120 Oscilloscopes
same as previously, but shunted to the
left
(on the screen) by one
sample position. The allocated RAM (random access memory)
storage locations are as at B, or after the next sample, as at C, etc.
After 512 samples, the original right-hand edge of the trace will
have walked across to the middle of the screen, while the left-
hand half of the original trace will have disappeared for ever.
Alternatively, a display 'window' 10.24 seconds long can be
considered as advancing along the waveform (see D in Figure
7.2). The important point to bear in mind is that each time the
trace is written to the screen of the c.r.t, the samples stored in all
1024 locations are displayed; starting at the left of the screen with
the oldest sample and finishing at the right with the sample just
taken.
Figure 7.2 shows at J the channel memory redrawn as a
circular track, with the store input (write) switch and store
readout switch of Figure 7.1 drawn like the hands of a clock. The
write switch will usually rotate continually at a constant speed.
The read switch will in all probability rotate at a quite different
speed and, as we have seen, not necessarily at a constant speed at
all. Each 'hand' would typically be an eight bit wide data bus: in
the case of a memory consisting of dual port RAM this would be
quite a good analogy. However, dual port RAM is expensive and
in practice ordinary CMOS, NMOS or ECL static RAM, or in the
lower-priced instruments, dynamic RAM, is used instead. This
has a single read/write data port, which is switched between the
two functions by an R/W control line. With the slow data rate in
the example just given, there would be no difficulty at all in
interleaving the write and read functions, even with relatively
slow, cheap, dynamic RAM. If the sample rate were slower than
the 100 s/s considered above, the screen trace would be rewritten
twice or more between each sample, to maintain a high enough
screen refresh rate to aw~id flicker. On the other hand, at much
higher sample rates, several or many samples could be taken
before rewriting to the screen.
Returning to the waveform in Figure 7.2, it is clear that at any
time there is a record of the last 10 seconds of the waveform in
store. This information can be frozen at any time if an event of
particular interest occurs- such as a dangerously high stress in the
Digital storage oscilloscopes 121
bridge due to an overloaded lorry in our fictional example. We can
set the DSO's trigger circuitry so that if the Y input voltage exceeds
a certain level, the sampling action is halted- the write hand in
Figure 7.2 J ceases to rotate. Furthermore, although the read hand
continues to rotate, thus continually displaying the stored trace on
the screen, since the trace displayed always starts at the left of the
screen with the oldest sample last taken, the trace is now
stationary. Like the flight data recorder in a crashed plane, the
trigger event has
terminated
the recording of data, rather than
initiating
it like the trigger circuitry in a conventional real-time
scope. Imagine the flight data recorder uses a loop of tape holding
data on just the last ten minutes of any flight, and the analogy is
perfect. This type of operation is known as 100 per cent pre-trigger
store and is illustrated in Figure 7.2 E. In the flight recorder
example, the trigger event was effectively the end of the world, but
in our DSO, we can arrange the circuitry to take some samples
after the trigger event before terminating the sampling process.
Another 512 samples, as in Figure 7.2 E will lose the oldest 50 per
cent of the pre-trigger information but store five seconds worth of
the waveform post-trigger. By suitable settings of the controls we
can, in principle, have any split we want between pre- and post-
trigger information, or set an even greater delay in terminating
sampling, as in Figure 7.2 E to H. In practice, DSOs usually offer the
choice of a small number of settings such as 100 per cent, 75 per
cent, 50 per cent and 25 per cent pre-trigger storage, while only the
more expensive instruments provide delayed (greater than 100
per cent post-trigger) storage.
