Fullscreen HTML5Med Res (??MB)HTML4

For laboratory-standard frequency measurements . . . By JIM ROWE A Rubidium Frequency Standard For A Song How would you like to have a precision rubidium frequency standard on your workbench to enable you to make laboratorystandard frequency measurements? It’s now possible and for a very low price – just buy a used Rb-vapour frequency standard on-line and build a simple power supply and buffer circuit. A S MENTIONED in the February 2014 issue of SILICON CHIP, used rubidium-vapour frequency standards are available via eBay from suppliers in China and elsewhere, for very low prices. But how easy is it to get one of these devices going again? This article explains what was involved in getting one up and running (it was really quite straightforward). When I was writing the February 2014 article, I discovered that quite a few ex-telecom Rb standards were being offered on eBay at very attractive prices – anywhere between $US99 and $US250 (plus shipping). So as mentioned in the February article, I took 66  Silicon Chip the plunge and ordered one. It arrived a couple of weeks later and I began planning how to bring it back to life. The unit I acquired was an FE5680A (see photo), originally made by US firm Frequency Electronics Inc. This seems to be one of the most common ‘retired’ Rb-vapour standards currently on offer, although if you search on eBay and elsewhere you’ll also find others like the LPRO-101 from Symmetricom/Datum/Ball/Efratom. I should mention that although you’ll find quite a few FE-5680A standards on offer, they’re not all the same (even those that look almost identical). In fact, the model name ‘FE-5680A’ seems to have been used for a plethora of Rb-vapour frequency standards. The many versions offer different options, such as (1) the output frequency (10MHz, 2.048MHz, 5MHz, 10.23MHz, 13MHz, 15MHz or adjustable between 1Hz and 20MHz); (2) whether the RF output is a sinewave or a square wave; (3) whether or not the RF output is brought out via a separate SMA connector or just via a pin on the unit’s main DE-9 connector; and (4) whether or not it can be controlled remotely via an RS232C serial interface. Further options specify the required power supply voltage(s), whether or siliconchip.com.au not it can be fine-tuned via an analog tuning voltage (0-10V) and the polarity of the ‘locked to rubidium’ logic output signal (ie, LOCK or LOCK-bar). So you need to be cautious in selecting an FE-5680A from those being offered. If you intend using it as a frequency and time standard, choose one that’s advertised as having a 10MHz sinewave output (available from either pin 7 on the DE-9 connector or from a separate SMA connector), can be controlled remotely via an RS-232C serial interface, has a LOCK-bar output (on pin 3 of the DE-9 connector) and needs both +15-18V and +5V supplies (this is the version I bought). Step 1: collecting info Although the FE-5680A I bought had a small label on the top of the case showing the main DE-9 pin connections and the supply voltages, it didn’t identify all the pins and their functions. So before attempting to fire it up, I decided to collect as much information on the FE-5680A series as I could. A quick search on the internet soon turned up quite a lot of useful information. Most of this came from the links shown in the ‘Handy Links’ panel at the end of this article, so I suggest you go to these first to save time. The Time Nuts mailing list archive is particularly informative, not just regarding the FE-5680A but for all kinds of stuff on time and frequency standards and their use. Most of the important information on the FE-5680A is summarised in Fig.1. Armed with this data, I was then able to knock up a suitable power supply on a breadboard. This comprised a surplus 18V/2.5A laptop PC power supply to provide the main 15-18V rail plus a simple 3-terminal regulator to derive a +5V logic supply rail. At that stage, I was simply going to power up the FE-5680A, so I didn’t provide anything else as I thought I’d be able to do all of the initial checking with a digital multimeter, digital scope and a frequency counter. Step 2: applying power When you first apply power to a rubidium-vapour standard like the FE-5680A, it draws a fairly substantial current from the main 18V supply (about 1.8A). That’s because it has to ‘warm up’ everything inside the ‘physics package’. It’s only after the siliconchip.com.au FE-5680A Series Rubidium Frequency Standard – Basic Information DE-9M Connector Pinouts (as viewed from