I've sent a set of corrections back to the ARRL for my Z90 QEX article. To provide for a uniform appearance, all schematics and figures are redrawn by the League's graphic artists and that opens the door to errors, of course. And, I found a few places where things changed between submitting the article and the current Z90.

I've shipped the Z90's for all new orders for which I have received payment.
 

I spent all weekend working on Z90 documentation, preparing 18 document sets. That might seem as if it would not take all weekend, but it does. To maintain the best reproduction quality and to keep the price down, I've printed all documents on my printers, an HP4000 laser for black and white with color documents and 11'x17" schematics on an HP1120C inkjet. A set of documents runs about 250 pages. In addition to printing, the print run pages have to be collated (in order to speed up printing, it is necessary to print 18 copies of Page 1, then 18 copies of Page 3, etc. Then flip them over and print the backside, as my HP4000 does not have automatic duplex printing. The output is thus 18 copies of page 1, etc. and has to be hand collated into 18 sets of page 1, page 2, etc.) Fold the oversize schematics, insert tabs, etc.

I dropped off the collated and tabbed sets this morning at Kinkos and will pick up the bound copies tomorrow.

I have sent out an E-mail notice to all non-waitlisted Z90 customers, but I have only received a few responses. If you intend to carry through with your commitment to purchase a Z90, please respond to my message.

I plan to ship several Z90 kits tomorrow.
 

Brent Crosby, the technician at Crystalfontz with whom I've been working, has solved the mystery of why a few Z90s have experienced stray characters appearing on the screen and also some display freezes. I've had two customers experience this problem, and I've duplicated it with some unshipped displays as well.

The problem is related to tying Pins 18,19 and 20 of the display to the Z90 ground plane. In particular, the display's Frame Ground (pin 18) is the problem. (Pins 19 and 20 are identified as "no connection" in the display data sheet.) Whether it was a ground loop, or stray coupling or some other problem may never been determined, but clipping these three pins on the display header solves the problem. And, since the problem manifested itself differently with different displays (and there is no problem with most displays) it was a difficult glitch to troubleshoot.

I've verified the fix with two displays that experienced the problem at intervals from a few minutes to a couple of hours. After cutting these three pins, both displays operated error-free for 24 hours each when I terminated the test. The Z90 and buggy display Crystalfontz was using also operates error-free after with the modification.

Instructions on how to make the modification are posted at the Construction Update page. If your display is not showing these errors, you do not need to make the modification.

Solving the display glitch also means that the remaining Z90 kits are ready to ship. I will send payment notices out over the next few days. I will probably continue to test the displays before shipping, including attaching the 20-pin header and making the pin 18/19/20 modification.

Crystalfontz has been extremely helpful in solving this problem. I'm not a particularly large customer of display screens, but I have been given great assistance by Brent Crosby and Brian Lilly at Crystalfontz. If you find yourself in need of LCD products, keep Crystalfontz in mind.
 

I received a pair of Z91 test panels  from Emachineshop today. Mechanically, both are fine, and the rear panel has good paint and silk screening. The front panel, as I was told, is in fact silk screened upside down.

The upside-down status of the front panel may not be clear in the top photo, but it is obvious in the bottom photo. The last information from Emachineshop was that the front panels will be stripped via sandblasting or chemical bath and then re-powder coated and re-silk screened.

I've also had a note from Crystalfontz suggesting a fix for the stray character problem. I've implemented it on a display that showed problems. In the "before" part of the test today, I confirmed that the display still had character glitches. I've just implemented the fix, and will report tomorrow on the results of an overnight run. I'm cautiously hopeful that the problem has been identified.
 

Mike, W2PY, has provided detailed instructions on how he modified the Z90's back panel to mount the Z10010 bandpass filter. I've added Mike's description and photograph to the Constructions Update page. If you plan to make this modification, originally developed by Doug, N6TQS, you will find Mike's information useful.
 

What frequency do you use for your K2 BFO in CW mode? I'm working on a tuning indicator and would like to understand the range of frequencies commonly used.

My BFO is set to 600 Hz, which is a bit higher than I like, but any lower and the K2's built in speaker response is down too much to be useful.

Drop me a note with your BFO setting, whether or not you use a K2.

I've added an AGC sweep for my Cubic R3030 receiver. The AGC study page can be seen by clicking here or via the navigation links at the top left of the page.

