SEM VCO buffer (was: RE: Japanese duals (was: Re: CA3046,SSM2210&2220 Replacements)))

[Martin Czech]

::::
:::	>I had a look at the SEM vco schemo today. What is the exceptional
:::feature there?
:::	>I can only see a temperature and maybe parameter compensated fet
:::buffer.
:::	>
:::	>Please some enlightenment.
:::	>
:::	>m.c.
:::
:::Yes, it's "only" a fast buffer.
:::
:::I do not know if, or how closely, the FETs are matched in the SEM.
:::But in general, without matched FETs, if you want to make a buffer
:::stage you normally have the choice between slow (opamp ...)
:::or high offset (discrete ...). If you want fast and low offset at the same
:::time,
:::a matched FET pair is very useful.
:::

I see. A sawtooth design needs a fast buffer for controlled discharge.

But we should not complain about slow high gain op amps if we don't need
high gain. Less gain means possibility of higher bandwidth.  There are
unity gain buffers with very high bandwidth (>=50MHz).  So far I've only
found bipolar types with large input currents (NS).  But if there are
CMOS versions, maybe one should have a look at them.

Another idea is to use a CMOS logic Schmitt trigger IC.  One of the faster
logic familys. The input currents are usually much smaller then speced,
the manufacturers simply don't want to use costly test equipment.
Of course this has to be checked for every individual IC. Maybe there
are now single SMD gates on the market, so you don't have to solder
in 6 triggers if you need only one. The temperature effect on threshold
is not too bad.

If one is looking for a high speed vco, this might be worth to evaluate.
If it works no op-amp buffer is needed (only pulse wave clock needed)

OTOH in a triangle osc a garden variety op amp as voltage follower has
bandwidth about 3MHz. This should give reasonable triangle response up
to 600kHz (5th harmonic).

Comparator speed is also an issue. Has anybody tried the NE521 dual high
speed comparator/flip-flop?

Btw.: I'm thinking about getting rid of the PNP-expo converter for a
triangle vco for obvious reasons.  This means no CA3080. Such current
mirror ideas are very old (s. Tietze-Schenk high speed triangle osc.). How
about using a (discrete) PNP current mirror for the "up" movement and
direct connection to the npn expo source for the "down" movement. This
would require a NMOSFET pair so that either the expo current is fed into
the mirror, or directly into the cap via the MOSFET switches.  Sure you
have to match the PNP mirror, but you can use high hfe types, and emitter
resistors/trims for improved matching, maybe even a Wilson mirror. A
CA3080  is basically the same, but has very poor PNP transistors due to
process reasons, and the current is mirrored several times (I think up
to 3 times) which doesn't improove accuracy either. The only advantage
is the thermal contact on chip.

So 2 npn, one opamp (LF411) , 2 pnp , 2 NMOSFET and one NE521 and
one buffer (CA3040 ?) would be the recipe. Reasonable part count.
Could be potted as submodule.

m.c.