OBSERVATIONS FROM SCOTLAND 18th July 2008 GM1SXX
In search of an accurate frequency standard.
If like me, you like to dabble in construction, at some point in time, you will feel the need for an accurate frequency reference. As the owner of no less than three Digital Frequency meters (DFM's), none of which actually agree, and the need to set up oscillators and the like at least *close* to their intended frequency, for me the time has come to beg, steal or buy some sort of accurate reference. By accurate, I mean that I'd like to be able to set a crystal oscillator/multiplier chain to better than 1Khz. Sounds easy, but it is not.
As an example, I use a 'times twelve' multiplication scheme to generate 1296Mhz from a quartz crystal. The crystal for a 1296Mhz oscillator needs to run at 108Mhz. A fifth overtone type series resonant crystal is used for this. In essence the quartz slab actually oscillates at a fifth of this frequency (at *around* 21.6Mhz) and excites an overtone at 108Mhz. It's important to realise that the quartz slab inside the crystal is really operating at around 21.6Mhz and *not* 108Mhz! Quartz can't move that fast. So we have a slice of rock oscillating at around 21.6Mhz exciting an overtone resonance at 108Mhz thats subsequently multiplied by 12 to reach the final frequency. Any error in the crystal will therefore be multiplied by 60... 5*12.
You will notice that I used the word 'around' a lot in the paragraph above. Overtones are not the same as Integer Harmonics, something I've made mention of before. Harmonics and overtones are different things and the makers of crystals work hard to make sure your crystal operates correctly at the wanted overtone frequency. Overtones are mechanical resonances. An error of 1Khz at the output of the oscillator/multiplier strip actually represents an error of around 16.7 Hertz at the crystal slab (1000/12/5) .
A quartz crystal is a peizoelectric resonator and as such, is sensitive to temperature variations. Just lightly blowing on the crystal in a 1296Mhz oscillator-multiplier strip causes a dramatic drift in frequency as the crystal slab expands by heating and the frequency drops. To avoid this, and to keep the frequency stable as the external temperature varies throughout the day and the seasons, it's common practice to use an ovened crystal or OCXO AKA...Oven Controlled Crystal Oscillator.
http://en.wikipedia.org/wiki/Crystal_oven
Digital frequency meters use a crystal oscillator as a frequency reference too. The better ones tend to use an ovened crystal oscillator or OCXO unit that provides a stable and fairly accurate reference that is used by the DFM's logic to calculate the frequency of the incoming signal. Many of these OXCO's run at 5 or 10Mhz and these units generally have a calibration screw that allows any residual errors to be corrected. The crystal oscillator is placed in a small 'oven' where it is heated to a constant temperature to provide a high degree of thermal stability. Ovened oscillators require a specified warm-up before they can be used with confidence.
An ovened oscillator consists of a crystal oscillator placed in a temperature controlled 'oven' with internal and sometimes external thermal insulation. Many of these OCXO's are constructed as a box-within-a-box using copious quantities of foam insulation to keep the crystal as a very consistent temperature. Frequently, some sort of proportional control system is used to precisely maintain the temperature in the oscillator unit.
The 'poor-man's alternative is a 'heat-clip'. This is a Thermistor (POSISTOR) with a Positive Temperature Coefficient (PTC) bonded to a simple metal clip that slips over the crystal to provide a self-regulating heat source. Such clips usually operate at 50C and the crystal must be specified for use at this temperature. I use heat-clips made by Murata and they are a lot better than just operating a crystal at ambient temperature and hoping for the best. I usually wrap the heat clip in a scrap of foam packaging material to protect it from drafts. Such heat-clip fitted oscillators are usually within an enclosure, frequently inside an outer enclosure to provide protection from weather, temperature changes and drafts.
Cheap DFM's such as the Watson FC130 don't have the luxury of an ovened oscillator but the FC130 does have a built-in signal-strength indicator, something I find very useful. The FC130 is actually a 300Mhz counter but can measure to 3000Mhz. It does this by having an internal divide-by-ten 'prescaler' chip for the higher frequency bands. So it has two ranges... 0-300Mhz and 0-3000Mhz, thanks to the prescaler. The prescaler is a simple digital divider with a bit of electronics in front to tweak the sinewaves to something more like a squarewave that the logic can cope with.
At home I have three DFM's. Two are made by RACAL and the other is an 'el-cheapo' Watson FC130. None of them actually agree. In the junkbox, I found a RACAL 5Mhz OCXO and decided to test it on all three DFM's. One of my RACAL DFM's was bought many years ago at a rally. It has probably not seen any calibration for the last 20 years. The other RACAL DFM was bought recently, second-hand but calibrated. It came with a very impressive-looking calibration certificate, having been bought from a certified calibration centre. The FC130 was bought new and came with no real claims to accuracy.
