I’ve actually been quite busy with Ham projects, and unfortunately have been shamefully neglecting my blogging duties. So, here are a couple of posts at the Turn Island Systems site.
I’ve just received another batch of QDX External Reference Interface boards. More info at Turn Island Systems
Also, I just sent out boards for a four-channel RF Combiner / Splitter, which takes four 1W (or lower) signals (80, 40, 20, and 10 meters), and combined them into a single output for feeding a multiband antenna. These provide at least 20dB port-port isolation, and less than 2dB loss. If all works as planned, these will be used for Ionospheric research.
And finally, here’s a 1W (approximately) power amplifier, using 74ACT04 digital integrated circuits to drive a wideband transformer. This takes a 3.3V digital input between 3 and 30 MHz, and produces a 1W square wave output. External filtering will be required, and the combiner shown above should provide adequate filtering.
There are “better” ways to make a 1W amplifier, but this is quick and easy, and requires no tuned circuits. I can’t wait to test these designs!
In the process of turning my three-output Si5351-based “ClockBox” into a three-band simultaneous FST4W beacon, I decided I needed a small power amplifier to boost the (approximately) 10dBm square-wave outputs up to a 1/4-watt clean sinewave. Rather than using FETs and transformers, or a Class-E design, I decided to try using the an 74ACT04 hex inverter as the power stage. Each inverter output can deliver in excess of 25 mA with about a 4.2V swing (+5V supply) so six of them in parallel looked promising. And I happened to have a full tube of old 74ACT04 parts in DIP form (remember those?)
So I built one, tuned for the 15 meter ham band:
I was worried about connecting the outputs directly, afraid that skew would result in excess output-stage current, but some tests showed low unloaded dissipation up past 30 MHz. Adding small resistors at each output might be a good tradeoff, but for now I decided not to waste the power.
I used a simple L/C impedance-matching network to transform the 50 Ohm load (antenna) down to about 3 Ohms which matched the ganged inverter output impedance. Instead of a simple series inductor I used a series L/C, calculated to provide the correct reactance at the design frequency, but giving a high impedance at higher or lower frequencies, and a bit of extra filtering. This also makes the design slightly more bulletproof — you can run any frequency into it without causing excess dissipation.
The output power of the test board was actually closer to 1/2 W, and the harmonics were slightly better than 40 dB below the carrier.
The amplifier efficiency is about 60% and the temperature rise on the buffer was well within comfortable limits. This is good enough that I am having some circuit boards made. The only frequency-sensitive components are two capacitors, and one toroid inductor. I would have liked to use small surface-mount inductors, and the board does provide for that option, but the toroid has much lower loss. The board uses two surface-mount 74ACT04 parts, and a small buffer to drive the 12 inverter inputs.
Someone pointed out this great blog post by Ryan Tolboom, where he digs deep into the “spread” measurement and calculations as done in the WSJTX program. Myself, I had only a hand-waving understanding until I read this:
Thank you Ryan!
I am having some QDX External Reference Boards being built, both the clock-multiplier and the direct interface versions. This should make it reasonably easy for anyone who wants to stabilize the QDX to connect an external 5, 10, or 25 MHz reference clock. I am also having some replacement QDX back-panels made, so we won’t have to drill a new hole in the existing panel.
I will have these back and tested within a week or two. If they work and fit as planned, they will cost under $20 for the adaptor / back panel set. Some assembly required!
If you are interested, contact me at Paul at [my callsign] dot com.
(All 3D images done with KiCad)
This is a presentation that looks at frequency stability — history and practice — including why and how to modify the QDX transceiver to use an external clock reference.
Freq Stability (pdf)
Transmitters and computers often don’t get along. Cables going to computers, routers, ethernet switches, USB hubs (and anything else with wires) can pick up RF and disrupt operations. This RF can be coming from your near-by antenna, from leaky coax, or from the transmitter itself. Usually when this happens the easiest fix is to add ferrite chokes to the cables. Sometimes it takes several chokes on a cable. You will end up experimenting.
The ferrites you will find on eBay (etc.) are probably designed for VHF signal suppression. They will work after a fashion on the HF bands, but you will have better results if you use ferrites made from a material optimized for HF frequencies. Cores using “31” material from the “Fair-Rite” company are excellent performers. These come in all shapes and sizes, but the clamp-on core shown below has proven to be quite useful.
Fair-Rite Products Corp
Part Number: 0431164181
31 ROUND CABLE CORE ASSEMBLY
Lower & Broadband Frequencies 1-300 MHz (31 material)
You can buy these from Mouser:
Here are some guidelines for ferrite core use:
- Put the core as close as possible to the equipment.
- If possible take two or three turns through the core. Up to a point, the suppression effect is proportional to turns² (turns-squared).
