From: Mike Monett
X-Mailer: Mozilla 2.02 (Win16; I)
Subject: Re: Binary Sampler
References: <3E244AFE.4F3E@sneakemail.com> <3E25A881.firstname.lastname@example.org> <3E25C839.77BB@sneakemail.com>
Date: Thu, 16 Jan 2003 12:22:50 -0500
NNTP-Posting-Date: Thu, 16 Jan 2003 12:22:16 EST
Organization: Bell Sympatico
John Larkin wrote:
> Special case. Not all noise is Gaussian.
Johnson and transistor noise are impossible to avoid, and present in
any signal that pertains to sampling.
>> Overall, the binary sampler represents a significant improvement
>> in cost, accuracy, and measurement time needed to capture usable
>> waveforms in poor S/N conditions.
> Humbly disagree. Getting one bit per sample isn't as efficient, or
> as linear, as getting 11 or so.
The 11 bit spec applies to dc, not rf. The typical sample rate of HP
and TEK samplers is 100KHz. This gives 1.1MBit/sec, regardless of
the actual signal frequency.
The binary sampler samples at the signal frequency. A 1 GHz signal
gives 1 gigabit/sec, which is a thousand times better. The situation
improves with frequency.
>> It does not suffer from the 3% to 5% ringing and overshoot in
>> typical sampling systems
> Whether a sampler rings or not is an implementation/layout issue.
> If a slideback sampler doesn't ring, it's probably because it's
Every conventional sampler has aberrations that are an unavoidable
part of the sampling process. PSPL shows pictures of ringing,
overshoot, dribble-up and dribble-down in their "Comparison of
Ultra-Fast Rise Sampling Oscilloscopes", an-2d.pdf (284 kB)
an-2c.pdf (284 kB) an-2b.pdf (171 kB) an-2a.pdf (947 kB) at
>> and can be made with parts readily available to anyone, not just
>> HP and Tektronix.
> A few-GHz dual-diode analog-feedback sampler can be made with
> about ten dollars worth of distributor-stock parts.
An example is one by Hubert Houtman, "1-GHz Sampling Oscilloscope",
Unfortunately, the technology doesn't scale well. 5 GHz would be
extremely difficult, 10 GHz would be impossible.
Contrast that with the binary sampler uisng off-the-shelf Motorola
GigaComm parts at 10GHz.
Another problem is the delay generator. High frequencies demand low
jitter. As you point out,
"My company makes small 'embedded' DDG boards with resolution in
the 10s of ps. If you don't mind being clock-synchronous, some
FPGA counters are all you need. If you can't stand the time
quantization and/or need picosecond resolution and jitter, this
becomes - trust me! - serious work."
The binary sampler is not limited to the 100KHz sampling rate of
conventional scopes. Instead, it samples at the signal frequency
using the heterodyne method, and does not even need a trigger if the
signal frequency is known. As --Mike-- has shown, low-jitter is much
easier to achieve in synthesizers than in variable delay generators.
> The slideback sampler is has been reinvented (and mostly
> abandoned) many times in the last 40 years or so. If you're going
> to go to the trouble to make a wideband sampling scope, it's not
> much more trouble or expense to build a true linear sampler... and
> it's a lot faster. More fun, too!
Yes, especially in the timebase.
> Granted, this idea is cute - I had a brief fling with it myself in
Yes, I seem to recall an app note you came across. We never could
find a copy.