Softrock Lite 6.2
Adventures in Electronics and Radio
Elecraft K2 and K3 Transceivers
April 2007 Archives
30 April 2007
Another installment of the Gateway computer saga occurred
today. The UNISYS contract technician showed up with a new speaker assembly
(damaged by the first UNISYS tech) and a new mother board, the second Gateway
has sent for my NX860XL laptop.
After installing the new mother board, the computer booted
Windows normally, but the video failed whenever the AC power module was plugged
To bring a conclusion to this saga, the technician
reinstalled the original mother board (which has worked flawlessly as long as
I've owned the computer). I told him to cancel any orders for yet another
replacement mother board. I'll take my chances on this one. (The reason for the
replacement board was a recall from Gateway, but my laptop has not experienced
the problems that are the reason for the recall.)
This is my last Gateway computer purchase.
I received the second prototype Z100 LED tuning aid
printed circuit boards today and assembled one. It works well and all the errors
I made on the first board layout are corrected.
I'll post photos of the Z100 tomorrow and add a Z100 page.
I plan to order parts and boards for an initial kit run over the next couple
days, so I should be shipping kits in about 10-12 days. I plan to bring some
with me to Dayton as well.
I also have additions to make to the Cohn
crystal filter page, but after dealing with the computer issue today, I
don't have enough energy to add the new material.
29 April 2007
I have one Z91 kit remaining, being held for a customer
wishing to wait until the final panel set arrives. That event is supposed to
happen within a few days, according to Emachineshop. All other Z90 and Z91 kits
After shipping nearly 40 kits, I've experienced what may
be the first "lost in transit" event. I shipped a kit to a customer in Austria
on the 9th of April and it seems to have vanished. It was shipped via Global
Express Mail, which is a 4-10 day delivery method, and the tracking number shows
nothing after acceptance at the Clifton VA post office.
I built an experimental Cohn-style crystal filter today,
nominal 3 KHz bandwidth, but both the simulation and measured filter are
narrower, about 2.3 KHz. I've added a page showing the filter design and
measured results here. I don't recommend a
Cohn filter for a swept frequency system, such as a panadapter, but it can be
used for that purpose.
A major advantage of a Cohn filter is that since all
coupling elements are identical, it is possible to obtain multiple bandwidths
with a single set of crystals, by switching shunting capacitors in and out. Or,
one could use varactor diodes for continuous tuning, perhaps driven from a D/A
converter. The main complicating factor to be dealt with in such an design is
that the filter's impedance changes with bandwidth. And, a crystal filter must
be operated with the correct source and termination impedance to avoid
potentially nasty passband ripples. Hence, in addition to switching coupling
capacitors, a multiple-bandwidth Cohn filter requires switched termination and
source impedances as well.
Why is a Cohn filter better suited for variable bandwidth
than a standard crystal filter. you may be thinking. The answer is that a Cohn
filter uses identical coupling elements, with no series capacitors between
crystals. Hence, it is simple to switch different capacitance values in and out,
without worrying about how to do this with floating small value capacitors.
(Actually, this statement is not quite right, as the Cohn filter is often
designed with series capacitors at the input and output, so that each mesh is
tuned to the identical frequency. But, it's simple to convert the series
capacitors to shunt capacitors via a series/parallel transformation.) See the
Cohn filter design page to see how this is accomplished.
25 April 2007
I've been out of commission with a computer problem for
the last several days. It's a long story, but the Cliff's Notes version is that
I received a telephone call from Gateway late last week, telling me that my
NX860XL laptop was the subject of a recall, to correct potential video problems,
and that I would receive a replacement mother board in the next day or two and
that a Unisys technician would be out to replace the board. Sure enough, the
board arrived Friday last week and the technician showed up Saturday.
Things went downhill in a hurry after this. My laptop
computer had been working perfectly--not a sign of any video problem the recall
was about, or any other problem. When the Unisys tech finished, I had a dead
computer. I watched him work a bit and found him to be not very gentle with
small connectors and parts.
After a series of telephone conversations on Monday,
Gateway thought the problem was a faulty inverter board that powers the CCFL
backlight, and promised to get one out and to send a second Unisys tech out to
install it. The second tech showed up this morning and found that the first tech
had damaged the replacement board, ripping off a small surface mount connector
and breaking the shell on a cable from the laptop lower frame assembly. He also
found a few screws missing. However, he was able to move the original mother
board back to the laptop and get it working again, although at least one
connectors is jury rigged, without the shell in place.
