inleiding
|
Een veel geziene GPSDO is het bekende aluminium kastje met groene printplaat voorkant. De Chinese experimenteel radio onderzoeker BG7TBL heeft een GPSDO ontwikkeld en geproduceerd in diverse uitvoeringen. Aan de hand van de datum codering op het frontpaneel is uit te vinden welk model het is. De toegepaste OCXO en GPS ontvanger kan per model verschillen.
Op het EEVBlog forum is ook veel informatie te vinden over de BG7TBL GPSDO modellen.
|
|
eigenschappen
|
specificaties
 elektrische voeding module: 11,7...12,9 VDC (12 VDC nominaal) / maximaal 1,5 A / maximaal 15 W
polariteit elektrische voeding: midden pen is + en ring is massa
afmetingen DC ingang: 5,5/2,1 mm stekker (5,5/2,5 mm is ongeschikt door slecht contact!)
GPS antenne voeding: 3,3 VDC (<50 mA) of 5 VDC (@50 mA) (soldeer jumper)
1 pps uitgang: 3,3 Vpp blokgolf
10 MHz uitgang: 1 Vrms (10...15 dBm) sinus (afhankelijk van uitvoering)
10 MHz uitgang: 3,3...4,7 Vpp blokgolf (afhankelijk van uitvoering)
RS232 aansluiting: GPS NMEA data 4800 N,8,1 (pen 2 is TXD, pen 3 is RXD, pen 5 is massa en pen 8 is 1 pps)
afmetingen: 107 mm breed, 55 mm hoog en 122 mm diep
voeding adapter: 110...220 VAC naar 12 VDC
omvang van levering: GPSDO module, actieve (5 VDC) GPS antenne met 5 m kabel (SMA) en netadapter
|
|
toelichting
|
pc verbinding
Door middel van een "rechte" kabel is het mogelijk om met een computer (of soortgelijk apparaat zoals een Arduino/Raspberry Pie) de GPS ontvanger data zichtbaar te maken. De uitgaande data (TXD) van de GPSDO is pin 2, verbind dit signaal met pin 2 (RXD) van de seriële poort van de computer. Verbind pennen 5 (massa) ook met elkaar. Nu is het mogelijk om met 4800 of 9600 baud (N,8,1) de data uit te kunnen lezen. Indien gewenst kan pen 8 van de GPSDO worden gebruikt om een 1 pps signaal te benutten.
opwarmen
De OCXO heeft ongeveer 30 minuten nodig om op te warmen en het 10 MHz signaal te stabiliseren. De stroom zal na opwarmen afnemen. Wanneer de GPS "gelocked" is (led aan) zal het waarschuwings led (WARN) uit gaan. Wanneer de WARN led uit is, is de nauwkeurigheid van het signaal beter dan 0,05 Hz bij 10 MHz. Als de GPS minimaal vijf uur "gelocked" is, zal het signaal nauwkeirger worden dan 0,005 Hz bij 10 MHz.
|
|
structurele afwijking van 2 Hz per 100 GHz
|
John; KE5FX (externe link) heeft ontdekt dat het 10 MHz signaal in de praktijk structureel 2*10^-11 te langzaam is. In de praktijk betekent dit dat dit resulteert in een afwijking van 2 Hz op een 100 GHz signaal. Dus het 10 MHz signaal is in de praktijk 9.999.999,999.8 Hz. Dus elk uur komt 75 ns te kort. John vermoed dat deze afwijking ontstaat door een afronding in een rekenkundige bewerking in de microprocessor. De kwaliteit van een oscillator is uit te drukken in stabiliteit en accuratesse.
stabiliteit
Stabiel wil zeggen dat de frequentie gelijk blijft gedurende de tijd. De variatie in frequentie is één criterium. Het beste is natuurlijk een rechte lijn in een frequentie/tijd grafiek.
accuratesse
Een acuuraat ofwel precies signaal wordt bepaald door hoe dicht het signaal bij het gewenste signaal is, gemeten over een langere tijd.
Voorgaande twee kenmerken kunnen samen vier mogelijkheden vormen, namelijk:
Stabiel en accuraat; Dit is de ideale combinatie. De gewenste frequentie blijft over lange tijd de gewenste frequentie en stabiel.
Stabiel en niet accuraat; Dit is niet de ideale combinatie, maar wel werkbaar. Dit is van toepassing op de BG7TBL GPSDO's. Het signaal is wel op lange termijn stabiel, maar niet de gewenste frequentie. Dit is wel een werkbare situatie en als de afwijking niet te groot is, hoeft het iet bezwaarlijk te zijn.
