Normally, a simulator is adequate to estabish confidence in what the hardware is doing, first, of course, because I verify modules within a circuit, and don't attach the LA unless the results I'm getting, as observed with the oscilloscope, are divergent from what the simulator predicts. When numerous observable signals are involved, the LA often makes it possible to isolate a fault, usually in timing, and "skull it out" and observe the details with the oscilloscope.

However, in order for the LA to be of any significant benefit, it must be able to sample at minimally twice the rate of the fastest signal I intend to observe. If you use a 100 MHz MCU, then the LA must sample at minimally 200 MHz. Sampling at a lower rate will make the signals "look strange" in that the transitions will sometimes look distorted, or, worst case, will be missed altogether. Most often, I rely on my oscilloscopes to track suspect signals. Now I can attach up to eight 'scope channels since I have two high-bandwidth analog 'scopes that can be synchronized. They aren't sampling, and they do reflect the signal timing as it occurs, and not distorted by a sample clock, as can happen with older digital 'scopes. Most of the time, when I'm "checking out" a function, I use a 250 MHz analog 'scope, with dual traces, mainly since I'm partial to manually controlled instruments, as opposed to computer-controlled types, since they require that I push buttons and turn knobs, both of which are generally "context-sensitive" while with the '70's-generation 'scope, I know which knob or button is which, just by "feel". With the totally computer-controlled LA, most of my attention is on the display, and the setup takes as long as the analysis, which can be quite some time. I use the 'scope to tell me where I should look with the LA, and then interpret the resulting "picture" to tell me what to examine closely with the 'scope.

Most convenient, of course, is when the sample clock is at an integral multiple of that fastest clock, provided the circuit is synchronous. If there are various asynchronous terms, things don't always look like what's on the 'scope, nor is it likely to look exactly like what the simulator shows you. It's also important to understand the difference between bandwidth, as associated with an oscilloscope, and sample rate of a logic analyzer, will create some confusion until one thoroughly understands the relationship between an asynchronous signal and a discretely sampling LA.

Using the logic analyzer with programmable logic and MCU's with internal program store is less easy without a high quality simulator, which is seldom available for low-level debugging and fault isolation. Unfortunately, the majority of simulators are concerned with step-by-step simulation of code, without concern for the sequence of physical event brought about by the execution of code. I view the MCU as "just another PLD" while most people view it as a computer. To me its just another piece of hardware. As a consequence, the way in which I use them can differ widely from what most people expect. I have more than one logic analyzer, one being portable and the other not. The latter is capable of creating stimuli to the circuit under test, at a substantial rate, and, sometimes coupled to the operation of the circuit under observation and sometimes not. In its usual configuration, it's capable of displaying many signals, at a pretty high sample rate, and it even has a couple of digital oscilloscope channels that operate at 500 MHz bandwidth (from a 2GHz sample rate). Both of them are pretty old, by now, but that's largely because it takes a while to learn to exploit all the features available. I suspect that a 150 MHz digital 'scope may somewhat limit the things that can be observed. That may not be a problem with a standard rate MCU, but anyone who's worked with faster MCU's can tell you that you can do more with more speed.

RE