On Tue, 20 Jan 1998 22:42:02 -0800 john blackburn <johnb@dircon.co.uk>
writes:
>Mike Keitz wrote:
>> You can get around the bias current limitation by providing another
>path
>> for the bias current to flow rather than through the input
>terminating
>> resistor.
>
>Nobody EVER does this kind of thing at the input stage of
>oscilloscope-type instruments. It's important to devise a circuit that
>doesn't send a voltage or current back into the source; you risk
>upsetting the circuit being tested or even stopping it working.

That is the purpose of providing another bias current path:  to prevent
voltage or current from coming out of the input terminal.  The input
terminal should appear to be a resistor to ground.  Even this can of
course upset the operation of circuits being tested.

Here's what I'm talking about.  A conventional setup has
         |\
   in-*--|A> - out
      Rt |/
      |
     GND
If the amplifier's input has a bias current (for the sake of argument,
let's assume it's into the amplifier), then with the input open a
negative voltage appears across Rt and thus at the input terminal.  This
voltage is amplified and causes a DC offset in the output.  With the
input shorted to ground, the bias current flows through the short, and
the output voltage changes.  This amplifier arrangement is unsatisfactory
unless the bias current or Rt are small enough to reduce the voltage
developed to a negligible amount.

Now consider adding another resistor from the input to a constant voltage
source.

      Vb
      |
      Rb |\
   in-*--|A> - out
      Rt |/
      |
     GND

With Vb properly adjusted, the current through Rb is the amplifier's bias
current, so it all flows into the amplifier and none through Rt.  The
open-circuit voltage at the input is zero.  To small input signals, the
input appears to be a resistor to ground of the parallel combination of
Rt and Rb.  Rb should be large enough that Vb is considerably more than
the largest input signal.  Then the current through Rb is approximately
constant regardless of the input voltage.  If the amplifer's bias current
is also constant, the input appears reasonably linear and resistive,
without stray voltage.

The major problem is to adjust Vb so that the bias current is exactly
cancelled.  If the amplifer only handles AC signals, automatic adjustment
via a "DC servo" feedback loop can be established.  If it has to handle
DC signals, then it is not possible to tell a changing input signal from
a change in bias current.  A fixed adjustment may or may not work
depending on the characteristics of the amplifier and the level of
performance required.

This circuit is a compromise.  Sean's oscilloscope, at least at this
stage, is still a hacker project rather than a hardcore high-performance,
general-purpose instrument.  There is a reason why good scopes cost
considerable amounts of money, and it's not just greed.  Using a single
op-amp as the front end is not optimal, but it is small and inexpensive.
It should have enough usefullness to explore other aspects of scope
design, then come back and revisit the front-end.

[JFET buffer]
>Not if it's done properly. There is a circuit that uses a matched pair
>of high-transconductance junction FETs (both in the same package)

I'm interested in how this works.  I have no doubt that it does, but
can't quite grasp it from your description.  I assume the second FET is
used to develop a voltage equal and opposite to the gate-source voltage
of the first one, so the output voltage is again zero with zero input.
Or does it just maintain a constant gate-source voltage by using a
precision current source, which is then removed by a level shift later
on?