front): LOCK (High = unlocked) +5V LOGIC SUPPLY GROUND (+15–18V return) 1 +15–18V SUPPLY 3 2 (drain typically 80mA) 4 5 GROUND (signal ground) (1.8A peak for cold start, 600–800mA after locking) 6 1PPS OUTPUT 9 8 7 RS232C TxD OUTPUT (1 µs pulse, only after locking) (sends responses to PC) 10MHz OUTPUT RS232C RxD INPUT (~1Vp-p into 50 Ω) (receives commands from PC) Basic Specification: Output Frequency: Waveform: Minimum amplitude: 10MHz sinusoidal 0.5V RMS into 50Ω Adjustment Resolution: <1 x 10 --12 --11 over range of 3.8 x 10 --5 Short Term Stability: 1.4 x 10 Drift: 2 x 10 Phase Noise: –100dBc <at> 10Hz; –125dBc <at> 100Hz; –145dBc <at> 1000Hz Input Voltage Sensitivity: 2 x 10 Frequency vs Temp: ±3 x 10 Spurious Outputs: Harmonics: Warm-up Time: –60dBc –30dBc <5 minutes to lock, at 25°C --9 (1 – 100 seconds) per year, 2 x 10 --11 --11 per day (15V – 16V) --10 (–5°C – 50°C) RS232C Serial Commands & Responses from the FE-5680A COMMAND FORMAT (hex) FUNCTION 2D 04 00 29 Request current Frequency Offset 2E 09 00 27 aa bb cc dd <dcsm>* Change temporarily to a new Frequency Offset 2C 09 00 25 aa bb cc dd <dcsm>* Change Frequency Offset to new value, Save in EEPROM (9600 bps, 8N1, no handshaking) RESPONSE BY FE-5680A 2D 09 00 24 aa bb cc dd <dcsm>* (Temporarily changes Frequency Offset to [aa bb cc dd] hex) (Changes Frequency Offset to [aa bb cc dd] hex, saves in memory) * <dcsm> = Ex-OR (bitwise) checksum of all four preceeding hex data bytes [aa, bb, cc, dd] Fig.1: here are the pin connections, the main specifications and the RS232C commands for the FE-5680A series of rubidium frequency standards. Used FE-5680A Rb-vapour standards are available on eBay from China for less than $150 including postage. electronic circuitry has been able to achieve frequency lock (to the frequency corresponding to the energy difference between the two ‘hyperfine split ground state’ levels of the rubidium atoms in the resonance cell) that the current begins dropping down to its ‘running locked’ level of 600-900mA. April 2014  67 These photos show the ‘works’ inside an FE-5680A rubidium standard with its mu-metal case removed. The upper shot is a top view, with the physics package and its shock-protective foam at the top. The quartz oscillator crystal is at lower right, with a silver-coloured thermistor above it. To its left is the small trimcap (C217), used to correct for long-term drift. The lower shot shows the underside of the assembly, with the underside of the physics package at bottom centre. This usually takes no more than about five minutes. During this time, the current drain stays relatively high until locking is achieved and if you monitor the RF output (at pin 7 on the DE-9 connector) with a counter, you’ll find that it swings up above 10MHz and then swings down below this again. Generally, it repeats this up-and-down sweeping a number of times, as the electronics ‘searches’ for the small dip in the photodetector’s output which corresponds to rubidium resonance. Then, when the dip is found, the output frequency is ‘locked to rubidium’ 68  Silicon Chip – ie, very close to 10.000000MHz. The internal logic also pulls down the voltage level of the LOCK-bar output (pin 3 of the DE-9M connector), while a 1μs-wide output pulse appears at the 1pps output (pin 6) once every second. But neither of these happens unless a lock has been achieved. When I first applied power to my FE5680A, I was monitoring the current drawn by pin 1 of the DE9 connector with one DMM and the voltage at pin 3 with another DMM. Sure enough, the current drain started off at about 1.85A and then began dropping – slowly at first and then somewhat more rapidly until it nudged below 800mA. This took about seven minutes but as the unit probably hadn’t been powered up for a few months I wasn’t unduly concerned about the time it had taken. What did concern me though, was that the voltage at pin 3 (LOCK-bar) remained high at about +4.9V, showing that the FE-5680A still hadn’t locked. There were no 1pps pulses appearing at pin 6 either – another sign that it hadn’t locked. The ‘clincher’ came when I started to monitor the FE-5680A’s RF output (pin 7) with my counter. It was still sweeping up and down between about 9,999,790Hz and 10,000,065Hz, every 10-15 seconds or so. Clearly it was going through the motions of searching for a lock but for some reason never finding it. I left it searching this way for an hour or so, in case it was especially slow on the uptake. However, when it still hadn’t achieved a lock after two hours, I turned