At the moment, I'm still waiting on Gateway to tell me how
they will make this good. I've asked for a replacement laptop that I can swap
the hard drive with and they can then deal with the problems they caused at
Needless to say, I find this irksome in the extreme. It's
one thing to have a problem not properly fixed, but to take a perfectly fine
working computer and damage due to incompetence it is something else. This is
the third Gateway laptop that I've owned and it will be the last unless they
convince me otherwise.
I finished revising the Z100 LED tuning bar PCB and because I made more
changes than I had originally
contemplated, I ordered three more prototype boards that I hope to receive
before the week is out. Assuming the prototypes are satisfactory, I will order
a production run on the boards and parts and will begin accepting orders in a
couple weeks. I hope to have a quantity of Z100 kits to take to Dayton. This
depends, in part, on retaining use of this computer between now and then, as
all my design software, etc. is installed on this machine.
I had planned to add pages for the Z100 LED tuning bar and
the Z1502 TDR-computer interface adapter over the weekend, but my web publishing
software is on the laptop.
20 April 2007
E-machineshop advises that the Z91 panels will be shipped about the 30th of
April. I hope these are acceptable.
I've also added more material to the lattice crystal filter
page. Andy, G4OEP, suggested that I might have used two crystals that were
particularly spurious-laden in my filter. A thought I had was that the spurious
level was increased due to overdrive, as the 8752B VNA default drive level is
-10 dBm. I've looked at both possibilities and documented the results at the
lattice filter page.
19 April 2007
I've done more work on the lattice filter, designing and building a bandpass
LC filter to reduce the spurious responses. The data is added at the bottom of
the lattice filter page.
17 April 2007
I've added a page presenting the
results of experiments with a simple lattice crystal filter. Although the
lattice design largely solves one problem (filter asymmetry), it introduces a
new, and worse, problem, spurious response feed through. To view it, click at
the link in the first sentence, or with the navigation links at the top of
16 April 2007
I finished building the LP100 wattmeter today. Go to
Telepostinc for details on the LP100.
For some unknown reason, I had a problem winding the voltage transformer in the
coupler. It's 26 turns no. 20 AWG magnet wire on an FT140-61 core, so winding it
is not exactly rocket science. Still, it took me three attempts to get it right.
I think the problem was related to uneven turns spacing and possible gaps
between the wire and core surface. From Larry's comments, my troubles with the
transformer are unusual to unknown. I've suggested some manual changes and
documented a process to (I hope) help other builders wind higher quality
After finishing the coupler, I calibrated the wattmeter,
which went smoothly. All the trims and offsets were well within specification. I
calibrated the power level using a Bird 100 watt 30 dB attenuator (measured at
29.95 dB actual) and a Tektronix TDS430A 400 MHz digital oscilloscope. The main
problem I found was that the K2's output power is not stable. It varies over
time by perhaps 10%, e.g., set for 20 watts nominal, it might start at 19
watts and then climb to 21 or 22 watts and then go down again. (This is a real
variation, as the K2's internal meter, the LP100 and my TDS430A all show power
climbing and dropping over time.)
All in all, a thumbs up for the LP100.
The LP100 coupler - voltage transformer is at the
left, current sample transformer at the right.
14 April 2007
E-machineshop says that the 3rd replacement set of Z91
panels are out of powder coating and are being silk screened. I anticipate
receiving the panels next week, assuming there are no further foul-ups. I will
mail replacement panels to Z91 customers and be ready to ship the one
remaining order being held pending final panels.
I've posted a comment on the Elecraft reflector
suggesting that if sufficient demand exists, I will consider a "son of Z90"
panadapter, with significantly simplified electronics, and packaged in an
Elecraft EC2 enclosure. I'll call this the "Z92." So, if you want such a
device, please drop me an Email and let me know.
My goal is to achieve a significant price reduction in
the electronics and still retain reasonable performance. Part of this
simplification will be an analog oscillator, instead of the AD9851 DDS used in
the Z90/91.This means that t he Z92 will have hardware to change for different
IF frequencies, although I might have a switching mechanism to provide two
Also, while a computer-only panadapter has some
benefits, about 75% of you purchased as Z90, not the Z91. And, that meant that
the unit cost for the custom Z91 parts (enclosure and panel) were way over
what I budgeted. So, if there is a Z92, it will be one model, with an LCD. The
LCD I have in mind will be quite a bit smaller than the current Z90, as it
must fit into the 3" (70mm) space available in an EC2 enclosure. It will
have a touch screen instead of mechanical switches, but will still be 320x240
pixels. This is, however, based on reading spec sheets and I have some concern
about the physical fit as the display height is exactly the amount of internal
space in the EC2.
