Text preview for : HP-Bench-Briefs-1974-07-08.pdf part of HP HP-Bench-Briefs-1974-07-08 HP Publikacje HP-Bench-Briefs-1974-07-08.pdf
Back to : HP-Bench-Briefs-1974-07-0 | Home
SI i l l
SERVICE INFORMATION FROM HEWLETT-PACKARD
JULY-AUGUST 1974
OSCILLOSCOPE Divider Probes
PROBES & MEASURE-
Let's talk about how to make mea-
MENT TECHNIQUES surements with a typical general
purpose scope-probe combina-
by Chuck Donaldson tion. Initially, we'll restrict the
discussion to passive probes work-
ing into a high-impedance oscillo-
scope. In order to appreciate the
Historically, the oscilloscope has problems involved, we might want
been used as a tool to make mea- to take a look at just what a probe
surements of amplitude versus is, and what it is intended to do.
time over a rather broad frequency The input circuit of an oscilloscope
range. Since the display is com- has traditionally been a parallel
pletely visual, the capability of RC circuit as shown in Figure 1.
deriving a great deal of qualitative The advantage of this type of cir-
information, e.g. waveform shapes, cuit is that Rin can be .made
perturbations, etc. as well as the high enough to present an insig-
quantitive values .of amplitude and nificant load to many circuits for those cases where the signal
time, has caused the oscilloscopeto (1 Megohm) while Cin can be he!d voltage is sufficient to drive the
become the engineer's "screw- to a low enough value to maintain oscilloscope after the 1O:l division.
driver". When a problem exists, or the desired system bandwidth. A simplified circuit of this type of
gross circuit characteristics are How does the probe affect the probe is shown in Figure 3.Looking
required, he has become accus- measurement? The simplest case into the probe tip, the circuit sees
tomed to reaching for his scope, is that of a 1:l transfer probe where resistance Of R probe + Rinput.
hooking it up to his circuit, and the probe simply looks like a small Typically, this would be 9M ohms
gathering information, without resistance (200 to 300 ohms) in + 1M ohm. More importantly, note
regard for such considerations as series with the scope input, and that the capacitors are in series
input and output impedances, rise a fairly large capacitance (approxi- and the effective capacitance is
times, etc. If a probe was required, mately 37 to 55 pf) in shunt with (Ccomd (CinDut).
the one which was mechanically the input of the oscilloscope. See Ccomp + Cinput
most convenient was typically the Figure 2. The total magnitude of R
one which would be used. remains essentially constant, but Typically the circuit sees a shunt cap-
the new value of C presented to the acitance of about 10 pf!
circuit under test is Cprobe +
Cinput, so we now have approxi- While this is a very dramatic im-
mately 1 Megohm paralleled by 57 provement, this concept can be
to 84 pf. This additional capaci- carried only so far since passive
tance will translate into a loss of probe capacitance cannot easily
rise time and bandwidth. We can be reduced to zero, and there is
measurably improve the situation certainly a point of voltage division
through use of a 1O:l divider probe, which becomes impractical.
TOOLS AND TECHNIQUES
Active Probes Measurement Accuracy Readinn the CRT, YOU answer
-
10 ns. Right? Maybe yes, and
0
Active probing devices are the Let's now consider measurement maybe no.
highest resistanceAowest capa- accuracy. Suppose you feed the Although 10 ns is the reading of the
citance probes available today. output of a pulse generator scope display, this value may not
An active probe with a 1 O : l divider through a probe to an oscilloscope. be the true rise time of the pulse.
can provide an input resistance of You adjust the scope time base Ignoring scope inaccuracies, what
1 Megohm shunted by about 1 pf for a fast sweep. (This is the typical other reason is there for suspecting
of capacitance. These active probes setup for measurement of rise that the measurement is not the
are typically used with oscillo- time). The result is the familiar real rise time?
scopes that have vertical amplifiers rise-time waveform shown in Basically, when the pulse passes
with a 50 ohm input impedance. Figure 4. If the time base is set for through the probe, the scope ampli-
a sweep of 5 nanoseconds per divi- fier, and even the scope CRT, it
sion, what is the rise time of the suffers rise time deterioration.
generator's output pulse (ignoring Thus, the displayed waveform
any inherent scope inaccuracies)? doesn't represent the true, original
signal.
Figure 4. Pulse rise time is measured between Although this is an unfortunate
the 10% and 90% points on the leading edge circumstance, it has one saving
of oscilloscope trace. in this case, rise time grace: it can be predicted. So if
is 10 ns. you erroneously said that the rise
time in the example was 10 ns, it
should be both interesting and
Active probes aren't usually informative to learn how to deter-
thought of as general purpose mine what the error might be.
probes, and because they are more True rise time can be determined
expensive, more care needs to be by the use of the tongue-twisting'
taken in their use. Suffice is to say formula known as the "square
that when working at very high fre- root of the sum of the squares."
quencies (above 300 MHz or so), In mathematical terms:
there are different probes and tech-
TRG= \ J T R D ~ T R S ~
-
niques available to obtain more
accurate results (e.g. sampling where TRG = true signal generator
osci Iloscopes). 5 nsldiv rise time
WWW. HPARCHiVE.COM
TOOLS AND TECHNIQUES -m
1
TRD = displayed risetime
~
whose true risetimes are 10 ns. Ten True Signal Risetime = 3.2
TRS = osciIloscope/probe system divided by 3.5 is 2.9. Using Figure 5, 3.5 ns
risetime. the error is 6%.
True Signal Risetime = (3.5)(3.2)
For a 50 MHz oscilloscope/probe Let's look at the problem another = 11.2 ns.
combination, the risetime is 7 ns. way. Suppose we want our mea-
Putting this into the formula, we surements to be within 5%. What Rearranging our formula,
get: is the fastest risetime we could
TRD = \ ~ T R G + TRS
view on the CRT without going
T R G = I / ~ = 7.1ns= ~ over 5% in error? Going to the 5% = 511.2 + 3.5 * = 11.7 ns
The actual risetime of the pulse point on Figure 5, the ratio is 3.2.
generator is 7.1 ns, not 10 ns. The For our 3.5 ns system (TRS = 3.5). Continued on Page 8
measurement was in error by about
41%.
graph of resistance versus fre-
The input capacitance of a scope quency, with capacitance as the
is held to a value such that the variable parameter, we can.
desired system bandwidth can easily determine the source re-
he measurement. How im- Figure A.
1 1.5 2 2.5 3 4 5 8 7
Figure 5. Measurement Error V
.I Signal/
System Rlse Time Ratio
It is important, therefore, to keep in
mind that the displayed risetime is
greater than the actual risetime.
The amount of error can be deter-
mined from the graph in Figure 5.
Here, percent of error is plotted
against the ratio of test system rise
time over true signal risetime. As
an example, suppose you are using
a 100 MHz oscilloscope/probe
combination that has a rise time of
3.5 ns. You are examining signals
WWW.HPARCHIVE.COM
vi.
*I
i .-.-