www.makarov.ca -

Analysis of the Lucent RFG-M-RB hardware and schematic

Simple mods improve quality of the reference RF signals produced by the Lucent rubidium standard

Rubidium frequency standards are used in the commercial applications as the frequency etalon, or reference. They are also available in the aftermarket at affordable price. Many amateur radio enthusiasts use rubidium frequency standards for their satellite and lab applications. I was in a need for a stable 10MHz frequency reference to verify the timebase of the test equipment that I have in my shack. Rubidium standards are available as the bare bone rubidium oscillators that need external circuitry, or as finished units. After some analysis of my needs and research of the available options I decided to go with a finished unit and purchased one at eBay. My RFG-M-RB unit is from a fairly known line of products made by Lucent and is shown in the following picture. It is a solidly built piece of equipment with a massive heatsink and only required a external 24V power source.

Fig 1. Lucent RFG-M-RB is in the right top corner on my Lab bench.

The device has outputs for 15MHz and 10MHz reference frequencies. After I played with it for a bit I noticed a few things that seemed strange to me and raised concerns about the quality of the generated RF waveforms. The frequency stability was just fine but the 10MHz waiveforms looked awful on my Tektronix scope. You can see one in the above picture. More on this later in the article. So I decided to investigate what may be wrong with the guts inside the device. I opened it and spent some time with a pen and paper to capture parts of the schematic to help me understand how it works. Click here for a copy of the schematics. While I make no guarantees it is 100% accurate, it should give you a very good idea how the thing works in case you have one of these units, but use this schematic at your own risk. In any case, chances are nothing better can be found anywhere else on the Net.

Fig 2. Lucent RFG-M-RB with the cover removed.

The shielded box on the right is the Efratom LPRO-101 rubidium oscillator. It is connected to the board that has an Altera microcontroller chip, filters, MMIC amplifiers, and the associated control circuitry. The output 10MHz and 15MHz reference frequencies are interfaced through the front panel SMA type connectors. The DB9 type connectors on the right side of the front panel are to connect the power, output the alarm signals, and communicate with other Lucent equipment (I believe it is the interface to the Lucent GPS-disciplined OCXO).

The cover that was removed and not shown in this picture is actually one big heatsink which the board and the Efratom oscillator are bolted to. Under normal operation the Lucent unit gets noticeably warm. Beware that when the cover is removed, heat removal is lost and the Efratom oscillator runs quite hot. A simple way to prevent overheating is to fully disassemble the cover which then splits to two halves, and bolt the oscillator assembly to the right half. Access to the board remains open.

Fig 3. The control and RF board.

Going from the left are the voltage regulators at the top left, below the regulators there is a timer that waits 10 minutes after the unit is powered on to allow the rubidium oscillator to warm up, and then enables the RF output. Then goes the Altera chip, to the right of which there is a watchdog circuit that senses the 15MHz RF output power and sends a fault signal to the microcontroller if there is no RF output. If this or other fault happens, the Altera chip turns on the Fault LED and sets optocouplers connected to the Alarm DB9 (J3) connector on the front panel, signalling a system fault.

The other DB9 interface connector is also wired to the Altera chip and is apparently implements a communication interface for connecting to other Lucent equipment. A possible setup that can be realized would be the rubidium frequency reference wired to the GPS-disciplined OCXO (oven controlled crystal oscillator). Such tandem would ensure both short- and long-term frequency stability for the output reference frequency. The interface may also be providing some other type of telemetry, but the protocol that this interface is using is unknown.

In the center of the board there is the 15MHz crystal filter which selects the 3rd harmonic out of the 5MHz square wave the Altera chip supplies via a LC-circuit L1C1 (C1 is the trimmer capacitor that you can see just below the filter's left side). The Altera chip produces a 5MHz signal out of the 10MHz reference supplied by the Efratom rubidium oscillator, by dividing the 10MHz by two. The 15MHz crystal filter is approximately 40kHz wide with 4db in-band loss, and the output of the filter goes through a 2-stage MMIC amplifier U3/U4 with about 36dB gain combined. You can see the U3 and U4 to the right of the crystal filter.

The 15MHz RF output from the U4 goes through the Minicircuit PLP-18-11 lowpass filter to suppress the higher order harmonics, and is then interfaced to the world through a SMA connector on the front panel. This output is the sinewave 15MHz reference frequency.

Two supplementary 10MHz outputs are taken from the Altera chip. One (marked TP J1 on the front panel) is always on and comes to the front panel via a emitter follower. The second one (J2, marked as 10MHz RF Out) is connected directly to the chip via a 200 Ohm resistor. None of these 10MHz outputs is sine wave. They are rather poorly shaped waveforms very much dependant on the impedance of the load as shown in the screenshots below.


