Seite 18 - Boards & Solution/ECE Magazine July 2013

July 2013 18
EMBEDDED DESIGN
Automatic white balance (AWB): sensors are
not good at recognizing colors. AWB adjusts
other colors in an image with reference to an
inferred white color in the image through a
so-called grey world algorithm. AWB deter-
mines white by examining the frequency (there-
fore wavelength) of incoming light, and renders
the image with natural colors.
Gamma correction: sensor pixels react to the
intensity of incoming light in a linear way. In
order to provide pixel data to common video
systems, such as a CRT tube with logarithmic
response, conversion to a non-linear value en-
coding may be needed. Gamma correction
provides this conversion.
High / wide dynamic range (HDR/WDR) pro-
cessing: this is the block responsible for map-
ping 20 pixels of sensor date to 8 bit RGB in a
way that renders both bright and dark areas of
the image visible in the rendered image. A
wide internal pipeline is required to ensure
that no detail in dark areas is lost even when,
for example, an intruder shines a flashlight di-
rectly into the camera lens. HDR, working in
close conjunction with a fast-auto-exposure
algorithm, can rapidly adjust exposure in
changing light conditions.
Table 1 shows the typical FPGA resources used
for implementing all the above ISP blocks in a
33K look-up-table (KLUT), low—cost, low-
power, FPGA. In addition to the ISP blocks al-
ready mentioned, the actual implementation
data incorporates a statistics engine that gen-
erates image histograms used by specific blocks
in the system, a Lattice Mico32 soft processor
for dynamic pipeline control, an I2C master
to control various signals, a HDMI PHY block
to drive HDMI signals directly off the FPGA
and even a graphics overlay of a logo. This
demonstrates that it is possible to fit an entire
image signal processing pipeline, plus HDMI
output, inside a low-cost, low-power FPGA
such as the Lattice ECP3-35. The internal
HDR pipeline is 32-bits wide, resulting in the
ability to provide 192dB (20 log 2**32) of
high dynamic range. In this real world imple-
mentation, a sensor with 120dB dynamic range
was used, limiting the HDR to 120dB – still
the highest of any FPGA implementation avail-
able. The actual implementation is capable of
processing 1080p images at 60 frames per sec-
ond while providing 120dB of HDR. As shown,
a simple low cost 33KLUT FPGA easily handles
a 1080p60 pipeline. The BOM for a 1080p60
HDR camera implemented with a Lattice
ECP3-35 consists primarily of the sensor, the
FPGA and associated clock oscillator, resistors
and capacitors, voltage regulator, a HDMI
connector and lens assembly. The implemen-
tation shown offers 120dB of HDR, 1080p60
performance, the fastest auto-exposure and
very high quality auto-white balance in the
industry. The LatticeECP3 is significantly
lower in static and dynamic power consump-
tion than competing FPGAs or DSPs. The
FPGA supports the use of DDR3. Manufac-
turers wishing to incorporate frame buffer
memory into their designs can take advantage
of this capability to utilize high-performance,
low-cost DDR3 memory in their camera de-
signs. A low-power SERDES-capable FPGA
enables manufacturers to implement a HDMI
PHY directly inside the FPGA, providing
HDMI functionality without the added cost
of an external HDMI chip. n
The current generation of electric vehicles
relies on lithium battery packs with an energy
range between 16kWh and 53kWh. A single
gallon of gasoline contains more than
36kWh of energy. For an electric or hybrid
electric vehicle, or any high power battery
system, to compete with an internal com-
bustion engine every bit of energy must be
squeezed out of the batteries. This means
each individual cell in the pack has to be
monitored and controlled.
n High power battery packs consist of a long
string of series-connected cells. Directly con-
nected to each cell is a battery monitor IC, re-
sponsible for accurately measuring each cell
voltage. This is no simple task, as the cells are
positioned at various points along a very high
voltage string that is subject to horrendous
electrical spikes and electromagnetic interference
(EMI). A battery management system (BMS)
combines the cell voltage with current, temper-
ature, and operating history, to continuously
assess each individual cell condition. It is a
tough challenge, but with accurate monitoring
and control, the driving range, reliability and
safety of the battery pack can be maximized.
Batteries in an HEV or EV are expected to last
10 to 15 years, and they are considered to be
at their end-of-life when they have lost 80%
of their original capacity. Battery lifetime and
reliability are maximized by restricting the op-
erating state of charge by not allowing them
to be fully charged or discharged. A typical
battery pack is operated in a restricted range,
such as 20% SOC to 80% SOC, where SOC is
the state-of-charge. These SOC limits could
be adjusted with age and operating conditions,
such as high temperature environments. As a
result of the limits, battery packs are not
utilized to their full capacity. For example, op-
erating a pack with 20% SOC to 80% SOC
limits the usable SOC range to 60%. The chal-
lenge for the BMS is to operate each cell as
close to the limits as possible, without exceeding
them. Amplifying the challenge, lithium bat-
teries exhibit a flat discharge curve over their
operating range. As a result, there is a very
small change in cell voltage over the operating
range and the battery monitor must make
very accurate measurements as part of the
SOC calculation.
To illustrate the importance of cell measure-
ment accuracy, consider the simplified lithium
battery discharge curve shown in figure 1.
This curve has a constant 5 mV/% (SOC)
slope across the operating region. A battery
pack operating within a 20% to 80% SOC
range and a similar discharge characteristic
will face a big penalty for poor cell voltage
measurement accuracy. As shown in figure 2,
if the battery monitor has a cell voltage meas-
urement error of ±10mV, a measured cell volt-
age of 3.75V could actually correspond to a
real cell voltage between 3.74V and 3.76V. This
corresponds to an actual SOC range from
Precision cell measurements add
value to battery management
By Greg Zimmer, Linear Technology