off the power and went back to the Time Nuts archive and KO4BB’s FAQs, looking for clues on how to tackle an FE-5680A that wouldn’t lock. One clue I found was that if an FE-5680A wouldn’t lock, it could be because the internal crystal oscillator had ‘drifted’ a bit in frequency. This could be enough to prevent the ‘searching for a lock’ sweeping operation from swinging sufficiently either side of the lock frequency (ie, above and below 10MHz). The solution was to open the unit up and adjust a small trimmer capacitor near the crystal (C217), to correct for the drift. However, I was dubious as to whether this was the cause of my particular unit’s problem, because it did seem to be sweeping above and below 10MHz by a comfortable margin. So I signed up to the Time Nuts mailing list/forum and posted a request for any further information that might be forthcoming from the experts. There were a few further suggestions but when I tried these out my unit still refused to lock. As a result, I removed the two halves of the FE-5680A’s mumetal case to reveal its ‘works’. It was then just a matter of finding trimmer C217, giving it a small nudge (clockwise at first, because there was no hint as to which would be the correct direction), then screwing on the two case halves again and testing to see whether it would now lock. siliconchip.com.au It still wouldn’t lock and when I subsequently used a counter to check the maximum and minimum frequencies while it was searching, these didn’t seem to be all that different. So perhaps I had picked the wrong direction for my initial nudge of C217? There was nothing for it but to open it up again and try giving C217 a slightly larger nudge, this time in an anticlockwise direction. It still refused to lock so I repeated this process a few more times but still without success. Then, deciding that the problem must be due to something else, like a worn-out rubidium lamp or a broken photodetector, I began looking around inside the unit and checked a few voltages and signal frequencies. By this stage I had discovered a partial schematic for the FE-5680A, which can be downloaded from the last link in the Handy Links panel. However, this didn’t turn out to be very helpful when it came to this particular problem, because it doesn’t include any details of what’s inside the ‘physics package’ like the lamp or photodetector. I was getting nowhere, so I contacted the eBay vendor I’d bought it from and he offered to replace the unit. I duly sent it back and the replacement unit turned up a few weeks later. When it was unpacked, it appeared to be identical to the first unit, apart from having a different serial number. I connected it up as before, monitoring the current from the +18V supply and the voltage at the LOCK-bar pin using two DMMs. As before, I also used my counter to monitor the output frequency as it searched for a lock after switch-on. What happened then was exactly the same as with the first unit. The scope showed that there were no pulses from the 1pps output and the counter showed that the RF output was just sweeping back and forth through 10MHz, without showing any signs of a lock. This continued despite leaving it on for another hour or so. By the way, I had previously read that rubidium standards like the FE5680A should not be allowed to run for very long without using a cooling fan, so I had pulled a small 12V fan from the junk box and rigged it up to keep the unit from getting too hot. A lucky breakthrough I went back to scouring the various siliconchip.com.au Fig.4: once the FE-5680A has ‘locked to rubidium’, it provides one of these 950ns-wide 1pps output pulses each second. reference sources, to see if I could find the answer. And after a while I found a note that the physics package in rubidium standards was quite sensitive to external magnetic fields – that’s the reason for housing them inside a mu-metal enclosure, after all. I then wondered if the difference between the Earth’s magnetic field in Sydney and that in Quangzhou might be just enough to result in a ‘failure to lock’ – despite the mu-metal enclosure or perhaps because the enclosure had somehow become magnetised. It occurred to me that one way to test this theory might be to turn the ‘new’ FE-5680A upside down, to roughly reverse the direction of the Earth’s field around it. So I turned it off, let it cool down, turned it upside down and then turned the power on again. Bingo! Within about three minutes, it found a lock and stayed locked for another few hours while I left it on to