- If possible, run power and ground together through the core.
- Before buttoning things back up, secure the heavy core to protect the wires and connectors. This advice is definitely appropriate for boat radio installations, but perhaps not necessary when in a stable environment.
As has been mentioned before, the transmit frequency stability of the QRP Labs QDX transceiver isn’t quite good enough for some of the narrow-band modes. It works fine for WSPR and FT8, but slower modes such as FST4W-300, -900, and -1800 require stability that the QDX’s internal TCXO just can’t provide. Here is a plot of a FST4W-1800 transmission, showing the frequency drift as the QDX heats up during the long 30-minute transmit cycle:
In the plot above, the actual transmit frequency is about 10 MHz, but my measurement system has mixed it down to about 36 Hz in order to get higher measurement resolution. Add 10.1402 MHz to frequency values shown on the plot. We can see that the transmit frequency initially rises by about +0.1 Hz, then eventually falls to -0.85 Hz.
FST4W-1800 has a FSK tone spacing of 0.089 Hz and does not track slow frequency drift, and so requires much better transmit stability than this. And remember, this test was done in the 10 MHz (30-meter) ham band, the drift will be double that at 20 MHz, and half that at 5 MHz.
I spent some time looking at the causes and possible solutions for this drift. First, I ruled out any software issues by measuring the output of the QDX internal 25 MHz TCXO. Here’s a plot of the TXCO frequency vs ambient temperature, using a homebrew temperature chamber (the QDX was not transmitting during this test):
The top chart is frequency vs temperature, the middle one is frequency vs time, and the lower one is chamber temperature over time. Again, here my measurement system is shifting the 25 MHz TCXO frequency down to 100 Hz in order to obtain higher resolution. We can see that over a 6C to 60C temperature range the TCXO shifts over a range of 4.3 Hz.
We can also see the intriguing phenomenon of “retrace”, where the oscillator frequency exhibits a kind of thermal hysteresis. All crystal oscillators exhibit this behavior to some extent, and this does limit the amount of correction that can be performed by measuring the temperature (one of the options contemplated for the QDX).
I attempted to slow the rate of frequency change by increasing the thermal mass of the TCXO (attaching a brass tab to the TCXO package), but the results were disappointing — the TCXO is too solidly attached to the circuit board and the heat transferred though the board from the QDX output transistors just couldn’t be escaped.
Setting aside any idea of a simple fix, I decided to bypass the TCXO entirely and provide an external clock. There was room (barely) on the rear QDX faceplate for a SMA jack, and room inside for a small circuit board, do I designed and built a little reference clock multiplier:
This simple circuit uses a PLL chip to multiply an external reference clock (either 5 or 10 MHz) up to the 25 MHz needed by the QDX. I will dig into more circuit detail in a later post. This circuit is powered by the QDX internal 3.3V supply, and provides a 0/3.3V logic-level output which is connected to the QDX clock buffer input (after removing the TCXO coupling capacitor):
Here’s the little board layout:
And as installed in the QDX:
I picked up +3.3V and ground from some testpoints on the QDX board, and having no microscopic coax, I used a twisted pair of #30 Kynar-insulated wirewrap-wire to connect the clock signal to the QDX board (two different-color wires would have made more sense, but I was in a hurry). The twisted pair just managed to squeeze through another testpoint through-hole, and was then soldered to the PCB:
The yellow oval shows where the TCXO coupling cap (C 54) used to be.
Using a Bodnar GPSDO as my 10 MHz reference I ran some tests. The 25 MHz output from my PLL was reasonably clean, as was the resulting signal from the QDX. Here are some plots of the transmitter, in the 20-meter band:
This is a 200 KHz span, 100 Hz resolution bandwidth scan, showing a bit of broadband noise and a discrete spur at -100 KHz (there is another near +100 KHz, not visible in this plot). This noise potentially affects the receiver performance, but initial measurements lead me to believe that other QDX noise sources are dominant. These measurements only show us the transmit signal, and to determine how much of this output noise is due to the PLL will require further testing. In any case, this noise will have no significant impact on the transmit performance. We can see how the noise falls off dramatically as we approach the carrier frequency — this is the characteristic of PLL noise under the loop-bandwidth frequency. Here is a close-in plot:
This level of noise is definitely acceptable.
And here are plots showing the transmitted FST4W-120 signal, transmitting on 14.09715 MHz:
As expected, the transmit frequency is rock-solid. And here is the modulation in FST4W-1800 mode:
On-the-air tests of the modified QDX, transmitting FST4W-120 show essentially perfect performance.
