We are using the TMS320C6748BZWT3 DSP with a 1Gb DDR2 SDRAM (Elpida EDE1116AEBG-8E-F)
When measuring the DDR Clock at the DDR device we noticed that the clock jitter is higher than the requirements for the DDR (see Clock Jitter table on page 19 of the DDR datasheet).
The jitter is measured with the following setup:
- Oscilloscope : Tektronix DSA72504D (25GHz)
- Probe : Tektronix P7313 (12.5GHz)
- Running DDR Analysis Tool
Results from jitter measurement (fails marked red):
Description |
Requirement DDR 800 |
Mean |
Std Dev |
Max |
Min |
p-p |
Max-cc |
Min-cc |
|
|
Min |
Max |
|
|
|
|
|
|
|
tJIT(per), CK |
-100ps |
+100ps |
NA |
NA |
93.711ps |
-87.383ps |
NA |
NA |
NA |
tJIT(duty), CK |
-100ps |
+100ps |
NA |
NA |
61.827ps |
-58.027ps |
NA |
NA |
NA |
tJIT(cc), CK |
- |
200ps |
2.0161fs |
29.286ps |
138.88ps |
-124.12ps |
263.00ps |
222.40ps |
-231.36ps |
tERR(2per), CK |
-150ps |
+150ps |
NA |
NA |
120.22ps |
-145.35ps |
NA |
NA |
NA |
tERR(3per), CK |
-175ps |
+175ps |
NA |
NA |
163.73ps |
-175.78ps |
NA |
NA |
NA |
tERR(4per), CK |
-175ps |
+175ps |
NA |
NA |
210.30ps |
-192.25ps |
NA |
NA |
NA |
tERR(5per), CK |
-175ps |
+175ps |
NA |
NA |
246.67ps |
-242.30ps |
NA |
NA |
NA |
tERR(6-10per), CK |
-300ps |
+300ps |
NA |
NA |
448.78ps |
-427.11ps |
NA |
NA |
NA |
tCH(avg), CK |
0.48..0.52 tCK (avg) |
501.11mtCK(avg) |
106.05utCK(avg) |
501.41mtCK(avg) |
500.73mtCK(avg) |
682.69utCK(avg) |
37.818utCK(avg) |
-31.491utCK(avg) |
|
tCL(avg), CK |
0.48..0.52 tCK (avg) |
498.89mtCK(avg) |
106.05utCK(avg) |
499.27mtCK(avg) |
498.59mtCK(avg) |
682.69utCK(avg) |
31.491utCK(avg) |
-37.818utCK(avg) |
- DRAM Clock is 150MHz
- The input clock of the DSP (40MHz) has a period jitter of approx. 29ps.
- The noise on the DSP and DRAM supplies are all < 60mV (including the PLL0 and PLL1 supply)
Do you have any idea what can cause the jitter?