The Z92 will also likely have
only one resolution bandwidth, around 2 KHz. This is because several
purchasers have asked for a wider span than the present 250 KHz. Widening the
RBW to 2 KHz or so offers the possibility of a 500 KHz span width.
It will interface with a USB port, or at least that's my
As far as pricing goes, I have
no idea what it will turn out to be. At the moment, I have the circuit laid
out in my head, and on the back of a piece of scratch paper, but I have not
priced anything yet. The EC2 enclosure and LCD are the cost drivers, as well
as the custom mechanical work in punching and silk screening. My goal is to
deliver a product with the LCD and EC2 enclosure for less than Z91 price, but
please don't hold me to that, as these two parts, plus the custom mechanical
work means well over $125 in parts cost with zero panadapter stuff.
The EC2 enclosure is way larger than needed for a Z92,
so it may be useful to add other options to it, such as a power supply, or a
Drop me a message if you are interested and also if you
are a Z90/91 customer tell me how much you use the 200 Hz RBW and narrow spans
versus 1 KHz and wider spans.
I've spent the day assembling an LP100 wattmeter kit from
Telepostinc. As I've mentioned, I've
known Larry, N8LP, Mr. Telepostinc, since we were both college students in the
Detroit Michigan area. The main board is finished and the coupler, and the major
remaining task is calibrating the assembled unit. I'll post a photo or two of
the build tomorrow or Monday.
10 April 2007
For comparison with the suburban Detroit broadcast band spectrum analyzer
sweeps shown on 09 April, I captured the same data here in Clifton VA this
morning, using my Advantest R3463 spectrum analyzer with identical settings. The
antenna is an 80 meter inverted vee.
The strongest station here (-23 dBm) is WDCT, 1310 KHz. WDCT is a three-tower
directional station, 5 KW output, located in Fairfax City, perhaps six or seven
airline miles from here, with the main lobe of the directional pattern in this
09 April 2007
My friend received the filter I discussed on 06 April and
sent me a spectrum analyzer sweep of the broadcast band with the filter in
place, as well as a "before" sweep.
The marker (green line) is at 1500 KHz, WLQV, a 50 kw
directional station about three miles from the receiving point. Before the
filter was in place, the signal level was about 0 dBm. It's now -73 dBm. WJR,
760 KHz, is not knocked down as much, but it's now a tolerable -48 dBm or so,
about a 38 dB reduction from the no-filter case.
After the filter is installed
Before the filter was installed
I received three prototype PCBs for the Tektronix 1502 TDR
analog/digital conversion module that I'm working on. I'll add a page describing
the module tomorrow or the next day, but for now one photograph will give an
idea of how it looks.
I've been working on a adapter
board and associated Windows software to capture the XY recorder module output
of a Tektronix 1502 TDR and it's at the point where I will soon have a unit
available for beta testing.
For those who may not be familiar with the term "TDR" and
how it works, TDR stands for "time domain reflectometer" and it can be thought
of as a "cable radar" in that it shoots a very fast rise time pulse down a
coaxial cable and then reads the reflected signal. You can see the impedance
bumps caused by connectors, where the cable has been compressed by clamps or
bends and other discontinuities, and read the distance to the discontinuity with
quite remarkable accuracy. If you wake up one morning and the SWR on your
antenna suddenly reads 10:1, a TDR can tell you whether the problem is at the
antenna or if it is a problem with the cable or connectors, and if the problem
is with the cable or connectors, it will locate the fault to within a foot or
Tektronix 1502s show up with reasonable regularity in the
surplus market and are not terribly expensive. Many come with a mechanical
recorder module that is often broken or unusable due to the cost of the
specialty paper. However, a reasonable percentage have an XY output module that
provide a buffered output that follows the trace data. My adapter module plugs
into the XY adapter and connects to a PC via the USB port.
The associated software gives a display similar to that
seen on the 1502's screen, but with the ability to print, save, etc. as well as
saving the data file for analysis by other programs. If there is sufficient
demand, I may develop a second version that will plug into the 1502 mainframe
directly, as a replacement for the mechanical recorder.
06 April 2007
I received the PDF galley proofs of my article on the
Z90's crystal filter design Wednesday, and returned it with a couple small
corrections yesterday. It will appear in the May/June QEX.
