TCXO Performance issues

  • It appears that Connor-Winfield have been experiencing some production issues with their D75 series TCXOs. A couple of years ago I used a 27MHz unit to replace the crystal in an Octagon OTLSO and the performance when compared to the standard crystal was transformed. Earlier this year I bought a batch of the same TCXOs from Digikey. On this occasion I was disappointed to find that several of the units had very poor short term stability.

    The problem appeared to be a control loop issue where, at near constant temperature, the output would ‘hunt,’ moving up and down over a period of a few minutes. Technically, the TCXOs were still well within their 500ppb tolerance and so returning them as ‘out of specification’ to Digikey was questionable.

    Fortunately, the Conner-Winfield TCXOs are a premium product and I contacted technical support at their HQ. After sending a lengthy description, a few graphs and returning a few unusable TCXOs they performed a failure analysis, acknowledged the fault and a couple of weeks later replaced the entire batch with some straight off the production line. – date code 2619


    As a test, you can assess your TCXO in about 10 minutes by looking at the CW beacon on QO-100.

    If you use an SDR on 739MHz, set up the system and let it all ‘warm up’. This is important as there is a TCXO in the LNB but another in the SDR which will also contribute to drift.


    1) Set the demodulator window to about 2kHz as a visual reference and the bandwidth across the screen to about 10 – 20kHz.

    2) Slow the waterfall down so it takes about 10 minutes to fill the available space. Add time markers to the plot if you can.


    Below I have shown 2 plots for comparison. The first is the plot from a faulty TCXO which over 10 minutes moves around its ideal frequency apparently unhappy to stabilise. The second plot shows the performance of a normal TCXO. In both cases a small amount of this drift will be due to the SDR’s oscillator at 739, but the majority of the drift will be from the 9750MHz VCO in the LNB.


    Thanks


    David G0MRF



  • I designed a board with one of these TCXOs on it and the stability was awful, I gave up on the idea and abandoned publishing the design .The TCXO kept hunting and was useless for SSB. This was last year.


    How can I tell if this is the same issue?


    Mike

  • Hello Heiner / Mike.

    The date code and markings on my batch from Digikey in January this year are:


    N71 RK5H The production lettering says 1713 So probably week 17 of 2013 (old stock?)

    My replacements are marked 2619, so it does look like MMYY in the date code.


    I am also aware that Michael DG0OPK has found similar problems with some marked
    N71 RK6L and 1713. This looks like a similar production lot.

    G0MJW I think if you have constant temperature and you can see 'hunting' then that is a product with a problem.

    To be fair, the first plot was one of the worst I had, however, even a 'good one' was showing 300 - 400Hz for no apparent reason.

    The batch marked 2619 however are an order of magnitude better and do not have any form of hunting that is visible in the bandwidths I've used. - Only long term drift with temp can be seen.....and that is better than the spec.


    David

  • G0MJW Interesting......... I can't tell if that is 4 weeks after my duff batch...or 4 years. Time to look at the data sheet. Do I see that as 25MHz rather than 27MHz ?
    That suggests that it's not a crystal issue, but something amiss with the ADC or the feedback loop used for the compensation. Amazing that this looks to be a parameter that is not tested. The comment from CW was


    "we were quite surprised. We are still investigating the cause at the crystal level,"


    Thanks


    David

  • Hi David, it will be 4 years, date codes are normally week number, & year so 1717 is week 17 2017, 2618 week 26 2018, 1713 week 17 2013. Didn't realise they have been around that long but looking at the data sheet on Digikey's site they have.

  • @PA3YFM A good question Remco. I am 95% sure it is a traditional crystal design rather than an NCO. (it even has SMD capacitors around the edge of the board) I was pointed towards it after trying the Fox series which have unusable digital steps as they compensate. A friend at High Sierra Microwave did the research and spoke to CW. He was looking for an analogue TCXO which had
    low phase noise for use in a 5.8GHz system. I wondered if there had been a change in technology between my original purchase in 2017 and the batch this year....but as the replacements work perfectly, that does not seem to be the case.

