September 8, 2015

CWTD Episode #74

Oven-Controlled Crystal Oscillator
... Controlling temperature in our VCXO project for frequency stability!


This is Part 2 in our series on the design of a "GPS-Disciplined Oscillator, and this time we focus on a circuit intended to maintain temperature of a Voltage Controlled Crystal Oscillator, and thus help to improve the stability of its output frequency. We'll be doing this by means of a control loop that drives a heating element keeping the local temperature inside the enclosure constant at about 45-degC +/- 1.

This latest project installment comes in the form of a small kit (parts + pcb) that also includes temperature hi/lo LEDs, voltage regulators, and a motherboard to hold the whole assembly that slides into a nice extruded aluminum (and insulated!) enclosure.

So come follow along and learn first-hand how to keep your oscillator's temperature under control, and thereby its frequency stable!

In October's grand GPSDO finale episode, we'll add the small GPS receiver module and the PLL circuits.

BREAKING NEWS! ... We'll also be describing a brand new (free) software load for a popular instrument that will enable you to do some wonderful things with the inexpensive little GPS receivers we see these days. Be among the first on the block to get our latest gadget for the shack ... and even use it on the family's entertainment console!

73, George N2APB and Joe N2CX

Listen to the Podcast

Text Window happing during the show ...
<19:57:09> "Dave - AD7JT": I guess I have been lucky, I have only had one bad experience on ebay. I have a 6T that came in a selr-contained white plastic disc for quite a few dollars more and it seems to work fine. The 6T has a second TimePulse output and has some additonal timing related features.
<19:58:49> "Mike WA8BXN":
<19:59:07> "Mike WH6YH": 6T, I'll go take a look. thank you.
<20:03:19> "Joe N2CX": GPS Disciplined Oscillator - GPSDO
<20:26:24> "Joe N2CX": Microchip temp sensor IC MCP9701
<20:46:17> "Al - N8WQ": what kind of coax did you use George from ref signal out to bnc connector
<20:47:30> "George N2APB": I have a bunch of thin UHF coax from flea markets, some of it is "armored" sheathing ... but it's just coax at the end of the day A. RG-174 will work just fine.
<20:48:08> "Al - N8WQ": rog
<20:48:23> "George N2APB": The SMA/SMB connectors are expensive though! So I try to collect as many of those from scrap circuit boards
<20:49:55> "Joe N2CX": What George is discussing is often called semi-rigid coax.
<20:52:03> "Mark NI2O": NMEA=National Maritime Electronics Association
<20:52:19> "Pat - W0BM": for the next session discussion, how do you get a GPS signal into the basement? How open does the antenna for the GPS need to be?
<20:53:18> "Mike WA8BXN": you probably will need an outside antenna for the basement --- many are available
<20:54:51> "Pat - W0BM": Mike, what kind of feedline? I'm probably 40 feet from the opening to the outside at 3 ft above ground.
<20:55:06> "Mike WA8BXN":
<20:55:54> "Mike WA8BXN": one antenna i have has 16 feet feedline that looks like rg-174 size coax
<20:56:06> "Mike WA8BXN": i might try quad shield RG-6
<20:56:33> "Mark NI2O": 2ghz CATV line amp works well
<20:56:46> "Al - N8WQ": Mike, can you link your gps antenna?
<20:57:41> "Mike WH6YH": Mark, NI2O can you post a link for a white paper on PPS steering you mentioned earlier? for inquiring minds
<20:57:58> "Mike WA8BXN": i used the following, not very carefully picked:
<20:58:01> "Mike WA8BXN":
<20:59:28> "Mike WA8BXN": down in the basement another option to consider is put the GPS/antenna outside and run the timing signal out of the GPS down to the basement
<21:01:01> "Mike WH6YH": IS the screen the same size as the SNA & NAT?
<21:08:53> "George N2APB": Yes it is the same size. In fact, it's the very same hardware platform!
<21:09:22> "George N2APB": In other words, if you had an NAT or SNA kit already, you already have the GDT!
<21:09:51> "Mike WH6YH": I do! It will be interesting to integrate this with the SNA
<21:10:19> "George N2APB": And if you don't yet have an SNA Kit, I created a subset of the SNA Kit, containing just the motherboard and display (i.e., none of the measurement and analysis stuff), and you can get the "GDT" platform, for a lesser price.
<21:15:58> "Mike WH6YH": of 9600 modem fame?
<21:16:24> "Mike WA8BXN": VE2ZAZ
<21:16:40> "Mark NI2O":
<21:17:44> "Mark NI2O":
<21:18:28> "Mike WH6YH": soldersmoke
<21:18:39> "Mark NI2O":
<21:19:59> "Mark NI2O":
<21:21:46> "Mark NI2O":
<21:22:39> "Mark NI2O": NTP servers
<21:23:06> "Mark NI2O": +/- some Special Relativity
<21:27:19> "Mike WA8BXN": I have information on the cables/adapters needed I will post later
<21:32:27> "Mike WA8BXN": The basic idea of a FLL (frequency locked loop) can be thought of this way: Assume you have a very accurate frequency counter and want to adjust the trim pot on the VCXO to put it on frequency. That trim pot varies the DC voltage on the varicap diode in the VCXO board, which controls the oscillator frequency to some limited extent.
Watching the frequency on the counter, you turn the pot one way or the other until the counter reads exactly 10.00000 MHz. A FLL does the same thing using a microprocessor of some sort. A frequency counter is implemented in that processor using the timing signal from the GPS. By using a very long measurement interval over which you count the cycles, you can measure the frequency very accurately. A digital to analog converter provides an output voltage from the processor board to adjust the VCXO frequency.
<21:35:51> "George N2APB": Nicely stated Mike. One of things to be careful of in using a microcontroller (like the Arduino or even a PIC) is the latency in the counting and interrupt circuits. Whenever even minor delays in included in the processing, they can ultimately affect the counting of locking scheme controlling the VCXO.
<21:36:52> "Mike WA8BXN": Yes, making it work exactly right is not so simple!
<21:37:11> "Dave - AD7JT": The tiny connector on the GPS board is a "MMCX" connector and adapters to other connectors are available at DigiKey, Mouser, and eBay. I was only able to find MMCX connectors on eBay.
<21:38:17> "George N2APB": Even with an FLL "cycle counting scheme" that is well-controlled and tied to hardware only (i.e., no software), there is a "1 count ambiguity" in the end count that will indeed provide some phase noise and jitter.
<21:38:35> "Dave - AD7JT":

