AW: frequency shifter - BFO problems

[Haible Juergen]

Thanks for all the answers - they were very helpful, and they all
go into the same direction, as did Martin Czech's private mail.
This list is really *great* - ask whatever question you want, you will
get knowledgable and helpful answers!

	>I've built the Electronotes PCC Freq. Shifter using AD633
multipliers, and
	>the same freq. multiplication techniques and have noticed the
same problem.
	>In my case, I've noticed some erratic output (Voltage vs. Time)
from each
	>multiplier near 0 Hz - even without doing and x/y comparison
between the
	>two...

	>i can't really offer practical advice as i've never built a 
	>quadrature downmodulated thing, but it seems that any miniscule
phase 
	>problem at 24kc will become truely detrimental at low
frequency.  

	>I've heard about similar problems before using 2 hf
oscillators. 

Yes, I am convinced now that it's a systematic error. 
Well, I could improove the jitter of the HF-oscillators (now
I know why Bode used a passive LC-filter after the sine shaper !!),
but at Zero the phase error always would go to the infinite.
So I will drop the whole concept of BFO. The only reason
why I wanted to use a BFO, was that I expected a smooth
zero crossing, while a base band quadrature oscillator with
thru-zero capabillities would produce some tiny artefacts 
during zero crossing. But now I think the later would make
less problems!

	>I've built one using a 90degree phase network (4 opamps),
Quadrature VCO
	>(LM13700) 2 multipliers (MC1495) and summing amp (1 opamp).
	>
	>Of course it's not possible to get a through zero shift with
this
	>arrangement, but it is possible to have a stable shift of 0.1
Hz or even
	>less, up or down, which I think is more usefull (?).

Good news: I think I have just figured out a circuit for a quadrature
VCO that *does* go thru zero.
One of the thru-zero VCOs in EN used two pairs of CMOS switches
to change direction of the integrator slope, and polarity of the
hysteretic switch, at the same time.
Now I could use my "triangle sin/cos" VCO (see Synthfool site;
I have got the idea form Tietze/Schenk, btw.), and do a similar
switching. Took some time to figure out how the switches would
have to work on the sin and cos slopes, but the results look
very promising.
The only thing that remains is the need of excellent triangle to
sine converters, as I can't have any filter after the oscillator to
smooth things: The filter wouldn't perform the phase reversal
fast enough, I think.
So I am looking for a good sine shaper. An overdriven ota or
differential pair would not be good enough. I think about
using a differential pair with degenerated emitter - which happens
to be built into one input of a MC1496 modulator already !!
Now this would be nice, going directly into the 1496 with the
triangle signals, and doing all the shaping right there. EN
claims to achieve 0.2% distortion with the degenerated emitter
method, which would be just ok.

Nothing prototyped yet - I don't know if I can do it bevore my moving.
This is still like a "shadow hanging over me" ...

JH.