10MHz TP (J1) signal into 50 Ohm scope input. The signal is produced by the Altera chip and the waveform is rather unusual, not sine wave or square wave. I tried different things thinking that my unit might be defective but it seems this is what the Altera chip is actually outputting.		10MHz TP (J1) signal into 1 MOhm 15pF scope input via a 3 feet coax. The emitter follower seems to oscillate with high impedance loads, and the oscillation peaks and valleys depended on the length of the coax that was connecting the TP port to the scope.

10MHz RF Out (J2) signal into 50 Ohm scope input.		10MHz RF Out (J2) signal into 1 MOhm 15pF scope input via a 3 feet coax.

So remember to switch to 50 Ohm termination your equipment that you connect to this Lucent unit's 10 MHz TP or RF outputs (the 15MHz RF one is sine wave and behaves fine under different loads). If your equipment does not have a 50 Ohm input, the equipment can be connected via a 50 Ohm passthrough, i.e. a T-connector with the stub shoulder terminated with 50 Ohm. Connect such passthrough on the Lucent side rather than your equipment side to eliminate oscillation and minimize waveform distortion.Alternatively, both the TP and 10MHz RF Out signals can be conditioned and converted to a clean sine wave as I am going to show further in this article.

I used my HP8753C network analyzer to scan the 15MHz crystal filter, and the results are shown below. If you have a non-working Lucent unit, the filter can be removed and used elsewhere, as it seems to be a nice part. For the test the network analyzer's RF output was connected to the left side of L1 (see the Schematic), and the input of the analyzer was connected to 15MHz RF Out (J4) on the front panel. So the test path included the matching circuit L1C1C17, the crystal filter, both MMIC stages, and the MCL lowpass filter. The VNA excitation signal was set low to not to overdrive the MMICs and the assumption was all those extra parts do not have sensible impact on the transmission and reflection characteristics of the crystal filter. You see gain in the transmission plots because of the amplifiers. The crystal filter loss in the passband was measured separately and was 4dB.

Fig. 4. The 15MHz crystal filter. Wideband scan. The 3dB bandwidth is 40kHz.

Fig. 4. The 15MHz crystal filter. Narrowband scan and input Return Loss.

After checking the stages of the RF channels I could see two problems with the design. First, all 15MHz RF stages including the crystal filter were badly overdriven, passing high harmonics content from one to the other. Second, it was a surprise to me that only the 15MHz reference was sinewave, the 10MHz ones were not. Third, the 10MHz outputs could not handle high impedance loads. One that had a emitter follower as the output stage was prone to oscillation even with a 1 meter long coaxial cable connecting it to the scope. The other one, which was unbuffered, had the rise and fall time badly changed if not terminated to 50 Ohm. May be the design assumption was the unit will always be working in a 50 Ohm environment, do not know. My advice, as I said above, is to use a 50 Ohm passthrough at the Lucent rubidium unit side if your referenced equipment does not provide 50 Ohm termination.

To clean the RF signals and improve buffering I came up with a few mods that I describe in the following sections.

The 15MHz channel mod

Click here for the part of the schematic that shows the 15MHz channel. The multiplied by 3 and divided by 2 reference 10MHz signal from the internal rubidium source is output via pin 5 of the Altera chip, passes through the 15MHz bandpass crystal filter FL1, amplifier stages U3 and U4, and low pass Minicircuits filter FL2 with 18MHz cutoff frequency and is output on the front panel via the 15MHz OUT connector J4 . It was found that all of the stages were badly overdriven with the signal coming out of the Altera chip pin 5. The mod described in this section was to drop the power level of the source signal entering the crystal filter stage. It is a very simple mod that produced effective results and only involves adding a resistor divider composed of the resistors Ra and Rb as shown in the following picture.

The results are shown in the following table in two columns. The left column shows the screenshots before the mod, the right column after the mod.


U1 pin 5 before the mod. The U1 is having problems hard driving the 15MHZ crystal filter input stage.		U1 pin 5 after the mod. The U1 is working in much lighter conditions.

Input into the 15MHz crystal filter before the mod. The waveform is distorted at the top probably because of the output stage of the U1 hard driving the filter input.		Input into the 15MHz crystal filter after the mod. The waveform is symmetrical and rings uniformly at the top and bottom.

Input into the first amplifier stage U3 before the mod. The waveform is pretty ugly. It seems the 15MHz crystal filter is having hard time because of the excessive signal power at its input.		Input into the first amplifier stage U3 after the mod. The signal amplitude is smaller, however is much much cleaner.

Input into the second amplifier stage U4 before the mod. Seems to have strong harmonic content.		Input into the second amplifier stage U4 after the mod. Clean 15MHz sine wave.

Output of the amplifier U4 into the MCL 18MHz lowpasss filter before the mod. The poor U4 MMIC is badly overdriven and pumping extra harmonic content into the LPF.		Output of the amplifier U4 into the MCL 18 MHz lowpass filter after the mod. Clean 15MHz sinewave.

Front panel 15MHz RF Out (J4) spectrum before the mod. Vertical scale 10MHz/, Horizontal scale 2MHz/. The 15MHz carrier is on the left side. The 3rd harmonic at 45MHz is -57dBc.		Front panel 15MHz RF Out (J4) spectrum after the mod. The 3rd harmonic is undetectable and is below -70dBc.

The 10MHz channel mods

Several options exist how to improve the 10MHz channel RF signal quality depending if square wave or sinewave is required. In case the goal is to have a TTL level digital 10MHz RF output, a simple way to have it done would be to remove the jumper W1 (see the schematic) and insert a stage built on a 74AC74 fast IC as shown in the following picture.

In case 10MHz sinewave is needed, there is a couple ways to do it. I have built, tested and measured two converters. They employ a combination of digital or analog circuits with a low pass filter. The converters have different level of harmonic suppression as I show further in this article.

The 10MHz filter that I used was taken from a old 10Mbit Ethernet LAN board. The part number I used was 20F001N (click here for the Datasheet), but I would think many other LPF filters from 10BaseT LAN boards can be used, as they all should have similar cutoff frequency. This is a fairly good filter for the purpose and after all it was free. The filter has two channels that are referred to as TX and RX (apparently for the transmitting and receiving channels of the LAN board) which seemed to have some difference in the frequency roll off characteristics. The converter board that I built using a pair of these filters is shown in the following picture.

The board has two circuits that I evaluated. One is based on a TTL buffer IC 7404 and has two branches each connected to the halves of the LPF as shown further in this article. Even a TTL IC was used here, the RF outputs from each of the branches are sinewaves due to the harmonics removal by the LPF. Same as in the proposed schematic in the above picture, the jumper W1 was removed, the input of this circuit was connected to R13 and the output to the front panel SMA connector J2.

The second circuit that I evaluated was based on a HF operational amplifier IC AD8041 that I had on hand. The output from the amplifier passes through the TX part of the LPF which seems to have better harmonics filtering than its RX counterpart. Not like the above described digital circuit, the input of this buffer amplifier I connected directly to the rubidium source, wiring it to C31. The circuit gain was then tweaked for minimum distortion. The reason I connected this buffer directly to the rubidium source was I wanted to produce as clean of a 10MHz reference as I could, bypassing the Altera chip - who knows how much jitter it introduced. This circuit had better harmonics filtering then the digital one, suppressing the second harmonic down to almost -60dBc and third harmonic down to almost 70dBc. The following few pictures provide the details for each of the circuits.

The circuit shown above has two branches. The upper branch drives the TX portion of the filter in push-pull configuration with two inverters, and as such outputs more power (+13dBm) than the lower branch (+8dBm) which drives the RX portion of the filter using a single invertor. The two branches have different level of harmonics suppression - click on the picture for details on the numbers. Anyway, the point is that with a single IC and one LPF filter you get two sinewave 10MHz reference outputs for a very low cost. The following picture shows the spectrum screenshots for the upper branch before and after the mod.


	10MHz output RF spectrum before the mod. Needs to be cleaned unless you want to use a harmonic.		10MHz output RF spectrum after the mod. Note the 5MHz and 15MHz spurs are also 10dB down compare to the no mod.

The schematics for the analog mod is given in the following picture. It can be used with or without the digital mod. The purpose of this analog mod was to produce a clean 10MHz sinewave reference signal. Same 20F001N Ethernet lowpass filter part was used here and worked great. The circuit outputs +14dB (~3Vp-p) into 50 Ohm. The second and third harmonics were measured 60dB and 70dB below the fundamental.

This analog mod has high level of harmonics suppression and produces possibly the best sinewave 10MHz reference that can be gotten out of this Lucent unit at a low cost.

Conclusion

The article provides some technical information about the Lucent RFG-M-RB rubidium frequency reference. Analysis of the parts of the schematics was performed and several circuits built to improve quality of the RF reference signals. This information may also prove useful for repair or restoration of the hardware. It can be suggested that other rubidium frequency reference hardware manufactured by Lucent may have similar architecture, in which case the information in this article may also be of a help.