[time-nuts] Thunderbolt versus Home made

Dr Bruce Griffiths bruce.griffiths at xtra.co.nz
Sun Feb 25 23:55:33 EST 2007


Didier Juges wrote:
> Dr Bruce Griffiths wrote:
>   
>> Didier Juges wrote:
>>   
>>     
>>> Angus wrote:
>>>   
>>>     
>>>       
>>>> Although that's not what I was talking about doing above, I think that
>>>> it's pretty much what some of the hardware GPSDO's actually do using
>>>> various types of oscillators (and to good effect too). As in these:
>>>>
>>>> http://www.jrmiller.demon.co.uk/projects/ministd/frqstd.htm 
>>>> http://www.frars.org.uk/cgi-bin/render.pl?parameter=&pageid=1285
>>>>
>>>> the loop filter is basically just an RC circuit. It's not quite raw
>>>> GPS, but is quite different to using a digital filter as used in a
>>>> Thunderbolt, Shera, etc., which can run to hours.
>>>>
>>>> When the N1JEZ board was mentioned, I assumed that it was this type of
>>>> controller that was meant - and I think it does need 10KHz unless it
>>>> gets much more of a mod than a couple of counters.
>>>>
>>>> Angus.
>>>>   
>>>>     
>>>>       
>>>>         
>>> Angus,
>>>
>>> Don't get me wrong, the James Miller design does a great job, 
>>> considering it's simple and elegant implementation and the fact that for 
>>> so little money you can have a frequency standard that beats anything a 
>>> ham could make in his or her shop.
>>>
>>> This will do a great job for most applications where you want to be 
>>> within a few Hz or so around 10 GHz, and as long as the GPS signals are 
>>> good. For the intended market, if the GPS signals are not good, you 
>>> simply wait or move the antenna.
>>>
>>> A lot of the money spent on commercial GPSDO's is spent on improving 
>>> hold-over performance, because many commercial applications simply 
>>> cannot postpone using the system until the GPS signal is good.
>>>
>>> I just wanted to point out that the 10 kHz output can be misleading. The 
>>> GPS timing is such that if you have a quality OCXO, a loop faster than 
>>> at least 20 minutes will actually do a disservice to the OCXO, 
>>> regardless of the PLL reference frequency. If you use an inexpensive, 
>>> not temperature stabilized VCXO, you can probably speed up the loop 
>>> quite a bit, simply because the VCXO itself in free running mode will 
>>> not be doing that great.
>>>
>>> If your frequency and time needs are modest (and that probably covers 
>>> 99% of applications, including anything I could dream of personally), 
>>> this design will be perfectly satisfactory.
>>>
>>> But, and this is a significant caveat, this is TIME-NUTS where most 
>>> people are looking at that like the holly grail 1% :-)
>>>
>>> Didier
>>>
>>>
>>> _______________________________________________
>>> time-nuts mailing list
>>> time-nuts at febo.com
>>> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>>>
>>>   
>>>     
>>>       
>> Didier
>>
>> There's no proof that the disciplined oscillator is in any way better 
>> than the undisciplined oscillator.
>> No before and after measurements seem to have been made.
>> The short term stability after disciplining with such a short time 
>> constant will invariably be several orders of magnitude worse than the 
>> undisciplined crystal.
>> Why bother?
>>
>> Bruce
>>   
>>     
> Well, if you use an inexpensive VCXO like the one used in the James 
> Miller design, disciplined to GPS, at least you can be reasonably 
> assured, unless a gross mistake was made in the implementation, that 
> over time, you will still be within 1e-9 or in the neighborhood. Maybe 
> the short term stability won't be great, but the aging should be taken 
> care of.
>
> Looking at the posts in ham radio groups, the stability of this design 
> is sufficient for microwave operation (a few Hz to 10 or 20Hz at 10 GHz 
> is more than sufficient in most cases), so it does the job for which it 
> is intended. Keep in mind a lot of hams operate their gear outside, 
> where temperature variations would seriously affect any oscillators.
>
> I also believe, and that was the intention of my post, that simply 
> because the GPS outputs 10 kHz, the time constant should not be 
> shortened. In most cases (reasonably good OCXO), the time constant 
> should be in the order of 20 minutes or more, from what I have gathered. 
> There would be absolutely no advantage, only degraded performance, in 
> running faster.
>
> However, as you suggest, if you do not use exactly the same parts and 
> implementation as used in the published design, you should do the 
> necessary measurements to make sure you have not done worse than running 
> the VCXO by itself (and that brings us to another thread: how many 
> oscillators do you need to make sure you have a good one).
>
> On the other hand, I have seen descriptions of many successful 
> operations with HP 10811 or the like, which are adjusted in the shack 
> against a GPSDO before being taken in the field, and as long as they 
> remain powered, will probably have more than sufficient stability for a 
> full day of operation outside, if they are reasonably well protected.
>
> I am not sure that this would be actually simpler or easier than using a 
> GPSDO like the James Miller design. The short term stability may be 
> better with a good OCXO, but the inexpensive GPSDO will be *sufficient*, 
> and you won't have to worry about drift.
>
> Just my $0.02...
>
> Didier
>
> _______________________________________________
> time-nuts mailing list
> time-nuts at febo.com
> https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
>
>   
Didier

The major objection to these designs is that there are no definitive 
before and after data provided. In fact very little performance data is 
provided.
Vaguely hand waving about a rubidium oscillator without providing its 
specifications (in particular Allan deviation as a function of averaging 
time) or at least the model so one can look it up, means that one is 
never sure how much of the instability is due to the standard.

Vague statements like "I couldn't detect any 1Hz sidebands" without 
specifying the measurement setup used leaves one with no confidence in 
the statement. Since the phase modulation is quasi random, one may 
expect a broadening of the phase noise floor close to the 10kHz carrier, 
not knowing the details of the techniques employed engenders little 
confidence in the sensitivity of the techniques employed.

I am also somewhat bemused by the seemingly insane desire to omit a $1 
or less opamp and consequent reduction in performance . Other examples 
of poor design abound in the circuit diagrams provided.

Bruce



More information about the time-nuts mailing list