About the external reference:
The reference multiplier input can come from any stable source of 5 or 10 MHz. I tested using a GPSDO, since I distribute that reference to much of my test equipment, but a stand-alone OCXO, or even stable TCXO will serve the purpose. Obviously, if you already have a stable 25 MHz clock available (the Bodnar GPSDOs can be configured to provide this) you don’t need a clock multiplier — just a connection to the QDX circuit board. Be sure to provide appropriate coupling and some type of input transient protection!
The 25 MHz clock not only drives the radio portion of the QDX, it also runs the microcontroller. Without this clock the QDX will not run, and any interruption will likely cause bad things to happen. The external clock source is the first thing I connect, and the last thing I disconnect.
While testing the newest QDX “Beta 1_06_005″ firmware, I hooked it up to the small/cheap inovato QUADRA linux box and ran some WSPR on 20 meters. This is what I got after a few hours with 4 Watts into a dipole (I’m currently in California, a bit north of San Francisco):
The QUADRA is a nice replacement for the virtually unobtainable Raspberry Pi 3:
A SMOS. That’s what we hardware engineers (and management, too) call it when something needs to be fixed and the hardware is already out the door. Fortunately, in this case it was indeed a software fix.
I’m talking about the QDX and the difficulty it had generating very small frequency shifts, as first noted by my friend Glenn (N6GN). I described the problem and my measurement techniques in these posts: http://wb6cxc.com/?p=244, and http://wb6cxc.com/?p=275
I became fascinated by this issue, and while adding some pretty useful features to my Time Interval Counter, the QDX gave me a great way to improve my measurement technique. In the process I re-connected with an old ham friend, and met some new ones who I really respect.
I also got to know the WSJTX mode FST4W, which uses much smaller/slower FSK than FT8 or WSPR. For example, a single transmission of FST4W-1800 takes a half-hour with a symbol period (Baud) of 11.2 seconds, using a 0.089 Hz tone spacing. This requires extreme accuracy and stability on the part of the transmitter and receiver. The FST4W (and WSPR) modes are being used to measure atmospheric Doppler shifts, among other things. Here’s a description: https://physics.princeton.edu/pulsar/k1jt/FST4_Quick_Start.pdf
Measuring FSK patterns is one thing, but analyzing the actual audio-to-RF frequency generation behavior called for a more deliberate approach. So, I set up my PC to generate a slow audio-frequency sweep (using the NCH Tone Generator program), and sent that into the QDX USB port. Actually, first I sent it to my Icom IC-7200 USB port and measured that output to make sure that my audio sweep (and measurement technique) were suitable. Here’s a 1000 to 1010 Hz sweep, transmitted on 20 meters:
I then tried the same sweep input with the QDX:
Here you can see the uneven and quite large steps, roughly 1.25 Hz each. This would work (barely) with the 1.465 WSPR tone spacing, but was completely unusable with the smaller tone-spacing modes. Here’s what FST4W-300 looked like (0.56 Hz tone spacing):
As you can see, one tone is being entirely skipped over, and the spacing on the others is incorrect. Not surprising given the sweep results.
I posted my measurements to the QRP Labs groups.io discussion (https://groups.io/g/QRPLabs/topic/qdx_fst4w_300_transmits_only/95257422 and elsewhere), as did Glenn and others, and Hans (QRP Labs owner and designer) dug right in to solving the problem. He quickly homed in on the algorithm that calculated the fractional divider values for the Si5351 clock generator chip used in the QDX, and he generously traded email with me, describing his progress and solution. It turns out it was a combination of limited floating point resolution and integer size. Hans also figured out a way to improve the algorithm by separating the carrier frequency calculations from the audio-offset calculations. The result was spectacular! Here’s the 10 Hz sweep using the new Beta 1_06_005 firmware:
And here is the QDX transmitting FST4W-1800 (0.089 Hz tone spacing):
This is perfect. The frequency-setting is now amazingly good.
And it was all just a Simple Matter Of Software!
But, there is still work to do if we want to use the QDX on the slower FST4W modes. The QDX transmit frequency drifts as the radio heats up during a long transmit cycle. This doesn’t really affect modes like FT8 or WSPR, but the small drift (about 1 Hz in fifteen minutes on 20 meters) will cause FST4W-300 and higher modes to fail. Here is the thirty-minute long FST4W-1800 transmit sequence (remember this is a tone-spacing of 0.089 Hz):
Probably the only reasonable way to get the needed stability is by using an external 25 MHz reference, probably a GPSDO (GPS Disciplined Oscillator). The QDX can be easily modified to accept an external reference, and I will be trying this fairly soon.
But all in all, this was a gratifying experience, and I salute Hans and QRP Labs for the speedy and spectacular work in solving a problem that most of the QDX users would never have even noticed. I now have two more QDX’s on order: https://www.qrp-labs.com/qdx.html