On Fri, 05 Dec 2008 06:51:11 +0000, Mike Monett wrote:

John Larkin wrote:

On Fri, 05 Dec 2008 04:13:50 +0000, Mike Monett wrote:

John Larkin wrote:

If you've got a r-r amp, one thing to do is get some gain still
referenced to the V+ rail, then diff-amp that. That makes the cmrr
of the diffamp matter a lot less. That's what I do with current
shunts... get some good gain *before* dealing with the common-mode.

Why should it make any difference? If you are trying to get gain at
the V+ rail, you are still dealing with cmrr. The opamp you are using
for gain is still taking a small difference between two large numbers.

I can't express it any simpler than I already have. A possible
schematic is attached.

The bottom op amp is still connected to the +Vcc rail. How does that
eliminate cmrr error?

Nothing *eliminates* cmrr error. But in a case like this, the big
error isn't usually from the opamp, it's from mismatch in the four
diffamp resistors. You can buy a superb opamp for a fraction of the
price of a single 0.01% resistor. This configuration improves the cmrr
of the second amp, again dominated by resistor mismatch, by the
single-ended gain of the first amp.


Also, the circuit you drew requires base current for the NPN. That
will add aother undefined and variable error to the output.

Use a high-beta transistor, like a BCX71. The error will be far below
1%, and that will be swamped by the resistor tolerances. Or use a fet
if you have to. If you really want to whine about something, consider
stabilizing this loop.

John


begin 644 DSCF1815.JPG

Attachment saved: C:\0DNLOAD\DSCF1815.JPG
`
end

Neither of your proposals reduce cmrr errors. Tossing more parts at it
doesn't help.

Stop thinking about the cmrr spec of the opamp and start thinking
about the overall circuit performance. If you weren't being so
p(r)issy, I'd show you a few more tricks.

John