Niet stabiel en wel accuraat; Dit is een onwenselijke situatie omdat de gegenereerde frequentie op het ene moment hoger en op eh tandere moment lager is. Over een lange periode is dit door gemiddelden niet merkbaar, maar als het signaal niet de gewenste frequentie heeft tijdens een meting, is het referentie signaal niet betrouwbaar.
Niet stabiel en niet accuraat; Een zeer onwenselijke combinatie omdat de signaal kwaliteit te laag is en doordat het "onvoorspelbaar" is, is het ook niet te compenseren.
En OCXO zal door veroudering af gaan wijken van de gewenste frequentie. Door deze aan het GPS referetie signaal te koppelen, wordt de frequenite verlaging continue gecompenseerd door het frequentie verloop te corrigeren. Stabiliteit van de OCXO is dus belangrijk om voor een goede korte termijn stabiliteit te zorgen. Het gebrek aan precisie van een OCXO wordt dus gecompenseerd door het ontwerp van de GPSDO. Als de veroudering (en dus frequentie verloop) langzaam is, is er voldoende tijd om de OCXO te corrigeren om tot een stabiel en precies signaal te komen.
|
There are a couple of revisions of the BG7TBL GPSDO. Likely due to the availability of the used oscillators , GPS receiver models and hardware/software improvements. My GPSDO is version "2016-05-31". Here are some specifications of my device shown:
oscillator
manufacturer: Bliley Technologies Inc.
description: Bliley NVG47A1282
type: OCVCXO
output frequency: 10 MHz
signal type: sinewave
singal level: +6...+8 dBm (+7 dBm typical)
output load: 50 Ohms
harmonics: -30dBc maximum
spurious: -80 dBc maximum
operating temperature: -5...+50 'C
frequency stability
temperature: 0,005 ppm over -5...+50 'C and 0,020 ppm over -30...+70 'C
aging: 1,0 ppb per day and 0,3 ppm per year
supply voltage: 1,0ppb for 1 % change
Frequency accuracy during warmup
1,0 ppm in 2 minutes (related to steady state at 30 seconds)
0,1 ppm in 2,5 minutes (related to steady state at 30 seconds)
0,03 ppm in 5 minutes (related to steady state at 30 seconds)
0,01 ppm in 15 minutes (related to steady state at 30 seconds)
Electrical Frequency Adjustment (EFC)
range: 0,5...1,1 ppm
sensitivity: 2,5...5,5 Hz per Volt
voltage range: 0...+4 VDC
slope: positive (higher EFC voltage > higher frequency)
input impedance: 10.000...100.000 Ohms
startup time: 500 ms
short term stability: 1*10^-10 for 0,1...1 second
Phase Noise
-120dBc/Hz at 10 Hz
-130dBc/Hz at 100 Hz
-140dBc/Hz at 1.000 Hz
-145dBc/Hz at 10.000 Hz
-145dBc/Hz at 100.000 Hz
Connections
pin 1: EFC input
pin 2: referenge voltage output (4,0 VDC +/- 0,2 VDC / 1 mA maximum)
pin 3: power supply (5 VDC +/-5%)
pin 4: 10 MHz RF output
pin 5: common ground
|
10 MHz signal
connector type: female BNC
type: sinewave*
10 MHz output @ 1 K Ohm: 10,2 Vpp / 3,47 Vrms
10 MHz output @ 50 Ohm: 3,15 Vpp / 1,10 Vrms
output impedance: 102 Ohms
There are sinewave and square wave models available. I don't have the suqare wave model, but I made some educated guess based on some reverse engineering how this can be selected. Near the 10 MHz output are some coils and capacitors placed. This seems to be an 7-pole low pass LC filter. It's likeky that the filter shapes the square wave into an sinewave. Next to L9 is a soldering jumper visible. It's likely that the jumper bypasses the low pass filter. Likely the output of the 10 MHz output will be a square wave after making this bypass connection. I validated the theory and it seems to work as expectged, but don't expect a nice clean squarewave. An adittional Schmitt trigger for cleaning up the waveform is advised to acheive a nive clean squarewave.
1 pps output
connector type: female BNC
voltage: 1 Vpp
pulse width: 100 ms (10% duty cycle)
overshoot @ 1 K Ohm: 53,59 %
9-pin sub-d
Pin 1; Not connected. (My modification: link to 12 VDC input for external purposes.)