make sure. The voltage at the LOCK-bar output (pin 3) stayed down at about +0.35V, while the scope showed 1μswide 1pps pulses coming from pin 6. What’s more, the counter remained steady at a reading very close to 10MHz, even when I changed to longer and longer gating times to achieve maximum resolution. Only when I went to a 1000-second gating time did I see that the FE-5680A’s output frequency was a whisker below 10MHz: 9,999,999.992Hz, in fact. At that stage, I hadn’t made any attempt to adjust the ‘offset’ by sending commands to it from a PC via the RS232C serial port. So the unit was still running with whatever offset figure had been stored in its EPROM way-back-when. Small wonder that it was locking to a frequency of ‘notquite’ 10MHz but just 0.008Hz (eight milliHertz!) short of it. By the way, the exact resonant frequency of the resonance cavity inside every rubidium-vapour reference depends on many parameters, some of them quite subtle. That’s why they need to be programmable in terms of the ‘offset’ that needs to be applied to their internal frequency synthesiser, to bring their locked output frequency to the correct figure. This offset programmability also allows them to be recalibrated from time to time, to correct for any long-term drift. It even allows them to be ‘locked’ to the GPS system, by comparing the timing of their 1pps pulses with those from a GPS receiver, but more about this later. Another surprise So why did the FE-5680A have to be turned upside down to achieve the lock? I could only guess that it was because of the slightly different flux density and orientation of the Earth’s magnetic field at my house. To check this theory, I turned the power off and let it cool down again. I then returned it to the right-sideup orientation and reapplied pow­ er. Much to my surprise, it quickly achieved a lock, this time in about two minutes and 25 seconds. And when I repeated this test a few more times, the same thing happened! April 2014  69 COOLING FAN 12V/130mA 47 Ω 5W +18V +5V OUT + – 22k E B 1 6 2 7 3 8 4 9 5 (DE-9M) +18V + – 1000 µF 25V CON5 ~1 µS 100nF IC1: 74HC14 +18V 1pps OUT GND 10MHz LOCK RxD +5V TxD 1 14 7 3 2 IC1a DE-9F 1pps OUT 6 8 IC1e 11 10 IC1f 13 1.5k 4 IC1d 9 GND CON2 IC1b IC1c 5 FREQUENCY ELECTRONICS FE-5680A RUBIDIUM FREQUENCY STANDARD K C 4.7k CON1 +18V IN 1pps OUT GND 10MHz OUT LOCK RxD IN +5V IN TxD OUT SIG GND A IN GND 220 µF Q1 PN200 FROM 18V/2.5A PLUG PACK ZD1 6.8V 1W REG1: 7805 12 CON3 CON4 10MHz OUT 1 6 RS-232C LINK WITH PC 2 7 3 8 4 9 5 DE-9F SC 20 1 4 ~780mV TxD A RxD LOCK NOTE: CONNECTIONS SHOWN FOR CON1 ARE THOSE TO SUIT MOST ‘USED’ FE-5680A UNITS CURRENTLY AVAILABLE. λ LED1 LED ZD1 A K K B K C A 7805 PN200 GND IN E GND OUT RUBIDIUM FREQUENCY/TIME STANDARD TEST RIG Fig.2: the circuit for the complete test rig, including the breakout board. A 74HC14 hex Schmitt inverter (IC1a-IC1f) is used as an output buffer for the 1pps output from the FE-5680A rubidium standard, while transistor Q1 buffers the LOCK-bar output to drive lock indicator LED1. An 18V plugpack supply powers both the FE-5680A and its cooling fan, while zener diode ZD1 and regulator REG1 derive the 5V supply for IC1, Q1 and the logic inside the FE-5680A. Somehow, whatever had initially prevented it from achieving a lock when it was the right way up had fixed itself and the unit was now able to lock reliably regardless of its orientation. Building a test rig It was now time to set the FE5680A up as a working frequency and time reference. The first step was to build a test rig using some pieces of 4mm-thick sheet aluminium (see photo). There’s a bottom plate to act as a heatsink for the FE-5680A, plus a back-plate to support the 12V fan and a small ‘breakout board’ for the power supply, lock indication and interfacing circuitry. There’s also a smaller front-plate to support a DE-9F serial port connector, plus BNC sockets for the 10MHz and 1pps (one pulse per second) outputs. It’s utilitarian but it works (you could also build it into a case, such as the Jaycar HB-5446). The circuit for the complete FE5680A test rig is shown in Fig.2. The 70  Silicon Chip FE-5680A itself is shown as a blue rectangle at centre left and all the connections to it are made via DE-9F connector CON1. As shown, the TxD and RxD connections on pins 8 & 9 are fed directly through to DE-9F connector CON4 on the rig’s front panel. CON4 is for connecting the test rig to a PC. However, PCs and laptops no longer have an RS232 interface, I have designed an RS232-to-USB interface and that device is described