[sdiy] Expo converters ( uh-oh, I started reading EN... )

[Grant Richter]

The reason for the difficulty in using the Rossum type with PNPs is  
because the transistor base which moves in the correct direction is  
grounded.
For NPN pairs, with the now preferred 56.2K over 1K tempco divider to  
the base, you can just series a diode with a 1 meg trim and run it to  
the center of the divider.
For PNP pairs, the correcting function needs to be routed back to the  
summing node of the input op-amp.
The purpose of the diode is to remove the Vbe drop of the base- 
emitter junction, so the correcting function zero intercept is  
referenced to zero volts since it normally has the Vbe voltage added  
in from where it is picked off from the joined emitters.

On Nov 29, 2005, at 8:52 AM, Grant Richter wrote:

>>
>> It's hard to match pnp/npn pairs -- that's the reason this  
>> configuration is not so popular any more.
>>
>>
>
> Also note: In the mixed THAT array, the NPNs are matched, the PNPs  
> are matched, but look at the offset spec between the two types.
> PNPs and NPNs arrive at a different Vbe even with identical  
> monolithic doping.
> The only way to get 1 mv. Vbe match from PNP to NPN is hand  
> matching. Very tedious and finicky.
>
> The three published types of high frequency compensation are (and  
> correct me if I misspell the names).
>
> Franco type: A resistor in series with timing cap. Actually  
> compensates for comparator reset time, but can "swamp" Rbe error  
> somewhat.
>
> Hemsath type: That's the one with the transistor driving the joined  
> emitters.
>
> Rossum type: This is the simpler one with a diode or diode  
> connected transistor. Not mentioned in regular Electronotes, but a  
> brief mention
> in one of the mid-month newsletters (how did Bernie do it?). Easy  
> for NPN pairs, more difficult for PNP pairs.