Building logic analyzer probes

The answer depends on a lot of stuff. Are you doing synchronous or asynchronous acquisition? How fast? Voltage levels, logic families, threshold voltages? How much circuit loading can you tolerate? Do you care if the data is correct?

Probe design is a big mass of problems. Signal fidelity/reflections/terminations/thresholds/sensitivity/noise immunity/cable repeatability/ channel skew Probe skew clock skew cable skew...cable length mismatches can be the biggest factor Pattern sensitive delays/crosstalks Ground loops between probes Input protection

Setup time DUT loading Crosstalk between channels Ground induced crosstalk at the source. Coupling crosstalk back into the DUT.

One rule of thumb: Take the inductance of the channel lead and the ground lead. Resonate that with the input capacitance. At that frequency, your probe is a dead short on your DUT. It ceases being a good probe well below that frequency. All that ground currrent is induced into your DUT and all the other acquistion channels. Now multiply all this by the number of channels you want to work reliably...

Making a probe that loads the circuit, blows up when you short it to VCC or static zap it and almost works sometimes isn't too hard. Making multiple probes that always give you the right answer is real hard. There's a reason why probes cost a LOT.

When you get fired, some administrator takes your logic analyzer and sends it to equipment disposal. The probes go to the wire box and are never seen again. Then the LA ends up on ebay with no probes and the buyer posts here. I recommend against. Cycle repeats...

Figger out what you're gonna do about probes BEFORE you buy an ebay LA. Finding probes without a matching LA is rare. Maybe buy a dead one with probes. mike Former logic analyzer designer.

If you're sending data, you have a few solutions. Use enough signal strength that the common mode noise doesn't bother you. Use differential signals with enough common mode compliance to ignore the ground loops. Or eliminate the path as you suggested with magnetic or optical coupling.

Now, what do you mean when you say crosstalk? If you're generating the signals that cause the crosstalk then it's easy...don't do that. If you use differential drivers and stuff the extra output into ground, you can often mitigate the ground loop problem by keeping the ground current independent of the logic state. But that's really a different issue from big noise on the power line.

Safety ground doesn't have to connect to signal ground. You can float the whole thing if you want. As I recall, all the safety guys care about is if you put big amps into a metal case, it goes out the safety ground. A relatively small resistance between safety ground and signal ground can make a lot of difference in induced noise.

Grounds are always an issue. I don't know of any magic formulae. mike