Moving Message Display System

Principles of operation

Please refer to the block diagram of figure 1.

An EPROM is used to store the message to be displayed on the LED display. Bits D0 to D6 from the EPROM output correspond to the seven LEDs in one column of the display. A high output turns an LED on, the LED display being a common cathode type. The MSB, D7, is used to reset the system when the end of the message is reached; when it is high the system resets.

In order to scroll the message across the screen, the cathodes (display columns) of the LEDs are scanned by the Cathode Multiplex block at a speed high enough so that the flickering cannot be seen. As each column is selected, the corresponding data for the LED anodes (display rows) is output from the EPROM and the Main Address Counter is incremented.

After a number of scans set by the Scan Counter, the Base Address Counter is incremented and the new value loaded into the Main Address Counter. This sequence gives the illusion of the message scrolling across the display.

The system has a single clock which is effectively divided down by each block.

PCB

Rather than draw the copper pattern for the PCB by hand, since it is repetitive, I used a computer drawing package and a laser printer to produce an image on acetate film. Unfortunately, the number of display modules had to be reduced to 12 due to the size of the screen drawing area and the size of the UV light box used to produce the PCB.

With the PCB of figure 3, the first seven connections (marked in red) are the LED anodes, and the other 60 are the LED cathodes.

Figure 3: PCB copper track pattern

Adding extra messages

The two extra address lines A11 and A12 are used to select one of four messages stored in the EPROM. Switches A and B select a message when set as shown below:

Switches A and B are SPDT types connected as shown in figure 5 so that each address line is either connected to +5V or 0V.

Figure 5: Switches A and B

Address counters

The address counters form the heart of the system as it is these that enable the scrolling of a message to take place. There are two counters which work together:

The Base address counter

The Base address counter is a 12-bit (only 11 bits are used) ripple counter IC (4040) which is reset to zero by bit D7 of the EPROM at the end of a message so that the first byte of data in a message is at the first EPROM address, 00000000000.

The counter then increments every time the Cathode Multiplex block finishes a scan, so that the display appears to scroll.
In the final design, the Scan Counter was added so that several scans of the display take place before the Base address counter is incremented.

The Main address counter

After the Base address counter is clocked at the beginning of a scan, a LOAD signal inputs the new address into the Main address counter. The Main address counter then advances from this address once each time the Cathode Multiplex block moves on to the next column.

The Main address counter needs to be 11 bits long and has to be programmable. Since this length of counter cannot be found in one IC package, three identical 4-bit programmable synchronous counters are used (74C161). These have common clock and common load lines, plus carry-in and carry-out pins which are connected as shown in figure 6.

Figure 6: Address counters

The whole block was tested when built using two Logi-boxes to monitor the outputs and provide CK and LOAD inputs.
I noted that the 74C161 counters do not place data from their P0–P3 pins immediately onto their D0–D3 when LOAD is pulled low, but actually do this when the next CK pulse is recieved. This does not affect the operation of the circuit in any way.

Select counter

Each of the 1-of-16 decoders has 4 address lines, A to D. Connecting the Q0–Q3 outputs from a 4-bit counter (4520b) to each of the decoder address lines makes each decoder pull each of its outputs low in turn when the counter is clocked.

Now, all that is needed is a method to making only one decoder do this at a time.

Enable counter and decoder

The 1-of-16 decoders have an enable input, which, when brought low holds all their outputs high, thus preventing any display column to be on.

One output from a 1-of-4 decoder (4556a) is connected to each enable input so that only one 1-of-16 decoder is active at any one time.

The 1-of-4 decoder needs a 2-bit counter (4520a) to supply its address and this is clocked every time the D3 output of the Select counter returns to zero (i.e., when a count of 16 has been reached and the Select counter returns to 0000). Because the 4520 is positive clocked, a NOT gate is required to invert the D3 output before it clocks the Enable counter.

The block was then tested using a Logi-box to show the 4556 outputs changing and then applying the fast system clock and using 4 LED displays to check that scanning worked at high speed.

To do this the anodes of each row of LEDs were tied high. However, as more rows of LEDs were switched on, the LEDs became progressively dimmer due to the 4515 IC not being able to sink enough current.

In order for the LEDs to all be the same brightness I decided that the Anode Multiplex block was needed.

Select counter

The Select counter is a 3-bit counter (4520) that provides the address inputs for the 4051 analogue multiplexers. The counter needs to count up to six, then reset on seven since there are seven LEDs in each column. This was achieved at first by using two AND gates wired as a three input AND as shown in figure 9[a]. However, when tested using a Logi-box, the counter reset on a count of four, rather than seven. This is because the signal is delayed in the first gate so that a glitch occurs when the output changes from 3 to 4, causing a brief reset signal to the counter, as shown in figure 10. To stop this, the inputs to the AND gates were swapped around as shown in figure 9[b].

