Hi Attila,

I gather you did not fully read the paper ?

In normal CMOS circuits, the higher the oscillator frequency the higher
the amount of current drawn to reach that higher frequency.  So, the two
oscillator system was used to keep time and "wake" up the higher
frequency oscillator (for example the 12.8 MHz) when the radio was in
operation.  When not in operation just the lower frequency oscillator
(32 KHz) was used to keep time and provide a "wake" of the
microprocessor and the higher frequency oscillator needed for the radio
operation.

This paper presents a circuit topography that allows the low current
operation at a high frequency (12.8 MHz) thus reducing complexity.  This
in turn allows the design and manufacture of a radio system using one
crystal oscillator at a frequency of 12.8 MHz (example in the paper)
with the low power advantage that previously required two oscillators.

Bill....WB6BNQ

Attila Kinali wrote:

On Sat, 10 Aug 2013 02:39:35 -0700
wb6bnq wb6bnq@cox.net wrote:

I did, but...

That's one advantage, and not a small one, but differential oscillators
have been in use earlier and even in places where power consumption did
not matter much. It pops up in crystal oscillator designs now and then
but without any mention why this architecture was choosen. So i started
to wonder whether there was any additional advantage than just lower
power consumption and being able to work with less headroom, like better
phase noise or better long term stability or less harmonics.

		Attila Kinali

--
1.) Write everything down.
2.) Reduce to the essential.
3.) Stop and question.
-- The Habits of Highly Boring People, Chris Sauve

On 08/10/2013 12:10 PM, Attila Kinali wrote:

Well, at least from this paper they have not analyzed that. Here they
only use it for it's benefits in power, which is obvious from the Abstract.

If you wish to know other benefits, they need to be analyzed separately,
which by itself might prove an interesting paper. Reducing current drawn
should be interesting, as this should reduce 1/f noise in the feedback
amp, which should make the 1/f^3 noise lower significantly, which should
be beneficial for the stability of the oscillator in noise terms,
however it might not be beneficial for the oscillator in systematic
frequency drift terms. As always, it's a balance thing.

It should not be too hard to build it, try it, measure it and learn from
it. Sounds like fun!

Cheers,
Magnus

Build it from discrete parts, of course, what frequency do you suggest
to try? 32768Hz, 1MHz? I have nothing in-between...

On Sat, Aug 10, 2013 at 5:04 PM, Magnus Danielson
magnus@rubidium.dyndns.org wrote:

Hi

I think you'll find that the low current amps in their schematic have pretty large 1/f noise.

Bob

On Aug 10, 2013, at 11:04 AM, Magnus Danielson magnus@rubidium.dyndns.org wrote:

Hi

I suspect that built from discrete parts you will simply have an audio / square wave oscillator. It's a classic multivibrator circuit….

Bob

On Aug 10, 2013, at 11:13 AM, Azelio Boriani azelio.boriani@screen.it wrote:

On 08/10/2013 05:13 PM, Bob Camp wrote:

True, but if you wanted to fool around a little and see what it could do.

For the intended application, it's probably good enough thought.

Cheers,
Magnus

Hi

Looking at the picture of the die, I suspect their radio has a VCO on it that they lock up through a (noisy) low frequency PLL. That would mean they really don't care a lot about phase noise of the reference.

Bob

On Aug 10, 2013, at 11:43 AM, Magnus Danielson magnus@rubidium.dyndns.org wrote:

On 08/10/2013 05:55 PM, Bob Camp wrote:

Agree. But I was arguing about looking at it outside of their system
limits and see if it could be practical approach otherwise. Then their
choice of transistor geometrics etc. is irrelevant. So, given that,
could it be potentially interesting?

Cheers,
Magnus

Hi

The stability with two active devices will be worse than with one. That's true of tc, ADEV, and phase noise.

Buffering out of the circuit is problematic for a precision application, so that's likely to add to the noise as well.

It's a reasonable way to do a cheap oscillator. It's probably not a lot worse than some inverter feedback clocks.  It's not a great approach for a TimeNuts stable part.

Bob

On Aug 10, 2013, at 12:03 PM, Magnus Danielson magnus@rubidium.dyndns.org wrote: