A High Performance Frequency Standard
- and the -
VNG in a Box

By Murray Greenman ZL1BPU

A trio of GPS projects, with LCD display and clock. Use an oven controlled oscillator for incredible accuracy. Time code generation.
 GPS Clock Core  VNG Time Code  Oscillator Control
 Telemetry & Commands  Phase Comparator  GPS Monitor/Display
 PC Software

INTRODUCTION

This design is an improvement on the popular GPSClock, and uses similar hardware. Most of the improvements are aimed at providing support for higher stability oscillators, especially the oven controlled type. The additions allow use of positive or negative slope OCXO references, provides time code generation, full holdover when GPS fix is lost, power failure lockout and default mode etc. This design is capable of Stratum 2 level Frequency Standard performance.

The design locks a 10MHz voltage-controlled OCXO to the GPS reference, in just a few minutes, and is capable of providing a low noise, accurate reference, with frequency performance at the level of a few parts in 1010 (1e-10) or better, typically close to 1 in 1012 (depends on oscillator quality and care with adjustment). With good holdover performance and typical variance figures around 10-11, the unit compares well with much more complex designs.

In addition to controlling the reference, the design provides a local (independent) but GPS synchronized time clock, generates a 1kHz reference, and provides a VNG format time code output. All these products are controlled to within 50us of UTC. 10MHz carrier accuracy is typically well within 0.1ns over 10 seconds.

Of course the reference can be any other frequency from 8MHz to 16MHz, and can be buffered and used to operate dividers for other reference frequencies. (For OCXO frequencies below 8MHz, use a frequency doubler to operate the micro.) Oh - and did I mention that this is essentially a single chip design?

This project can be used in three ways:

(The picture on the right is a natural light screen shot of the PLED clock display - it is blurred because of the long exposure time.)

In this discussion we'll call the device the VNGBOX, but unless otherwise mentioned, the comments apply to the 'High Performance Frequency Standard' and the 'Phase Comparator for Frequency Standards' as well.



The GPS Display (left) and VNGBOX (centre) prototypes in a recycled cabinet
New life for an old HP5326A counter!

The clock and reference continue to operate accurately when GPS is lost, and although the reference frequency will then drift as the oven oscillator ages, it should hold to within 109 for at least 24 hours (depends on oscillator quality). The controller is held digitally at the current operating point until GPS returns. One hour holdover has been measured (using an HP 10811A oscillator) at 2us.

Most GPS units continue to send 1pps pulses even when the GPS fix is lost, and these can be sufficiently inaccurate to cause the much higher accuracy OCXO to be forced off frequency. With this unit, the GPS fix can be monitored (by a separate micro, called the GPS Monitor/Display) and the unit put into HOLD if the fix quality is poor, or reception is poor, thus preventing the unit from wandering off following the inaccurate pulses when the fix is lost. It is best used with a Motorola Oncore or Navman Jupiter T GPS engine with TRAIM capability.

When the two micros are combined in a single unit, the GPS data and the telemetry from the VNGBOX can be combined into a single data stream for PC display of both GPS and local reference!

The clock is essentially bullet-proof and has perfect accuracy, even if GPS is occasionally or frequently lost. On repower after power failure, where local time will be lost, the device returns to a safe condition, and automatically locks to GPS with the time code is disabled. The time code is restored when the clock is reset and the preferred mode set. The unit should be used with a backup UPS power supply capable of 50VA or so. A single supply 24V design would be practical, given suitable power supplies. The total power requirements are about 10W (depends on the OCXO).

Here is a summary of the special features of the VNGBOX:

* Longer integration times give lower noise for better quality oscillators. If the oscillator wanders more, shorter integration times allow the unit to track the changes more quickly.



The VNGBOX 16 x 2 PLED display explained

The display shown in the photograph is a 16 x 2 unit. At the top left is the VNGBOX (locally set) time, in 24 hour format. The information shown on the right of the display is the reference phase (top) and feedback term, related to oscillator Electronic Frequency Control (EFC) voltage (bottom).