[Note:  In the coming days I will be adding a significant amount of commentary, captions and better-looking diagrams to this "photo journalistic"
 approach taken with the material at show-time ... de n2apb]

Our Approach ....

Hans Summers, G0UPL ...

From Hans' website ...

The control circuit is the tricky part. In the simplest kind of control circuit, the measured temperature is just compared with the desired target temperature. Then the heater is switched on if the oven is too cold, and off if the oven gets too hot. This kind of controller is common (getting les common) in the mechanical room thermostat, or the thermostat in your kitchen oven or fridge, or the air conditioner in your home or office. The disadvantage is that the thermal mass of the oven takes time to heat and cool, and this means the temperature can vary quite considerably as the heater (or cooler) cycles on and off.

The right photograph above, shows the crystal frequency (received on an HF receiver and plotted in Argo spectrum analysis software on a laptop). The cycles of the heater on/off are very clear here. The cycle duration was around 38 seconds, of which the oven was switched on for 17 seconds and off for 21 seconds. This kind of temperature cycling would be completely unacceptable for QRSS or WSPR operation such as in the U3 kit.

So we come to proportional oven control. There are several good examples of homebrewed ovenised oscillators on the web, but two of my favourites are: Andy G4OEP, about half way down this page and Des M0AYF's extensive page on his QRSS ovens. These two are my inspiration. Both of these ovens use a proportional controller. There's a heater, a temperature sensor, and a control circuit consisting of an op-amp with relatively low gain (not an on/off comparator). The limited gain of the op-amp makes it possible for the heater to be partially on. If the thermal characteristics are right, and the gain is correctly matched to them, then the proportional control circuit makes it possible to control the frequency without the characteristic "hunting" cycles of an on/off oven.

