which oscilloscope?

Then you may be better with TWO scopes. Get a good, cheap (second hand?) analog one with as many MHz as you can afford, and then get a Digital Multi-channel one, with some analog features, but 25MHz digital rates are not great. [ie 100MHz-200Mhz Digital/Analog with a 350MHz analog model, is likely to be cheaper than a 350Mhz bells-and-whistles model.] Plus, you have TWO instruments, which removes a lab bottle neck.

Use the Analog one once per PCB design to verify clean signals, and the Digital features you can expect to use during development.

The better digital-leaning ones have more digital channels, and they can also store EDGES-by-time, rather than have a fixed sample rate. That means at 25MHz, you can define each edge to 5 or less ns, and catch rare events. It is relatively rare these days, to have many digital tracks all over a PCB - that tends to all swallow into one FPGA+uC, but you can expect to trouble shoot signals between boards, and things like protocal margins.

-jg

w

lly, we

We have a few very high end LeCroys for deep memories and a bunch of low end Tek's for everyday work, but they really skip on memory, for some reason they charge $1000's more for a decent size memory, no one here likes the Agilent's

Good idea. Thanks. If my highest-frequency digital signal is 25 MHz, do you think a 100 MHz digital scope is sufficient? What is a digital-learning scope?

Please do not top-post. You have lost all connection with the previous articles. Your answer belongs after (or intermixed with) the quoted material to which you reply, after snipping all irrelevant material. See the following links:

(taming google)

--
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: 
            Try the download section.


** Posted from http://www.teranews.com **

I'd see if you can find a 100msps+ Analog, with 200Msps+ Digital (5ns), that should be at the 'bottom end', as that's quite do-able with mainstream FPGA and ADC parts.

The Digital Model can also be a PC based one (it's all pixels), whilst the analog model, clearly is not.

As an example, this is one I use :

[- but the edge-storage feature is still coming as SW upgrade...]

-jg

For serial protocols a logic analyser will be much for useful, and have a

*MUCH* deeper memory. You can get good value multichannel PC based analysers for cheaper than a decent scope.

The analyser will either sample the (digital) waveform with an internal or external clock, much like a scope, or, more usefully only timestamp the waveform transitions. This allows best use to be made of the devices memory.

Yes.

John

Ca you recommend a manufacturer of PC based logic analyzers ? (surely I can do a search, but it is better to hear from a person who knows.)

Personally I almost exclusively do FPGA design and the company I work for has the cash to supply professional analysers. Currently I use an Agilent

16900 series which as configured is about $30,000 worth - however - must of the debugging tasks I encounter could easily be achieved with a much simpler PC based alternative.

You really need to look at your memory depth (bearing in mind the analyser only needs to record edge locations in the digital domain), sample rate (generally not a problem if you synchronise the sampling with your system clock via an external clock input), and finally the number of channels you need to observe. From what you say 8 channels would be more than sufficient, which is really the bottom end of LSA capabilities.

Have a look at:-

You really need to decide your requirements, but from what you have said so far you need to spend nothing like the $7000 you were quoting.

HTH

Icky

I was trying to find one test equipment unit that will do everything I need. Now I am leaning more towards an analog oscilloscope, to watch my analog signals, and a PC based logic analyzer. I looked at the links you gave me. That is very good. Thanks.

To throw a different tack on the subject: what probes do you use?

When trying to find glitches, faulty wave shapes and such in fast cpu busses, I find the probes being the bottle neck rather than the scope BW.

I found myself creating probes based on e.g.

...allowing me to see things on a lowly 100 MHz scope that would not be revealed on any high grade scope with (good quality) 1/10 or 1/1 probes.

SW sequence related issues are usually solved by clever (written by SW engineers, not by me :-) SW routines and debug ports. (never needed a logic analyser).

If everything is OK, but the system stops for no reason, the scope + appropriate probe comes out. YMMV - so far I get away with it :-)

For regular digital/embedded stuff, on a 100 MHz scope, any probe will work. When the fuzzies start to matter, a good fet probe is magical.

The way-out extremes are the Tek SD-14 sampling probe, 3 GHz bandwidth and something like 0.3 pF at the tip. And the HP 54006A passive probe,

6 GHz bandwidth.

Me neither. The best logic analyzer is in your head; re-read the source code.

Last week I had a first-article product where the uP wouldn't run code. Two hours of frustration condensed into one sentence: one of the data lines must be pulled down by \RESET to tell it to boot in 8-bit mode, and the schottky diode that does that had a bad solder joint, and probing the most handy node, the diode lead, fixed the problem temporarily, and the diode is *under* the eprom.

John

I disagree. The human mind is a terrible bottleneck! ;-)

I've spent many unproductive hours debugging, I should know. (But it's those five highly productive minutes at the end that really matter...)

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

You could check out Rigol or Insteck. I think that they are the underlying manufacturers for Agilent and Tektronix.

Depends on the hardware on hand, I have found (not the easy way, I might add :-), that the act of putting a std probe on a wayward adress / data line, changes the conditions enough to make the measurement invalid. Sometimes the fault does not occur with the probe in place (which in it self tells you something), sometimes the device stops working completely.

Or all seems OK. But when the probe is removed, the system crashes after 5 minutes..

I had good results with the 1K resistive low cap probe, showing artefacts

"Talal Itani" wrote in news:t317k.112$cv5.70@trnddc01:

Bring in a sales rep to demo different lines. Tek might not be the best choice, but it is safe, in that when it breaks, nobody can say "Why didn't you buy a Tek?" The warranty is pretty good, too.

--
Scott
Reverse name to reply

For that, your best tool is a protocol analyzer. If you want to use a hardware-based protocol analyzer, consider renting one until your project is up and running.

As an alternative, there are software protocol analyzers all over the net. Google, and you'll find more than you need.

Good luck!

Tom

You can buy used ones pretty cheap, might even be cheaper than renting:

FWIW, if you're going the used route, my all-time favorite protocol analyzer is just about anything from the HP4951 family. Very capable, small and light as those things went back in the day, and simple to use. Suitable for both field and engineering use.

Tom