25 May 2008

I've spent the last week working with a color LCD display. I have a few projects in mind  that may benefit from a color display, possibly with touch screen input.

The image on the right shows 12 nested boxes overlaying a line sequence from a test program I wrote.

The display has a 5.7 inch diagonal and is near quarter VGA  resolution, in this case 320 x 234 pixels versus the 320 x 240 pixels of a true QVGA display.

Getting the data transfer hardware protocol working was difficult, and I was close to packing the display away for another day. I'm driving the display with an 18F4620 PIC, running Swordfish BASIC. The data transfer requires 8 data lines and four control lines, plus a few other control and power lines.
 

I've downloaded Mozilla's Firefox 3, Release Candidate 1 earlier  today and see that the two "pointing finger" symbols in the Sticky Updates page are not rendering correctly, although shown correctly with Internet Explorer 7.0.

For a test, you should see a pointing finger on the line below. A capital "F" is incorrect.

F

I believe Wingdings is a standard Windows installation font. Regardless, it's on my computer and if it's available to Internet Explorer, it's available to Firefox.

I've turned this anomaly into Mozilla's reporting system. It's entirely possible, of course, that it's a problem with my HTML coding. I use Microsoft FrontPage for this site and don't try to tweak or modify the code by hand. I'm not an HTML expert and have no desire to become one, so I rely upon FrontPage to translate my layout into acceptable HTML code.

I've used Firefox as my primary browser for a couple years, along with Thunderbird for E-mail, and I like both programs over their Microsoft counterparts, Internet Explorer and Outlook.
 

I've taken another look at signal generator phase noise, improving the dynamic range of my test setup with a notch filter. The results can be seen at my signal generator phase noise page.
 

The last few days have found me working on computer networking. I decided to add a Raid-1 redundant network attached storage (NAS) box to my home computer network. After researching the available options, I purchased a Netgear ReadyNAS Duo. As the name "Duo" suggests, it has a maximum capacity of two drives. It's available with 500, 750 and 1000 GB drives, and I went with a 500 GB, the model RND-2150. It ships with one drive, with the purchaser responsible for buying and adding the second drive, which I did. With the second drive and shipping, the NAS price was about $450.

I configured it as "Raid-X," Netgear's name for Raid-expandable. With only two drives possible in the box, it's  the functional equivalent of Raid-1, where the data is mirrored in both drives. The result is 500 GB disk drive space, but redundant so that if one drive fails, no data is lost. In fact, the drives are hot swappable, so that it's possible to unplug one drive and remove it or to install a new drive without turning the power off. Installing  the second drive took all of five minutes. It's a matter of opening the hinged door on the NAS, unlatching the empty disk caddy and using four screws to mount the second 500 GB drive in the caddy. Then, slide the caddy back into the enclosure and close the door. The NAS's firmware recognized that a new drive had been added and it proceeded to format and set up the new drive for mirroring. When this process was completed, a corresponding message popped up on my computer screen. (I did install the NAS control software, of course, without which the message would not have been displayed, although the automatic disk recognition and formatting would have occurred whether or not a PC was attached.)

This addition has caused me to rethink and revise how I load and save data. With relatively few exceptions, I have been using one computer mostly as a file server, so all my documents, spread sheets, etc. were on that computer's hard drive, backed up to a second hard drive in the same PC. (I currently have six computers in my home network, which even I admit is a bit excessive. The network uses a 4-port Netgear 10/100 MB wired router, a LinkSys 802.11.g wireless transceiver and a Buffalo Wireless Bridge in the basement shop. Two computers have 802.11.g wireless cards. It's grown up like Topsy, over time and could be simplified with a combination wired/wireless 8-port router.) 

Where this process failed is  that I habitually stored programs I wrote in a sub-directory of the compiler. Hence, all my Swordfish 18-F series PIC programs were stored on the Gateway laptop in the c:\program files\Swordfish directory. Likewise for Delphi programs. While this speeds up disk access, my shortsightedness became apparent when the Gateway laptop was stolen in our house burglary this spring. I had made some backup of Swordfish files, but almost none of my Delphi programs. Although I kept the Swordfish and Delphi installation disks, I lost all my Delphi work and substantial parts of my Swordfish code.