elsewhere in this issue. This can not only be used with the rubidium frequency standard but can be used anywhere a legacy instrument with an RS232 interface needs to be hooked up to the USB port of a current-model PC. Getting back to Fig.2, the FE-5680A’s 10MHz output from pin 7 of CON1 is fed directly to CON3, one of two BNC output connectors on the front panel. The rest of the connections on CON1 are taken to the external breakout PCB at the back of the test rig. As shown, pin 1 is connected to a +18V supply rail which is fed into the breakout PCB via CON5 (ie, from the external +18V plugpack). And pin 4 is connected to the +5V supply rail provided by REG1, a 7805 3-terminal voltage regulator which is fed from the plugpack via series zener diode ZD1. ZD1 is used to drop the input voltage by about 6.8V so that REG1 doesn’t dissipate too much power. The 12V cooling fan is powered from the +18V rail via a 47Ω 5W series dropping resistor. There are two ‘signal processing’ circuits on the breakout PCB, both quite straightforward. One is a simple buffer using PNP transistor Q1 to drive LED1 as a LOCK indicator. As shown, Q1’s base is connected to pin 3 of CON1 via a 4.7kΩ resistor, so that the transistor is held off whenever the FE-5680A holds its LOCK-bar output high. Conversely, when it pulls this output low to indicate that it has locked, Q1 turns on and LED1 lights. The remaining circuitry on the PCB involves IC1, a 74HC14 hex Schmitt siliconchip.com.au Rb FREQ REF PN200 A 1.5k LED1 LOCK 14150140 100nF GND GND 74HC14 1PPS LOCK GND 1PPS Q1 4.7k 22k K FAN POWER + + 4102 C 220 µF BUFFERED 1PPS OUTPUT 1000 µF + 6.8V ZD1 (P) REG1 7805 IC1 +18V +5V GND – 47 Ω 5W CON1 18V DC IN + + inverter. This is used as a buffer for the FE-5680A’s 1pps output which appears at pin 6 of CON1 when the unit is locked (note: a buffer is needed because the FE-5680A’s 1pps output has very little drive capability). One of the six inverters is used at the input to minimise the loading and this then drives the five remaining Schmitt inverters in parallel to provide increased drive capability while also re-inverting the pulses. This double inversion inside IC1 does introduce a small propagation delay but this is no more than about 40ns, so it doesn’t matter. The output pulses from IC1b-IC1f are fed directly to CON2, the second BNC socket on the test rig’s front panel. So there it is: a simple test rig which can be used to bring a retired FE-5680A rubidium-vapour frequency and time standard back to life. By making some relatively minor changes (eg, to cater for different connector pin-outs), it could also be used to resurrect other rubidium vapour standards. Y C NEUQERF MUIDI BUR DRAD NATS E MIT D NA DRA O B TU OKAER B Fig.3: install the parts on the breakout PCB as shown here, taking care to ensure that all polarised parts are correctly orientated. Note that REG1’s tab must be fastened to the metal end panel of the test jig for heatsinking. Building the breakout PCB The breakout circuit is built on a PCB coded 04105141 and measuring 102 x 38.5mm. Fig.3 shows the assembly details. No particular order need be followed when installing the parts on the PCB – just be sure to install the polarised parts with the correct orientation. It’s a good idea to mount the 5W resistor slightly proud of the PCB, to allow the air to circulate beneath it for cooling. Once the assembly is complete, it can be attached to the rear aluminium end panel using a couple of right-angle brackets. REG1’s tab is then fastened to this panel using an M3 x 10mm machine screw, nut and lockwasher, to provide heatsinking. The fan is also fitted to this end panel (after making a matching cutout) so that it blows air across the FE-5680A rubidium standard mounted on the base. The DE-9F connector and the two BNC output sockets go on the front plate of the test jig, as stated previously. Tuning offset As mentioned earlier, even when a rubidium-vapour standard like the FE-5680A warms up and ‘locks to rubidium’, its output frequency will not be exactly equal to 10MHz. That’s besiliconchip.com.au Here’s a closer view of the author’s breakout board, mounted above the cooling fan on the rear plate of the test rig. REG1’s mounting tab is screwed to the rear plate as well, to provide adequate heatsinking. cause of the many complex parameters controlling the resonant frequency of the resonance cell – close to the theoretical figure of 6.834,682,612,8GHz but not exactly so. And the actual frequency also very drifts very slowly with time. Because of this