I've received an update from Crystalfontz concerning the odd behavior of some of the "C0" graphic display boards. It's possible that they have found the reason some displays sporadically show extra characters. I'll expand on this when more definitive tests have been concluded.  [The latest, as of the afternoon of the 17th is that this was a false alert - the problem remains unresolved.]

I also have copied and expanded the K2 frequency stability test data of 11 January 2007 to a new page, and added cold (20°F) start to room temperature data. You can view the page by clicking here, or via the navigation links at the top left of this page.

I've also added a page presenting the results of AGC operation measurements made on three receivers:

• Elecraft K2
• Drake R7
• Racal RA6790/GM

The data shows audio output versus RF input over the range -130 dBm to 0 dBm. The curves show where the respective designers have placed the AGC threshold and also show how effective the AGC is at holding the audio output constant for variations in signal level. You can view the page by clicking here, or via the navigation links at the top left of this page.

Also, this particular panel shows three white horizontal ghost lines. I've inspected all the panels that I shipped out today and found one more with similar ghost lines. I've kept those for my own use and sent out panels without this problem.
 

I received the replacement Z90 and Z91 panels this morning. I've only unpacked the Z90 panels and looked at four or five, but so far all look good.

These panels are the original color I requested, matte charcoal with white and yellow lettering. It should match the black cabinet covers and your K2 better than the current light paint / black silk screen color scheme.

I plan to package up the panels and ship them out to customers who have built their Z90's and Z91's first. I should be able to get that done early next week. I will also notify customers waiting for the re-done panels to arrive that their kits are ready. I plan to ship the Z91 kits first, as I am still working with Crystalfontz on the random character issue and they have almost all of the display modules I was reserving for the Z90 deliveries.

Here are the Z90 panels. The finish is a matte charcoal/dark gray. The back panel lettering is OK, but the flash reflections make it look odd in the photograph. This afternoon, I'll add a photo of a completed Z90 with the panels.
 

I bought a used Racal Dana model 1992 counter, with IEEE-488 and high stability oven options and it arrived a couple days ago. To the best of my ability to check its calibration against WWV, it's as close to right on as I can determine. I used an indirect method, using the internal counter function in the R3463 spectrum analyzer, to measure WWV at 20 MHz. In the 1 Hz resolution mode, the R3463 reads 20.000000 MHz, which says that the R3463's 10 MHz time base must be quite close to correct; certainly with an error of less than 0.5 Hz at 10 MHz. Measuring the R346's 10MHz time base output with the 1992 counter shows 10 MHz within 0.05 Hz. Unless I can snag some time with a Rubidium or GPS-disciplined standard, I'll consider both devices close enough to right for my purposes.

I've used the 1992 counter to run two stability tests on my K2. I set the K2 to USB mode and locked the dial at 9999 KHz, and connected it to the 1992's 10 MHz output port through a 30 dB 100 watt attenuator. The counter input is connected to the K2's audio output. If my K2 were perfectly calibrated,  the result would be a 1000 Hz output; in fact it's pretty good in that regard, but that was not the primary purpose of the test. Rather, it is to see how stable the K2 is, both from a cold start and as it changes temperature cooling down from a transmit cycle and during a transmit cycle. I collected data from the 1992 via the IEEE-488 interface bus and a Prologix USB-to-IEEE-488 adapter. http://prologix.googlepages.com/. This card looks like a serial COM port to the computer and as a full bi-directional IEEE-488 port to the device it is attached to. If you need an inexpensive ($120 or so) IEEE-488 adapter, I highly recommend it. It is not fully compatible with LabView and most other software written for a National Instruments card, so if that's your need, take a look at Prologix's FAQ page for more detail. I use it with PrintCapture to grab plotter and print images from several instruments. It can be used with your favorite language to write control programs via mimicking of a standard serial COM port. In this case, I used Terminal to capture the 1992's output data, which I then reformatted in Excel and plotted with Origin 7.5.

My K2 resides in my basement workshop, and the temperature stays in the range 70-75 degrees this time of year, so this is not a test over a wide range of temperature. It is, however, representative of typical ham shack environments, at least for those of us who do not take the K2 along on winter camping trips, or operate mobile.

