6
Sampling oscilloscopes
In addition to the ART (analogue real-time) oscilloscopes at
which this book has looked so far, there are other types of great
importance; in particular sampling oscilloscopes and DSOs-
digital storage oscilloscopes. The latter have gained wide accept-
ance as the limitations of the early models have been overcome,
new techniques to extend their capabilities being introduced with
almost bewildering speed. Chapter 7, then, is devoted entirely to
DSOs. But we will look first, in the rest of this chapter, at
sampling oscilloscopes. There are four reasons for doing things in
this order.
First, historically speaking, sampling oscilloscopes predate
DSOs by the best part of two decades. Second, an important class
of DSO - the digital sampling oscilloscope - uses exactly the same
technique for capturing a repetitive, very high frequency wave-
form as that used in the traditional sampling oscilloscope
described in the remainder of this chapter. This
technique
is thus of
extreme contemporary importance, even though the type of
sampling oscilloscope described below is no longer in current
manufacture.
Third, it will enable us to clear up one phenomenon - aliasing
-
before tackling the increasingly complex digital storage scene.
And fourth, although they no longer feature in oscilloscope
manufacturers' catalogues, there are of course many
analogue
sampling scopes, as distinct from the later digital sampling
oscilloscopes, still in use.
Both sampling scopes and DSOs look at an input signal at
discrete 'sampling' instants, rather than continuously like an
analogue real-time scope. They are therefore only aware of the
state of the signal at these instants and are completely ignorant of
what happens in between the samples. This ignorance is the basic
cause of aliasing, as will become apparent shortly.
Analogue sampling oscilloscopes, which I shall call simply
sampling oscilloscopes from here on, offer certain advantages
over ordinary real-time scopes but, as is always the case in
Sampling oscilloscopes 89
electronics (as indeed in life itself), these advantages are not
obtained without some accompanying limitations. Sampling
scopes were introduced in the late 1950s and offered unheard-of
bandwidth compared with real-time oscilloscopes of the day. In
the latter, by using a 'distributed amplifier' consisting of many
valves effectively harnessed in parallel, and restricting the c.r.t.'s
Y deflection range to just four divisions against the eight provided
as standard nowadays, a bandwidth of 85 MHz was achieved. In
contrast, the Hewlett-Packard model 180 sampling oscilloscope
boasted a bandwidth of no less than 2 GHz (2000MHz), more
than twenty times that of the best real-time scopes of the day.
Subsequently, following great advances in the design of
cathode ray tubes and using advanced solid state circuit tech-
niques, real-time oscilloscopes with a bandwidth of 500MHz
became available from a small number of manufacturers. The
state of the art was represented by the now discontinued
Tektronix 7104 oscilloscope, with a bandwidth (via the u
amplifiers) of 1000MHz, or in excess of 2000MHz for signals
connected directly to the Y plates of the cathode ray tube.
Corresponding advances in sampling oscilloscopes led to
instruments with bandwidths of 14 GHz in the early 1970s, and
latterly to the Tektronix l1801B DSO. This digital
sampling
oscilloscope (as distinct from an ordinary digital storage oscillo-
scope) has pushed the bandwidth of such instruments of 50 GHz.
Thus there is much the same ratio between the maximum
bandwidths of real-time and sampling oscilloscopes as prevailed
in the 1950s.
So how do sampling scopes achieve their notably superior
bandwidth? And what are the limitations which were mentioned
earlier? Clues can be gained from the block diagram of a basic
real-time scope, see Figure 2.1. The bandwidth limiting factors
there are the input attenuator, Y amplifier, Y deflection stage and
of course the c.r.t, itself. The techniques used to maximize the
bandwidth of the attenuator and amplifiers are discussed in
Chapter 10 whilst the corresponding techniques in the case of the
c.r.t, are covered in Chapter 9. The sampling oscilloscope avoids
all these limitations at one fell swoop, by simply not attempting to
deal with the whole signal in real time. Instead, it takes samples
90
Oscilloscopes
of the instantaneous voltage
of
the input signal on succcssivc
cyclcs and asscmhlcs these samples
to
forrrl
a
pict.ure
OI
the
comp1er.e wavdorrn.
It
can
mly
operate in this way if the signal
goes
on
repeating from cycle to cycle for as lorig
as
it
takes
lo
build up the display.
Herice
the
sampling oscilloscope
is
limited
to
displaying repetitive wavcforrns. This is one limitation. Another
results from the omission
of
input attenuator and input amplifier.
The size
of
the largest input signal which a sampling oscilloscope
can handle is quite restricted, only a few volts peak to peak
-
including any d.c. component. Fortunately, when using a
sampling scope we are often interested only in the a.c. behaviour
of
the circuit under investigation.
So
a.c. coupling can be used
to
prevent any
d.c.
level present eating away at the usable a.c. input
voltage range, whilst for handling larger signals, a
xl0
attenuator
can be used. Likewise, the omission
of
an input amplifier limits
the usable range in the other direction, the smallest signal swing
viewable being limited by sainpling noise
-
the irlevitable small
sample-to-sarnple voltage variations which occur even when the
input voltage itself is not varying.
Thus
rhe
main requirt?ment for
;I
sampling oscilloscope
is
a
circuit capable
of
accurately sampling
the
iriput wavelorrn
al
i-rgular inlrrvals.
