCB layout and ordered three prototype board this afternoon. I have two volunteers for build verification and user experience feedback.

The Z90 PCB set was done with ExpressPCB. While I had no problems with the schematic capture software, the associated board layout software is short several useful features, such as design rule checking for missing connections or short circuits between nets, and it does not have an auto-router, although autorouters are about equally useful and dangerous. And, of course, the ExpressPCB software is proprietary and can only be used with their manufacturing service. (I understand they will generate standard Gerber files for an extra fee.) Larry, N8LP, has also used ExpressPCB for his LP100 wattmeter and other projects.

Both of us decided to shift to a different schematic capture and layout software package. After looking over the reasonably priced alternatives, and downloading and trying several demo versions, we've selected DIPTrace, whose home page may be found at I have previously used layout software from Ivex (now out of business) and Electronics Workbench (worked, but expensive and buggy) and Eagle (either you will love it or hate its quirky user interface and I'm, in the dislike category). I also looked at a couple of the freeware schematic capture and layout packages.

After doing the Z100 schematic and layout with DIPTrace, I can give it a qualified OK. It's relatively easy to learn and use, and it follows normal Windows conventions. I did not find any major bugs, and the program seems to have responsive user support. (The developers are in The Ukraine and are active on the DIPtrace forum.) DIPTrace is missing a few useful features that are scheduled to be implemented over the next few months. It has a reasonable pricing schedule and an upgrade path that lets you purchase capacity as necessary. I bought a "DIPTrace Lite" license, which works for double sided boards, up to 500 pins for $145. The Z100 board has about 194 pins. I don't know how many pins the Z90 PCB would have under DIPTrace's computation mechansim, but I'm sure it would be well over 500.

No response from Emachineshop on the Z91 front panel problems. I'll have a telephone chat Monday and see what they propose to do for the most recent problems.

15 February 2007

I've added a note to the Construction Notes page providing a trouble shooting methodology for mixer, crystal filter and log amp problems. I developed this approach when running down a problem with the 1 KHz filter in a Z91 and it worked well. You might wish to print this note and add it to the troubleshooting section of your manual.


14 February 2007

I've taken a short break from Z90/91 work today and worked on a Liberty Basic program to read and control my Advantest R3463 spectrum analyzer via a Prologix USB-GPIB interface card. I spent several hours trying to understand why a code sequence would work well when sent using the keyboard and Terminal program to the R3463, but would fail when sent from inside the program.

After solving the mystery, I realize that I should have found it earlier. After running a sweep, the R3463 responds by returning a 0 value when a status register is queried. When set to 0 span to look at signal strength, a useful sweep period typically runs between 10 and 100 seconds. During the sweep period, the R3463 holds the CTS/RTS line but does not send data. Hence, it's necessary when initializing the COM port to set a CTS/RTS time out of longer duration than the sweep period. The default timeout in Liberty Basic is 1 second. After increasing that to 30 seconds for a 20 second sweep period, things finally made sense.

In case you are wondering why all this serial com port discussion for a device controlled over a GPIB port (IEEE-488/HPIB are other terms for the GPIB port), the Prologix adapter can be run as a serial com port interface over the USB port. This lets you use Terminal for debugging and also avoids specialized DLL calls for GPIB communications. Rather, the GPIB device may be addressed just as if it were a serial communications port in your software.

The graph below shows received signal strength levels for a period of 400 seconds. The signal measured is the strong shortwave broadcast station at 9975 KHz, WWCR, and the R3463 was set for 200 Hz resolution bandwidth, 30 Hz video bandwidth. These settings wash out almost all the AM modulation. The R3463 was connected to an 80 meter inverted vee antenna with an apex 80 feet above ground level.

The data shows two periods of strong fading, with a center period of relative stability.

As interesting as this sort of plot is, a much more informative analysis looks at the probability of signal levels. Strictly speaking, this should be done as the probability of a fade of X dB above or below the mean signal level. I did not take the time to compute the mean signal level and subtract the value from the individual data readings, but this adjustment only shifts the curve up or down. Its main usefulness is that it allows one to compare  data with different means, of course. If I have the time, I'll do more data capture and analysis using the fade from mean methodology and add a page with the statistics.

