SERIES V850 DIGITAL DELAY GENERATOR
JITTER
The series V850/V850 VME digital delay generators are intended for use in precision timing measurement systems which often require very low levels of pulse-to-pulse jitter. Production V850's are specified to have RMS jitters of less than 50 picoseconds +- 1E-9 of delay, equivalent to 50 ps for short delays, increasing to 217 ps RMS at 167 milliseconds, the maximum available delay. In both engineering and production test, the DDGs are tested using a Hewlett-Packard model 5370B Time Interval Counter, about the lowest jitter time measuring instrument available.
During DDG testing, it was noted that indicated jitter seemed to be below 60 ps RMS over the entire DDG timing range, except for a peak jitter of 80-90 ps for delays around 4 milliseconds. Since nothing in the V850 design suggested a mechanism for this jitter peak, an experiment was planned to better characterize DDG jitter and to attempt to correct for the inherent measurement error of the HP5370.
The difficulty in measuring jitter over long timebases is that the measuring instrument (typically an oscilloscope or a high-resolution time-interval counter) will have its own internal timebase jitter, and this may well be worse than the instrument under test. A technique commonly used to measure frequency-domain phase noise is to use two presumably identical DUTs and measure the noise between them. A similar technique may be used to measure the longterm jitter of two delay generators by programming them to generate nearly identical delays and measuring the jitter of the resulting time jitter.
Two normal production-line V851 modules (6-channel delay generators) were configured for simultaneous external triggering, and the HP5370 was used to measure the DIFFERENCE in time between the two DDG outputs. The DDGs used their internal crystal clocks and were not phase locked to an external source or to each other. The trigger source was a non-synchronized pulse generator, Phillips PM5715 with a V858 SRD edge sharpener.
The DDGs were programmed for various delay times, as shown in the table below. In most cases, the second V851 was programmed to have a delay of 1 us longer that of the first, with the HP5370 used to measure the mean time difference and jitter of the 1 us nominal difference. The HP counter was independently determined to have a baseline jitter of about 27 picoseconds for delays in the range of zero to 1 us.
The table shows programmed times for the two V851's, measured mean difference, difference error, measured jitter, and computed V851 jitter. The V851 jitter was computed by removing the 27 ps HP jitter, then correcting for the fact that two DDGs were contributing to the jitter measurement. If H is measured (HP) jitter and V is estimated single-V850 jitter, we calculate V in picoseconds as...
V = 0.707 * sqrt(H^2 - 27^2)
It must be noted that the use of this computational method may produce serious errors when DDG jitter V is much below the HP error of 27 ps.
PROGRAMMED
DDG1 TIME
|
PROGRAMMED
DDG2 TIME
|
MEASURED
DELTA -T
|
MEASURED
dT ERROR
|
MEASURED
JITTER H
|
MEASURED
JITTER V
|
| 0 |
0 |
1.191 n |
1.191 n |
33 p |
13.4 p |
| 0 |
1 u |
1.001210 u |
1.210 n |
37 p |
17.9 p |
| 10 u |
11 u |
1.001075 u |
1.075 n |
41 p |
21.8 p |
| 100 u |
101 u |
1.001076 u |
1.076 n |
42 p |
22.7 p |
| 1.000 m |
1.001 m |
1.000921 u |
0.921 n |
42 p |
22.7 p |
| 4.000 m |
4.001 m |
1.000504 u |
0.504 n |
41 p |
21.8 p |
| 10.000 m |
10.001 m |
0.999616 u |
-0.384 n |
38 p |
18.9 p |
| 100.001 m |
100.001 m |
0.996035 u |
-3.965 n |
52 p |
31.4 p |
| 160.000 m |
160.001 m |
0.992343 u |
-7.657 n |
74 p |
48.7 p |
The data shows an initial delay difference of about 1 nanosecond between the DDGs, most of which can be attributed to the cable which daisy-chains the trigger from module 1 to module 2. At longer delays, the 1 usec expected time difference between the modules can be seen to steadily decrease; this is largely due to a difference in the DDGs clock oscillators of about 0.05 PPM, verified as actual. (These oscillators are new and have not been aged, so such drift is not unusual. No attempt was made to align crystal oscillator frequencies between the modules).
The calculated DDG jitters range from 13 picoseconds at zero delay, to 49 ps at the maximum delay of 0.16 seconds. The prieviously observed jitter excess at 4 milliseconds appears to be an artifact of the HP5370.
We conclude that typical V850-series VME delay generators have base jitters below 20 picoseconds RMS and remain below 50 ps over their entire delay range. Typical units are considerably better than the HP5370B which is used for production testing. |
