[time-nuts] 10 MHz OCXO recommendations
>From the current crop of ebay OCXO's available, what have you actually used that you would recommend for the following? Sine output, either 5V or 12V, better than 0.2ppm stability after warmup. I plan on putting things in a sealed box, so there shouldn't be too wide a temperature fluctuation. Everything will be located indoors. I want to use it as a standalone reference to PLL a 14.4 MHz VCXO. With the right divisors, I can get both the 10 MHz (/10/10) and the 14.4 MHz (/16/9) down to 100 KHz. I could use a GPSDO, but that means needing a GPS antenna, plus costing a lot more. Joe Gray W5JG ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
jg...@zianet.com said: > I want to use it as a standalone reference to PLL a 14.4 MHz VCXO. With the > right divisors, I can get both the 10 MHz (/10/10) and the 14.4 MHz (/16/9) > down to 100 KHz. You can run at 400 KHz by dividing by 5/5 and 4/9. > I could use a GPSDO, but that means needing a GPS antenna, plus costing a > lot more. You said "0.2ppm stability". What do you need for accuracy and/or how are you going to calibrate your setup? -- These are my opinions. I hate spam. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
> You can run at 400 KHz by dividing by 5/5 and 4/9. True. Either will work. > You said "0.2ppm stability". What do you need for accuracy and/or how are you going to calibrate your setup? I will initially calibrate the OXCO against my HP GPSDO. Idealy, it will only need checking against the GPSDO once a year. Is that asking too much for it to be stable over that length of time? I could simply look up the specs for the various models that I find on ebay. However, I know that a lot of the stuff that is sold has been through the mill and may not meet spec any more. Thus my request for recommendations for models that someone here has actually purchased and tested. Joe Gray W5JG ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
Joe It is not clear to me what your goal is. Simplicity, cost or performance. A PLL based on 1 pps is not optimum. I call this approach GPSPLL versus GPSDO. Thanks to work done by Karen in Moscow I found out that you can program a $14 ublox M7 to frequencies above 1 KHz. If you use 200 KHz for 5 MHz XO and 400 KHz for 10 MHz you can make a GPS PLL with one 390, one 86 and an op amp. If the XO output is sine wave you have to add a Schmitt Trigger. I recommend Wenzel, AC14 or SN75176. I prefer 75176 it is 8 pin DIL. I do have a board layout if interested. Our Australian team members will introduce it to the HAM community. Not exactly a time nut product. We have done a GPSPLL using a M7 at 1 KHz and tests show better than 1 E-10 at 1 second after 10 seconds GPS takes over it gets better following GPS performance. I have a few boards and op amps left over if interested contact me off list. Bert Kehren In a message dated 11/11/2014 1:05:49 A.M. Eastern Standard Time, jg...@zianet.com writes: > You can run at 400 KHz by dividing by 5/5 and 4/9. True. Either will work. > You said "0.2ppm stability". What do you need for accuracy and/or how are you going to calibrate your setup? I will initially calibrate the OXCO against my HP GPSDO. Idealy, it will only need checking against the GPSDO once a year. Is that asking too much for it to be stable over that length of time? I could simply look up the specs for the various models that I find on ebay. However, I know that a lot of the stuff that is sold has been through the mill and may not meet spec any more. Thus my request for recommendations for models that someone here has actually purchased and tested. Joe Gray W5JG ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
Hi Ok, let’s back off a bit. If 0.2 ppm over a narrow range is the objective, that sounds like a TCXO application. If you really are only after the 14.4 MHz, that’s the quick / easy way to go. Everything running on 3.3V, total current may be below 10 ma. Far fewer parts involved. A lot depends on just how that spec gets interpreted. Surplus OCXO’s are well … surplus. A significant percentage of the OCXO’s I’ve seen on the surplus marked are pulls off boards and are busted. I’ve bought quite a few from “tested / guaranteed / 100% good / perfect buy” sellers and found issues with then when you carefully test them. My conclusion is that in most cases, tested = has some sort of output. Tested may not include “does not catch fire if left on for an hour”. I’ve bought them from the same people everybody else has. There are a couple ways around this. The first is to buy parts that have never been on boards. You can be pretty sure that a part with long leads has not been on a board. Finding OCXO’s at 10 MHz that way is not very easy. The next