Italian - ItalyEnglish (United Kingdom)
Ulti Clocks content

Online Users

We have 2 guests online

Counter

mod_vvisit_countermod_vvisit_countermod_vvisit_countermod_vvisit_countermod_vvisit_countermod_vvisit_countermod_vvisit_counter
mod_vvisit_counterToday36
mod_vvisit_counterYesterday127
mod_vvisit_counterThis Week374
mod_vvisit_counterThis Mounth3292
mod_vvisit_counterAll Days430311
GPS synchronized 10MHz oscillator Print E-mail
Written by IK0OTG   
Friday, 10 September 2010 14:03

 GPS synchronized 10MHz oscillator             GPS_TITOLO

 

Introduction

 On the network (eBay) and at HAM fairs, it’s easy find

 GPS receivers and its antennas at an affordable price.  

They present themselves as small cards with SMD components, an SMB antenna type connector and a multi pin connector for the information output and for the power input. There are many brands and many models with specific features for each model, but almost all, in addition to typical GPS information (coordinates, height and speed) give also a 1pps signal  (1 pulse per second) synchronized with GPS signal, which derives from a cesium atomic oscillator.

 At Friedrichshfen fair of 2009 (Germany), for 25.00 €, I purchased the model RESOLUTION T of TRIMBLE, inclusive of its pre amplified antenna ( Fig.1).

 

GPS_10MHz_FIG1

 

Fig.1

In Figure 2 you see the same type of card, which I have housed inside a metallic box approximately 9 cm to 3 cm, connected to 5-pin connector (microphone type).

 

GPS_10MHz_FIG2

 

Fig.2

The GPS antenna must be placed outside in position to receive at least 3 or 4 satellites. Typically these pre amplified antennas have a coaxial cable 2 or 3 metres long. Not included extensions on this cable. Made of appropriate length, the cable that connects your GPS with the controller.

 

Again, on network and fairs are sold, for about ten Euros, plus any delivery charges, 10 MHz Quartz Oven Oscillators, with excellent stability in the short term and whose oscillation frequency can be varied via an internal trimmer, or through an external voltage (Vcontrol). The range of frequency through Vc, typically is ± 0, 5 ppm (i.e. ± 5 Hz for a 10 MHz oscillator).

          There are two types
  • with negative slope (i.e. increasing or decreasing Vc decreases or increases the frequency)
  • with positive slope  (i.e. increasing or decreasing Vc increases or decreases the frequency)

These oscillators are of many brands and models, typically smaller and more expensive they are.

 

GPS_10MHz_FIG3

 

 

 Fig.3

 

The one shown in Figure 3 (negative slope) is old (39th week of 1993) and not just small, (8x6x4 cm) but fully functional.

 

Most of the OMs employ measuring instruments, found on the market of the surplus, whose precision depends on their age, state of use and price. The quality and reputation of the manufacturer alone are not enough to ensure the accuracy of the instruments, in fact, due to ageing components, all oscillators, including atomic Rubidium and excluding only cesium type, moving frequency over time. Manufacturers of oscillators indicate this shift with the term Aging Stability and is expressed in parts per year. Finally, as all oscillators, even those rubidium types, have a trimmer for the correction of frequency, after some years, the false contacts on this trimmer can cause appreciable frequency variations.

To make an accurate calibration of frequency meters, frequency generators, spectrum analyzers, etc. you need to have a generator to 10 MHz precise and stable over time. Yyou get this generator synchronizing a quartz oven oscillator, which ensures a good short-term stability, with the high precision signal of GPS that maintains the frequency at its nominal value.

           

That's what I did on this system.

 

 

The technique

 

The best-known system to keep an oscillator coupled to another is the PLL (Phase Locked Loop), but in this case to lock 10 MHz to 1 Hz (1pps) of GPS, it would be necessary to make a fairly complex circuit with at least 3 or 4 loops and, above all, you would get a worsening of the phase noise on 10 MHz Oscillator.

There are some projects that use the PLL, but they use a particular GPS that besides the 1pps, delivers even a signal at 10 kHz and with this signal they engage the PLL. These GPS are more expensive, not easy to find and still remains, even if in some cases reduced, the problem of phase noise.