The graph below shows on the vertical axis the percentage probability that the received signal is at least as strong as the corresponding X axis value.  For example, 1% of the time, the received signal level was at least -68 dBm. If we compute the mean of the data set (20,200 data points) we find it to be -51.06 dBm. This means that there's a 1% chance that the signal will experience a fade of at least 17 dB from the mean.

Looking at the probability of fade distribution, the fit to a log distribution is remarkable. In fact, the R value is 0.99, an excellent fit.

The fade probability data can tell us a great deal about the path over which the signal arrives. I've been extensively involved in signal strength measurement and analysis for VHF and UHF paths over the last 30 years but this is my first detailed excursion into HF fading. In  the VHF/UHF mobile radio world, almost all propagation is via scatter as the mobile station seldom has a direct line of sight path to the serving base station. A scattering path is best described by Rayleigh statistics. If the mobile station has a line of sight to the base station, Rician statistics may provide a better fit to the observed data.

HF propagation, as I understand it after a cursory reading of the literature, is often described via a two-ray model, i.e., a dominant skywave path and a second skywave path, with the receiver detecting the sum of the two paths. As the relative strength and phase relationship of the two paths change, the sum of the two signals will differ. If we think about it, it's also apparent why signals fade down far more than they fade up. Assuming each path has 50% of the signal power (not necessarily realistic, of course) if the  two paths arrive with 180 degree phase shift, the result is cancellation and no received signal. If the two paths arrive with 0 degree phase shift and thus mutually reinforce each other, the net power increase is only 3 dB over receiving one path only. Hence, on the up-side we see 3 dB and on the down side, infinite dB, so it's obvious that fades will be biased to reduced signals.

When I have the time, I'll sort through a bit of the HF fading literature and gain a better understanding of the data I'm capturing.

13 February 2007

The latest installment in the Z91 panels from Emachineshop has arrived. To recap, the re-done order resulted in the front panel silk screen being applied upside down. After some discussion with Emachineshop, we agreed they would have the front panel paint removed, chemically or sandblasted, and then repaint and re-silk screen the bare metal panels.

UPS delivered the Z91 panels today. Here's the result.


The photos show a "ghost image" of the old, upside down, silk screening. A ghost image of the white line is visible in both images and in person, I can read the old text as well. If you look carefully in the top photograph, you can see the Clifton Laboratories logo under the paint as well. In my opinion, the panels were not taken down to bare metal, but rather a second coat of black paint was applied. (The white dots are from the Styrofoam packing material.)

I've sent Emachineshop an Email asking what we are to do next. To say that I am disappointed is a major understatement.

09 February 2007

I've been intrigued by the "Luxembourg Effect" since first hearing about it many years ago. The "Luxembourg Effect" is named after the early high power LF broadcast station in Luxembourg where the effect was first noticed in the 1930's. The LE is cross modulation caused by a high power transmission due to non-linearity in the ionosphere. In other words, one station's modulation is superimposed on other stations (albeit at a rather low level) as a result of the ionosphere's non-linearity.

Outside of specialty literature, little is available on the LE, but with the BBC scanning copies of their Research Department Reports from the 1960's onward, a pair of papers on the Luxembourg Effect are now available over the internet. The more interesting paper is BBC Research Department Paper 1972/23 and may be downloaded at

A complete list of the BBC's older research department papers can be found at The actual papers are linked in this document and can easily be downloaded.

Speaking of BBC research, there are several BBC White Papers worth reading, including those on dipole antennas and coaxial cables. Both of these White Papers are fairly mathematically, but it's possible to skip the math bits and still benefit from reading the white papers.

The White Paper index may be found at and is worth perusing for other worthwhile reads, including several studies on Broadband over Power LIne (BPL) and their potention for harm to shortwave reception.

We live in a part of Fairfax County VA that is reasonably undeveloped, due to watershed conservation zoning. Consequently, we have a large wildlife population including fox, deer, raccoon, woodchucks and the like. The deer, in particular, are quite common. I took the two photographs below two days ago.


