thru zero VCOs (was: Re: [sdiy] Temperature Compensated VCO attempt - help?

[Magnus Danielson]

From: jhaible at debitel.net
Subject: Re: thru zero VCOs (was: Re: [sdiy] Temperature Compensated VCO attempt - help?
Date: 3 Feb 2003 13:49:56 +0100,Mon,  3 Feb 2003 13:49:56 +0100

Hi Jürgen,

> 
> > Actually, no... I never separate the sign and the absolute value like you
> > do.
> > I let them all the way into the 4-quadrant multiplier that the Gilbert-cell
> > supplies. This comes from the observation that a Gilbert-cell actually has
> > three input signals, the standing current, and the two diffrential inputs.
> > The standing current of the Gilbert-cell comes from a normal expo-setup
> > where
> > as the two different input signals each apply it's polarity switching magic
> > on
> > the current. The beauty of this design decission is that signals requiring
> > less
> > precission enters the Gilbert-cell on the base inputs (which has the
> > annoying
> > distorsion) and the really sensitive signal, the expo-current, enters the
> > common emitter current path for which the other transistors just sets
> > current
> > ratios for.
> 
> 
> Ah, clever!

Yeah, thanks! That came from the right place and meant alot! ;O)

> I had no idea you were referring to a 4 quadrant multiplier. (I should read
> whole threads, not just single mails (;->) - do you have schematics posted?)

Or you could have picked up on my habbit to write Gilbert-cell in synonymous
for a 4-quadrant multiplier.

> > What I didn't do however was to describe how a phase quadrature oscillator
> > is designed. 
> 
> Well _if_ there are problems with this concept, I expect them to come
> from offset voltage drift. In a quadrature VCO, you'd need two such
> 4 Quadrant multipliers, and what if - due to offset drift - one of them
> sees a tiny positive offset, and the the one a negative offset?

Well, that may be one of the things you really have to look out for, but this
issue it share with other solutions too, so I don't think it is unique or made
worse by this particular solution, now is it?

> And even in the simple (not quadrature) case, what would happen when 
> your sum of external voltage and offset voltage changes sign just 
> when you're close enough to a comparator threshold? 

Ultimately it depends on noise, but since the bandwidth of a Gilbert-Cell may
be very high (have you actually LOOKED at the BW parameters for a MC1495?)
in relation to what we "need" the time/voltage differance can be very small and
it still will turn properly. As said, the noise may force it to turn early, but
also equally probably to turn too late, so that will be how phase noise enter
into the system, but it is no different from how reset/schmitt-trigger noise
modulate phase in other solutions, so as long as noise levels is kept down I
think your home safe.

Just as with the discussion on compensation for q/kT linearization is
benefitial, but it also helps to defeat temperature dependence so it should be
applied properly.

Cheers,
Magnus