and regardless of whether the standard has been languishing unused on a shelf or running for many months, it’s generally necessary to program the standard’s internal frequency synthesiser. That’s done to bring its output frequency as close as possible to the magic figure of 10,000,000.0000Hz. With the FE-5680A and most other Rb-vapour standards made in the last 15 years or so, the internal frequency synthesiser is a DDS (direct digital synthesiser). As previously stated, this is programmed by sending an offset number to it from a PC via its RS232C port. The offset is generally a 32-bit binary number, which is sent to the standard as a signed 32-bit integer April 2014  71 The two BNC sockets and the DE-9F socket are mounted on the front end-plate. You will need to build the USB/ RS232C Interface described else­ where in this issue to connect it to your PC so that the tuning offset can be adjusted. The main window of VID’s RS-232 Hex Com Tool v6.01, a serial terminal application which runs under Windows but lets you transmit and receive data in hexadecimal – as you can see from the characters in the upper Transmit box. It’s easy to set up and use to send commands to a rubidium standard like the FE-5680A. (usually as eight hexadecimal digits), preceded by a short command. In the case of the FE-5680A, there are two commands to change the offset – one to do so temporarily and the other ‘permanently’ by saving the new offset in its EPROM. As shown in the command table at the bottom of Fig.1, there’s also a third command which allows the PC to request the FE-5680A’s current offset figure. The fact that the offset programming number is a signed 32-bit integer means that the number can have any hexadecimal value between 7FFFFFFF (= +2,147,483,647), through zero (00000000) and down to 80000000 (= -2,147,483,647). And since the significance of a single bit of the offset programming number is stated as 1.7854 x 10-7Hz, this becomes the setting resolution. In other words, the frequency offset can be programmed to any figure between +383Hz and -383Hz, in increments of 1.7854 x 10-7Hz. That’s a pretty good method of fine tuning, isn’t it? Of course, just how closely you’ll be able to coax the output frequency to the ideal 10.000,000,0000MHz will depend mainly on the accuracy and resolution of your measurement setup. If you can only measure down to 0.1Hz, that will be as close as you’ll be able to go. It’s a good example of the old adage that you need a really accurate clock to check another really accurate clock. In my case, I was able to use the 12-Digit 2.5GHz Frequency Counter described in the December 2012 and January 2013 issues of SILICON CHIP. This can measure the frequency with a resolution down to 0.001Hz (1mHz) using the internal gating ranges and down to 0.0001Hz (100µHz) using an additional external ÷10 timebase divider (see the Circuit Notebook pages in this issue). But this is only feasible because I also have a source of 1pps timebase pulses which have excellent accuracy and stability in both the short and long term: a Stanford Research Systems PRS10 Rubidium Standard, as shown on page 38 of the February 2014 article. I bought this a few months ago. Because I’m now running it ‘disciplined’ by the GPS 1pps pulses from my GPS-Disciplined Frequency Standard (SILICON CHIP March-May 2007), its own 1pps output pulses (and 10MHz output) are accurate to within ±5 parts in 10-11. By using my 12-digit counter with this fancy external timebase set-up, I was ready to begin searching for the correct offset to program my FE5680A, so that its output would move as close as possible to 10MHz (you may recall that when I first got it to lock, its frequency turned out to be 9,999,999.992Hz, ie, 8mHz low). I had to do two things before this could be done, however. The first job was to make up a USB/RS232C Issues Getting Dog-Eared? Keep your copies of SILICON CHIP safe with these handy binders REAL VALUE AT $14.95 PLUS P&P* Order now from www.siliconchip.com.au/Shop/4 or call (02) 9939 3295 and quote your credit card number or mail the handy order form in this issue. *See website for overseas prices. 