Posts by PA3FYM

    SV1BDS ... Good question.


    From my quick & dirty analysis it doesn't make a difference as the whole thing is edge triggered.


    I have to dig deep into the ATmega328p datasheet to find out if the 'delay' between the 20% vs 50% duty cycle generates an intrinsic 'phase offset' (which will be compensated anyway).


    Code
    1. _____|----|_____ (50% duty cycle)
    2. ________|-|_____ (20% duty cycle)
    3. . .
    4. . .
    5. |dT| Question: delta T matters ?

    Adding some words of my own to Mike words. First of all I appreciate A75GR 's efforts to simulate the feed. However, we do not live in a simulated world. This implies compromises. Like Mike said, during Christmas last year we had to take some decisions concerning 'what is good enough'? The result is a feed which performs well enough, even better than (most criticists) expected (!)


    One simulates 3 dB AR, somebody else simulates 0.1 dB AR ...


    Concerning the double dip, this was simulated and also measured in practice.

    The trick is to get Z=1 (50 + j0) in the centre of the 'balloon' and that took some (trail and error) efforts to find the right feedpoint (several attempts/prototypes have been constructed).


    Not preaching to my own parish, the result is an ultra simple solution to dual feed a single dish with relative good efficiency, easy to construct with simple metal working tools.


    Two sheets of brass, a piece of copper pipe (with some nylon) compared to a solid piece of aluminum with four ports, (phase) couplers etc ...


    So, I also challenge people to come with a better solution given the simplicity and ease of the design: two plates, a piece of copper pipe (and maybe some nylon (or other dielectric)).

    A75GR Yep, the properties of Nylon are known to me and the lens is not an 'integral' part of the publication, but merely, like Mike stated, an alternative when people don't have 'Rocket LNB lenses' and want to illuminate a dish with f/D > 0.3 - 0.4 like standard BC-dishes with an f/D = 0.6.


    I tried to find out (hitherto with no success . . ) from which material these Rocket LNB lenses are made. It must be very cheap material because the overall prices of these LNB's are very low.


    Ad c. Yep, that complicates the simulation because you don't know which part of the parabola is considered. I did my experiments with a Triax offset dish (f/D = 0.6) and found max SNR with lenses generating some under illumination (G = >11 dBi). The lens proposed in the article -with the knowledge at that time- was modeled for exact 90° -10 dB opening angle (for a f/D = 0.6 dish). In practice it doesn't perform that bad, but with todays knowledge there are better alternatives.


    Considering pros and cons, of course feel free to mention them, but also accept that people react ; -)

    @G0MRF Depends what reasonable is ; -) You did not specify the R-divider so your PFD is 10 MHz , which forces the whole thing into FRAC-N mode.

    First decide if you want FRAC or INTEGER mode and then derive the PFD from it.

    As a rule of thumb, for lowest phase noise your PFD has to be as high as possible

    using INTEGER mode. But these chips have much sophisticated options to improve P/N-performance in FRAC-N mode.


    I use 500 kHz PFD for 1967.5 MHz , so my R-divider with 10 MHz reference is 20 and I am a fan of the 4/5 prescaler mode (iirc it produced a better signal on the analyser).

    Output here is always 5 dBm (although I never reached 5 dBm with my cheap Chinese boards)


    Depending on what ADF435x board you have, there are a few routes to Rome and you've to experiment with it (in conjunction with the loop filter) and look on the spectrum analyser for the best spectral impact. E.g. you can experiment with the charge pump current etc.

    A75GR Rasto, thanks for the review.

    Like Mike said, Willi HB9PZK also had a positive second opinion.


    Concerning the 'cons', it all matters how you look at it.


    a. Lens: at the time of writing the article (I believe early February) 'amateur lens science' was not as developed as nowadays ; -) I now use another design which performs as good as my 'Rocket LNB' lens. Note: RL (S11) seems not the most important thing. See picture below, picture taken bij PE1CKK.



    b. UV resistance: ?? (why didn't you took 'colour' or 'shininess' as performance parameter? ; -)


    c. Phase point: in practice it's very easy to find the phase point (focal point) with (offset) dishes. Just tweak on engineering beacons, like millions of satellite viewers do. As a result the S-band part is also in focus (is 4x less critical).