Figure 9: Resetting: [a] (A.B).C [b] (B.C).A

Using the spare outputs

The problem was solved by re-designing the resetting of the system.

Since there are 60 columns of LEDs used in the system, there are 4 spare outputs from one of the 1-of-16 decoders, each one being taken high when it is selected, and they are selected one at a time by the Cathode Multiplex block. I decided to use three of these to do the following tasks:

1. Clock the Base address counter. The length of the pulse means the counter has time to settle into its new state before the Main address counter loads in the new value.
2. Load in the new value of the Base address into the Main address counter.
3. Reset the Cathode Multiplex block since there is one output still left unused.

When tried, it was found that the Base address counter was being clocked twice each time output 61 went high due to a glitch. When a logic probe or an oscilloscope was connected to output 61 to try to observe the glitch, it disappeared. This must have been because of the load placed on the output by the test instruments, so I placed a variety of resistances between the output and +5V to try to stop the glitch. When this failed, I tried a 0.2nF capacitor (equal to the capacitance of the oscilloscope) instead cured the problem. After this, the design worked perfectly because the clock and load pulses are now longer, guaranteeing that the Base counter will be clocked and the Main counter will load in the new value from the Base counter.

Figure 13: Using NOT gates to delay the pulse

De-coupling capacitors

Six 0.1μF capacitors were placed near to many of the ICs to avoid problems with glitches causing unwanted clocking, etc. (The circuit behaved erratically when capacitors near the 555 IC were removed.)

Larger 100μF capacitors were also placed on each breadboard to smooth the power supply.

Darlington pairs

Nine ULN2004 ICs were used at the LED cathodes to sink the higher current. Each IC has seven identical Darlington Pair drivers as shown in figure 15. These can sink a maximum of 550mA, so are perfect for this application.

However, the Darlington Pairs are inverting, meaning that the 4515 ICs needed to be swapped for 4514 ICs (they both have the same pinout!) so that bringing an output high allows a column of LEDs to be switched on, since the output of the corresponding Darlington Pair will become low.

Figure 15: ULN2004 Darlington driver IC

Transistor drivers

For the LED anodes, the ULN2004 could not be used since current needs to be sourced from the driver circuit rather than sunk. Therefore, the transistor switch circuit shown in figure 16 was used. When the base voltage is 5V, the transistor is fully switched on, allowing a larger current to flow through the transistor into the LED.

Figure 16: Transistor driver

LED resistors

Because of the transistor driver, 0.7V is lost from the 5V supply, meaning that there is 4.3V across the LED and resistor. Because the display is being scanned at high speed, the LEDs are only on for a brief period, meaning that they will appear dim unless a higher current than they would normally use is pushed through them. The low resistance of 47Ω gave a bright display, meaning that the current through the LEDs is 90mA.

Because the brightness of the display does not now depend on the amount of current the ICs can sink and source, the Anode Multiplex block was removed, cutting down the number of ICs in the system from 27 to 23.

Suggested additions to the circuit

• Two rows: To enable lower-case letters to be shown properly, the display could be doubled in size to 14 dots in a column. This would also allow different sizes and styles of writing to be used, and better graphics to be produced. It should be fairly easy to add an extra row if a block similar to the Anode Multiplexing block were introduced to run each row alternately. The message length would of course be halved, two bytes of data being needed for each column (figure 20).
• Pause: If two bytes for each column were used (as above), there would be two spare bits. One would still be used for resetting, but the other could be used for an effect such as pausing the display. Pausing can be already achieved if the clock pulse to the Base address counter is stopped. The extra bit could trigger a monostable to cut off the clock signal (figure 21).
• Dual or full-colour display: The same technique as used to create two display rows could also be used for a dual-colour or full-colour display.

Figure 20: Two row display — simplified block diagram

Figure 21: Pausing the display — simplified block diagram

The display module PCBs

Using Microsoft Paint I designed a PCB that could accommodate a matrix of 16 x 7 standard 5mm LEDs spaced 1cm apart.

This PCB includes two ULN2803 darlington pair ICs and one 4514 1-of-16 decoder IC. Up to four PCBs can be arranged side-by-side to create a large display.

