How-to design LED signage and LED matrix

displays, Part 3
edn.com/how-to-design-led-signage-and-led-matrix-displays-part-3/

September 3, 2014

Part 1 of this three-part series introduced the technical and

engineering steps necessary to design an LED display system from
individual LED lamps. Part 2 provided the remaining steps needed to
implement a basic LED display. In the third and concluding
installment of this tutorial, we'll explore some of the issues which
affect the image quality and reliability of LED displays. We'll also
become acquainted with the technologies and design techniques
commonly used to deal with them.

Anti-ghosting/ghost-canceling/pre-charge FETs

Ghosting, spike noise, or phantom noise are unwanted lighting

effects caused by Anode gate “float” which can occur in time-
multiplexed LED driver. Since LED lamps (PN junction of diodes)
have relatively high levels of capacitance, their residual charge can
keep triggering capacitive charge transfers between the floating
nodes. And every time there's forward electron flow through a PN
junction

The situation where this phenomenon is most is a diagonal line

image. Figure 1b shows an example of so-called “ghosting” caused
by anode float. Modern LED driver ICs, such as the TLC59283,
employ so-called “pre-charge FET” circuits which eliminate these
ghosting effects (Figure 1a ). As explained earlier, the root cause of
ghosting is stray charges on the LED's anode which forward-bias its
PN junction and cause it to light at unwanted times. These pre-
charge FETs are designed to insure the LED lamps remain reverse-
biased and unlit except when the driver circuit is actually on.
Figure 1 Pre-charge FETs (right) and the ghosting (left).
Blank bands, black bands and Enhanced spectrum PWM

LED display designers face several other challenges as they strive

to produce ever-larger products which deliver the pest-possible
image quality. One of the biggest issues is eliminating the blank
bands which can occur when capturing the image of an LED display
on a camera. As we discussed in Part 1 of this series, this is caused
by “slow-synching” between the display and the camera. This can
be avoided by using a faster frame refresh rate (FRR). Unfortunately,
larger displays require faster FRRs. As a result, it becomes
increasingly difficult to achieve an FRR that's sufficiently high to
avoid slow-synching effects as display size increases.

Another issue is black bands which appear when a camera captures

a display image image at the moment some of its LEDs are OFF.
This can be avoided by keeping the LED lamps ON during a camera
scan period but, as the following example will show, that's not
always possible.

Black bands become a more significant problem as PWM control

LED ICs grow to control larger, higher-quality displays where the
length of their PWM operation cycle time grows longer. For
example, the latest 16-bit PWM control with a 25 MHz reference
clock requires 2.6 ms = 216 bit / 25 MHz, which is a frame refresh
rate of 381 Hz. Here, a gray scale code of 128 for a total of 216
clock cycles generates 5.1 us (= 128 / 25 MHz) of ON time, and 2.6
ms minus 5.1 us of OFF time. The camera captures LED lamps in
the OFF state during this 2.6 ms period.

Black-banding can be mitigated using a technique called enhanced-

spectrum PWM (ES-PWM), a method for PWM generation which
divides one long PWM cycle into shorter sub-PWM cycles. In the
above example, if 128 clocks of the ON period are divided into 16
periods of 8 clocks each, creating an effective FRR of 6 kHz (= 381
Hz x 16). At 6 kHz, the refresh rate is high enough to avoid black
bands with most cameras.

An original PWM code cannot always be equally divided. In this

situation, the ES PWM function splits one ON period into rounded
integers. For instance, to divide a gray scale code of 100 into 16
pieces, the ES-PWM circuit generates twelve of 6 clocks and four of
7 clocks to maintain a total gray scale of 100 (= 6 clock x 12 + 7
clock x 4).

Detecting LED open, LED short, & output leakage conditions

Many LED display systems are controlled remotely, making it

difficult for an operator to detect any failures. Because the human
eye is sensitive to a faulty lamp that remains constantly ON or OFF,
the failure of even a few lamps can degrade the quality of a viewer's
video experience. As a result, many displays implement ways to
detect open and shorted LEDs, as well as output leakage conditions
which can cause LEDs to malfunction.

An LED open detector (LOD) function monitors LED lamps for open-
circuit failures. Under normal circumstances, a driver IC’s constant-
current output terminal stays at the head room voltage required by
the constant-current circuit. When the constant-current circuit's LED
fails and becomes an open circuit, the constant-current circuit
drives its output terminal to almost zero voltage. The LOD function
detects these telltale voltage changes and generates an error
signal.

Similarly, an LED short detection (LSD) monitors the LED lamp for
conditions which indicate the LED, and/or its driver are short-
circuited to its anode's supply voltage. When the LED fails in a
shorted mode, its output terminal reverts from its normal bias state
to the full voltage applied to the anode. The LSD function
distinguishes this voltage difference and generates an alarm signal.
An output leakage detection (OLD) differs slightly from the first two
safety functions. It's designed to detect conditions which arise
when an LED is forced into its ON state due to debris forming a
conduction path from an output terminal to the ground. When this
occurs, the LED is turned ON – no matter what the output of its
constant current-circuit driver happens to be. The OLD element
produces a small amount of current at its output terminal node
which it uses detect any leakage path by monitoring the terminal
voltage.