The rectangles (bottom left) flash in time with the time code, or are replaced by an error message when there is a fault condition. To the right are displayed the operating mode, and the flag byte (helps diagnose problems) and the feedback integration timer.

The basic VNGBOX is simple to build - two ICs and a couple of transistors. Very likely the GPS engine and reference oscillator will cost more than the rest of the parts. The LCD display is a standard type used everywhere. The unit is also fully functional without the display. A new organic LED or PLED display (as shown) can also be used with no circuit or code changes.

The unit can be this simple, or made as complex as you wish, by adding the extra features - a better OCXO, better regulated power supplies, active filters and amplifiers in the feedback path, and of course the GPS Monitor/Display module (second micro).

Precision time keeping and the precision 10MHz reference are achieved through the use of a Voltage-Controlled Oven Controlled Oscillator. For reduced performance, but lower cost and power consumption (say for a portable VNGBOX), a Temperature Compensated Crystal Oscillator (VTXO) is suggested. These can cost $30 new, but can often be found on the internet at good prices. Just about any frequency from 8 to 15 MHz can be used, but unless you plan only to use the device as a clock and phase comparator, you MUST use a voltage-controlled oscillator. These devices usually have a control range from about 2 parts in 107 to 1 in 106. The design will handle both negative and positive control senses (but only positive voltages). Bipolar devices can be controlled using an offset technique.

The clock transmits serial telemetry data every second, in addition to providing the LCD display. The serial data can be used for time logging, or setting other clocks. The phase information can be used for calibration or long-term tracking of the reference oscillator performance. A special companion monitoring program has been designed for this unit.

SOFTWARE  

The firmware and PC software for this project is well tested, versatile (any reference frequency can be used), comprehensive (includes software for both for micro and PC) and is inexpensive. See the Micro Projects page for purchasing details.

A completely new PC program, REFMON4, has been developed for this project. While the telemetry is compatible with the older software (RECORD2F), the new software has a number of improved features.



Oscillator performance monitoring with REFMON4
Click on image to view full size.

This recording was made with a 4MHz NDK OCXO, adapted for voltage control, output divided by two and then multiplied by five to operate the micro at 10MHz. (Here's another example using the HP 10811A oscillator). The long term nature of this recording also demonstrates the very low thermal phase shift in the oscillator and multiplier. You would expect to see a daily variation in the EFC (feedback voltage), and there is no obvious daily effect. Over one week you also see very little effect of oscillator ageing, which would cause a linear trend in the EFC voltage.

Main Graph Display
The vertical scale of the main graph represents 0 - 25µs with 100ns resolution. The horizontal scale is seconds, marked every 10, 60 and 3600 seconds (one hour). The horizontal resolution is one sample interval.

The display of all graphical elements is updated every second, even at very slow chart speeds (so each horizontal position contains sample rate samples). All events recorded are overdetermined (new pixels are plotted every second, even though the graph horizontal speed is much slower). In this example there are 1200 samples at each location - any noise on the oscillator will show up very quickly. The main graph is scrolled to the left when the right margin is reached.

Phase and Feedback
The green trace is the least significant eight bits of the second-by-second phase measurement. It has a resolution of 100ns, hence the vertical scale 0 - 25.5µs. In the above example, the oscillator phase has not varied by more than 1us (1e-10 or 1 part in 1010) in that time. - That's equivalent to less than 1 milliHz movement at 10MHz in over a week!

The purple trace is the EFC feedback voltage, the eight most significant bits of this 12-bit value, and represents the feedback voltage 0 - 5V (read the scale as V and divide by 5.12). The latest versions of this software also display the least significant eight bits as a pale grey trace.

These phase and feedback measurements are calculated into real units (µs and Volts) and displayed at the bottom right (in the same colours) every sample interval.

Offset and Variance
Offset and P-P Variance are also calculated and displayed every sample interval. Note the values for offset and variance calculated by the program from this week-long long recording (bottom right). Pretty impressive! Note also that the colours match the plotted offset and variance traces. The offset calculation can result in an error (divide by zero) if there has been no phase difference, and so instead the worst-case accumulated offset (100ns over the sample interval) is displayed instead.

The pink trace is the p-p variance accumulated from the start of the recording. This is measured by detecting the peak positive and negative differences from a 10-second mean of the phase (this acts as a low pass filter to remove offset from the calculation). The value given is the P-P Variance over an ever increasing sample period, and naturally gets better with time. The 1 sec or 10 sec Tau Allan variance can't be calculated on the fly, but will be about the same order. Variance is a scalar, and has no sign. This graph is plotted with a logarithmic scale - read the vertical axis as powers of 10.

The pale blue trace is the accumulated phase offset from the start of the recording, averaged over 10 sec to remove variance spikes (low pass filter). Use the reciprocal as a measure of frequency error. Of course the error expressed in powers of 10 will be the same for phase and frequency. The value is unit-less (i.e. sec per second or Hz per Hz), and the sign represents the direction of offset. Low frequency (increasing phase) is shown as negative. When there is no measureable phase error, no point is plotted. This graph is also plotted with a logarithmic scale - read the vertical axis as powers of 10. In the above graph the result is impressive - 6.89 parts in 1013!.

GPS Status
Just under the main graph (inside its border, just below the zero line) is a GPS status line. This is normally white, but will show blue if the GPS signal is missing, red if the GPS fix is lost, and green if either the GPS telemetry is lost or misunderstood by the PC. On the slow PC used for this recording, this latter effect clearly happens at least once in most groups of 1200 samples. If your RS232 link is poor, you'll see a lot of green.


The sub-graph and repeater boxes

Fast Sub-graph
Under the main graph is a small sub-graph which operates at higher horizontal speed (one sample per second). The vertical axis is the same for phase, but has full resolution (1mV) for EFC feedback. The vertical axis and horizontal time scale are limited in size, so the data wraps around. Under the graph is a single dot which moves along to indicate the current recording point. The horizontal range is just under five minutes. This graph is not scrolled - the data simply restarts plotting at the beginning, and overwrites the previous data.

This display is very useful for initial oscillator setup, and you can also quickly spot any tendency for the oscillator to move when you open the door and walk into the room! The sub-graph is fascinating to watch, as you can observe the inner workings of the differential/integral error feedback algorithm as it measures phase comparisons across 64 seconds.

LCD Display Repeaters
Finally, under the sub-graph are the two 'LCD repeater' displays. The left small box shows a repeat of the VNGBOX display. It shows local clock time (time code time) at the top left and oscillator phase top right (in hex). The bottom line shows the operating mode, flags and integration time and EFC feedback (in hex).

To its right is the repeat of the GPS Display. It is not copied from the GPS Monitor/Display display (if you see what I mean), but is reconstructed from the GPS telemetry. This means that you don't need the GPS Monitor/Display micro for this to work. The upper line shows the current GPS time (tends to be slightly delayed behind the VNGBOX display). On the lower line of green text the date, latitude, longitude and fix quality (DOP, satellites in fix / satellites in view) are displayed in rotation. If no GPS telemetry is multiplexed with the VNGBOX data, this box does not show. If the GPS data and the VNGBOX telemetry are not correctly multiplexed, the whole program will not work correctly.

Software Summary
In summary, the new features of REFMON4 include:

Presettable information (default chart speed, operating frequency, baud rate, com port) is stored in a setup file. When using REFMON4, every valid telemetry sample is used (i.e. every second), even at very low plot speeds. This means that the graph is overdetermined, and any noise will show as a thickening of the plotted phase line.

 GPS Clock Core  VNG Time Code  Oscillator Control
 Telemetry & Commands  Phase Comparator  GPS Monitor/Display
 Processor Schematic  RS232 Schematic  Oscillator Schematic
Copyright © M. Greenman 1997-2005. All rights reserved. Contact the author before using any of this material.
home HOME