Programming by computer

In order to program messages into the system, the EPROM was replaced with a RAM and tri-state buffers to control the data flow. The RAM has a rechargeable backup battery. A ‘run/program’ switch sets up the system to either run the selected message or wait for bytes received on the parallel port. A circuit from Electronics — the Maplin Magazine was used to simulate the responses a printer would make to a PC on the parallel port.

Messages are written using the MMDS Editor software which I wrote for the Amstrad NC200 Notebook computer.

Appendix 1: Moving Message Display Simulator program

The following software was written in BBC BASIC for the Amstrad Notebook NC200 so as to judge the size of display needed to give a good effect.

 10 REM MMDS Simulator program
 20 REM Programmed by Tim Surtell
 30 :
 40 mess$="Moving Message Display System"
 50 delay=15
 60 dlen=12
 70 mess$=STRING$(dlen," ")+mess$+STRING$(dlen," ")
 80 :
 90 CLS
100 FOR I=1 TO LEN(mess$)
110 PRINT TAB(1,1)MID$(mess$,I,dlen)
120 TIME=0:REPEAT UNTIL TIME>delay
130 NEXT
140 GOTO 100

Note that the real display system scrolls column by column, not character by character.

Appendix 2: EPROM data program

This program was written in BBC BASIC for the BBC Micro by John Hewes to allow a message stored in the ‘M$’ variable to be viewed on the screen and printed as it will appear on the LED display. The resulting list of bytes can then be dumped to the EPROM programmer.

Please note that character definitions are given for the supplied test message only.

  10 REM EPROM data program
  20 REM Programmed by J. Hewes
  30 REM *** Warning! Do not use RENUMBER (due to DATA statements)
  40 :
  50 MODE 0
  60 DIM M% 1000
  70 :
  80 M$=STRING$(16," ")+">>> MOVING MESSAGE DISPLAY SYSTEM >>>"+STRING$(16,"*")+"@"
  90 :
 100 IF LEN(M$)>1000DIV6 PRINT "Message too long!":END
 110 :
 120 FOR I%=1 TO LEN(M$)
 130 C%=ASC(MID$(M$,I%)):VDUC%
 140 RESTORE C%+1000
 150 FOR N%=0 TO 5:READ B%:?(M%+6*I%+N%-6)=B%:NEXT
 160 NEXT
 170 :
 180 REM Check on screen
 190 :
 200 FOR I%=0 TO 6*LEN(M$)
 210 IF ?(M%+I%)=0 GOTO 260
 220 X%=2*I%:Y%=40
 230 FOR B%=0 TO 7
 240 IF ((?(M%+I%))AND2^B%)>0 PLOT69,X%,Y%+4*B%
 250 NEXT
 260 NEXT
 270 :
 280 PRINT "Print ? (Y/N)":IF GET$<>"Y" END
 290 :
 300 REM Check on printer
 310 :
 320 VDU 2:PRINT M$
 330 VDU 1,27,1,42,1,1,1,(6*LEN(M$))MOD256,1,(6*LEN(M$))DIV256
 340 FOR I%=0 TO 6*LEN(M$)
 350 VDU 1,?(M%+I%)
 360 NEXT
 370 VDU 3
 380 END
 390 :
1000 REM Character data (Line No. = 1000 + ASCII code)
1032 DATA 0,0,0,0,0,0 :REM Space
1042 DATA 127,127,127,127,127,127 :REM Block (*)
1062 DATA 0,8,8,62,28,8 :REM Right arrow (>)
1064 DATA 128,128,128,128,128,128 :REM End marker (@)
1065 DATA 0,63,68,68,68,63 :REM A
1066 DATA 0,127,73,73,73,54 :REM B
1068 DATA 0,127,65,65,65,62 :REM D
1069 DATA 0,127,73,73,73,65 :REM E
1071 DATA 0,62,65,65,69,127 :REM G
1072 DATA 0,127,8,8,8,127 :REM H
1073 DATA 0,0,65,127,65,0 :REM I
1076 DATA 0,127,1,1,1,1, :REM L
1077 DATA 0,127,32,16,32,127 :REM M
1078 DATA 0,127,16,8,4,127 :REM N
1079 DATA 0,62,65,65,65,62 :REM O
1080 DATA 0,127,72,72,72,48 :REM P
1082 DATA 0,127,72,76,74,49 :REM R
1083 DATA 0,50,73,73,73,38 :REM S
1084 DATA 0,64,64,127,64,64 :REM T
1085 DATA 0,126,1,1,1,126 :REM U
1086 DATA 0,112,12,3,12,112 :REM V
1089 DATA 0,96,16,15,16,96 :REM Y

Appendix 3: Test messages

Four test messages are available to download and can be read by the MMDS Editor software.