In
a riulshell, this
is
the strohoscopic tech-
nique used to slow down the motion (or I'requericy)
of everits
which are
too
fast
lo
observe by conventional means.
If
we
want to study some mechanical cvcnt like the turning
of
gears
which rotate too fast
for
the eye
to
see, we can illuminate them
with
a
stroboscope.
If
they are repeatedly briefly
lit
once per
revolution, or once
after
several complete revolutions, they will
present
a
stationary image. But
if
after each revolution
(or
group
of
revolutions) we light them
up
a
small
amount
of
time
later (say
At
later), then the
eye
sees
samples
at
successively
later positions,
and
if
this happens continuously, the eye can be
deceived into seeing continuous (albeit much slowed down)
mor.ion
-
rhe same
effect
wrhich
in
a
movie rnakes the spokes
of
a
wheel sccm
to
bc tiirriirig slowly
or
even turning back-
wards whcn the vehicle
is
in fact travelling forwards rapidly.
This
is a
direct
analogy
of
'sequential sarnpling', the most
co
ni in on
r
r
ch n
i
q
11
e
11
siid
i
n
sa
m
p
I
in
g
scopes.
Sampling oscilloscopes 91
signal
I
trigger ~ At
~! $I ~"
~ I $ ' ~ I ~
" ~ '~ ',~
$
', , , ,
, , , ', , : I :
real !'_'"",,"A=L'I,'"I .... ' l,,,,I .... ~ I,,,,I .... '!,,,,I .... '1 .... i .... I .... 1 .... J,',,,I .... I,,I,! .... I,,,i1,,,,I,,,,1
time 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100105110 ns
samples e- _--t- - --o- .... ~ -- -'e- ---e-
taken 9 ----'e'---e'--- --"~----e
oscilloscope
display
[ -;- 9 1
equivalent
1 .... 1 ....
time
0 5 10 ns
Figure 6.i When viewing a low frequency, a sampling oscilloscope may take
one sample per cycle of the input waveform, as here
Figure 6.1 illustrates the basic process. A signal is applied to the
vertical input of the sampling oscilloscope, and also (internally or
externally) to its trigger circuitry. Assume that the negative tip of
the waveform just causes triggering, and that the first sample of
the signal voltage is taken at that instant. On the oscilloscope a
dot appears in the correct vertical position. On the next signal
cycle a trigger pulse again occurs at the same point on the
waveform, but this time circuitry in the scope delays the taking of
the sample by the time increment At. This second dot will appear
at an appropriately higher level on the c.r.t., and it must also be
displaced to the right by a distance representing the time delay At.
Subsequent samples build up a dot representation of the
complete waveform.
Figure 6.2 shows the same procedure, except that instead of
taking a sample from every cycle of the input waveform, here a
sample is only taken from every nth sample. So now, the
timescale on the oscilloscope screen does not represent, as in
ordinary real-time scopes, the actual or real time at which the
sample was taken (a little over 11 ns after the first sample), but
represents instead the time equivalent to the distance between
the two samples, had they been on one and the same signal cycle
(0.2 ns). The user of a sampling oscilloscope will not usually be
92 Oscilloscopes
aware of, nor want to know, how much real time elapses
between samples; he or she is only concerned with the timescale
of the reconstructed image. In practice, the sampling rate is
seldom as high as Figures 6.1 and 6.2 might suggest. Typically, the
sampling rate is around 100 ks/s (kilosamples per second). So the
500 MHz triangular wave shown in Figure 6.2 would in fact be
sampled on every 5000th cycle.
To briefly recapitulate. The c.r.t, display is built up of discrete
dots whose vertical positions correspond to the signal voltage at
the time of sampling and whose horizontal positions correspond
to the time delay between the beginning of the next sampled
waveshape and the moment when the sample is taken. To
recognize the beginning of the waveshape, the instrument uses
trigger circuitry much like that of a conventional oscilloscope.
The trigger circuitry will be preceded by a divider stage and
gating, to limit the sampling rate to 100ks/s or so. High speed
logic circuits such as ECL (emitter coupled logic) can cope up to
1 GHz or higher. For triggering from very high-frequency
signals, e.g. from several GHz upwards, the design may employ
a tunnel diode or similar specialized circuitry, providing 'trigger
countdown'.
signal
l I
trigger ~At
! ~ ~ ~
~t ~I
~t i~ a ~J~ ~I ~l
0.2 --~i4 ,nsl o , , , ' i '
real
, I ' , I ' ' I
....
time Im"""~"'lJ"lJl'"Jll"'l'l"l'l'liJ .... 1 .... l~,,,lh,,,l,,,,l,,,jll~,l,l,,,,l,u,,,l .... l,,i,l,,,,l,
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100105110ns
samples _~__-o- -_ _._
taken ~--4---~ e .... 4---~--- ~ ---i -----'~
oscilloscope display
[ ";'" 1
lee
o%
equivalent
time 0 1 2 ns
Figure 6.2 A sampling oscilloscope may take one sample per cycle of the input
as in Figure 6.1, or, more typically, one sample per n cycles of the input, as here.
In this example n = 5, but in practice n could equal, say, 50, 500, 5000 or any
larger number
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.