Pin 2; 9600 baud (NMEA) output data from the GPSDO (TX). (For an external display for example.)
Pin 3; Data input the GPSDO (RX). (Usually not relevant/used.)
Pin 4; Not connected. (My modification: link to alarm led signal for external purposes.)
Pin 5; Common ground.
Pin 6; Not connected. (My modification: link to run led signal for external purposes.)
Pin 7; Unknown purpose. This signal is connected to the RS232 driver chip.
Pin 8; Is the one pulse per second (1 pps) signal output.
Pin 9; Not connected. (My modification: link to GPS lock led signal for external purposes.)
|
I measured the 10 MHz RF output of the GPSDO by my Rigol DSA-815TG and the results are shown below. The signal was attenuated by an external 10 dB attenuator. I forgot to correct the measurment, so the readings are 10 dB higer for real. Therefore the 10 MHz output is approximately 14 dBm (25 mW @ 50 Ohms). There are some harmonics shown, but not too bad an nothing to worry about. If a clean 10 Mhz signal is desired, an additional Low Pass Filter would be a good idea. I guess the signal is clean enough for regular 10 MHz timing input of devices.
|
The 9-pin female sub-d connector has a GPS data output available.
Pin 2 is 9600 baud (NMEA) data from the GPSDO (TX). (For an external display for example.)
Pin 3 is data to the GPSDO (RX). (Usually not relevant/used.)
Pin 5 is common ground.
Pin 8 is the one puls per second (1 pps) signal.
A "straight" cable can be used to connect the GPSDO to a personal computer. Every second, the gps data is transmitted over the serial data connection. The settings are 9600 baud, N,8,1.
Here is one sample of the data set which is transmitted every second. Due to privacy reasons I replaced my home position by question marks...
$GPGGA,160459.00,????.?????,N,?????.?????,E,1,11,0.78,13.6,M,46.0,M,,*6E
$GPGSA,A,3,01,23,03,19,31,06,09,11,14,17,22,,1.55,0.78,1.34*0F
$GPGSV,4,1,13,01,35,143,20,03,75,073,30,06,25,306,22,09,29,209,22*76
$GPGSV,4,2,13,11,13,158,18,12,06,340,,14,14,045,26,17,40,257,23*7E
$GPGSV,4,3,13,19,39,284,28,22,54,086,33,23,58,186,29,25,03,015,*7D
$GPGSV,4,4,13,31,26,059,32*40
$GPGLL,????.?????,N,?????.?????,E,160459.00,A,A*62
$GPRMC,160500.00,A,????.?????,N,?????.?????,E,0.113,,290916,,,A*73
$GPVTG,,T,,M,0.113,N,0.210,K,A*23
Every NMEA sentence contains a lot of information. Every NMEA sentence starts with a typical $GPxxx code. Here is shown which sentence type contains what kind of information. If this information is processed by an Arduino controller for example, it's possible to show the location, time or gps fix information on a connected display module...
GPGGA = fix quality, time, position, number of satellites, altitude,
GPGSA = DOP and active satellites
GPGSV = GPS satellites in view
GPGLL = position, time
GPRMC = time, position, speed (knots), course, variation
GPVTG = speed over ground (knots), speed over ground (km/h)
I made an sample of the 1 pps signal and GPS data burst together. The view below repeats every one second. The 1 pps pulse width is 100 ms. The time synchronisation is done by the rising edge of the pulse. The 9600 baud data burst can be seen on the other chanel of the oscilloscope. The shown data burst contains all the information shown above.
The image below shows the eye pattern of the data signal. The edges are sharp and the signal quality is very good!
The Rigol DS-1054A oscilloscope has some special features like the decoding feature for RS232 signals. I tested the feature and it works great. The $GPGSV characters are shown on the display decoding line.
|
|
buying
|
At september 2016 I ordered my BG7TBL GPSDO from Moncss' store at eBay. Unfortunately the seller turned out te be a scammer... The seller soled a lot via eBay and the reviews were rather good, but apparently something went wrong and a lot of negative comments followed. And my packaged also never arrived... Luckily the money was refunded by Papyal.