A related issue is backup software. Netgear supplies backup software with the NAS, but it's not suitable to back up Windows operating files and program files. It's fine for backing up working files, such as documents, spreadsheets, etc. This lead me to explore the world of backup software. Backup software divides into two categories; one aimed at making disk images with a snapshot of all  the bits on a disk including Windows and program files, and the other for backing up user files, such as the documents and spreadsheets. There's clearly overlap between these two categories, and a copy of the complete disk obviously grabs all the user files as well as the system and program files. And programs aimed at backing up user files often include a "disaster recovery" module to save and restore the system and program files. After reading PC Magazine's reviews of backup software and after reading on-line manuals and downloading and using trial copies of the best ranked programs, I've decided that Genie Backup Manager Pro 8.0 is the best suited for my use. Acronis True Image is a good alternative. Genie is more a user file backup program with disk image option whilst Acronis is more a disk image program. I think either would work well for me, but I found Genie a bit easier to use during my experiments. But, the difference between the two was slight.

In order to make the backup software and NAS work, I've created a directory (share in NAS terminology) on the NAS disk to hold nothing but my user files, programs, documents, spread sheets, graphs, raw data, the source files for this web site and the like. It's 20 GB in size, although there's probably some duplication in the files, as  they have also grown like Topsy. I need to rearrange my Delphi and Swordfish files to move their code-related files to the NAS box as well.

I should add that although Raid-X means that a single hard drive failure causes no loss of data, it's still possible for both drives to fail (lightning strike, for example) or for the NAS to be stolen or damaged in a fire. Hence, there's still need for a true backup. At the moment, that is to be via a separate USB-attached hard drive plugged into the NAS. Once I have my files sorted out and properly arranged on the NAS, I can then back them up to the USB drive. The USB drive can be disconnected except when the backup occur and kept locked in the fireproof safe I added after the break in. It's a matter of discipline to keep the backups going, and that's something I'm not all that good at.

I should also mention that the NAS has three USB ports, one of which I'm using for a USB-connected hard drive. The other  two USB ports are usable as printer connections, with the printers appearing as attached to the NAS. This means that it's no longer necessary to keep one PC running as a print server. Both the printers I use pre-date the era of direct USB-printer connectivity, so I attach them to the NAS with Belkin USB-to-parallel adapters. My HP Laser Jet 4000 appears to the network as a LJ 4000, but my Desk Jet 1120C color inkjet printer comes up as "unknown printer." It was simple enough, however, to match it with Window's stock 1120C driver. Both printers work nicely through the NAS box.

I've collected more analog phase detector data comparing the Thunderbolt with the H5065A Rubidium standard. The plot below shows 28 hours of data. It should be obvious that the two frequency standards grew significantly closer between hours 7 and 10. The beat frequency dropped, with the period increasing to about 1.25 hours.

It's considerably easier, of course, to see this jitter in the FFT analysis.

I've given further thought to measuring the frequency offset between the 10 MHz outputs of the Trimble Thunderbolt GPS receiver and the HP 5065A Rubidium standard.

At the minute difference in frequency between these two standards (the difference is around 0.0005 Hz), it's easier to measure phase difference. I've used a Racal 1992 frequency counter in phase measurement mode, with the results presented on 07 May. I understand the preferred methodology is to look at the 1 pulse-per-second GPS output, but for my purposes the frequency offset between the two 10 MHz outputs is the most useful.

Analog phase detectors come in a variety of forms, such as an XOR logic gate. Another phase detector is the common frequency mixer, provided that the mixer has an output that extends to the essentially DC frequency we expect to see here. I decided to use a Minicircuits ZP-1MH mixer operating as a phase detector and an Agilent 34401A digital multimeter to look at the mixer output. I've set the 34401A to automatically output a DC voltage reading once every 10 seconds, with the data being captured over  the GPIB interface and saved to a disk file.

The 34401A is the center instrument in the photo at the right. The ZP-1MH output has a range of about 625 mV plus and minus when driven by the 10 MHz signals. Since the phase detector output varies so slowly, we measure the output with a voltmeter operating in DC mode, not AC.

Another way of extracting the frequency error from this phase data is to perform a fast Fourier transform and look at the spectral lines.

The figure below shows an FFT analysis of the analog phase data. We see a spectral peak at 4.394x10^-4 Hz, quite close to the average figure we came up with by just looking at the time difference between 9 cycles of the phase data. The FFT data, however, also shows by the peak width how stable the frequency offset it. We note the width is rather broad and other peaks can be seen, although of significantly lesser magnitude. We can also see the change in offset by carefully looking at the raw phase voltage data plotted above. The time between the 2nd and 3rd peaks, for example, looks to be very close to 0.5 hours, or 1800 seconds. In contrast, the last two peaks are closer to 0.75 hours or 2400 seconds apart. Thus, the estimated frequency offset of 4.4x10^-11 is not constant and varies considerably over time. This variation can be a function of many things, including changing quality of the satellite signals as more or fewer satellites are visible to the Thunderbolt.

