Digital O'scopes: Sampling Rate vs. Sample Storage Space

[[email protected]]

What is the relative importance of sampling rate versus sample memory
in a digital oscilloscope?

I'm looking at two candidates: one has a sampling rate of 2GS/sec, but
only 2.5K of sample memory, while the other samples at only 400MS/sec,
but has 1M of sample memory.

For debugging microcontroller systems with typical peripherals (I2C,
SPI, ADC, LCD), which is more important, sampling rate or sample
memory size?

 

[Rich Webb]

What is the relative importance of sampling rate versus sample memory
in a digital oscilloscope?

I'm looking at two candidates: one has a sampling rate of 2GS/sec, but
only 2.5K of sample memory, while the other samples at only 400MS/sec,
but has 1M of sample memory.

For debugging microcontroller systems with typical peripherals (I2C,
SPI, ADC, LCD), which is more important, sampling rate or sample
memory size?

For your stated purpose, I'd say that a better candidate would be a
logic analyzer. Whereas PC-based, USB-interfaced o'scopes tend to be a
bit underpowered, similar logic analyzers are often cost effective and a
good fit for a home/hobby lab. Hit Google for "usb logic analyzer" for a
long list of candidates. My personal favorite (which gets a lot of use
developing professional products as well) is the Intronix logic analyzer
from http://www.pctestinstruments.com/. This one includes "interpreters"
for RS-232, I2C, SPI, and CAN protocols; very handy.

For o'scopes, I'd recommend that you also look at the digital scopes
from Instek, which are pretty well thought of around here (I don't have
one but I am looking for an excuse...). One source is over at
http://www.tequipment.net/InstekGDS-2202.html

 

[Joerg]

For your stated purpose, I'd say that a better candidate would be a
logic analyzer. Whereas PC-based, USB-interfaced o'scopes tend to be a
bit underpowered, similar logic analyzers are often cost effective and a
good fit for a home/hobby lab. Hit Google for "usb logic analyzer" for a
long list of candidates. My personal favorite (which gets a lot of use
developing professional products as well) is the Intronix logic analyzer
from http://www.pctestinstruments.com/. This one includes "interpreters"
for RS-232, I2C, SPI, and CAN protocols; very handy.

A logic analyzer cannot easily detect marginal line conditions such as
overshoots, reflections, marginal levels, bus contentions. I'd go with a
scope.

For o'scopes, I'd recommend that you also look at the digital scopes
from Instek, which are pretty well thought of around here (I don't have
one but I am looking for an excuse...). One source is over at
http://www.tequipment.net/InstekGDS-2202.html

Yep, got a GDS-2204, happy with it. Bought it via Newark because they
had the best price:

http://www.newark.com/jsp/search/browse.jsp?N=500003+1001455&Ntk=gensearch_001&Ntt=Instek&Ntx=

Yesterday I had to work with one of those TI lunchbox-size scopes at a
client. Oh man, what a step back after being used to the Instek.
Especially when I hit the 2.5K brick wall at the end of its memory bank
where the Instek got 25K.

 

[Scott Seidman]

Yep, got a GDS-2204, happy with it. Bought it via Newark because they
had the best price:

I had to instrument a whole lab class. I was looking long and hard at
Instek, and if it were for my personal use, that's what I would have went
with. That said, because they (I ordered 18) will serve multiple people
and multiple classes, I went with Tektronix -- I didn't want to hear "Why
didn't you go with Tek" the first time one broke.

I did go w/ Instek for everything else-- bench power supplies, bench DMMs,
bench function generators. I'm very happy with the power supplies, and
actually prefer them to topwards because they have a button that disables
output. I'm a tad concerned about the function gens, but they are a whole
bunch cheaper than Tektronix, and will be less painful to replace when
students blow up the output stages.

 

[[email protected]]

I had to instrument a whole lab class.  I was looking long and hard at
Instek, and if it were for my personal use, that's what I would have went
with.  That said, because they (I ordered 18) will serve multiple people
and multiple classes, I went with Tektronix -- I didn't want to hear "Why
didn't you go with Tek" the first time one broke.

Is the 2.5K buffer on the Tek large enough to look at protocols such
as I2C and SPI?

 

[Rich Webb]

Is the 2.5K buffer on the Tek large enough to look at protocols such
as I2C and SPI?

Like a lot of things: It depends. In this case, it depends on what you
want to see when you "look at" them. You can certainly grab several
characters and check that your voltage levels and slew rates are okay.
That's a good and useful thing. It's even possible to do some counting
of 1s and 0s on the screen (try to avoid fingerprints) and decode a few
characters. Been there, done that, and it gets old really quickly --
especially when the characters are in a continuous stream.

 

[Joerg]

IMHO it's not enough. I had the "pleasure" to do SPI debug at a client
and all they had was a TDS-220. We had glitches in 64-bit packets and it
was a bear to eyeball small timing aberrations. The Instek GDS-2204 does
that with ease.

BTW I won't see responses from gmail until someone else answered and
quoted you. Best would be to use some better domain.