    I am curious to watch the 2nd episode of the opera : -)

    OK, strange, I have good results with these converters. Measured the output voltage 'unloaded' first (check!) , connected my amplifier (check!), it worked (and still works!) flawlessly and could discard my seperate 28V power supply.

    GI8RQI I know somebody using that clock multiplier but did not get satisfactory results concerning phase noise & jitter (have to ask him precisely what 10 MHz ref he used).


    The only thing which works fine* with me is the setup as published in my first posting ever at this forum: an ADF4001 locked in conjunction with a low noise 25 (or 27) MHz VCXO GPS locked.


    *fine = low phase noise, stability, precision AND accuracy, see picture below :


    PA3FYM I searched for PL904 and I see that no one selling them. For Lars gpsdo there is no need to use a NEO7 I think a chiper NEO6 is suitable as it produces 1 pps. Also I will try to lock a 25 MHz VC-TCXO using one more /5 divider on the HC390 to provide the 5MHz and 1MHz required. As you know the software is any reason that it will not work?

    Of course any (proper) 1 PPS GPS-rx can be used. There is always jitter on the 1 PPS (due to propagation/atmospheric issues etc) but that is 'smoothened' inside the GPS-rx. Also in the Lars software the PPS jitter is handled.


    Concerning your 25 MHz VC-TCXO, the software does not know you are applying another 5 MHz source (i.e. 25 / 5) but I think you need some fiddling concerning the time constants and 'gain' for the software PID-loop. I reckon your 25 MHz VC-TCXO has a higher 'k' (i.e. Δf/ΔV) than 'the average' OCXO. But these parameters can be entered in the software. Bear in mind the software (read: the filtering) was designed for a reference source with intrinsic short term stability (like an OCXO) but I am very curious how it'll perform with your 25 MHz setup.


    Important tip: solder the RC-network (3k9 / 1 nF) as close as possible at the analog pin. I soldered the 1 nF directly on the Nano board

    SV1BDS I was not satisfied with the results in my contraption (i.e. Philips 10 MHz TCXO with LS86 phase detector @1 MHz using a Ublox producing 1 MHz and cleaned with a dedicated jitterblocker).


    On receive jitter was still present and a LOT of phase noise. Most signals sounded like there was aurora ; -)


    Weighing the ingredients time, homebrewing, components, result and satisfaction I went for the Lars GPSDO.


    Less time, less components, cheap, good result and satisfied : -)

    The level (formally) needed is around the forward voltage (for Si diodes 0.6 - 0.7 V, so around 5 mW).


    Cf the datasheet the 4351 may deliver +5 dBm (approx 3 mW) but I never realized that with these cheap Chinese boards. However, with my signal generator 3 mW is enough to let (partially) saturate the diode, and hence, produce harmonics.


    With +7 dBm the 11th harmonic of 953.6 (10489.600 MHz) is utterly strong, reducing the 953,6 MHz level also reduces the amplitude of the 11th harmonic. However, there will be a moment that the signal is 'suddenly' gone, because the P/N transition in the diode will not conduct anymore.


    It's worth a try anyway : -)

    Is this the K3TZ patch? If yes, it looks perfect to me. Tweak a little with the patch corners to have the center (i.e. the 'up bump' between the two 'dips') on 2400 MHz

    I am a member of that forum and posted some in this specific thread. I first want to fiddle with it some more before I release the code. The Lars GPSDO is now a dedicated QO-100 reference and I moved my other reference back into the shack to feed the measuring equipment : -)

    - ADC7 is now sampling port

    - speeded up ISR (int serv routine)

    - ADC clk_div = 16 (instead of default 128)

    - speeded up PWM DAC code

    - changed PWM mode to 'phase correct' PWM and F_PWM = ca. 31 kHz (= F_CPU / 510)

    - changed EEPROM write and read routines to Little Endian

    - (tried to) added TC converging scheme

    - experiment with EMA and DEMA (double exponential moving average)

    - keyboard input is read once


    What my GPSDO does now: It starts up with a low time constant (TC = 16) and when a lock is achieved the TC increases periodically and converges to the (from what I understood from the whole EEVblog-thread) optimal TC for the configuration.


    The TC itself is also part of a seperate PLL in a sense it calculates the DEMA of the 'ns values' (most left column). The DEMA has to be within certain limits. When the DEMA falls 'out of lock' the TC is decreased.