approach is to buy a working device (like a GPSDO or test gear) and pull it’s OCXO. This works much better if you bought the thing 15 years ago and its not been of any use. You could buy parts brand new, you then run into the fact (been true forever) that surplus is pennies on the dollar. The final solution is the most common - buy lots of OCXO’s and sort them out / repair them yourself. If you are going the sort / repair route, the Morion MV-89 parts are out there in bulk and are pretty cheap. There are no published schematics, so it’s a trace it out and debug thing to fix one. You also need to be a bit handy with something like a torch to get them open. The trick is to do it in a way you can re-seal the parts. The “Trimble” OCXO’s are not quite as common. Welded package parts are going to be a sort only thing, Popping them open and then re-welding them - not so much. That Jackson Labs GPSDO on eBay is sounding better and better isn’t it? Bob > On Nov 11, 2014, at 1:05 AM, Joseph Gray wrote: > >> You can run at 400 KHz by dividing by 5/5 and 4/9. > > True. Either will work. > >> You said "0.2ppm stability". What do you need for accuracy and/or how > are you going to calibrate your setup? > > I will initially calibrate the OXCO against my HP GPSDO. Idealy, it will > only need checking against the GPSDO once a year. Is that asking too much > for it to be stable over that length of time? > > I could simply look up the specs for the various models that I find on > ebay. However, I know that a lot of the stuff that is sold has been through > the mill and may not meet spec any more. Thus my request for > recommendations for models that someone here has actually purchased and > tested. > > Joe Gray > W5JG > ___ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
> If 0.2 ppm over a narrow range is the objective, that sounds like a TCXO application. Please point me to such a TCXO at 14.4 MHz that is affordable. Joe Gray W5JG ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
Hi I have not had much luck with ocxo's sourced from the usual auction site. By far and away the best value for money for me has been found in some packaged frequency standards containing OCXO's and a non auction site purchase of a "BVA" OCXO. I find the concept of occasionally adjusting a good OCXO which in turn is used as a reference works well for me.I have some that haven't needed adjustment for over 2 years (they are still well within one part per billion of being on frequency.) That being said the MV89's were at least fairly in expensive. Mark Spencer On 2014-11-10, at 10:05 PM, Joseph Gray wrote: >> You can run at 400 KHz by dividing by 5/5 and 4/9. > > True. Either will work. > >> You said "0.2ppm stability". What do you need for accuracy and/or how > are you going to calibrate your setup? > > I will initially calibrate the OXCO against my HP GPSDO. Idealy, it will > only need checking against the GPSDO once a year. Is that asking too much > for it to be stable over that length of time? > > I could simply look up the specs for the various models that I find on > ebay. However, I know that a lot of the stuff that is sold has been through > the mill and may not meet spec any more. Thus my request for > recommendations for models that someone here has actually purchased and > tested. > > Joe Gray > W5JG > ___ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. > ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
Connor-Winfield has a line of TCXO's with 0.1 ppm stability that can be bought from Digi-Key in single unit quantities for about $25. Search for M100F. They don't have a 14.4 Mhz, but they do have 10.0 Mhz, which you indicated could also work for you. Regards, John AC0ZG On 11/11/2014 11:20 AM, Joseph Gray wrote: If 0.2 ppm over a narrow range is the objective, that sounds like a TCXO application. Please point me to such a TCXO at 14.4 MHz that is affordable. Joe Gray W5JG ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
Hi This is one of several that come up when I do a search at Mouser: http://www.mouser.com/ProductDetail/ABRACON/ASVTX11-14400MHZ-T/?qs=sGAEpiMZZMsBj6bBr9Q9aTQ72VsxIq1y8x3GpFIJ334%3d That’s the link, to the full reel. At one piece they are $5.89 each. Temperature spec is +/- 2.5 ppm over -30 to +75C. Runs off of 3.3V at low current. Many of these parts are compensated to be pretty flat past room temperature. If it was a straight line (which it most certainly is not) you might get 0.25 ppm over 10C. If you are after a normal room environment, it’ll swing 2 to 4C. Bob > On Nov 11, 2014, at 1:20 PM, Joseph Gray wrote: > >> If 0.2 ppm over a narrow range is the objective, that sounds like a TCXO > application. > > Please point me to such a TCXO at 14.4 MHz that is affordable. > > Joe Gray > W5JG > ___ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