Yesterday, I designed and built a low pass filter to help
a friend in suburban Detroit MI with a major AM broadcast intermodulation
problem receive signals below 500 KHz. I started with the excellent filter
design program from AADE. (Freeware, available for downloading at
Well worth downloading and installing if you have even a passing interest in
filters.) Here's the AADE computer designed filter:
I made the two 19.3μH inductors by winding 33 turns no. 26
AWG magnet wire on a pair of T50-3 powdered iron cores. (-3 mix material is
suitable for frequencies between 10 KHz and 5 MHz.) Both measured within 5% of
the design value and showed a Q of 160. The capacitors are parallel
combinations of standard value 5% polystyrene capacitors. I built the filter on
an extra Z10010 bandpass filter board.
The measured response shows that it knocks down the
main interfering signals at 760 KHz (WJR), 1310 KHz (WDTW) and 1500 KHz (WLQV)
by more than 35 dB, with about 1 dB loss in the desired passband. Data taken
with an HP87510A gain/phase VNA.
The filter design is a 1 dB Chebyshev, so there should be 1 dB passband ripple,
which the following expanded sweep demonstrates. The data noise below 4
KHz is a measurement artifact, as the 87510A is only rated to a minimum
frequency of 100 KHz. In fact, it works pretty well down to about 3-4 KHz.
03 April 2007
Yesterday was occupied by having the driveway replaced. This spring's wet
weather, combined with a propane delivery via heavy truck finished off about
half the old driveway. H&H Paving removed the damaged part of the driveway,
installed a gravel bed and then an asphalt top, including re-topping the
undamaged part of the driveway. In digging up the damage, it turned out
that the original paving was done without a gravel bed, which explains why it
If you live in Northern Virginia and need similar work,
I can recommend H&H Paving. They were the most expensive bid, but showed up on
time, finished on time and did exactly what they said they would do. The crew
that worked on my job never stood still -- they were in motion from start to
finish. Also, their machinery was in good shape and clean. Even the dump truck
hauling the hot asphalt looked like it just came out of the wash rack,
including tire blacking.
The photo below shows a load of asphalt being loaded
into the spreading machine.
The finished job, showing the road entrance. I had a pad added for the trash can
pick-up point, and also increased the entrance width. Spring has arrived in
Northern Virginia, but the trees are still budding out.
The garage area of the driveway. I had two white stripes put down showing the
garage width to help me when backing into the garage.
As usual, the prior month's data has been moved to the
archives, March 2007 in this case. The March 2007 page may be accessed by
clicking here, or via the table links at the top of
A week or so ago, I was fortunate to acquire an HP 5065A
Rubidium vapor frequency standard, recently calibrated against an HP 5061A
Cesium primary standard. I used the 5065A to adjust my Racal 1992 frequency
counter time base and then to see how much the 1992's time base drifts.
Rubidium standards are "secondary" standards, in that
although they are highly stable, their absolute frequency must be occasionally
recalibrated against a higher accuracy standard, one which depends upon the laws
of physics for its absolute frequency, such as a Cesium standard. This is a
highly simplified view of frequency standards, and for more information I
suggest reading HP's (now Agilent) Application Note 52, Frequency and Time
Standards, currently available at
http://cp.literature.agilent.com/litweb/pdf/5989-6171EN.pdf and Application
Note AN-52-2, Timekeeping and Frequency Calibration, available at
http://cp.literature.agilent.com/litweb/pdf/5989-6247EN.pdf. If these links
do not work, try a Google search for the Application Note titles.
The 5065A Rubidium standard has a rated stability of ±5x10-13
averaged over 1000 seconds, while the 1992's high stability oven time base
(Option 4E) is rated at ±5x10-10 over 24 hours, so any observed error
will be dominated by the 1992's time base.
Before starting stability measurements, I calibrated the
1992's time base against the 5065A, and found rather odd behavior. After
adjusting the fine adjustment, the 1992's time base made an instantaneous
frequency move (as expected) but then made a slow (period of hours, or days)
drift back towards the original value. This made calibrating the 1992's time
base challenging, as it is necessary to guess the degree of long term shift and
correspondingly offset the adjustment. After a couple of days, I had the error
at zero. (The resolution in my calibration and data collection setup is 1x10-10.)
The plot below shows one week's data. As can be seen, the
first adjustment reveals the increased drift period mentioned above. After two
more adjustments, the stability improved to be well within Racal's specification
of ±5x10-10 over 24 hours. In fact, the data shows it remained within
±5x10-10 over a five day period.
these numbers into prospective, an error of 5x10-10 corresponds to
0.005 Hz at 10 MHz. This is perhaps 100 times more accurate than one can measure
WWV's 10 MHz signal off-the-air due to Doppler and other ionosphere
The vertical spikes in the data are due to the collection