Mother deer
Her two young. Old enough to have lost their spots but still together as a family group.
08 February 2007

My Z90 QEX article is wrapped up, with the last edits submitted. It will appear in the March/April edition, and should be in your mailbox�what, you are not a QEX subscriber? Sign up now, it's a great magazine for the technically oriented ham�in a few weeks.

I've also finished breadboarding a CW zero beating aid, the Z100. The concept is not original, and dates back at least to HAL Communications' SpectraTune sold 25 years ago, but advances in electronics make the design much simpler. My plan is to offer this as a simple kit, that can be assembled in a couple of hours. It requires +12V and has two 3.5 mm (1/8") stereo jacks in the back. You plug a jumper from one into the K2 speaker or headphone output and the second jack is paralleled for your headphones or external speaker.

The Z100 has 24 LEDs, with each LED corresponding to 25 Hz in frequency. To correctly match your transmit and receive frequency, you tune until the two green LEDs in the center flash. If you are a bit off in frequency, the yellow LEDs on either side of center will flash, and if you are way off, the red LEDs will illuminate.

Breadboard model on top of K2, with a CW signal correctly tuned for zero beat on transmit. I had to hand hold the camera with a slow exposure to make the LEDs and K2 display show up, and I did not hold it steady.

The rubber band holds the top and bottom case together and let me have easy access for measurements and software programming.

Breadboard version. The final kit will have a printed circuit board, of course, but will use the same Hammond 1553B plastic enclosure.

The trim pot lets you set the center frequency to match your CW tone choice. It's settable from 400 Hz to 3000 Hz, so that it can be used for RTTY or other tuning purposes, although the days of requiring a separate RTTY tuning aid have largely dispappeared.

Displaying a 170 Hz shift RTTY signal. When properly tuned in, the two red LEDs on either side of the two yellow LEDs are illuminated.

The LEDs are not in very evenly, of course, which is a function of (a) lack of craftsmanship on my part; and (b) using a pre-made prototype board instead of a custom made PCB.


It's way too early for anything definitive on delivery or pricing, but my hope is to have a price point similar to Elecraft's mini-kits. The Z100 kit, however, will include the enclosure. I'm working on a PCB layout now and hope to have a first prototype PCB version running in a couple of weeks.

If you are a CW operator and would like to build a prototype and provide feedback on its usefulness, please contact me via E-mail. There's nothing K2-specific about the Z100, of course.

02 February 2007

January 2007 is now in the archives and may be read by clicking here or via the link at the top of this page.

My QEX article The Z90 and Z91 Digital Panadapters is now on its second revision, and I believe that the errors have been fixed, but I have not yet had a chance to go through the second proof copy in detail.

I'm caught up on all orders for which payment has been received. To recap:

  1. Z91 orders for customers electing not to take interim enclosures�I'm waiting on Emachineshop to re-make the re-made front panel. It's out for refinishing and re-silk screening, but I do not have a estimated delivery date. However, I believe these should be in my hands before the end of February.
  2. Z90 orders for customers electing not to take interim enclosures�I've notified all customers that their kits are ready to ship upon receipt of payment and have shipped half a dozen or so Z90 kits in the last week. Several customers have not responded to this notice, however. I will make one more contact and if there is no response, I will start releasing kits to those customers on the waiting list.
  3. Z90 replacement panels�All Z90 customers that received interim panels should have received replacement panels a couple of weeks ago. Any Z90 customer not receiving a replacement panel should contact me via E-mail.
  4. Z91 replacement panels�As reflected in (1), I am still awaiting acceptable Z91 front panels. As soon as I receive them, all existing Z91 customers will receive replacement front and rear panels.

Buffer Amplifiers�I made 50 buffer amplifier kits, and I am nearly out of stock. If you do not really need two buffer amplifiers, I would greatly appreciate it if you do not order two. The economics of scale make it very difficult for me to make 10 or 15 more buffer amplifier kits at the current price. I will meet orders as long as I have remaining stock, but when these are gone, that's likely to be it. Or, if I make a run of 10 or 15 additional buffer amplifiers, the price will increase to $35 each.