We may still wish to capture an event which triggers the scope,
but with greater time resolution than provided by the 100 sis in
the roll mode example above. But at 100 ks/s, say, giving a time
resolution of 10 ~s in the stored trace, the waveform would be
rushing across the screen so fast as to present a meaningless
jumble to the observer. In this case, triggered storage mode, also
known as single sweep or single shot, is more appropriate. The
DSO operates in exactly the same way as in roll mode except that
the waveform being acquired is not displayed until the trigger
event stops the sampling clock. An 'armed' indicator is often
provided- this is illuminated to indicate that the scope is
122 Oscilloscopes
Figure 7.3 The Hewlett-Packard Infiniium range of oscilloscopes includes this
model 54845A, with its 1.5 GHz bandwidth. It samples at 4 Gs/s simultaneously
on all four input channels, or at 8Gs/s in two channel mode (reproduced by
courtesy of Agilent Technologies, the new name of Hewlett-Packard Measure-
ments Division)
continously acquiring the input signal. When the trigger event
occurs, the armed light goes out and the sampling clock is
stopped, either immediately or when the desired amount of post-
trigger information has been stored - a 'stored' indicator (if
provided) then lights. A reset or release button is provided to
rearm the system, ready to stop on the occurrence of another
trigger event. Operation is very similar to that of the single shot
mode in a conventional scope with camera or an analogue (tube)
storage oscilloscope, with the big difference that with these one
cannot capture pre-trigger information.
Refresh mode
It was mentioned earlier that when the sample rate (the
equivalent of timebase speed in ordinary scope parlance)
becomes too high, the display in roll mode is no longer useful. An
alternative to single shot operation in this case is
refreshed
or
recurrent
mode; unfortunately the terminology relating to this
mode, as with other modes and functions of DSOs, varies from
Digital storage oscilloscopes 123
manufacturer to manufacturer. This mode is particularly useful
when the waveform of interest is repetitive, or very nearly so.
With it, the DSO produces a stable, triggered display looking very
like the display on an analogue scope. The waveform is, however,
still being acquired continuously, so that whenever sampling is
stopped, a segment of the waveform preceding that instant is held
in store.
Of course it is unlikely that the screen display will correspond
exactly to an integral number of cycles of the input waveform
(see Figure 7.4(a)), where the screen is shown as displaying
about 1~ cycles. So here, half a cycle or so is not being
displayed each time the trace is written on the screen. This
seems to contradict the earlier statement that the input signal
is being acquired continuously. But acquisition and display are
f
trigger
level
7-1
b c
(a)
C
increasing
~
acquisition store
(b)
Figure 7.4 Refresh mode
1023 0
display store
124 Oscilloscopes
not the same thing. The signal is acquired continuously in a
'cyclical' acquisition memory like that shown in Figure 7.2 J but
the display is fed from a separate display memory. This is
indicated in Figure 7.4, where (a) shows the display while (b)
shows how the memory transfers are organized. The digitized
waveform is fed continuously into the acquisition store so that
at point c the data overwrites that previously stored there at
time a- time has been drawn increasing as a spiral so that this
can be seen more clearly.
If the trigger circuit has been set to detect the positive-going
edge of the waveform at a, then each of the next 1024 samples
stored in the acquisition store will be read out again immediately
and be stored also in the display store. The display store then
stops accepting data from the acquisition store and retains a
snapshot from a to c of the recurrent waveform. The trigger
circuit would have detected the positive-going edge at b, but it
was ignored as there was already a 'sweep' in progress.
Now, the 'readout switch' transferring data out of the acquisi-
tion memory to the display store continues to 'rotate' in
sympathy with the acquisition store write switch, indeed (except
in the case of dual port RAM) they are the same thing- the
acquisition memory read/write bus.* So when the next trigger
event occurs at time d, the following 1024 samples are stored also
in the display store as previously. However, as the display store
address counter stopped clocking up after filling location 1023 on
the last 'sweep', the 1024 samples starting at position d in the
acquisition memory are stored in positions 0 to 1023 in the
display memory. Thus although the segment of waveform c to d
was not displayed, it
was acquired. A separate trigger at a higher
level may have been set to stop acquisition on, or shortly
following, a positive- or negative-going glitch. If such a glitch
occurred in an undisplayed portion of the waveform such as c-d,
it would duly appear on the frozen display when the last
*The acquisition memory read and write 'switches' can in fact be in different
positions, or even 'rotate' at different speeds as described in the section on roll
mode, simply by supplying different memory addresses (which may also be
incremented at different rates) depending upon whether the R/W line is at logic
1 (high) for a read or logic 0 (low) for a write cycle.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.