Text preview for : DDR4 TdiVW VdiVW Bit Error Rate Measurement or Understanding Bit Error Rate 5991-1761EN c20140918 [5 part of Agilent DDR4 TdiVW VdiVW Bit Error Rate Measurement or Understanding Bit Error Rate 5991-1761EN c20140918 [5 Agilent DDR4 TdiVW VdiVW Bit Error Rate Measurement or Understanding Bit Error Rate 5991-1761EN c20140918 [5].pdf
Back to : DDR4 TdiVW VdiVW Bit Erro | Home
Keysight Technologies
DDR4 TdiVW/VdiVW Bit Error Rate
Measurements or Understanding
Bit Error Rate Measurements in DDR
Application Note
02 | Keysight | DDR4 TdiVW/VdiVW Bit Error Rate Measurements or Understanding Bit Error Rate Measurements in DDR - Application Note
Historically, DDR (double data rate) has defined its timing specifications with a
belief of a zero bit error rate (BER). While a zero bit error rate is statistically not
possible, timing budgets had enough margin to justify the method of specifica-
tion and measurement. With each generation of DDR Synchronous Dynamic
Random Access Memory (SDRAM), speeds increase, package sizes decrease,
and power consumption decreases. (See Table 1). Added challenges come with
these improvements of decreased design margins, signal integrity, and interop-
erability. Latest DDR technology offers data rates of 3.2Gb/s or higher. Each
picosecond now matters and can be the difference in passing and failing bits.
At these high data rates, BER measurements are important in order to under-
stand the true reliability of a system. Noise and jitter affect the signal integrity
and its overall reliability. Measuring and understanding the components of noise
and jitter can enable designers to minimize them on the overall system design
to ensure valid and accurate data transfer. In this paper, we are going to take
a look at the measurements of the timing data input valid window (TdiVW) and
the voltage data input valid window (VidVW).
Table 1. DDR technologies and key JEDEC specifications
DDR standard DDR LPDDR or DDR2 LPDDR2 or DDR3 LPDDR3 or DDR4
mobile-DDR mobile-DDR2 mobile-DDR3
Specification JESD79E JESD209 JESD79-2E, JESD209-2B JESD79-3C JESD209-3 JESD79-4
JESD208
Operating voltage 1.5 - 3.3 V 1.8 V 1.8 V 1.6 B 1.5 V 1.2 V 1.2 V
Clock frequency 100 - 200 MHz 100 - 200 MHz 200 - 400 MHz 100 - 533 MHz 400 - 800 MHz 667 - 800 MHz 800 - 1600 MHz
Data transfer rate 200 - 400 MT/s 200 - 400 MT/s 400 - 800 MT/s 200 - 1066 MT/s 800 - 1600 MT/s 1333 - 1600 MT/s 1600 - 3200 MT/s
Package type This Small Fin Ball-Grid Fin Ball-Grid Fin Ball-Grid Fine Ball-Grid POP Fine Ball-Grid
Outline Package Array (FBGA) Array (FBGA) Array (FBGA)/ Array (FBGA) Array (FBGA)
(TSOP) POP
Package size x4, x8, x16, x32 x16, x32 x4, x8, x16 x16, x32 x4, x8, x16 x16, x32 x4, x8, x16
Backward No Yes, with DDR No Yes, with DDR2 No No No
compatibility
03 | Keysight | DDR4 TdiVW/VdiVW Bit Error Rate Measurements or Understanding Bit Error Rate Measurements in DDR - Application Note
Impact of Noise and Sources of jitter in a system include inter-symbol interference (ISI), cross talk,
and duty cycle distortion (DCD) and can limit the performance and the interface
Jitter on Data Transfer between the memory controller and the DRAM.
Jitter is defined as the deviation of a signal transition from its ideal time. As data
rates increase, timing budgets decrease and each pico second of margin gained
becomes more important. A small amount of jitter can easily close the data valid
window of the read or write data at high speed data rates. This will ultimately
increase bit error rate and data sampling error.
