On Aug 17, 2015, at 7:29 PM, Dave G4UGM <dave.g4ugm
at gmail.com> wrote:
-----Original Message-----
From: cctalk [mailto:cctalk-bounces at
classiccmp.org] On Behalf Of
drlegendre .
Sent: 18 August 2015 00:15
To: General Discussion: On-Topic and Off-Topic Posts
<cctalk at classiccmp.org>
Subject: Re: test equipment / Re: Z80 / Z84C Swap (Doh!)
Hey Dave & All,
Could you give a little quick kick-start guide to bit depth & sampling rate on
DSOs? It's always kind of stumped me, not that I've ever read deeply into it..
but how is it that you can get any kind of (vertical, right?) resolution out of 8
or even 12-bit samples?
Well the 8-bits is around 5mv on a 1volt signal. It draws a thin line that goes up in
big(ish) steps. I am off to bed now but will upload some screen shots to my OneDrive so
you can see them.
It depends on the screen resolution. My Tek 602, a rather early (and very large/heavy)
DSO, does just fine with 8 bits because the screen is fairly small. Compared to the
screen resolution of a typical analog oscilloscope screen, 8 bits is a pretty good
choice.
You can occasionally find DSOs with higher resolution, but there?s a substantial penalty
in sample rate.
You need at least twice the sample rate as the maximum
frequency you want to plot, but really to be useful I recon 4 x . So the 6022 samples at
48M samples/second and has 20Mhz bandwidth amps so you probably get usefull displays up
10mhz (ish).
Nyquist rule: you need 2x the sample rate of the highest signal frequency. Or more
accurately, the bandwidth. That?s the same thing if you start at 0 Hz, but if you?re
sampling a piece of spectrum, you only need 2x the bandwidth of the band in question, not
2x the frequency of the upper edge. Note also that you have to consider filter skirts, so
you need 2x the frequency at which your input filter has ?high enough? attenuation.
To pick an extreme example, the current record holder DSO has 100 GHz (!) bandwidth and
240 Gs/s sampling rate.
>
> Example line of thought - 8 bit sample = 256 possible vertical positions.
> Even if the screen is low-end (640 x 480) that's almost 2X more height in
> pixels than samples in an 8-bit sample. So each increment is like 2 pixels tall
> and seems like it would be awfully blocky and imprecise. Things would seem
> to get even worse if you try to do maths functions..
>
> I must be viewing this quite wrong?
You would not display them as 2 pixels tall. Typically the display machinery will
interconnect the sample points by lines, and the result usually looks pretty convincingly
like an analog scope display (unless you crank the zoom way up). As for math, that
depends. If your math involves filtering, the effective bit count goes up (the
?processing gain? DSP people speak of).
paul