72  Silicon Chip siliconchip.com.au Parts List 1 FE-5680A rubidium frequency standard (see text) 1 PCB, code 04105141, 102 x 38.5mm 1 12V cooling fan 2 DE-9F D-sub female sockets (CON1, CON4) 2 panel-mount BNC sockets (CON2, CON3) 1 2.5mm PC-mount DC power socket (CON5) 2 3-way right-angle locking (polarised) headers, PC-mount 2 3-way locking header plugs 1 6-way right-angle polarised pin header, PC-mount 1 6-way locking header plug 1 14-pin DIL IC socket 1 18V 2.5A plugpack supply Aluminium panels and brackets to make test jig Semiconductors 1 74HC14 hex Schmitt inverter (IC1) 1 PN200 PNP transistor (Q1) 1 7805 regulator (REG1) 1 6.8V 1W zener diode (ZD1) 1 green 5mm LED (LED1) Capacitors 1 1000μF 25V electrolytic 1 220μF 10V electrolytic 1 100nF MKT ceramic Resistors (0.25W, 1%) 1 22kΩ 1 1.5kΩ 1 4.7kΩ 1 47Ω 5W 10% Miscellaneous Machine screws, nuts & washers, hook-up wire, cable ties interface so that I could hook up the FE-5680A to the PC that I was going to use, as my PRS10 standard was already connected to the PC’s one and only legacy RS232C port. That’s one of the reasons why I developed the USB/ RS232C interface described elsewhere in this issue of SILICON CHIP. Once that had been done, I then needed a ‘serial terminal’ program that would run on Windows XP SP3, communicate via a USB virtual COM port and preferably also allow me to send and receive messages in hexadecimal to make things easier (hex is a lot easier than straight binary). After spending quite a bit of time downloading and trying out a number siliconchip.com.au Here’s the ‘tweaked offset’ output frequency of the author’s FE-5680A rubidium standard, captured on the 12-Digit 1GHz Frequency Counter using a home-brew time period divider to extend the gating time to 10,000 seconds. As you can see, the reading is 10,000,000.0000Hz, although the decimal point is not in the correct position. DB-9 Or DE-9: Which Is Correct? A “DE-9” D-sub 9-pin connector is often mistakenly referred to as a “DB-9” connector. The “E” refers to the shell size. A “DB-25” connector has a “B” size shell but the common 9-pin connector is smaller and has an “E” size shell. This connector (regardless as to what you call it) is used for a variety of purposes. Two common applications are RS232/EIA-232 (serial) connections (including UPS cables) and a variety of video interfaces on the IBM PC. of freeware terminal programs, I finally settled on a program called ‘RS232 Hex Com Tool v6.01’, written by a firm called Virtual Integrated Design (VID). A free demo version of this can be downloaded from VID’s website (see the links panel) but it closes down after about three minutes of operation and must be started up again if you want to use it for another three minutes – a bit irritating. After you have used it for a short time, you’ll probably want to purchase the full registered version. This is sold online for about $US40, via another firm called SWREG Inc (see the links panel again). About the only thing that this terminal program doesn’t do for you is work out the special ‘exclusive OR checksum’ that the FE-5680A needs after the data bytes are sent to it in the two change offset commands. Still, if you’re only sending change offset commands with fairly small numbers (as we are here), it’s not all that hard to work out the checksum yourself. Doing it At this stage, I was ready to try reprogramming my FE-5680A with an offset which would bring its output frequency as close as possible to 10MHz. First off, I hooked everything up and launched the RS232 Hex Com Tool program. Then I sent the FE-5680A the command to discover its current offset, as stored in its EEPROM. As you can see from Fig.1, this command is ‘2D 04 00 29’ and it must be terminated in a carriage return (0D hex). The FE-5680A immediately responded by sending back ‘2D 09 00 24 00 00 00 00 00’, terminated in another carriage return (0D hex). This showed me that the offset currently stored in its EEPROM and being used to set the DDS was zero – ie, 00 00 00 00, with a data checksum at the end of 00. So now I knew that this particular April 2014  73 Why Not Purchase A New Rubidium Standard? A month or two before I wrote the article on rubidium standards for in the February 2014 issue of SILICON CHIP, I had made a decision to try to acquire one for my workshop. At that stage, I wasn’t aware of the really low-cost surplus units available via eBay, so I searched around and came up with a short list of just two units that seemed to be just within my modest test gear budget: the Quartzlock E10-MRX (February, page 36) and the Stanford Research Systems PRS10 (February, page 38). Both these units were available for less than $2300 including GST, so it was a matter of choosing between them. In the end, I decided in favour of the PRS10 because it was capable of being ‘disci- plined’ by the 1pps pulses from a GPS receiver. The E10-MRX didn’t seem to offer this feature and I wanted to be able