I ordered "a second one" which did arrive after a couple of weeks! As shown below, the package was heavily damaged but the contents are luckily undamaged.
|
The PCB is a nice quality PCB. The board is marked with component numbers and the solder quality is very nice.
|
My GPSDO is equipped with a rather well known (professional grade) Ublox GPS receiver model NEO-7M. This is a well known GPS receiver module. There's an SMA connector soldered to the PCB for linking to the antenna connector at the frontpanel.
The antenna power supply voltage is set to 3,3 VDC. By moving the 2 Amps SMD fuse to the 5 V position, the GPS antenne can be fed with 5 VDC instead of 3,3 VDC.
Indication LED D2 is connected to the time pulse output of the GPS receiver for diagnosis.
There are three sodlering jumpers; COM0, COM1 and GPS0. The first are for setting the baudrate of the RXD and TXD wiring. By default the speed is 9600 baud. By soldering COM0 jumper, the speed is changed to 4800 baud. This can be convenient since some devices use 4800 baud by default for NMEA GPS information. The purpose of GPS0 jumper is unknown to me. Even in the hardware integration manual of the receiver, the GPS0 function not described. This jumper should always be "open".
Pin 1 of the receiver is reserved and should be "open". D7 and R1 (top left) are therefore nog populated by components. Since the pin function can change in the future by some software update, the board is prepared for addition of a led for example. Nice thinking ahead!
GPS receiver
descriptiuon: uBlox Neo-7M GPS receiver
power: 2,7...3,6 VDC / 33...111 mW
features: UART, SPI, USB 2 (12 Mbit/s) en DDC (I2C compliant)
sattelite channels: 50
update speed: 5 kHz
cold starttime: 27 seconds
warm start time: 1 second
cold start sensitivity: -147 dBm
warm start sensitivity: -156 dBm
time pulse: 0,25 Hz...1 kHz (probgrammable and defauls set to 1 Hz/ 1pps)
feature: RTC oscillator for fast restart
dimensions: 12,2 × 16,0 × 2,4 mm
|
|
power in and timing outputs
|
Below is a detailed view of the power input and timing signal output. The board is equipped with two BNC connectors, but there's also space for (probably right angle) SMA connectors. It's a nice idea to prepare the board for two possible connector types. Nice.
The board is equipped with diode D1 to protect the circuit against reverse power polarity. The power input is very nicely decoupled by some SMD capacitors. Also placment of inductor L6 is a good idea to filter the DC signal in both directions. EMC fundamentals are clearly integrated in this design!
A note has to be made that the version dat on the board is not the same as the dat eon the frontpanel, but that doesn't bother me. Maybe some component is upgraded for the latest released and some overstock of the former design is used...
|
|
indicatoren
|
ALM (alarm)
Rode led ter indicatie dat het 10 MHz OCVCXO signaal onvoldoende nauwkeurig is. Led brand niet bij normaal bedrijf. Led brand wel gedurende opstarten omdat het systeem nog moet opwarmen/stabiliseren.
GPS LOCK
Groende led ter indicatie dat er voldoende satellieten zijn gevonden met voldoende signaalsterkte voor een betrouwbaar referentie signaal. Bij normaal bedrijf brand deze led continue. Bij het opstarten wordt er naar satellieten gezocht, dus zal het even duren voordat de led permanent groen gaat branden.
RUN
Bij normaal bedrijf knippert de groene led langzaam. De GPSDO is gereed voor gebruik. Wanneer de groene led niet brand, dan is er iets niet in orde aan de GPSDO. Aannemelijk is het dat de centrale microprocessor een storing geeft.
|
My BG7TBL GPSDO is equipped with a Bliley NVG47A1282 10 MHz OCVCXO. The OCVCXO has an operating voltage of 5 VDC and an Electronic Frequency Control (EFC) range of 0...4 VDC. Since the operating voltage is 5 VDC, the 12 VDC input voltage is reduced by an LM7805 voltage regulator. Since the oscillator draws 0,6 Amps ind cold situation and the power loss in the regulator is (12 - 5 =) 7 VDC, the loss (and therefore heat generation) is 4,2 Watts. During normal operating temperatures of the oven, the loss is still (7 VDC * 0,2 A = ) 1,4 Watts. In noticed that the device is getting quite hot and not only by the oven. Since there are also 12 VDC operated OVCXO's, using this the voltage regulator isn't loaded unneccessary.
