Masking

Would it be possible to allow the user to graphically enter a waveform mask and trigger when the signal waveform exceeds the masked bounds? This would be useful for triggering on spikes and drop-outs, distorted waveforms, etc.

Would it be possible to allow the user to graphically enter a waveform mask and trigger when the signal waveform exceeds the masked bounds? This would be useful for triggering on spikes and drop-outs, distorted waveforms, etc.

Dear Idlewise,
it would be very easy to include triggering based on going outside a range (min to max) or entering a range (min to max), or entering then leaving the range in time which might be less than, in a time duration, or greater than a time duration. Do you think this would achieve what you want?

Dear Idlewise, it would be very easy to include triggering based on going outside a range (min to max) or entering a range (min to max), or entering then leaving the range in time which might be less than, in a time duration, or greater than a time duration. Do you think this would achieve what you want?

I was thinking more of using the Windows GUI and let the user draw with the mouse a mask that the single should not intrude into, either above, below, or above and below the nominal signal. Consider a nominal square wave with distortion: the user could draw a mask such that jitter, over- and under- shoot, as well as noise and ripple could all cause a trigger, whereas a signal within the nominal waveform window would not.

I was thinking more of using the Windows GUI and let the user draw with the mouse a mask that the single should not intrude into, either above, below, or above and below the nominal signal. Consider a nominal square wave with distortion: the user could draw a mask such that jitter, over- and under- shoot, as well as noise and ripple could all cause a trigger, whereas a signal within the nominal waveform window would not.

Hello Idlewise,
This is very tricky - especially at full rate (ie at 10ns intervals). I do not see us doing it any time soon.

Here are the steps you have to take:
1. Use triggered capture to capture one cycle of waveform, and stick it into a buffer. Start capturing the next cycle immediately.
2. Compare each suceeding sample in the captured waveform with the mask window. If outside the window signal a trigger.

Issues:
1. Small time variations between the mask and the actual waveform would mean that edges wouldn't be where you thought they should be. (We're talking 10ns here). So you need a way of normalizing the captured waveform so that it is the same duration as the mask. The amount of normalization would equate to the maximum slop in time you were prepared to accomodate (eg 3% would be entirely acceptable for RS232 waveforms). To do the normalization means manipulating the memory at the same rate as one cycle. If you were looking at USb symbols say, you'd have 83 ns. Not very long.
2. If you have a long cycle time (eg 9600 baud RS232) then some form of time compression is required, otherwise the calculation and normalisation time would be very large (eg your'e looking for glitches on RS232. Conceivably they could be 10ns glitches. Each symbol for RS232 takes 937.5 usecs (for an 8 bit byte with start character). That's 94k samples you have to manipulate. Quite tricky to do, while data is still coming at you at 100 MSa/sec for the next symbol.

So sorry, unless you can think of a better way, this is something that will have to wait.

Hello Idlewise, This is very tricky - especially at full rate (ie at 10ns intervals). I do not see us doing it any time soon. Here are the steps you have to take: 1. Use triggered capture to capture one cycle of waveform, and stick it into a buffer. Start capturing the next cycle immediately. 2. Compare each suceeding sample in the captured waveform with the mask window. If outside the window signal a trigger. Issues: 1. Small time variations between the mask and the actual waveform would mean that edges wouldn't be where you thought they should be. (We're talking 10ns here). So you need a way of normalizing the captured waveform so that it is the same duration as the mask. The amount of normalization would equate to the maximum slop in time you were prepared to accomodate (eg 3% would be entirely acceptable for RS232 waveforms). To do the normalization means manipulating the memory at the same rate as one cycle. If you were looking at USb symbols say, you'd have 83 ns. Not very long. 2. If you have a long cycle time (eg 9600 baud RS232) then some form of time compression is required, otherwise the calculation and normalisation time would be very large (eg your'e looking for glitches on RS232. Conceivably they could be 10ns glitches. Each symbol for RS232 takes 937.5 usecs (for an 8 bit byte with start character). That's 94k samples you have to manipulate. Quite tricky to do, while data is still coming at you at 100 MSa/sec for the next symbol. So sorry, unless you can think of a better way, this is something that will have to wait.