Like a lot of things: It depends. In this case, it depends on what you
want to see when you "look at" them. You can certainly grab several
characters and check that your voltage levels and slew rates are okay.
That's a good and useful thing. It's even possible to do some counting
of 1s and 0s on the screen (try to avoid fingerprints) ...

And _don't_ use a pen, probe tip, screwdriver or any of that either ;-)

... and decode a few
characters. Been there, done that, and it gets old really quickly --
especially when the characters are in a continuous stream.

There is a classic movie about a loner type guy who becomes a math
genius, can decipher all kinds of codes just by looking at the screen
and then goes completely crazy ...

 

[David L. Jones]

What is the relative importance of sampling rate versus sample memory
in a digital oscilloscope?

I'm looking at two candidates: one has a sampling rate of 2GS/sec, but
only 2.5K of sample memory, while the other samples at only 400MS/sec,
but has 1M of sample memory.

For debugging microcontroller systems with typical peripherals (I2C,
SPI, ADC, LCD), which is more important, sampling rate or sample
memory size?

By FAR the best option in this case is the 1M of sample memory.
2.5KB is not much use at all in analysing serial data packets.
1MB will allow you to capture and post analyse long and/or widely
dispersed packets in the one capture.
In this case 400MS/s will be more than enough sample rate for say
50MHz clock rate systems (analog bandwidth dependent of course).

You'll never ever regret getting a digital scope with a large sample
memory, but you'll often regret one that only has 2.5KB.

One could of course argue that a logic analyser might be more useful
here as an additional tool, but that's another ball game.

A mixed signal scope (analog + logic analyser) is also a superb
option. Rigol make some good low cost ones.

Dave.

 

[David L. Jones]

A logic analyzer cannot easily detect marginal line conditions such as
overshoots, reflections, marginal levels, bus contentions. I'd go with a
scope.

I'd second that.

Logic analysers are handy tools, but NOT as good a scope. You can't
beat seeing the true probed digital signal instead of a 0/1 that the
logic analyser thinks it sees.

Dave.

 

[David L. Jones]

Is the 2.5K buffer on the Tek large enough to look at protocols such
as I2C and SPI?

It's usable, but often not enough in my experience.

Dave.

 

[neon]

i DON'T LIKE EITHER ONE OF THEM. with either one of them the timing is artificial not real. a sampling scope is ok if everything is ok. Can you find a periodicly clictch that occurs at ramdom maybe. a storage scope is better only if you can tell when there is a problem and catch it. third choise analog scope then you may find your problem faster. Now here it is the problem at hi speed tektronik want your house and car fot it.

 

[Scott Seidman]

On the other hand, if you ask nicely, Tek might be willing to donate
or subsidize equipment for schools -- Oregon State University
(Tektronix is in Beaverton, Oregon) had almost all of their
undergraduate labs outfitted with Tek equipment through such means.
(And guess what? OSU has an undergraduate "robotics" program named...
TekBots!)

Let's just say there was very liberal educational pricing. You pretty much
start with a pretty good discounted price, and on top of that, one in four
scopes is free. I did spend some time trying to track down the right
person at Tek for a better deal, and couldn't find that person.

 

[Rich Webb]

Logic analysers are handy tools, but NOT as good a scope. You can't
beat seeing the true probed digital signal instead of a 0/1 that the
logic analyser thinks it sees.

No arguments here; o'scopes are necessary tools. On the other hand if
the threshold is set correctly then the 0/1 that the logic analyzer
thinks that it sees is the same 0/1 that the other digital devices think
that *they* see, and that can be a useful thing to see.

 

[Joerg]

Logic analysers are more for the "next level up" where you're confident that
electrically all the signals are OK and you're debugging some complex protocol
on some big bus. Many people never do anything complex (or buggy?) enough to
need one... and perhaps ironically, the need for general purpose logic
analyzers is decreasing as large multi-drop parallel busses have fallen out of
favor to lots of high-speed point-to-point serial busses all flowing into
switch/routing ICs.

Yep. I have to turn mine on regularly just to replenish its backup
battery. It even went from the bench to a storage area.

 

[JosephKK]

I'd second that.

Logic analysers are handy tools, but NOT as good a scope. You can't
beat seeing the true probed digital signal instead of a 0/1 that the
logic analyser thinks it sees.

Dave.

It is not that simple. Capturing 32, 64 or more channels
simultaneously at speed with decent record lengths can tell you things
about software - hardware interaction that a scope never will.
Protocol analyzers and bus analyzers also have their place.

 

[David L. Jones]

It is not that simple. Capturing 32, 64 or more channels
simultaneously at speed with decent record lengths can tell you things
about software - hardware interaction that a scope never will.
Protocol analyzers and bus analyzers also have their place.

I was of course talking in electrical (signal integrity) terms, not
multi-channel. Logic analysers can be a big trap for young players.

Of course a logic analyser comes into it's own when you want/need to
view a large number of channels and/or need complex data triggering
options. That's why you can still buy'em.

High end logic analysers these days allow both digital and analog
capture on the same probe, e.g. Tektronix iCapture/iView
This is the best of both worlds, real techno-porn.

Dave.