Another system to synchronize two signals is the FLL (frequency-Locked Loop). With a precise frequency counter it is measured periodically the signal to be monitored and, if its frequency differs from the nominal value, the voltage Vc is adjusted to bring the frequency within the limits. The quality of the results obtained with this system depends on the quality of the frequency counter and from short-term stability of the oscillator to control.

In the data sheet of the oscillator, the manufacturer must declare, to the model used in this implementation, a short-term stability < 0.1x10 -10/10 seconds, (in 10 seconds the 10MHz oscillator moves less than 0.0001 Hz) and therefore excellent for an FLL.

Now we need a precise frequency counter, but since accuracy of a frequency counter depends on precision of his time base, (the clock that, at predetermined times, starts and stops counting of impulses that arrive at the entrance of the meter), we can build one using, as the precise time base, the 1pps signal provided by GPS.

If we achieve a counter with time base of 1s (1pps) we will be able to appreciate variations of 1 Hz (we will have 10 MHz ± 1 Hz), but if we use a time base of 20 seconds we will be able to measure 10 MHz ±0.05Hz, which means that we can detect variations of 5 cents of Hz on the 10MHz signal and then vary the value of voltage control (Vc) correcting this minimal error. We will get so, a 10 MHz oscillator with a short-term stability < 0.1x1010/10 seconds and a precision of ±0.05Hz.

The counter

 

To achieve the counter I used a 4013 (dual D Flip Flop) and a 16F684 PIC that has the peculiarity of having a 16 bit Timer/Counter TMR1 with a gate (called T1G) outside accessible. T1G enable counting when is zero and disables it when is one. 16F684 also contains a PWM generator that I used to produce the voltage control (Vc) to control the frequency of the 10 MHz Quartz Oven Oscillators.

 

A D type flip flop, transfers on exit Q, the logically level present on input D (Data) at the instant that a rising edge arrives on input Clk (Clock).

GPS_10MHz_FIG44

 Fig.4

 

 

Referring to the block diagram of Figure 4 we will have:

 

· 1pps signal of GPS arrives on the Clk pin 11 of 4013

· The output Q of 4013 (pin 13) is connected to the PIC T1G (pin 3)

· The entrance Data of 4013 is connected with the exit door A0 of PIC

· The 10 MHz arrives to counter clock-in TMR1 (T1CKI pin 2)

· By PWM output of PIC comes out a series of pulses whose width depends on the error of 10 MHz

 

Let's assume that everything is already working properly and that at instant T0 the situation is the following:

 

· The GPS is halfway between an impulse and another

· The door A0 and the Data of 4013 are to 1

· The output Q of 4013 and T1G of PIC are to 1 and the counter is stopped

· The TMR1 counter is 0 (zero)

 

Ø At T1 the SW put at zero the door A0, therefore also the Data of 4013 goes to zero, but everything else is unchanged because at the Clk of 4013 hasn't arrived the rising edge of GPS.

Ø Instant T 2 comes the rising edge of 1pps and

 

1. The 4013 transfers the data level on exit Q putting it low because the Data is at zero

2. The T1GI becomes zero and the counter starts counting

3. The program keeps low port A0 for about 19.5 seconds and continues the counting

 

Ø At instant T 3 program put again port A0 to 1, but everything remains unchanged because at the Clk of 4013 isn't arrived the rising edge of GPS.

Ø At instant T 4 comes the rising edge of 1pps and

 

vthe 4013 transfers the data level on exit Q which goes high because the Data is 1

vthe T1GI becomes 1 and the counter stopped counting

vthe program keeps height port A0 for about 2 seconds

- Reads counter value TMR1

- Calculates value of 10 MHz

- if different from 10MHz ± 0.05 Hz, adjust the duty cycle of PWM generator square waves, so that, after the low pass filter, we get a direct voltage Vc that corrects appropriately the frequency of the 10 MHz Quartz Oven Oscillators  

- Transfers on display the calculated value of frequency

- Reset TMR1 counter

 

Ø  At instant T5 program calls to zero the door A0 and the data of 4013.

Ø  At instant T6 comes the rising edge of a new 1pps and begins a new cycle

 

You may have noticed that who starts and stops the counter is always the 1pps signal of GPS and this ensures the accuracy of the count.