to experiment along these lines to see if I could avoid having to send it away for calibration every year or two. So I went ahead and ordered a PRS10 from the Stanford Research website, together with their small ‘breakout board’ which makes it easier to connect everything together. It arrived a few weeks later, complete with an operation and service manual in a neat 3-ring binder. Shortly after it turned up, I also bought a 24V/2.7A switchmode power supply (the PRS10 runs from 24V DC) and gave the PRS10 a quick ‘test run’ to see how it performed. Handy Links Frequency Electronics official FE-5680A product page: http://www.freqelec.com/rb_osc_fe5680a.html KO4BB’s FAQ for the FE-5680A Rubidium Frequency Standard: http://ko4bb.com/dokuwiki/doku.php?id=precision_timing:fe5680a_faq Time Nuts mailing list archives: https://www.febo.com/mailman/listinfo/time-nuts and also at http://www.mail-archive.com/time-nuts<at>febo.com/info.html FE-5680A Series Option 2 Technical & Maintenance Instructions: www.ka7oei.com/10_MHz_Rubidium_FE-5680A.html www.guido-speer.de/Pub/images/Rubidium/5680_TECH_MANUAL.pdf Another good source for info on precise frequency & time: http://leapsecond.com/ To download a free demo version of RS232 Hex Com Tool v6.0: http://www.rs232pro.com/ To buy and download a full (registered) version of RS232 Hex Com Tool v6.0: Either go to http://www.rs232pro.com/ and click on the ‘registered’ link, or go directly to https://usd.swreg.org/soft_shop/47653/shopscr6.shtml Partial digital schematic (V0.3) of the FE-5680A: http://www.rhodiatoce.com/pics/time-nuts/FE-5680A/FE-5680A_schematics_ v0.3.pdf FE-5680A, locked to rubidium and connected to my test rig, was providing an output of 9,999,999.992Hz with an offset of zero. That meant that I now had to try sending positive offset numbers to the FE-5680A to nudge its output frequency up by close to 0.008Hz, to reach the magic number of 10MHz. This actually took longer than you might think, mainly because to get the measurement resolution, I had to have the counter set initially for a gating 74  Silicon Chip time of 1000 seconds (16.67 minutes). And that meant sending a new offset number and then waiting for just over half an hour (2 x 16.67 minutes) to see the result. When I got close to the ‘finish line’, I then had to use the external timebase divider to give me a gating time of 10,000 seconds (2 hours and 47 minutes), to bring the counter resolution down to 100µHz. This now made for really slow progress because after The results were so impressive that I immediately decided to build both the PRS10 and the power supply into an instrument case, together with a cooling fan at the back of the case and a small wideband distribution amplifier for the 10MHz output. This PRS10-based frequency and time standard has been running for about three months now, disciplined by the 1pps pulses from my GPS Frequency Reference. Its own ‘even more stable’ 1pps pulses are being fed into the external timebase input on my 12-digit counter, to give it even better accuracy than when I was just using 1pps pulses from the GPS reference or Deluxe 1pps Time Receiver. sending a new offset number, I then had to wait nearly six hours to see the result. In short, you have to be patient when adjusting the offset! After some trial and error, I was finally able to find the offset number (00 00 02 F8) which brought the FE-5680A’s output frequency to 10,000,000.0000Hz – as you can see from the photo of the frequency counter. After that, it displayed this impressive figure for over a week – with just the occasional appearance of a ‘1’ in the least significant digit. Of course, most readers won’t have a second rubidium standard to serve as a timebase for the frequency counter. In that case, the best approach is to use the 1pps signal from a GPS receiver, eg, the GPS 1pps Timebase described in February 2013 or the Deluxe GPS 1pps Timebase described in April 2013. These are sufficiently accurate over long time periods to do the job. A final word So that’s the story of how I was finally able to get a low-cost ‘retired’ FE-5680A rubidium vapour standard going again and set up as a very useful frequency and time reference. It turned out that both units I purchased on-line were OK straight out of the box and there was no need to go though any of the tedious disassembly of the physics package or rejuvenation of the rubidium-vapour discharge lamp. I don’t know whether I was fortunate or whether this would normally be the SC case with used units, though. siliconchip.com.au