When Your Sample Rate is Much Faster than Your Edge Rate...

If your signal has a slow edge relative to the sample rate (meaning the 20-80% risetime of the signal is 6-8 or more sample intervals long), then over-scaling the signal and using linear interpolation is a good technique for minimizing front-end noise and quantization error in jitter measurements.

However, there are certain rules you must follow:

a) You must capture at least one voltage point above and one voltage point below the threshold level on every single edge that are not over-scaled. In practical terms, this means that, with the threshold level exactly in the middle of the screen vertically, the edge must be not overdriven for a period of at least two full sample intervals. For example, if you are sampling every 50 ps, then you must be able to see more than 100ps of the edge on the scope's screen.

b) You must understand how the display scaling and the A/D scaling interact on your scope. Some scopes, such as the HP54720, have a variable-gain preamp and therefore the display scaling is exactly the same as the A/D converter scaling. That is, A/D code 0 is always at the bottom of the scope graticle and A/D code 255 (assuming an 8 bit A/D converter) is always at the top of the scope graticle. Most scopes do not have a variable-gain preamp, and so the scaling seen on the screen will be different from the scaling in the A/D converter unless you are at a standard scaling value. In general this means that A/D converter codes 0 & 255 are off-screen rather than at the edges of the screen. If your scope does not have a variable gain preamp, for simplicity, we would recommend sticking to the standard values if overdriving.

c) As edge speed increases, the amount by which you would be able to over-scale decreases and M1's Time Optimal (TO) filter run on a non-over-scaled signal will become superior to using the linear filter. Exactly where this occurs can depend on many factors, but as a rule of thumb, if you have 3-4 or fewer sample points per edge, we would recommend using the TO filter on a non-over-scaled signal.

 d) Since the TO filter looks at data points quite a ways upstream and downstream of the actual threshold crossing point, over-scaling the signal as discussed above presents a violation of Nyquist. As a result of the over-scaling, you have introduced a sharp corner into the signal, which will generally contain energy well above the Nyquist limit. This will create aliasing errors when the data is passed through the TO reconstruction filter.

Note: Currently M1™Oscilloscope Tools supports only the Time Optimal filter. Older versions of M1™ allowed the user to choose the type of filter that M1™ used.


Learn More

Download A 7-Day Demo of M1 Oscilloscope Tools

About ASA

Press Contact

+1-413-596-5354  voice
+1-413-596-9686  fax
info@amherst-systems.com

M1™, M1 Oscilloscope Tools™ and SEEj™ are trademarks of ASA Corp of Amherst, MA, USA.