However, I wasn't quite comfortable with proportional control circuits. An ideal control circuit would maintain the oven at a constant temperature regardless of the ambient, environmental temperature. But a proportional circuit does not. The oven is kept at a reasonably constant temperature but there is still some variation, as the environmental temperature varies. Even if everything else is perfect, the circuit inherently produces an error term. To stop "hunting" oscillations, the amplifier gain needs to be low, but the lower the gain, the larger the error term becomes.

So to PID controllers (Proportional Integral Derivative) which are used in industrial automation. Surely the subject of many text books and the nightmare of Electronic Engineering students everywhere. There is plenty to read on this topic, and it seems to quickly get complicated enough. Software programs are nowadays used for the control logic. But here, we use a simple op-amp. The quick summary (after an awful lot of reading and research): it turns out that the "Derivative" term isn't necessary in a simple temperature controller. The "Proportional" term we have covered already, such as in the circuits above. Best of all, the "Integral" term can be implemented simply by putting a capacitor in the feedback loop of the op-amp; and the "Integral" term removes the error that is inherent in "Proportional" only circuits. Simple and neat, and does everything we need!

Full information on the "partial kit" that Hans is supplying for us is at ...










The CWTD "GPS Receiver Module"

NEO-7M module



Announcing ... The Midnight GPS Display Terminal ("GDT")

Handheld terminal that dynamically interprets and displays the live NMEA serial data feed coming from any GPS receiver module! 

The "GDT" is a new software load for the Midnight SNA hardware platform, enabling anyone with an SNA to easily upgrade to the GPS Display Terminal.

(A separate, lesser-priced "GDT Kit" is available for those not yet having an SNA.  The GDT Kit does not have the specific parts for SNA measurement, like the AD8307, DDS-60, BNC connectors, etc.)

The serial port into the SNA is 5V “digital” TxD/RxD.  One will need (for now) to use an RS232-to-TxD/RxD adapter. (Inexpensive from Sparkfun, Adafruit, Parallax, et al.)

The GDT interfaces with all GPS receiver modules to dynamically display GPS data, including additional screens showing raw NMEA "sentences" (raw data). 

Additionally, the GDT controls GPS receivers in order to configure Timepulse frequencies, duty cycles, and other operational parameters. (Initially being done with the u-Blox product line (LEA, NEO, M8F families.)

GDT Firmware  ... is ready-made to load right onto your NAT and SNA instruments!  Just right-click and save to your local computer, place onto the NAT/SNA SD Card and use the UPLD command to load it.




Interested in Building the CWTD Oscillator Oven Controller?

Purchase the CWTD Oven Control Kit ...  See the GPS-DO Project page



1)  John Miles, KE5FX site ... <> ... A great assemblage of links having to do with high precision, accuracy and stability techniques can be found on the. You can literally spend days reading up on techniques, software, measurement results and the history of timing and frequency control here. Just plan on spending an afternoon to browse it.

2) “Resources for Precision Timing, Stability, and Noise Analysis” ... <> ... If you want in-depth knowledge this is THE place to go. Plus, you can download plenty of extremely valuable application notes and technical papers for future reference. 

3) PLL reference papers that are a good read ...


AN-885 Introduction to Single Chip Microwave PLL's (TI’s LMX1501A)


MM74HC4046 CMPS Phase Lock Loop ...


Experiment #4 - CMOS 4046 Phase-Locked Loop...


CMOS Phase-Locked-Loop Applications Using the CD54/74HC/HCT4046A and CD54/74HC/HCT7046A ...


CD4046B Phase-Locked Loop: A Versatile Building Block for Micropower Digital and Analog Applications ...


ADF4110 PLL …


ADF4001 PLL …



4) And a simple Phase Detector chip + a simple VCO ...


Phase/Frequency detector ...


Supposedly "goes well" with ...


VCO with vvc control ...