Instrumentation & Measurement Magazine 24-4 - 11

A Review of Streamline
Calibration Approaches for
Digital Storage Oscilloscopes with
Time-Interleaved Channels
Francesco Bonavolontà, Mauro D'Arco, Egidio De Benedetto,
Dominique Dallet, and Annarita Tedesco
T
he digital storage oscilloscope (DSO) has a vital role
in measurement practice, since its resourcefulness
fits both general purpose aims and requirements
of more specific applications. A deep knowledge of DSO operation
principles and available features is fundamental for
technicians to best exploit the possibilities of the instrument.
Operation principles regard digitization, acquisition, processing
and visualization of the input waveform and concern
a number of issues. For performant DSO digitizing systems,
one of the most relevant issues is the calibration of the timeinterleaved
structure at the very heart of the instrument.
Unfortunately, the calibration techniques adopted by the manufacturers
are never disclosed in plain form with application
notes or demos, and limited information can be read from patents,
if available. In this article, the state-of-the-art of DSO
technologies is reviewed; a general framework to face the calibration
issue in time-interleaved systems, which is based on
the generalized sampling theorem, is presented; and the case
of some DSOs, which require special calibration approaches,
implemented through the analysis of the input-output crosscorrelation
of the individual acquisition channels, is discussed.
Concluding remarks underline the value of the contribution,
that provides a significant example of how more ideal acquisition
channels can be realized by the combination of analog
circuitry with digital processors.
State-of-the-art DSO Technologies in a
Nutshell
DSO technologies have evolved and differentiated in a variety
of alternative solutions that permit, in different ways,
bandwidths in excess of 100 GHz and sampling rates up to
250 GS/s. All utilize time-interleaving (TI) and cope with very
large bandwidth requirements by exploiting either digital
bandwidth interleaving (DBI) [1], [2] or asynchronous time interleaving
(ATI) approaches [3], [4].
The systems based on TI approaches are hardware intensive
and represent the brute-force solution to improve
bandwidth and sampling rate. They include M analog-to-digital
converters (ADCs) that acquire the input signal at the same
rate, with a time-offset equal to 1/M the acquisition period, so
June 2021
that the aggregate throughput is M times the sample rate of
the individual ADCs. The analog bandwidth of the system is
determined by an M-way track-and-hold amplifier, deployed
in the front-end of the system. The track-and-hold uncouples
the ADCs between each other and prevents dynamic errors. TI
technology is characterized by a bulky use of hardware including
several digital signal processors aimed at improving and
calibrating the DSO response. Nonetheless, its real-time bandwidth
performance is, at present, limited to at most 35 GHz,
that is far beyond the bandwidth offered by DBI and ATI.
DBI systems divide the input signal spectrum into a baseband
portion and one or more high frequency portions by
means of analog filters. The high frequency contents are then
down-converted via heterodyning and acquired separately
from the low frequency content. A final representation of the
input signal is gained thanks to digital signal processing,
which consists in equalization, digital up-conversion of the
high frequency portions to the original frequency band and recombination
with the low frequency portion. Differently from
TI architectures that use M identical channels, DBI technologies
exploit asymmetrical channels to deal with the low and
high frequency contents of the input signal. Also, they require
massive processing to accommodate gain and relative phase
differences, especially in the band where adjacent portions of
the frequency contents are joined.
ATI architectures are made up of two symmetrical time-interleaved
channels, each of which includes a sampler, a low
pass filter and an ADC to acquire bandlimited versions of
the input signal. The samplers operate in a time interleaved
mode, asynchronously from the ADCs, and allow folding the
high frequency band upon the low frequency one. The bandwidth
requirement for the ADCs, in charge of digitizing the
output of the samplers passed through the low pass filters, is
thus halved. The successive processing allows distinguishing
the low frequency band from the high frequency one, upconverting
the latter to the original frequency position, and
recombining them in a digital representation of the input signal.
ATI solutions are characterized by compact architectures
and low noise. In fact, processing the data streams, separately
acquired on two independent channels, allows averaging the
IEEE Instrumentation & Measurement Magazine
1094-6969/21/$25.00©2021IEEE
11
Instrumentation & Measurement Magazine 24-4

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 24-4
