Hi folks,

When running WSPR I noticed my QDX transmissions would consistently get reception reports of drifting a few Hz. So I purchased a small package of 1" square 3M thermal pads (type 5590H) and placed one over the TCXO on the rear of the board. Ever since I did that the drift reports have been exactly zero every time. Well, a few are still +-1 but I attribute those to the reporter ;-)

73, Elwood, WB0OEW

On 11/21/22 14:58, Elwood Downey wrote:

... I purchased a small package of 1" square 3M thermal pads (type 5590H) and placed one over the TCXO on the rear of the board.

From 3M
3M? Thermally Conductive Tapes are used to bond heat-generating components to heat sinks and cooling devices

I appear to be missing something, you only applied the tape and no heat sink or did you
skip that part?

--
Linux Home Automation Neil Cherry kd2zrq@...
Main site
My HA Blog
Author of: Linux Smart Homes For Dummies KD2ZRQ

On Mon, Nov 21, 2022 at 12:18 PM, Neil Cherry wrote:

I appear to be missing something, you only applied the tape and no heat sink or did you
skip that part?

It sounds like the thermal pads are acting as a heatsink of sorts.? They *are* thermally conductive, after all, and a 1" square pad is *much* larger than the TCXO.?

The next question is why does this help?? Perhaps the heat path to the TCXO is via the board, and reducing the thermal resistance to ambient (by attaching the pad) reduces the TCXO temperature variation (just a guess.)
?
--
Paul Elliott - WB6CXC

On 11/21/22 19:41, Paul WB6CXC (tech-blog: wb6cxc.com) wrote:

On Mon, Nov 21, 2022 at 12:18 PM, Neil Cherry wrote:
I appear to be missing something, you only applied the tape and no heat sink or did you
skip that part?
It sounds like the thermal pads are acting as a heatsink of sorts.? They *are* thermally conductive, after all, and a 1" square pad is *much* larger than the TCXO.

Thanks for the serious answer, I was very curious. I was wondering about the
conductivity too.

The next question is why does this help?? Perhaps the heat path to the TCXO is via the board, and reducing the thermal resistance to ambient (by attaching the pad) reduces the TCXO temperature variation (just a guess.)

I don't know, does seem odd as the TC part would try to keep it at a
certain temperature.

--
Linux Home Automation Neil Cherry kd2zrq@...
Main site
My HA Blog
Author of: Linux Smart Homes For Dummies KD2ZRQ

Optional 25MHz TCXO module is pictured above (diagram only, as yet). In tests, the frequency change over the range 4C to 65C was +/- 0.25 ppm (parts per million). This is a tiny 14.8 x 10.5mm PCB with TCXO, SMD all ready-assembled, which is installed onto the main QCX+ PCB in place of the 27MHz crystal and 0.1uF capacitor C2. The TCXO gives very good frequency stability if you intend to use WSPR operation, where frequency drift is critical. It is highly recommended for WSPR operation, particularly if you are using a 17m or 20m radio.
Why are you applying thermal pads???
TCXO module already has high stability for WSPR. Should not require thermal pads.

Charles Brown N4SO

On Mon, Nov 21, 2022 at 08:21 PM, Neil Cherry wrote:

does seem odd as the TC part would try to keep it at a
certain temperature.

No, you're thinking of an OCXO, with a built-in heater.? A TCXO, like the one in the QDX, has compensation circuitry that adjusts the capacitance (usually) around the crystal, as a function of the measured temperature.? I used to design TCXOs with a mix of temperature-stable (NPO) and temperature-variable (PTC and NTC) capacitors, or thermistors providing bias to varicap diodes.? Some TCXOs in addition to passive compensation also use digital compensation of some sort.? ?You can improve the thermal frequency drift with TCXO techniques but you can't eliminate it.? It's a fascinating subject and I could go on...

In any case, reducing the temperature variations on an OCXO will reduce the frequency variations.
?
--
Paul Elliott - WB6CXC

On Mon, Nov 21, 2022 at 08:40 PM, Charles Brown wrote:

Optional 25MHz TCXO module is pictured above (diagram only, as yet). In tests, the frequency change over the range 4C to 65C was +/- 0.25 ppm (parts per million). This is a tiny 14.8 x 10.5mm PCB with TCXO, SMD all ready-assembled, which is installed onto the main QCX+ PCB in place of the 27MHz crystal and 0.1uF capacitor C2.

Yes, but that's for the QCX, not the QDX.? The QDX already has a "High precision 25 MHz TCXO" on-board.

I haven't checked the TCXO specs, but the thermal drift I have seen on the QDX is definitely measurable.? I don't know if it's bad enough to cause WSPR problems -- probably not.

?
--
Paul Elliott - WB6CXC

The performance of the TCXO in the QDX is impressive.

Attached are graphs of measurements made here with the QDX in receive only from cold start and in receive, followed by two back-to back transmissions of FST4W-300 (so 600 seconds transmit) at 5 W output followed by receive.

