This is my first data acquisition project.  So I have a million
questions.

Some background first.

My task is to reverse engineer a hall effect sensor that is actuated
by some spinny stuff.  I first must characterize how the sensor
behaves in situ. When I redesign this sensor, and the changes that I
make will cause some phase shift of the waveform, and changes in the
dwell of the waveform, from how it would have occurred on the original
part.

Consequently, I must be very certain of the phase shift and the dwell
of the original parts, and of the redesigned parts.

Specifically, I am interested in buffering signals.  I am hesitant to
bring any signals straight into the computer which are not buffered,
even if they are the appropriate voltage.  I just hate it when I let
the smoke out.  Just to get my juices flowing I built a unity gain
follower.  I used a 741 op amp.  I input a 100 kHz sine wave.  The
output of the op amp lagged the input by .18 microseconds. At the
speed which the mechanical system will run, this equates to a shift of
1/25th of.a degree.  I want this shift to be less than 1/100th of a
degree, if possible.  Or a lag of less than 45 ns.

I think this is reasonable, or at least that I really need it.  I need
to know at what position, to the 10th of a degree, the sensors trip.
So I don't want to introduce any errors greater than a 10th of a 10th
of a degree.

So here are some of the questions that I have,

Is it ever appropriate to place unbuffered signals into the computer?

Should I make any distinction between the inputs to a counter card and
the inputs to a DAQ card, with respect to buffering or not?

If I buffer one signal but not another, how do I correct for the shift
in the one, to keep it synchronized with the other?

Do I need to qualify any buffers that I use and correct the timescale
in the VI?  Or, must I use buffers which are all essentially the same
with respect to lag time?

Some of the signals will require attenuation, any recommendations
regarding fast, attenuating buffers?

I was reading some literature from a strain gauge manufacturer, they
recommended the use of an analog low pass filter. The idea is to
prevent aliasing of frequencies faster than the sampling rate of the
data acquisition cards.  So they say pass only frequencies 1/2 or less
than the sampling rate.  Is this reasonable in all cases?

Sorry to go on so, I have plenty more questions. The writing of posts
is sometimmes helpful in thinking a problem through.  Hence, my
verbosity..

Mike

Some comments:
- The ancient 741 is probably not the best choice in your application.
Select an opamp with higher GB-product to decrease the delay.
- Besides of limiting bandwith, an electronic lowpass-filter
introduces delays that you try to avoid.
- If you know the delay-times, in my opinion you could recalculate the
original position (lineair to the frequency of the spinning object).
Only accurately possible if sampled at high enough frequency.

Hope some of this helps.

Serenity systems design

Thanks for the comments.  I was just using the 741 as jumping off
point.

I am not an EE so I am learning many new things.  I see the GB product
in the spec sheet, but don't know its significance.  Can you explain?

After studying spec sheets, I had selected a Linear Technology op amp,
LT1809, to use for further experiments.  LT1809 has GB product of
160MHz typ.

I haven't found a spec sheet on the 741 I was using (I bought it years
ago form Radio Shack for entertainment purposes).  I have no way of
guessing its GBP.

I was attracted to the LT1809's high slew rate (350V/us), and low
distortion.

Can I find better?

I don't have one yet so I haven't had the opportunity to see how much
it is better.

Mike