In theory, we should be able to see the frequency with which the Thunderbolt applies individual "nudges" to the disciplined 10 MHz oscillator. This should be around 1 Hz, but the data has far too much noise to show  this level of detail.

I've made changes to the Z10000 buffer amplifier construction and operation manual. The changes were suggested by Jerry, W4UK, to clarify the differences between the -K2 and -U versions by adding photographs and to provide a photograph of the board keyed to the post-construction resistance checkout test points. The revised manual may be downloaded by clicking here or from the documents page.

I've written about lamps and their use as feedback elements at my Bill Hewlett and his Magic Lamp page. The classic HP audio oscillators, such as the 200CD, use small (10 watt or less) tungsten filament lamps as amplitude stabilizing elements.

The standard incandescent lamps used for illumination use tungsten as a filament material for several reasons. One of the more important ones is tungsten's self-stabilization effect. As the filament temperature increases, the resistance increases. Thus, there's a negative feedback mechanism here as well; if the line voltage increases, the power drawn by the lamp is less than it would be should the filament be made from a material with a relatively constant resistance material.

Edison (and his contemporary, Joseph Swan in Great Britain) used carbon filaments. Carbon was relatively easy to make into filaments, although a great deal of practical engineering work was necessary to develop a usable carbon element.

It took twenty years until a practical tungsten filament became available around 1900.

Carbon, unlike tungsten, has a negative resistance temperature coefficient, i.e., as the temperature increases, the resistance decreases. Ron, K8AQC, recently sent me two old Western Electric model 2W switchboard lamps, with carbon filaments. These are slide-base lamps, about 1.75" long, with a nominal operating voltage of 24 volts. The identifier is "2W" which does not mean the lamp is a "2 watt" device.
 

As the plot shows, the lamps start with a rather high resistance, between 2 and 3 Kohm, and rapidly drop as the applied voltage increases and the filaments warm up, at which point the resistance becomes nearly constant.
 

I purchased a used Trimble Thunderbolt GPS-disciplined 10 MHz oscillator unit a couple weeks ago. It arrived Monday and I completed an enclosure and power supply for it this afternoon.
 

The Thunderbolt version I bought is an OEM special, with an enclosures but without the +24V input power option commonly found. Instead, the board requires +12V, +5V and -8 to -12V. The switching power supply takes care of the +12V, and I installed a 7805 three-terminal fixed regulator to reduce the +12 to +5 for the +5 requirements. The -8V is, fortunately, only required for the oscillator steering DAC controller, with a current  requirement of only a few mA.

This suggested a flying capacitor voltage inverter which I built on the small piece of perf board visible in the second photograph near the IEC AC input socket. The voltage inverter starts with a 78L09 regulator to knock the +12V supply down to +9V. The actual voltage inversion is accomplished through a Microchip TC1044S charge pump voltage inverter. The TC1044S works by charging a large value capacitor (I used a 33 uF Tantalum) across the +9V supply during one half of the 10 KHz switching waveform. During the second half of the 10 KHz period, the capacitor is inverted, which yields -9V. The negative output voltage is stored by a second large value capacitor (another 33uF in this case) to smooth the output voltage.

If much current is drawn, the ripple increases, but at the 2 or 3 mA required by the Thunderbolt, the ripple easily meets Trimble's standards.
 

The plot below shows the phase between the Trimble's 10 MHz output and that of the 5065A, measured with a Racal 1992 counter, operated in phase measurement mode. The data is taken once every 10 seconds and the measurement range is 0-360 degrees, so as the phase difference exceeds these bounds, it "wraps" around 360 degrees. Hence, the sharp jumps represent phase wrapping.

The phase error is not constant, indicating that the two oscillators differ in frequency.
 

The data presented will be more meaningful after the Thunderbolt has operated a few days without interruption.

As usual, the preceding month's Updates are now archived and may be read by clicking here or via the archive link table at the top of this page.

I've been busy  getting a few orders for assembled Z10000 amplifiers and Z10010 filters out in the last couple days so this page has not been updated.

I've also taken a more detailed look at a few audio transformers, including harmonic distortion and transformer modeling in LTSpice. I've added a page with the details, viewable  by clicking here, or by clicking on the Audio Transformer Data and Modeling link in this site's navigation menu.