The QDX was receiving 10 MHz from a GPSDO and the frequency analysis was done using the WSJT-X frequency calibration option.

Gwyn G3ZIL

Here are some QDX frequency drift measurements (latest HW and SW version, no modifications), running WSPR on 20 meters.? The QDX is driving a dummy load, and I am using my tiny frequency counter with a stable external 10 MHz reference clock.?

The first chart shows WSPR, with WSJTX set for a 50% transmit duty cycle.? This results in two transmit cycles back-to-back, followed by two idle cycles.? The QDX had been running this "50%" cycle for about 30 minutes before I started the measurements.? I am using the internal NCO/mixer in my counter.? The NCO is set for 14,095,600 Hz (the receiver frequency for typical 20 meter WSPR), which puts the mixer output in the 300-3000 Hz region.? The counter is showing the "audio" frequency, and 1 Hz drift here is equal to 1Hz drift at the transmit frequency.? You can see slightly more than 1/2 Hz drift between the start and end of the two-WSPR-cycle transmission.? It's interesting to note that the WSPR tone spacing is reasonably accurate -- this is not always the case and appears to depend on the WSJTX audio frequency settings..


The next chart shows the drift (using WSJTX "Tune") starting from a warmed-up but idle QDX, and then transmitting for about 25 minutes.? The counter NCO has been set to 14,097,000 Hz giving a difference frequency of about 150 Hz, and there are about 1.5 measurements per second.? You can see about a 3 Hz drift, with an interesting initial frequency "bump".



My QDX is built for 12V operation, and the power supply delivers about 12.5V? The transmit power is initially 4.56W, and settles down to 4.38W.? The case is fairly toasty, but not painfully so.

--
Paul Elliott - WB6CXC

On Tue, Nov 22, 2022 at 07:43 PM, Paul WB6CXC (tech-blog: wb6cxc.com) wrote:

It's interesting to note that the WSPR tone spacing is reasonably accurate

Looking more carefully, the tones still have a problem.? The middle two tones are too close together.
?
--
Paul Elliott - WB6CXC

开云体育

I believe this is the TCXO that QRP-Labs utilizes:

Hans, I'm not implying any criticism of the TCXO -- it looks like a good one.? No oscillator is perfect and the shifts I'm measuring shouldn't matter in "normal" use.? And?as you mentioned,?many TCXOs are much worse.? For example, this inexpensive surface-mount Fox TCXO (chart below) uses digital correction steps of about 100ppb (1 Hz @ 10 MHz), and yes, would be useless for modes like WSPR.? Here is the Fox as the temperature drifts through an inflection point in the compensation curve.? The small 1/10 Hz steps are due to the resolution of my measurements, but the larger steps are the TCXO corrections.? I was looking at this Fox TCXO for use in my counter and realized that even though it met its specifications it was completely inappropriate.? The chart shows freq vs time.

In comparison, here's a fairly inexpensive surface-mount Abracon TCXO.? Here, I am swinging the ambient temperature somewhat randomly between +16C and +60C through a few cycles (I may have been using my modified toaster-oven in a cold garage for the temp chamber).? You can see that there is a smooth freq/temp curve, but also what looks like a hysteresis in the compensation curve, where the heating vs cooling frequencies are on different tracks.? I believe that the TCXO used in the QDX has a tighter tempco spec than this one does: (chart is freq vs temperature)

After a bit more testing and some research, what I had originally thought to be thermal-delay hysteresis turned out to be "retrace", a well-known (just not to me!) phenomenon in crystal oscillators.? A simple junk-box xtal oscillator also showed retrace, as did a SiTime MEMs oscillator.? Even an OCXO showed retrace.? Retrace also affects the eventual frequency of oscillation, which can be anywhere within the retrace envelope.? I had been planning to do a secondary digital temperature compensation of the TCXO in my projects, and I can still do this and get a slight improvement, but retrace limits the ultimate accuracy.? Here's another retrace curve, showing many temperature cycles (11C to 52C, using my modified "Reptile Incubator" TEC thermal chamber):


And another, this time with a TCXO from ECS:


I'm using a GPSDO as my frequency reference in these tests.

So if anyone wants to improve the frequency stability of the QDX, be prepared to either at least double the price and power-consumption of the radio, or hack on an SMA connector for an external 25 MHz reference clock.? And ask yourself why you want this increased stability -- most of us, including me, don't need it.

If anyone cares, I can dig up the part #s of the TCXOs I've shown here.
--
Paul Elliott - WB6CXC

I love the QDX, but some modes, such as FST4W-300, require more frequency stability than the stock QDX can provide.? After much study I ended up adding a tiny "reference clock multiplier" board inside the QDX, with a little SMA jack on the rear faceplate (there's room for it, just barely).? I feed in a 10 MHz stable clock (or a 5 MHz one) and the QDX transmit frequency is rock-solid.


I have a blog-post about it here:?
--
Paul Elliott - WB6CXC