From: email@example.com (Tom Bruhns)
Subject: Re: Wideband gain measurement
Date: 11 Oct 2002 10:45:31 -0700
NNTP-Posting-Date: 11 Oct 2002 17:45:31 GMT
It occured to me that the linearity, and therefore the relative
accuracy, of the Agilent 89410, is good enough that you don't need a
precision attenuator to do this measurement. The accuracy I see in
actually doing the measurement as described below is much better than
the OP wanted.
1. Set the input ranges on the analyzer to two values about 25dB
apart. Don't let it autorange and keep them set to these values. Let
the instrument warm up, do an autocal, and turn autocal off; the rest
depends on time stability of the instrument, so be sure it's warmed up
thoroughly and isn't in a changing environment. Enable averaging: use
as heavy averaging as you can wait for. 100000 msmts averaged only
takes a few minutes anyway.
2. Connect the source to both channels, and set it up for a periodic
chirp, just below full scale of the more sensitive channel. Set
appropriate frequency span, rbw and window. Make a frequency response
3. Store the freq response in a data register. Set up a math
function for frequency response/that data register. Display the math
function. Set the scaling for 0dB reference level, centered at 50%,
0.1dB/division, and you should see a straight line at 0dB. The math
function cals it to be zero dB.
4. Connect the amplifier between the inputs, so the one set to higher
full scale measures the amp's output. Put the chirp signal into the
input at a level to prevent output overload. Make your averaged
When I used a precision-calibrated 50dB attenuator as the "amplifier"
in a measurement done this way, I saw attenuation within about 0.1% of
what I expected based on the attenuator calibration.
This measurement makes use of the very good linearity, and therefore
relative amplitude accuracy, of digitized signals. The ADC in current
production 89410's is only 14 bits, but because of linearizing
techniques, you can measure signals 120dB below full scale with very
decent amplitude accuracy. The measurement described above operates
the channels within the range from full scale to about -26dBfs, where
they are extremely linear.
It's very helpful in making accurate measurements if you understand
very specifically the limitations...and the strong points...of the
equipment you have available. Often, you can design setups that take
advantage of the strong points, and reject (cancel out or ignore) the
weak points. In this case, I'd never trust the internal attenuators
in the 89410 to be as accurate what I can get out of the ADC system
over a moderate range like the suggested setup takes advantage of. If
I didn't have this particular instrument but had something else, I'd
look for ways to similarly take advantage of its particular strengths,
while avoiding the limitations.
firstname.lastname@example.org (Tom Bruhns) wrote in message news:<email@example.com>...
> "John C. Price" wrote in message news:...
> > Folks,
> > I have a measurement problem that I have not encountered before. I want to
> > measure the gain vs. freq. of an amplifier that works from 1 Hz to 1 MHz,
> > and I want measurements accurate to 2%. The thing has a nominal gain of 50
> > and the input impedance is 8 pF in parallel with more than a G-Ohm. How do
> > I do it? No problem making rough measurements with a scope but I doubt that
> > 2% is possible. Should I get an HP 3400? I have an FFT spectrum analyzer
> > that seems to do the job below 25 kHz, but what about above that?
> As someone else suggested, a network analyzer; but also a precision
> attenuator to check the analyzer's absolute accuracy. I don't believe
> I'd have any trouble making the measurement with my Agilent 89410.
> The rated absolute accuracy is +/-0.5dB, but you can easily check that
> at zero dB "gain" by putting the same signal into both channels, and