All three DFM's disagree but the FC130 reads closer to 5Mhz when connected to the Racal Frequency reference than either of the RACAL counters. In search of another 'reference', I remembered that my RACAL RA1772 receiver has a RACAL 9420 'frequency standard' fitted and the output goes to a rear-panel BNC.
The 9420 reference is a high stability plug-in type and quite sought after by experimenters.
Using this 'reference', after an appropriate warm-up period, the three DFM's of course still disagreed, but yet again the Watson still read closest to 5Mhz. So much for cheap and nasty.
You are probably 'getting warm' now..... a reference means exactly that.... something you can refer to. It may not be exactly correct, but it should at least be consistent. Consistently 'wrong' beats consistently variable! Most frequency references can be adjusted to cope with crystal ageing, but the big issue is that you need something that's orders of magnitude better to set them against.
Enter the atomic clock.
Rubidium standards are one way to do this. Rubidium atoms within a lamp can be used as an accurate timing reference. These standards, many made by EFRATOM, turn up from time to time on the second-hand market, and on EBay and are snapped up by those seeking to obtain an accurate frequency reference. Racal made the 9475 Rubidium reference and these turn up from time to time on EBay and elsewhere although the EFRATOM rudidium standards seem to be a bit more common.
I believe the first 'atomic clocks' were Cesium based. These used a cesium lamp as their reference. http://tycho.usno.navy.mil/cesium.html
Both Cesium and Rubidium are lightweight and highly reactive alkali metals. The GPS satellites use cesium clocks made by Symmetricom.... http://www.symmttm.com/cesium-clock.html
The forthcoming European Galileo NAVSATS will use Rubidium clocks.
usno.navy.mil/ptti/ptti99/PTTI_1999_627.PDF
So for home use as a frequency and time standard, it's now becoming feasible to use GPS as a precision reference. GPS satellites operate on 1575.42Mhz and carry very accurate atomic clocks on board. It's possible to use the signals from the GPS system satellites as an accurate frequency or time reference and several papers by various authors have been presented on the World-Wide-Web on how the radio amateur or other interested parties can employ GPS to 'discipline' a crystal oscillator as a precision frequency reference. This allows a far higher accuracy than an ovened oscillator alone can provide. The GPS system is used to fine-tune or 'discipline' an ovened crystal source via a phase locked loop.
Rubidium lamps have a problem.... life expectancy. You don't really want to leave a Rubidium system running constantly. GPS on the other hand is always available, faulty satellites get replaced, to keep the system operational and a GPS receiver can now be found fairly cheaply that can be used to discipline an oscillator.
The Trimble company in the US have an even better 'toy', a ready-made frequency and time reference called the Trimble Thunderbolt. Much though I like to build stuff, life is short and I have lots of interests so I've opted to purchase a second-hand Trimble Thunderbolt from an EBayer in China. The Thunderbolt is a self-contained solution to providing an accurate time and frequency reference. It too uses a GPS disciplined OCXO to provide a 10Mhz reference plus a 1PPS timing pulse. The original cost of these units was high but they are now starting to appear on the second-hand market on EBay and elsewhere. The complete Thunderbolt kit consists of the GPS and PSU in a single cabinet plus some coaxial cable and a GPS antenna. The antenna should be mounted in the clear as far away from obstructions and trees as possible.
I didn't buy the complete system, just the board carrying the receiver, OCXO and logic, the board on the right of the photo. I'll make a GPS patch antenna and a suitable three-rail PSU to drive the unit. The unit runs from +5, +12 and -12V. I've already found and downloaded the manuals and the software to 'talk' to the unit over it's serial interface cable.
The Thunderbolt should give me a precise 10Mhz frequency reference with which to calibrate other equipment. Such a reference has many other uses for the radio amateur, one being the means to lock a radio beacon, receiver or other equipment to the reference for excellent long-term stability.
The original purpose for the Trimble Thunderbolt was as a precise time and frequency reference for controlling wireless networks. I'm sure more of these units will turn up on Ebay and other places and they should not be overlooked with anyone with an interest in calibration, frequency-locking or accurate timekeeping. The seller I bought from had another 20 of these for sale. Although based in China, it seems his name is Bob, not I suspect a very common name in China. The 'Thunderbolt' is on it's way here by slow-boat ;-)
The Thunderbolt manual is at.. ThunderBoltBook2003
The Thunderbolt of course would make an equally good timepiece with the addition of a PIC microcontroller, an LCD display and a few lines of software. I need to make a PSU capable of supplying 5V at 700ma and + and -12V at 25ma. I'll think about running it on solar power, like much of my other station equipment.
References.
http://www.trimble.com/tmg_thunderbolt.shtml
http://www.trimble.com/tmg_thunderbolt_ts.asp?Nav=Collection-2356
http://en.wikipedia.org/wiki/Atomic_clock
http://www.hanssummers.com/radio/gpsref/index.htm
73 Al.
GM1SXX