I've had a Trimble 34310-T OCVCXO "lying around" which would be a nice replacement I guess. This oscillator uses 12 VDC instead of 5 VDC eleminating the strain of the voltage regulator. I've read that the Trimble 34310-T OCVCXO seems to have a double oven and that the )schort term) stability is a factor 10 better. The board is prepared for more OCVCXO models and in this case also for the 34310-T. This oscillator looks similar to the Morion MV89A which is used in earlier BG7TBL GPSDO's.
voltage selection
"Reverse engineering" the board, I found that there's a 0 Ohms jumper placed on position R5. This links the 5 VDC regulated voltage to the power input circuit of the oscuillator(s). By removing the jumper from position R5 and placing this to position R4, the 12 VDC input voltage is coupled to the oscillator(s). Therefore I moved the jumper from position R5 to position R4. The ocillator power circuit on the board is now connected to the (unstabilised) 12 VDC power input as shown below.
oscillator replacment
The next step is to replace the oscillator. This is a rather straigtforward job. After desoldering the smaller Bliley oscillator I soldered the Trible 34310-T in position. I kept approximately 1 mm clearance between the board an the metal housing preventing possible short circuits om the board. The operation is shown below.
the result
The result is quite nice I guess. I'll place some insulation material around the oscillator in the future to reduce heat loss.
EFC range
Before replacing I checked the EFC voltage ranges. Since the're rather similar, I thougt this wouldn't be a problem. And it turned out to be fine indeed. After finalising "the operation" I found an interesting video of Damon Stewart the same evening scrolling around the internet. He pointed out some very interesting information about the EFC range which can be set. It turned out that the EFC rance from the processor is 0...3,3 VDC. The operational amplifier is amplifying this signal 1,5 times by default to create a 0...4,95 VDC voltage range. There are two jumper positions "X2" and "X3" at the bottom of the board as shown below. By placing the jumpers, the amplification factor can be set to two or three times instead of 1,5. This results in a voltage range of 0...6,6 VDC for X2 and 0...9,9 VDC for X3. The Trimble 34310-T OCVCXO has a voltage range of 0...6 VDC and the EFC voltage for 10 MHz is approximately 3,2 VDC. At a cold start the EFC voltage peaks at 4,5 Volts. A voltage range of 0...6,6 VDC would be logical in theory, but the smaller the range, the higher the resolition as Damon also mentioned. Since the EFC voltage probably never reaches 4,95 VDC, I kept the factor to 1,5 times. If the oscillator reaches 4,95 VDC EFC, the range can always be changed afterwards.
Mission completed succesfully. ;-)
|
|
EFC voltage logging experiment
|
I wondered what the EFC behavour is for this Trimble 34310-T OCVCXO. If the EFC is out of the limits, the GPSDO would not lock and the 10 MHz signal wouldn't be stable. Therefore I made some measurments. I connected a voltage logger to the EFC voltage of the oscillator and I connected (for the second measurement) also one wire of the logger to the ALM led. The test setup is shown below.
The results of two measurments is shown below. During both cases the GPS locked almost instantly.
The blue line represents the first measurment. The EFC voltage starts around 3,3 VDC and increments each 1 second in small steps. Near 4,5 VDC the voltage decreases in 1 second steps. After some undershoot, the EFC voltage stabilises around four minutes after startup. During startup de oscillator crystal oven is cold and the EFC tries to compensate by "speeding up the pace" by increasing the EFC voltage. It's likely that the oven temperature is near normal operating temperature after 1:45 minutes. Since the crystal is warming up and changes frequency, the EFC voltage is lowered to compensate the temperatur change. Four minutes after startup the system seems to stabilise. The measurment is stopped at nine minutes after initial startup. After six minutes the EFC voltage didn't change anymore, so the results are not shown in the graph.
After dismanteling the test setup I thought it would be more logic to also record the ALM led state and the warm start behaviour. So I rebuild the test setup with logging of the ALM led state. The red line shows the results of the "warm start". The "off time" between measurments isn't recorded, but approximately a couple of minutes maximum. The red line shows that the EFC voltage doen't rise as much as during a cold start. It's logical that's because the oven is still rather warm and the oscillator crystal is still near the desired temperature. Again after four minutes the EFC voltage is rather stable. The red ALM led (indicating that the system is locked an stable) is switched off by the processor at five and a half minutes. Altough not recorded, based on the EFC voltages during a cold start a lock after five and a half minutes is quite logical. Altough the EFC range of the used Trimble 34310-T OCVCXO is 0...6 VDC, the "normal" 0...4,95 VDC range is good enough. This concludes this EFC logging experiment. ;-)
|
|