The data below represents 14 hours of measurement, starting around 5 PM and concluding this morning around 7 AM. There's about a 60 Hz excursion from cold start to stabilized, with around a two hour stabilization period, although the worst of the drift is over in less than one hour.

At the end of the test period, I disconnected the counter and did a key-down transmit sufficient to warm the hottest spot on the heat sink to 145F, as measured with a Fluke 62 Infra-red thermometer. The K2's internal fan was running at maximum speed at this point. Upon return to receive, the frequency had dropped about 30 Hz, and it started to recover when I terminated the data set and started a second data set. The downward spike at the transmit interval represents a partial counter reading and should be disregarded. Note that the K2 continues to drift downward after returning to receive mode, as the heat sink's energy is conducted to the rest of the chassis.

I am still waiting on an answer from Crystalfontz on why some displays have stray characters pop up whilst most work without a problem. The results so far are negative -- they believe the data timing I built into the Z90 firmware is OK. I'm not surprised by this, as my timing values are extremely conservative, and well within the data sheet values.

I've added a new page to the site presenting some quick phase noise measurements I made at 13.5 MHz on a variety of signal generators available to me, including the Z90 in signal generation mode. It comes as no surprise to find that the venerable HP8640B is the hands down winner of the five signal sources I tested. Click here to see the data or go to the oscillator noise page via the navigation links.

I've also been experimenting with low frequency antenna systems a bit, and took advantage of the abnormally warm weather over the weekend to refurbish my old home brew active (voltage probe) antenna.  I believe there's significant room for improvement in power-line related noise reduction through better isolation and grounding, but it's still possible to receive WWVB at 60 KHz with a standard receiver (Drake R7) connected to the active antenna during the day with decent signal-to-noise.

Here's WWVB as seen on my Advantest R3463 spectrum analyzer, in a zero-span amplitude versus time plot, with 200 Hz RBW and 10 Hz VBW. WWVB uses an amplitude-level digital modulation system, with one level at full power (around 50 KW ERP) and the other level a -17 dB reduction. It looks as if the 17 dB drop puts the signal right at my current noise level.

The plot below uses a vertical scale of 5 dB/div to better show the difference.
 

According to Emachineshop, the replacement panels are now scheduled for 15 January 2007. I've revised the status report accordingly.

I exchanged a flurry of E-mail messages yesterday with the ARRL's editor working on my Z90 article. I should receive a proof copy in the next few days. It's not clear if the second article, describing the Z90's Gaussian crystal filters, will appear in the Mar/Apr issue or if it will be held for a later edition.
 

I've added a new page showing how you can use the Z90 to check the calibration of your signal generator's FM modulation using Bessel nulls. You may view the page by clicking here, or from the navigation menu.
 

QEX's "In the next Issue" paragraph says that my Z90 article will appear in the Mar/Apr issue. I have not yet received proofs and hope they arrive soon.
 

I received a report that a Z90 printed circuit board had poorly etched pads associated with the 1 watt, 100 ohm surface-mount resistors (the bias resistors for the Gali-74 amplifiers). The pads were  not fully separated from the ground plane. Although this problem will normally be detected by the stage resistance and voltage checks, it's always better to find a problem before parts are installed than discover them afterward. Hence, if you have yet to assemble your Z90, it may be worth checking these pads for an etching problem. A magnifying glass may be useful, or you may make ohmmeter checks to verify that isolated pads are not tied to ground by poor etching.

This is the first PCB production problem report I've received, so I believe it is an isolated incidence. But, boards are normally etched in large panels and the sheared into their final size. This means that there may be another board or two with etching problems. I'll add this information to the builder's update page.

I recently purchased a new (to me, at least) Advantest R3463 spectrum analyzer. I've been trying to understand how it works and have a basic understanding of which buttons to push. I expect to receive an operation manual in a few days, which will help, I'm sure.

Here's an image from the R3463 showing the entire two meter band, 144-148 MHz, in "peak hold" mode, whereby the trace shows the strongest signal received during the time peak hold  has been engaged. This screen capture shows about three hours of activity. The strongest signal is the 146.79 MHz repeater in the City of Fairfax, perhaps six or seven miles airline from here. The spectrum analyzer is connected to a 25-1300 MHz discone antenna about 75 feet above ground.

The R3463 has a color LCD display but I've selected black-and-white mode for data save as the traces are easier to interpret without the color distraction.