Mark wrote: I find the concept of occasionally adjusting a good OCXO which in turn is used as a reference works well for me.I have some that haven't needed adjustment for over 2 years (they are still well within one part per billion of being on frequency.) A few of us have advocated this approach on the list, and there is good reason for it. A GPSDO offers two advantages: (1) it is self-adjusting, therefore easy to own and use; and (2) it has better stability at long tau than the OCXO alone. The price you pay for those advantages is poorer stability at low tau than the OCXO alone, which can be anywhere from slight with a good design (e.g., Thunderbolt, Z3801) to shockingly bad with a bad design (including many DIY attempts). If one does not need the very best performance at long tau -- and most time-nuts do not -- a free-running OCXO that you adjust manually every now and then can be the best reference available to the average time nut. ("Long tau" can be anywhere from 100 seconds to several thousand seconds, depending on the particular OCXO.) Plus, not spending money on GPS discipline allows you to spend more on the OCXO to get better stability at low tau, and a more extended upper limit on "low" tau (say, better than GPS all the way to 2000 seconds instead of 200 seconds). Personally, I do use GPS discipline to keep my best OCXO in "perpetual adjustment," but that is mostly for convenience. Usually, I turn disciplining off when I'm taking data. Only when I'm doing something where the data are averaged for longer than about 3000 seconds do I leave it on (3000 seconds is based on the stability of my particular OCXO). Remember, GPS has a well-defined stability floor, and is not better than a good OCXO at averaging times (tau) less than 100 or even 1000 seconds -- so GPS discipline cannot do anything to help the stability of a good OCXO at shorter tau than that. (Yes, it may be able to help a lousy OCXO or TCXO at lower tau -- but you can get a better OCXO than that for $20, so why bother?) There is so much focus on GPSDOs that I think many time nuts do not realize this fundamental fact. A few rules of thumb: -- An OCXO is the best low-tau reference most amateurs can afford -- GPS discipline cannot help at low tau because it is noisy -- Most of us do not need extreme stability at long tau And some general conclusions: -- Get the best OCXO you can find -- Enclose it (thermally isolated from the enclosure) -- Don't try to whip a so-so OCXO into shape with GPS discipline Finding a really good OCXO may take some effort. Some models are more likely to be "really good" than others (like the BVA that Mark mentioned, and some others that have been vetted in large numbers), but even then there can be large differences from sample to sample. So, one may need to sort through a number of them to find a "really good" one. If one doesn't have access to a clearly better oscillator for comparison, using the "three-cornered hat" technique with one's best oscillators is probably the best method available to the amateur time nut. Note that quartz oscillators tend to exhibit best stability if they are left on continuously, and stability may improve for a long time (months, perhaps even many months) after they are turned on, depending on how long they were off and how much trauma they received before being powered up again). The point is that GPS discipline is not always (and maybe, not usually) the best way to get the best stability possibile over the range of tau that is most important to amateur time nuts. Further, it takes very well-designed GPS discipline to improve things at long tau without making them worse at shorter tau, so GPS discipline can easily be a net negative (particularly since most of us do not need extreme stability at very long tau). So, a good OCXO that is manually adjusted from time to time as required will likely have the best stability most amateur time nuts can obtain, at the range of tau that is actually important for the applications to which it will be put. Best regards, Charles ___ time-nuts mailing list -- time-nuts@febo.com To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
Re: [time-nuts] 10 MHz OCXO recommendations