Low gray scale enhancement

The human eye has more sensitivity to darker light sources than
brighter lights. In other words, it recognizes which of two dark light
sources emits more photons. However, when the human eye is
saturated with bright light from two different sources, it cannot
distinguish the difference.

For handling video image, low gray scale data requires more
attention. Here a technique like gamma correction is widely used.
As for LED display systems, software programming can implement
a gamma correction function with both ON/OFF and PWM control
drivers.

Recent LED drivers, like the TLC5958, integrates more proactive

improvements on low gray scale handling. A common problem is
that red LED lamps are stronger than green and blue with dark white
image output, even though red, green and blue all have the same
low gray scale data. This occurs because red LED lamps can turn
ON longer than green and blue lamps due to its lower forward
voltage. A low gray scale enhancement (LGSE) function can correct
this difference inside the IC. Figure 2a has no correction while 2b
has been corrected.
Figure 2 Two examples of low gray scale enhancement with both
showing dark white image data.
Regarding this low gray scale concern, LED current PWM pulses
need very sharp turn-ON and turn-OFF times, or rise and fall times,
TR and TF. If TR and TF are slow, low gray scale problems can get
worse.

“First line” issues and integrated SRAM

As mentioned earlier, ES-PWM control speeds up FFR. By using ES-

PWM with the time-multiplexing anode control, the first line of time-
multiplexing gets darker. Figure 6a has two lines that appear to be
more reddish than the others (very top and middle). All other lines
look to be more white. This first line issue is caused when the green
and blue lamps are not fully turned ON.

A solution to the root cause of the first line issue can be found by
integrating static RAM (SRAM) bits to store gray scale PWM codes
for the entire frame, thus avoiding data transfer time lag. For
example, the TLC5958 integrates 48 k bits of SRAM on-chip for up
to 32 times of multiplexing.

Design tips for display systems and driver ICs

Inrush current control

In general, an LED display system handles huge amounts of current.

For example, eight pieces of 48-output LED driver ICs controls 25
mA each. The total current is 9.6A. The biggest problem with an
LED display system is that this 9.6A of current keeps turning ON
and OFF at very high frequency with fast TR and TF.

Many LED driver ICs come with noise reduction features such as
delay between each output. Because a system handles 10 MHz
order of digital signal on its PCB, noise management is an
important design factor early into the project.

Thermal error flag/pre-thermal warning

As stated, an LED display system handles huge amounts of current

– which translates into huge amounts of heat. This excessive heat
can cause thermal shutdown and unexpectedly stop LEDs from
working. It is a major issue when the entire display stops working,
but viewers might think that the system is simply turned OFF.
However, in most cases, only a partial module stops working and
viewers can see that something is wrong (Figure 3 ). Because of
this, many LED driver ICs do not come with a thermal shutdown
function. Instead, they come with a thermal error flag (TEF) or pre-
thermal warning flag (PWF) function.

Figure 3 LED display with some modules inoperative.

These flags are generated by a circuit similar to thermal shutdown
detectors. Instead of stopping an IC when temperatures get hot, hot
temperature condition flags are sent to an image processing
controller. Upon receipt of a flag, the controller cools down the
system by reducing screen brightness, showing darker images, or
simply stops the system for a moment.

48-output driver

PCB layouts can be nightmarish on a typical LED display module

design. We compare system concept sketches utilizing one 48-
output driver (Figure 4a ) and three 16-output drivers (Figure 4b ).
Both diagrams are a 16 x 16 RGB matrix, which equals 768 LED
lamps. It is clear that a 48-output driver like the TLC5958 can
simplify your PCB design.

(a)

(b)

Figure 4 PCB layout comparison between 48- and 16-output drivers.

With the numerical example specification, key points in IC data
transfer calculations are reviewed as a final step of the LED display
system building, which is a continuation of the discussion started in
Part 1. We also visited how various LED display driver features
improve video output quality, plus we examined some tips for
making system design easier and more efficient.

References

How-to design LED signage and LED matrix displays (Part 1),
EDN, July 30, 2014
How-to design LED signage and LED matrix displays (Part 2),
EDN, July 30, 2014
More information about LED signage is available on Texas
Instruments’ web site.
Download the datasheets of the devices referenced in this
article: TLC59283, TLC5958

About the Author

Masashi Nogawa is a product marketing engineer for Texas

Instruments’ Power Management group where he is responsible for
the SWIFT product line. Masashi received his BSEE and MSEE
degrees from the University of Electro-communications, Tokyo, and
he holds six US patents. Masashi can be reached at
ti_masashinogawa@list.ti.com.