|
Homebrew Equipment A place to show, design and discuss the weird and wonderful electronic creations from the hands of individual members. |
|
Thread Tools |
11th Feb 2021, 10:45 am | #1 |
Nonode
Join Date: Aug 2010
Location: Resolfen, Wales; and Bristol, England
Posts: 2,588
|
Comparing two similar frequencies?
I am working on building a GPSDO at the moment which uses a signal from a GPS receiver, compares it with a signal from an internal OCXO and uses the result to tune the OCXO. The two signals are 10kHz and applied to the inputs of an XOR gate, the output of which should provide a varying DC level depending on the phase match of the two 10kHz signals.
I have two issues right now. One is that I want to change the frequency of the compared signals to 1Hz, since it has been suggested that signal from the GPS receiver is more stable than the 10kHz, the other is that I have a confidence issue with the DC output, in that I'm not sure that it accurately follows the input phase shift. Has anyone any comments or ideas?
__________________
Richard Index: recursive loop: see recursive loop |
11th Feb 2021, 12:02 pm | #2 |
Dekatron
Join Date: Aug 2003
Location: near Reading (and sometimes Torquay)
Posts: 3,086
|
Re: Comparing two similar frequencies?
This is not so simple I think, because the GPS outputs are not instantaneously accurate, only on average. If you have a 10kHz signal then that is already a tracked oscillator output to give the GPS timing accuracy. So it could perhaps have a lot of phase jitter as it tracks the GPS.
So I would expect that you have to do this in a more long-range way, so not quite like a typical PLL. |
11th Feb 2021, 12:40 pm | #3 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,798
|
Re: Comparing two similar frequencies?
As you divide the frequencies down, the sensitivity of the phase conversion process worsens.
Keep the frequencies as high as you can at the point of comparison. Have a look on the surplus market for an HP vector voltmeter. This does good accurate phase measurements from 1MHz to 1GHz. It's what we used for comparing our Rubidium and Caesium standards against off-air receivers and GPS. If you can't get the frequencies to be the same from both sources, use a lissajous pattern on a scope. chinagraph pencil on the screen so you can look for phase shift. XOR gate detectors have accuracy issues due to the variation in supply voltage and temperature affecting the amplitude of the pulses being averaged. Another method is to clock two DDS, one from each reference to create two notionally identical frequencies and to steer the tuning of one DDS to keep the outputs in phase. The tuning history of the variable DDS can be integrated to give a measure of the difference. David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
11th Feb 2021, 3:08 pm | #4 |
Guest
Posts: n/a
|
Re: Comparing two similar frequencies?
As David said, keep the frequency as high as possible. E.g. at 10kHz the "drift" frequency is 1000th of a Hz for 1 in 10^8 precision, about once in 20 minutes, that is useable. 1Hz, a cycle in about six months. Extra noughts soon add up (logarithmically), you only need another 15 to make it the age of the universe.
All sums done is head, errors my have crept in! |
11th Feb 2021, 10:36 pm | #5 |
Heptode
Join Date: Mar 2009
Location: Portsmouth, Hampshire, UK.
Posts: 674
|
Re: Comparing two similar frequencies?
It might be worth looking at JRMiller's design. There have been some measurements published of his later design.
I wanted a GPSDO for occasional use that would settle to around 1 in 10^8 in a couple of minutes from cold rather the usual higher accuracy in a much longer time. My OCXO was not that good (1 in 10^7) so I needed a fairly fast loop. I ended up with a similar design using ATtiny85 to set-up the GPS module (NEO-7M) and then perform the divider function. Output from GPS module is 0 - 3.3v so I used a 4046 with ‘Signal in’ AC coupled to allow 5v supply to be used. A quick calculation of PLL time constant (T): ‘Gain’ of (exOR) phase comparator (Kp) = 10V/Cycle (5v in 180') For 10MHz OCXO used ‘Gain’ (Ko) = 25Hz/V For a 10kHz comparison frequency, divider ratio required (N) = 10k/10m = 1000 This gives T = N/(Ko*Kp) = 1000/(25*10) = 4 sec. Dropping to a 1Hz comparison frequency would increase T to 40,000 seconds which is a bit outside the realm of analogue PLLs. Jim |
12th Feb 2021, 5:07 pm | #6 | |
Nonode
Join Date: Aug 2010
Location: Resolfen, Wales; and Bristol, England
Posts: 2,588
|
Re: Comparing two similar frequencies?
Quote:
The plan is to keep this GPSDO powered up permanently, the location above my workbench has a good, unimpeded view of the sky and I expect to have a stable 10MHz reference. I've decided to take an additional 1MHz output through a buffer to lock my Marconi 2955 radio test set.
__________________
Richard Index: recursive loop: see recursive loop |
|