Hi The only gotcha is accuracy. If that is part of the equation, then even a pretty dumpy OCXO properly GPSDO’d will beat one that is a very good OCXO indeed. A darn good OCXO will age down in the 1x10^-11 per day range. In a bit over 10 days you may be past +/- 1x10^-10. An OCXO based GPSDO that holds 1x10^-10 “some of the time” is one that gets yelled at and kicked a lot . It all depends on what you are after. Bob > On Nov 11, 2014, at 7:26 PM, Charles Steinmetz wrote: > > Mark wrote: > >> I find the concept of occasionally adjusting a good OCXO which in turn is >> used as a reference works well for me.I have some that haven't needed >> adjustment for over 2 years (they are still well within one part per billion >> of being on frequency.) > > A few of us have advocated this approach on the list, and there is good > reason for it. A GPSDO offers two advantages: (1) it is self-adjusting, > therefore easy to own and use; and (2) it has better stability at long tau > than the OCXO alone. The price you pay for those advantages is poorer > stability at low tau than the OCXO alone, which can be anywhere from slight > with a good design (e.g., Thunderbolt, Z3801) to shockingly bad with a bad > design (including many DIY attempts). > > If one does not need the very best performance at long tau -- and most > time-nuts do not -- a free-running OCXO that you adjust manually every now > and then can be the best reference available to the average time nut. ("Long > tau" can be anywhere from 100 seconds to several thousand seconds, depending > on the particular OCXO.) Plus, not spending money on GPS discipline allows > you to spend more on the OCXO to get better stability at low tau, and a more > extended upper limit on "low" tau (say, better than GPS all the way to 2000 > seconds instead of 200 seconds). > > Personally, I do use GPS discipline to keep my best OCXO in "perpetual > adjustment," but that is mostly for convenience. Usually, I turn > disciplining off when I'm taking data. Only when I'm doing something where > the data are averaged for longer than about 3000 seconds do I leave it on > (3000 seconds is based on the stability of my particular OCXO). > > Remember, GPS has a well-defined stability floor, and is not better than a > good OCXO at averaging times (tau) less than 100 or even 1000 seconds -- so > GPS discipline cannot do anything to help the stability of a good OCXO at > shorter tau than that. (Yes, it may be able to help a lousy OCXO or TCXO at > lower tau -- but you can get a better OCXO than that for $20, so why bother?) > There is so much focus on GPSDOs that I think many time nuts do not realize > this fundamental fact. > > A few rules of thumb: > > -- An OCXO is the best low-tau reference most amateurs can afford > -- GPS discipline cannot help at low tau because it is noisy > -- Most of us do not need extreme stability at long tau > > And some general conclusions: > > -- Get the best OCXO you can find > -- Enclose it (thermally isolated from the enclosure) > -- Don't try to whip a so-so OCXO into shape with GPS discipline > > Finding a really good OCXO may take some effort. Some models are more likely > to be "really good" than others (like the BVA that Mark mentioned, and some > others that have been vetted in large numbers), but even then there can be > large differences from sample to sample. So, one may need to sort through a > number of them to find a "really good" one. If one doesn't have access to a > clearly better oscillator for comparison, using the "three-cornered hat" > technique with one's best oscillators is probably the best method available > to the amateur time nut. Note that quartz oscillators tend to exhibit best > stability if they are left on continuously, and stability may improve for a > long time (months, perhaps even many months) after they are turned on, > depending on how long they were off and how much trauma they received before > being powered up again). > > The point is that GPS discipline is not always (and maybe, not usually) the > best way to get the best stability possibile over the range of tau that is > most important to amateur time nuts. Further, it takes very well-designed > GPS discipline to improve things at long tau without making them worse at > shorter tau, so GPS discipline can easily be a net negative (particularly > since most of us do not need extreme stability at very long tau). So, a good > OCXO that is manually adjusted from time to time as required will likely have > the best stability most amateur time nuts can obtain, at the range of tau > that is actually important for the applications to which it will be put. > > Best regards, > > Charles > > > > ___ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instr
Re: [time-nuts] 10 MHz OCXO recommendations