    Unfortunately, it is likely that while the current stock at the factory is OK, the distibutors may all have faulty oscillators.

  • I just finished to modify my second VENTON LNB for portable use with a D75F this morning and just gave it a test run on the 40cm portable dish....its a 25MHz type that I just bought a few weeks ago from a DL ham on eBay.


    I was very astonished to find its also drifting very slowly up and down within a 1-2 KHz range over a period of lets say 1-2 minutes....its just drifting in this narrow range, so obviously some kind of "regulation" is still happening.


    But I have another perhaps related problem:


    I generate the 3.3V for the TCXO from the LNB PLL chip voltage via a series resistor and a 3.3V zener diode. I used a wrong resistor value, because initially I thought the internal LNB supply voltage to be 6V, but only realised later that it actually has 5V.


    So as a matter of fact now my input voltage at the D75F is below spec, around 2.9V....


    I need to find a suitable SMD resistor the next days to modify again and obtain a supply of around 3.3V to remain within the specified limits.


    So now after reading this brandnew thread here I am totally uncertain if my issue is a faulty D75F or if it is just related to my too low voltage supply - any opinion on this anyone?


    Unfortunately, I cant read the markings anymore, since I mounted it upside down, stuck to the PCB. (would require total dismounting)


    73 de Oscar DJ0MY

  • After making bad experiences with the frequency stability of the much-praised Goobay LNB, I got a Golden Media LNB Type 201+ (about $ 10 on ebay) and replaced the 25MHz crystal with a D75F-TCXO. As you can see in the picture, I connected the TCXO directly to the 6V of the 78L06 (which he tolerates according to the data sheet, hi). I connected the RF output via a 1nF-c to the lower pad of the quartz. Not only has the new LNB proven to be very frequency stable after this mod, but it also provides 5-8dB stronger signals at the PlutoSDR RX input.

  • DJ7TH Glad to know your D75 is working fine. After this experience I think if I needed a 25MHz TCXO it would be worth looking at the TXC 7N series. Although I haven't tried them, there are some notes on the forum suggesting they work very well (280ppb). They are available from Mouser, who have a distribution centre in Europe.

    DJ0MY The D75 data sheet recommends 3.3V. I suspect any variation from this is likely to give a small offset in frequency rather than random drift.

    73 David G0MRF

  • It would be great to find an official way on how we could have faulty D75F's replaced

    (I guess Connor Winfield will not talk to non-professional users?)


    And/or to find and list here suitable / working TCXO alternatives for 25MHz that could be used instead.


    Those TXC 7N look interesting. Were they already confirmed to give smooth analog compensation instead of jumpy digital compensation? (but suggestions for any other working types would be highly appreciated by many hams I guess)


    73 Oscar DJ0MY

  • Hi,


    I now fixed my supply issue and restored the 3.3V and also added a missing 10nF at the TCXO 3.3V pin to GND. But unfortunately still moving around with similar symptoms.


    I just removed it from the LNB and will try one of those TXC types.

    In the attachement you see a picture of my faulty Connor Winfield TCXO.

    Maybe that helps to narrow down the faulty batches.


    Seems that coding is more likely to be Year / Week.

    (therefore opposit as to the proposed above)


    73 de Oscar DJ0MY

  • Now D75F was replaced by an TXC - 7P-25.000MBP-T

    (7N types were not on stock at Digikey)


    This is the CMOS version. Just added a 10nF at the supply voltage pin against GND and a 1nF cap for coupling into one of the the PLL chip Xtal traces.


    Works like a charm and is very stable. Just had it running for the last 30 minutes on the CW beacon frequency with my Airspy SDR and had hardly around 70Hz very slow drift during that time.


    I am very satisfied now with the alternative TCXO.


    Thanks to David G0MRF for hinting us into this direction.


    73 de Oscar