When operating at a high data rate, the data valid window is very small. The
noise in the system or crosstalk from adjacent electrical signals, or even elec-
tromagnetic interference could easily distort the signal integrity which could
result in data sampling error. Real-time eye diagram provides eye height and
eye width measurements for signal integrity checks and estimates the data valid
window. However, simply measuring a data eye with real time eye diagram mea-
surement does not give full insight into the data valid window and expectations
of a bit failure rate. Figure 1 shows a write data eye with over 100,000 unit inter-
vals (UI) measured. This initial measurement of over 100,000 UI would result in a
measured TdiVW of 353.75ps (or a margin of 56.25ps to the defined mask) and
a VdiVW of 377mv (or a margin of 71.1mv). Figure 2 shows how the (determinis-
tic jitter (DJ)) and (random jitter (RJ)) lines show the impact of jitter over a long
period of time and the actual data valid input window after 1 trillion UI
(a 1e-12 bit error rate or BER). Deterministic jitter is generally bound and
predictable and can be correlated to the data stream; such as inter symbol
interference and duty cycle distortion. Random jitter is generally Gaussian and
is unbounded. As with any Gaussian distribution, as a population increases, so
does the peak-to-peak value of the distribution. Therefore, total jitter is DJ plus
a BER multiplier of RJ. Note that the data valid input window is made on write
data. You can see how important it is to make BER measurement calculations to
form a statistical measurement of total jitter in order to understand the de-
sign's data valid window result and to properly understand at what rate you can
expect an error within the design. Additionally, understanding the components
and sources of jitter can enable designers to reduce jitter in their designs and
ensure better data performance.
Figure1. Direct Measurement with 100,000 UI Figure2. Dj and Rj impact on the data eye
04 | Keysight | DDR4 TdiVW/VdiVW Bit Error Rate Measurements or Understanding Bit Error Rate Measurements in DDR - Application Note
Making a Data The data valid input window is the window of valid write data to the respective
clock edge. The time interval error (TIE) is measured for each data edge in rela-
Jitter Measurement tion to the clock edge. Two TIE trends can be measured; one for data transitions
before clock and another for data transitions after clock to calculate TdiVW.
Once a TIE trend of data is established, jitter decomposition of deterministic
jitter and random jitter can be determined. A total jitter for data on each side of
DQS can be calculated, and a TdiVW is determined for a desired bit error rate.
Figure 3 shows the same eye diagram from Figure 1 now with contour points at
a BER of 1e-12 (or 1 trillion data transitions). Directly, measuring 1 trillion data
transitions-, would be extremely time consuming. By extrapolating different
BERs, we can begin to understand the failure rate to the mask that is expected.
At 1e-12, the TdiVW measurement is 173ps (with a margin to the defined mask
of only 5ps). The VdiVW measurement is 259mv (with a margin of 27mV to the
defined mask). This is a significant difference to the initial measurement for
TdiVW and VdiVW. At a BER of 1E-12, there is little margin to the mask, but we
can say that we can expect better than a 1e-12 failure rate (or 1 in 1 trillion).
Figure 4 shows the contour of 1e-15 which fails the mask. We know that we can
expect bit failures to this mask within 1 quadrillion data transitions.
While there is no specific data valid window specification for read data, it is just
as important to understand the read data being presented to the controller from
the memory device. Measurements can be made in the same manner for read
data so that memory controller designers can understand the impact of jitter
and noise on the data being sent from the memory to the memory controller
and create design expectations to develop a robust memory controller.
Figure 3. Data Eye with 1E-12 BER Figure 4. Data Eye with 1E-15 BER
Summary As memory technology reaches higher data rates, the signal amplitude and data
valid window becomes very small, resulting in increased data sampling error.
It is clear that simply making a measurement of an undefined quantity, may
falsely indicate great margin. Measuring jitter and noise to compute statistical
measurements of data valid windows helps you to quantify the bit error rate of
your system. Understanding and making these measurements is important in
developing high quality designs.
05 | Keysight | DDR4 TdiVW/VdiVW Bit Error Rate Measurements or Understanding Bit Error Rate Measurements in DDR - Application Note
myKeysight For more information on Keysight
Technologies, Inc. products, applications or
www.keysight.com/find/mykeysight
services, please contact your local Keysight
A personalized view into the information most relevant to you.
office. The complete list is available at:
www.axiestandard.org www.keysight.com/find/contactus
AdvancedTCA