Sorry a few more points to mention. If for some reason I am particularly concerned about the stability of an OCXO reference I will compare it to another OCXO and on occasion to a GPSDO as well while measuring a "Device Under Test." This gives me some comfort that if am looking at the performance of a particular "Device Under Test" that any drift in the OCXO I am using as a reference would have been detected. (It also gives me a reason to keep my stack of HP5370 and HP5335 counters running.) I don't expect this approach to give me absolute certainty of picking up drift or jumps in my reference but it does give me some comfort. While I suspect this approach would not go over very well in a commercial lab vs buying a high performance cesium standard or H Maser (: for hobby use it seems to work for me. Timelab is also useful for collecting analyzing the data from the various counters. I also trigger all the counters from the same 1pps source. I typically compare my "best" OCXO's to my "best" GPSDO on a more or less continuous basis (from a time nuts perspective it's of some interest to look at their long term drift.) From time to time I also cross check my "best" GPSDO against another GPSDO (: Regards Mark Spencer Sent from my iPad On 2014-11-11, at 4:26 PM, Charles Steinmetz wrote: > Mark wrote: > >> I find the concept of occasionally adjusting a good OCXO which in turn is >> used as a reference works well for me.I have some that haven't needed >> adjustment for over 2 years (they are still well within one part per billion >> of being on frequency.) > > A few of us have advocated this approach on the list, and there is good > reason for it. A GPSDO offers two advantages: (1) it is self-adjusting, > therefore easy to own and use; and (2) it has better stability at long tau > than the OCXO alone. The price you pay for those advantages is poorer > stability at low tau than the OCXO alone, which can be anywhere from slight > with a good design (e.g., Thunderbolt, Z3801) to shockingly bad with a bad > design (including many DIY attempts). > > If one does not need the very best performance at long tau -- and most > time-nuts do not -- a free-running OCXO that you adjust manually every now > and then can be the best reference available to the average time nut. ("Long > tau" can be anywhere from 100 seconds to several thousand seconds, depending > on the particular OCXO.) Plus, not spending money on GPS discipline allows > you to spend more on the OCXO to get better stability at low tau, and a more > extended upper limit on "low" tau (say, better than GPS all the way to 2000 > seconds instead of 200 seconds). > > Personally, I do use GPS discipline to keep my best OCXO in "perpetual > adjustment," but that is mostly for convenience. Usually, I turn > disciplining off when I'm taking data. Only when I'm doing something where > the data are averaged for longer than about 3000 seconds do I leave it on > (3000 seconds is based on the stability of my particular OCXO). > > Remember, GPS has a well-defined stability floor, and is not better than a > good OCXO at averaging times (tau) less than 100 or even 1000 seconds -- so > GPS discipline cannot do anything to help the stability of a good OCXO at > shorter tau than that. (Yes, it may be able to help a lousy OCXO or TCXO at > lower tau -- but you can get a better OCXO than that for $20, so why bother?) > There is so much focus on GPSDOs that I think many time nuts do not realize > this fundamental fact. > > A few rules of thumb: > > -- An OCXO is the best low-tau reference most amateurs can afford > -- GPS discipline cannot help at low tau because it is noisy > -- Most of us do not need extreme stability at long tau > > And some general conclusions: > > -- Get the best OCXO you can find > -- Enclose it (thermally isolated from the enclosure) > -- Don't try to whip a so-so OCXO into shape with GPS discipline > > Finding a really good OCXO may take some effort. Some models are more likely > to be "really good" than others (like the BVA that Mark mentioned, and some > others that have been vetted in large numbers), but even then there can be > large differences from sample to sample. So, one may need to sort through a > number of them to find a "really good" one. If one doesn't have access to a > clearly better oscillator for comparison, using the "three-cornered hat" > technique with one's best oscillators is probably the best method available > to the amateur time nut. Note that quartz oscillators tend to exhibit best > stability if they are left on continuously, and stability may improve for a > long time (months, perhaps even many months) after they are turned on, > depending on how long they were off and how much trauma they received before > being powered up again). > > The point is that GPS discipline is not always (and maybe, not usually) the > best way to get the bes
Re: [time-nuts] 10 MHz OCXO recommendations