 

 

Referring to the block diagram of Figure 4 we will have:

 

· 1pps signal of GPS arrives on the Clk pin 11 of 4013

· The output Q of 4013 (pin 13) is connected to the PIC T1G (pin 3)

· The entrance Data of 4013 is connected with the exit door A0 of PIC

· The 10 MHz arrives to counter clock-in TMR1 (T1CKI pin 2)

· By PWM output of PIC comes out a series of pulses whose width depends on the error of 10 MHz

 

Let's assume that everything is already working properly and that at instant T0 the situation is the following:

 

· The GPS is halfway between an impulse and another

· The door A0 and the Data of 4013 are to 1

· The output Q of 4013 and T1G of PIC are to 1 and the counter is stopped

· The TMR1 counter is 0 (zero)

 

Ø At T1 the SW put at zero the door A0, therefore also the Data of 4013 goes to zero, but everything else is unchanged because at the Clk of 4013 hasn't arrived the rising edge of GPS.

Ø Instant T 2 comes the rising edge of 1pps and

 

1. The 4013 transfers the data level on exit Q putting it low because the Data is at zero

2. The T1GI becomes zero and the counter starts counting

3. The program keeps low port A0 for about 19.5 seconds and continues the counting

 

Ø At instant T 3 program put again port A0 to 1, but everything remains unchanged because at the Clk of 4013 isn't arrived the rising edge of GPS.

Ø At instant T 4 comes the rising edge of 1pps and

 

vthe 4013 transfers the data level on exit Q which goes high because the Data is 1

vthe T1GI becomes 1 and the counter stopped counting

vthe program keeps height port A0 for about 2 seconds

- Reads counter value TMR1

- Calculates value of 10 MHz

- if different from 10MHz ± 0.05 Hz, adjust the duty cycle of PWM generator square waves, so that, after the low pass filter, we get a direct voltage Vc that corrects appropriately the frequency of the 10 MHz Quartz Oven Oscillators  

- Transfers on display the calculated value of frequency

- Reset TMR1 counter

 

Ø  At instant T5 program calls to zero the door A0 and the data of 4013.

Ø  At instant T6 comes the rising edge of a new 1pps and begins a new cycle

 

You may have noticed that who starts and stops the counter is always the 1pps signal of GPS and this ensures the accuracy of the count.

 

 

 Circuit diagram

 

 

GPS-10MHz-FIG55

 

Fig.5

 

           Top left we find the GPS board in which, through a 5-pin, microphone type connector, we bring power (+ 5V-GND) and take signals 1pps and port RS232. GPS  requires a voltage of 3 .3V ± 0 .3V, but the pre amplified antenna operates on 5V, then the 5V arrives on leg 1 (antenna power in) and between pin 1 and pin 2 (GPS power in) I have soldered two diodes in series that causing a fall of 0.7 + 0 .7V bring voltage to 3 .4V.

Via port RS232, and a SW downloadable from the website of Trimble, you can read the geographical coordinates and other information provided by GPS, I expected the link but I never tried this functionality.

           

From left are U8 that is the D filp flop 4013, 16F684 PIC (U6), a connector to connect the display, the integrated U3 LMC6482 constituting the low pass filter and low gain amplifier, U4 LTC1485 that is a buffer to not load the output of 10 MHz Quartz Oven Oscillators and finally U1 MC3487 used as a10 MHz signal splitter amplifier, so we can power 4 equipments (50 Ω) at some time.

 

Down, the supply chain that provides + 14 .3V, + 8V and + 5V.

PCBS

 

GPS-10MHz-FIG6

Fig.6

GPS-10MHz--FIG7

Fig.7

In Figure 6 and 7 are single sided PCB trace and components lay-out, everything seen top.

  Get it from download section (.BMP file)

GPS_10MHz_FIG_8A

Fig.8

GPS_10MHz_FIG_9

Fig.9

Figures 8 and 9 are photos of PCBs with soldered components

Note the two pins where to connect power (18Vac).

The board marked APEX is the display, seen from behind, and below, where the three wires are welded, there is the card of the LEDs seen from behind. Figure 8 board is a little different from traces of Figure 6, because I had planned it to have two exit at 10 MHz (a BNC front and a SMA rear) and two commands for a particular application that is not used in this project. To enable the other two exits to 10 MHz, before solder it, I cut pins 13 and 14 of U1 socket (MC3487) and, copper side, I did two jumpers between the pins 2-14 and 3-13.

