[ARC5] BC453 FET BFO Success!!!

Mon Dec 18 15:59:03 EST 2017

Hi

Ok, the 5 MHz “reference design” drifts 20 ppm / hour. ( 100 Hz at 5 MHz is 20 ppm).
If drift in Hz is proportional to frequency squared, then drift in ppm would be proportional 
to frequency. Cut frequency by 100:1 and drift in ppm drops by 100:1. Thus, as you go
very low in frequency, your oscillator gets *very* stable. 

If one builds a practical 1 Hz LC oscillator, you would then expect it to drift 20 / 5,000,000 = 4 ppt 
per hour. In more conventional units, it would be just a bit under 1x10^-10 per day. Yes, there’s a 
case of your choice of beer (delivered) behind that  bet not being true. :)  

Some rules of thumb work over a range, but then … not so much. The question becomes -
what are the practical limits to some these sorts of rules? 

The first point is that we are talking about a radio here. The oscillator needs to fit in the
radio. That places some basic limits on component size. We need to buy the parts 
so that rules out superconductors or other crazy stuff. An oscillator that is the size of a
house or that costs a million dollars isn’t a practical answer to the question of what to 
put in a BC-453. 

Another obvious gotcha is that Q of an L/C tank does have some practical limits. Over some 
range, you might get higher Q as frequency goes down. That will not go on forever. For 
rational sized / reasonable.value coils it stops working well above audio frequencies. Indeed
if Q alone is the answer, a VHF cavity is a better bet than an LC tank ….. Since it *does* 
matter to some degree, it’s part the reason we don’t design RC oscillators into radios. 
(the other part is phase noise, but that’s getting a bit further afield…). 

Another limit might be that if you look at a some frequency ranges, you can get better
temperature stability coils at a lower frequency. Caps … not so much. Over some limited
range, as you go lower in frequency coil stability may improve. That sort of works down through 
HF, but not very well below that. 

Aging in capacitors isn’t going to be any different at 1 MHz vs 10 MHz. The packages and
the materials are all the same. Coil wise, some of the same material differences that impact 
temperature impact aging. Ceramic forms with air in them are very stable. Put in powdered 
iron and they get less stable. Don’t even think about ferrite materials if you are concerned 
about aging on a coil ….. If you constrain coil size and the value keeps going up and up …
that’s a problem. 

Sustaining stage variation in capacitance could be an issue. If you have a tube that works 
well into the VHF range, it’s not struggling much at any of the frequencies we’re talking about.
In terms of capacitances, they are small. To the extent they change, it’s as a percentage. Small
is good in this case.  Gain wise, it’s no more an issue at one frequency than at another one. 

Can you design / build an oscillator poorly? You most certainly can. What is the “best” for this or 
that may not be best at a different frequency or with different components. One person’s 
idea of good construction techniques may not be the same as somebody else’s idea.  If 
one of us is working at 100 KHz and the other at 100 MHz, tough to compare one to the other.
All of this sort of stuff *will* play very much into the result and thus a comparison of A to B. 

So where do I send the beer? 

Bob

> On Dec 18, 2017, at 2:57 PM, Kenneth G. Gordon <kgordon2006 at frontier.com> wrote:
> 
> On 19 Dec 2017 at 4:49, Leslie Smith wrote:
> 
>>   Jim,
>>   The general rule about LC oscillator stability is this:
>>   Stability at a higher frequency decreases with the ratios of square of frequency.
>>   As an example an oscillator at 4000kHz will drift 4x more than the same oscillator at 2000kHz. 
>>   (F2/F1)^2.
>> 
>>   Inverting this formula, I get stability at a lower frequency increases with the reciprocal of the 
>> square of the two ratios.
>> 
>>   This, of course, is a very general statement and assumes a lot about the contribution to drift by 
>> coil and capacitor, but speaking generally I have found the formula to be generally correct. 
>> 
>> Taking this as a rule, and knowing that a well built oscillator at 5MHz may drift 100Hz in one hour 
>> then your circuit should have a stability of 100 * (85/5000)^2.  That's my prediction. 
>> This is about 28mHz.  milli-Hertz, not Mega-Hertz!
> 
> That is 0.028 Hz/hour, gentlemen. If this bears out in actual practice, then Mac's "apparent" 
> drift is very, very excessive, and there must be other issues at play there. We need details.
> 
> Even including the drift of the HFO combined with the drift of the BFO, the drift, on average, 
> should be something less than 0.06 Hz/hour.
> 
> We need to test this. However, in my case, I sincerely doubt that my equipment can reliably 
> show such minimal drift.
> 
>> I've seen others use a linear relationship for drift and frequency.  If that holds (in my book it's 
>> wrong)
> 
> Correct. Again, in actual practice, that curve is, at the very least, highly irregular. As I 
> mentioned, there was at least one, and probably several, articles which were somewhat 
> recently printed in Electric Radio Magazine which detailed such drift in LC oscillators of 
> several designs.
> 
>> then your circuit will drift not more than 2Hz.  I predict the drift will be less than I can 
>> measure with my (very-ordinary/old) Fluke counter.
> 
> Likewise here.
> 
>> I'll watch your result with interest.  Of course at 85kHz an RC oscillator may prove a second 
>> option.
> 
> Yes. With its own, somewhat different, drift characteristics.
> 
> Ken W7EKB
> 
> ---
> This email has been checked for viruses by Avast antivirus software.
> https://www.avast.com/antivirus
> 
> ______________________________________________________________
> ARC5 mailing list
> Home: http://mailman.qth.net/mailman/listinfo/arc5
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:ARC5 at mailman.qth.net
> 
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html