Just to add a note on the original question : there are some brand new never used Vectron 8091s available for a reasonable price now and it's drift and jitter have been on par with my Wenzel. I am not yet set up to measure it's phase noise or other general rf characteristics but according to its cut sheet they're quite good. On Tuesday, November 11, 2014, Mark Spencer wrote: > Sorry a few more points to mention. > > If for some reason I am particularly concerned about the stability of an > OCXO reference I will compare it to another OCXO and on occasion to a GPSDO > as well while measuring a "Device Under Test." This gives me some comfort > that if am looking at the performance of a particular "Device Under Test" > that any drift in the OCXO I am using as a reference would have been > detected. (It also gives me a reason to keep my stack of HP5370 and > HP5335 counters running.) I don't expect this approach to give me absolute > certainty of picking up drift or jumps in my reference but it does give me > some comfort. > > While I suspect this approach would not go over very well in a commercial > lab vs buying a high performance cesium standard or H Maser (: for hobby > use it seems to work for me. Timelab is also useful for collecting > analyzing the data from the various counters. I also trigger all the > counters from the same 1pps source. > > I typically compare my "best" OCXO's to my "best" GPSDO on a more or less > continuous basis (from a time nuts perspective it's of some interest to > look at their long term drift.) From time to time I also cross check my > "best" GPSDO against another GPSDO (: > > Regards Mark Spencer > > Sent from my iPad > > On 2014-11-11, at 4:26 PM, Charles Steinmetz > wrote: > > > Mark wrote: > > > >> I find the concept of occasionally adjusting a good OCXO which in turn > is used as a reference works well for me.I have some that haven't > needed adjustment for over 2 years (they are still well within one part per > billion of being on frequency.) > > > > A few of us have advocated this approach on the list, and there is good > reason for it. A GPSDO offers two advantages: (1) it is self-adjusting, > therefore easy to own and use; and (2) it has better stability at long tau > than the OCXO alone. The price you pay for those advantages is poorer > stability at low tau than the OCXO alone, which can be anywhere from slight > with a good design (e.g., Thunderbolt, Z3801) to shockingly bad with a bad > design (including many DIY attempts). > > > > If one does not need the very best performance at long tau -- and most > time-nuts do not -- a free-running OCXO that you adjust manually every now > and then can be the best reference available to the average time nut. > ("Long tau" can be anywhere from 100 seconds to several thousand seconds, > depending on the particular OCXO.) Plus, not spending money on GPS > discipline allows you to spend more on the OCXO to get better stability at > low tau, and a more extended upper limit on "low" tau (say, better than GPS > all the way to 2000 seconds instead of 200 seconds). > > > > Personally, I do use GPS discipline to keep my best OCXO in "perpetual > adjustment," but that is mostly for convenience. Usually, I turn > disciplining off when I'm taking data. Only when I'm doing something where > the data are averaged for longer than about 3000 seconds do I leave it on > (3000 seconds is based on the stability of my particular OCXO). > > > > Remember, GPS has a well-defined stability floor, and is not better than > a good OCXO at averaging times (tau) less than 100 or even 1000 seconds -- > so GPS discipline cannot do anything to help the stability of a good OCXO > at shorter tau than that. (Yes, it may be able to help a lousy OCXO or > TCXO at lower tau -- but you can get a better OCXO than that for $20, so > why bother?) There is so much focus on GPSDOs that I think many time nuts > do not realize this fundamental fact. > > > > A few rules of thumb: > > > > -- An OCXO is the best low-tau reference most amateurs can afford > > -- GPS discipline cannot help at low tau because it is noisy > > -- Most of us do not need extreme stability at long tau > > > > And some general conclusions: > > > > -- Get the best OCXO you can find > > -- Enclose it (thermally isolated from the enclosure) > > -- Don't try to whip a so-so OCXO into shape with GPS discipline > > > > Finding a really good OCXO may take some effort. Some models are more > likely to be "really good" than others (like the BVA that Mark mentioned, > and some others that have been vetted in large numbers), but even then > there can be large differences from sample to sample. So, one may need to > sort through a number of them to find a "really good" one. If one doesn't > have access to a clearly better oscillator for comparison, using the > "three-cornered hat" technique with one's best oscillators is probably the > best method available to the amateur ti