 

 

The container

 

The container used for this oscillator is a metal box of company HIFI2000, (www.hifi2000.it) model E551815 (18x15x5.5 cm). The bottom of the box is the heat sink for the three voltage regulators that, especially during the first 15-20 minutes, when the thermostat of the oscillator is always connected, warm very much.

The SW

 I wrote SW using the Proton IDE (PIC Basic).

            In the download section you can find the 

  • .bas of the version with negative slope, 
  • .hex of the version with negative slope
  • .hex of the version with positive slope.

The program is fully annotated and it is therefore easy to understand its functioning. It is appropriate to explain just how works Timer1, because it is capable of counting only up to 65,536 impulses, then reset and start over.

To count 200 million impulses (exact 10 MHz per 20 seconds) TMR1 counts up to 65536 for 3051 times, the 3052nd time in its memory (register TMR1H and TMR1L) there will be the value 49664 (65536 * 3051 + 49664 = 200000000). Since our frequency, on regime with oven warm, will vary, at most, only a few Hz, SW simply check the number 49664 and, if different, it vary proportionately the duty cycle of PWM generator whose frequency, rectified, filtered and amplified (Vc), will correct the 10 MHz Quartz Oven Oscillators. Red and green LEDS inform us of the situation:

 

· Green:  counter = 49664 ± 2, corresponding to 10 MHz ± 0,05 Hz

· Red:     counter ≠ 49664 ± 2, corresponding 10 MHz with error greater than   ±0,05Hz

 

Into download section of this site, you will find  .HEX file to program the PIC directly.

Display

 

In the first row, we find the value of the duty cycle that, at regime, it indicates how much the system has corrected the 10MHz Quartz Oven Oscillators for having 10MHz ±0.05Hz. If oven oscillator, without any correction, oscillates exactly at 10 MHz, we would have a duty = 127. Broadly speaking, we can consider that for every correction of 0,05 Hz the duty will vary by 1 unit. Values between 20 and 234 normally be tolerated, but much depends on the type of oven oscillator.

In the second line we find the value of the measured frequency.

 

Switch on

 

After checking that everything is assembled and welded as expected, without installing the IC and without connecting the GPS, connected power supply (18Vac  1A) and check  + 14 .3V (cathode to D7), + 8V (pin 8 of U3) and + 5V ( pin 1 of U6). The Red led will be on

Switch off everything, let discharge capacitors, install the ICs, connect the GPS and again switch on. For a time variable between zero and xx minutes, (satellite acquisition time from the GPS), on the display will appear a row of black rectangles like figure 10 (if you don't see either these adjust the contrast).

GPS_10MHz_FIG_10

 Fig.10

If, after the time limit indicated in the data sheet of the GPS, the screen is again like  Figure 10, check that on resistance R9 (47 kΩ) arrive impulses to 1pps, If not arrive, Check the circuit of your GPS. If they arrive, check the right connectors attach and welds on PCBs.

When at the resistance R9 will arrive one pulse per second (1pps), the display will appear

 

· Duty = 127

· frequency with random values

.

Wait for 15 – 20 minutes (time depends on how much time the oven oscillator needs to arrive in temperature) and the frequency, slowly (display refreshes every 22 seconds), will reach nominal value; When it will be 10 MHz ± 0, 05Hz the Green led light on. It is possible that within few minutes, due to the thermal overshoot of oven, the Red LED might light-on a couple of times, then, if everything is OK, we will have the Green LED always on and display will indicate a frequency that will change between 9 999 999,95 Hz and 10 000 000,05 Hz (most of the time fixed at 10′000′000,00 Hz).

Conclusions

 

The system has proved to be very effective and easy to realize, especially using PCBs. Needs no calibration and allows to have, at a modest cost, a frequency generator of quality and characteristics that, until a few years ago, for amateurs, were unthinkable.

Available, by e-mail ( This e-mail address is being protected from spambots. You need JavaScript enabled to view it ), for questions and clarifications,

 

A friendly 73 IK0OTG Pietro

Last Updated on Monday, 24 October 2011 14:37