64x16 Pinball DMD

[Gab3Ru55]

I am trying to build a pinball machine, and the design of the LED display is giving me grief.
I plan on mounting 1024 LEDs in an acrylic sheet in a 64x16 array, and using 8 595s to drive the columns and 2 595s to drive the rows. My problem is I need to find a good power transistor/MOSFET to drive the rows, as each LED used 20 mA, for a total of 1.28 A. I had been thinking of using 16 SI2312BDS, but I would really prefer not using SMT. I will be using a chipkit to drive the system.
Any recommendations for the transistor or any other part of the display would be very helpful.
Thanks!
Gab3

 

[WTP Pepper]

ULN2803 with external BD137 transistor on each output would be a good start.

 

[Gab3Ru55]

ULN2803 with external BD137 transistor on each output would be a good start.

What would the advantage be to using the 2803? Could I just use the transistors with the shift registers?
Edit: To calculate the resistors used, should I just use the values for a single LED, because the resistors would be on the columns, and only one led per column would be on at a time?

 

[dougy83]

What would the advantage be to using the 2803? Could I just use the transistors with the shift registers?

ULN2803 has 8 transistor outputs and is designed for use with TTL levels, which is what I assume you're using. Yes it can connect directly to your CMOS shift registers.

Edit: To calculate the resistors used, should I just use the values for a single LED, because the resistors would be on the columns, and only one led per column would be on at a time?

Yes. You will need to take into account the voltage drop of the high-side driver, the LED forward voltage and the ULN2803 "saturation" voltage.

There are other transistor arrays available, albeit not as easily obtained as the ULN2803. e.g. 4AK21 is an octal MOSFET, which would have much less voltage drop than the ULN2803. http://www.datasheetcatalog.org/datasheet/HitachiSemiconductor/mXrzswt.pdf

 

[Gab3Ru55]

Okay, I drew these paper schematics of the control setup (just to see if I was arranging things properly)
this is the row driver setup
this is the column resistor with a single LED
Do these look right? Thanks for all the help!

 

[dougy83]

Yes, what you've drawn would be correct except that the the ULN2803 can't handle 1.2A which is why WTP suggested using a BD137 on its output. Sorry, I missed that in his post. Also, you might want a buffer on the column driver as the shift register won't necessarily be able to supply 20mA to all of its outputs at once (you might also like to increase the 20mA as each row is only driven for 1/16 of the time).

My suggestion is to use either a discrete MOSFET on each column and row, or MOSFET arrays.

 

[Gab3Ru55]

Yes, what you've drawn would be correct except that the the ULN2803 can't handle 1.2A which is why WTP suggested using a BD137 on its output.

Okay, so I use a MOSFET to boost the current (Fairchild has the BD13716S that can handle 1.5A)

Also, you might want a buffer on the column driver as the shift register won't necessarily be able to supply 20mA to all of its outputs at once (you might also like to increase the 20mA as each row is only driven for 1/16 of the time).

I think my shift registers can handle it, NXP's 74HC595N,112 can output 35mA per pin, with a 750mW max overall (using the DIP package). Would I take 20mA times the LED voltage of 2.5, or the output of 5 to get the wattage?

 

[dougy83]

Okay, so I use a MOSFET to boost the current (Fairchild has the BD13716S that can handle 1.5A)

Yes, use a single MOSFET on each row and column.

I think my shift registers can handle it, NXP's 74HC595N,112 can output 35mA per pin, with a 750mW max overall (using the DIP package). Would I take 20mA times the LED voltage of 2.5, or the output of 5 to get the wattage?

The maximum ground or supply current for the IC is specified at 70mA.

 

[Gab3Ru55]

What would happen if I tried to run the LEDs under their normal current?

 

[dougy83]

What would happen if I tried to run the LEDs under their normal current?

They will be less bright.

 

[Gab3Ru55]

They will be less bright.

I don't think that will be as much of a problem, the LEDs I'll be using are super bright 800-1000 mcd ones, so having them be dimmer might actually be a good thing.

 

[Gab3Ru55]

Does this look correct? I have the BD137's base attached to the output of the ULN2803, the collector to the matrix, and the emitter to the ground.
The display will be driven by a propeller or more likely a chipkit, so I'm confident that I can send the serial data fast enough for all 64 outputs.

 

[dougy83]

Does this look correct? I have the BD137's base attached to the output of the ULN2803, the collector to the matrix, and the emitter to the ground.

No, it won't work as BD137 is NPN and the ULN2803 output is open collector.

Just use a mosfet on each output of the '595.

 

[Gab3Ru55]

No, it won't work as BD137 is NPN and the ULN2803 output is open collector.

Just use a mosfet on each output of the '595.

I thought that the BD137 was the mosfet, should I choose a different one?
I know the base goes to the 2803, should the collector or the emitter be connected to the ground?
Thanks!

 

[dougy83]

If you're using a N-channel MOSFET, then you don't need the uln2803. You also won't be connecting to the base, collector or emitter, but instead the gate, drain and source. The gate connects to the '595 output, the source to ground and the drain to the cathode of the leds.

 

[Gab3Ru55]

Would the NTD5867NL (through hole) work? The gate threshold voltage is a max of 2.5v.
I'm assuming that it means that at most it takes 2.5 volts to make the mosfet allow current between source and drain, is that correct?

 

[dougy83]

Yes, the 2.5V is the maximum voltage required to get a drain current of 250uA. In the datasheet (https://www.electro-tech-online.com/custompdfs/2013/03/NTD5867NL-DPDF.pdf), Figure 3 shows the resistance between drain and source vs. the gate-source voltage. At 5V (which is what I assume you're using), the resistance is 31 milliohms; this is essentially fully on in your circuit.

 

[Gab3Ru55]

Okay, so the cathodes of the LEDs are connected to the mosfets, which are connected to the ground, and the anodes are connected through the resistors to the 8 shift registers that provide power.
The mosfets are connected to the two shift registers driving the columns, and both sets of shift registers go to a chipkit.

Thanks for taking the time to explain this, I'm relatively new to all this (plus learning about conventional vs. actual current when I was young messed up my thought process). All of these light bulbs are going off in my head, and all of the crazy numbers on the datasheets now make sense.

 

[dougy83]

Okay, so the cathodes of the LEDs are connected to the mosfets, which are connected to the ground, and the anodes are connected through the resistors to the 8 shift registers that provide power.
The mosfets are connected to the two shift registers driving the columns, and both sets of shift registers go to a chipkit.

Sounds good to me. The two sets of shift registers can be connected as a single string of shift registers as you'll need to update all registers for each column drive subcycle.

Thanks for taking the time to explain this, I'm relatively new to all this (plus learning about conventional vs. actual current when I was young messed up my thought process). All of these light bulbs are going off in my head, and all of the crazy numbers on the datasheets now make sense.

Conventional current makes more sense when reading schematics due to the direction that the arrows are drawn in the symbols. It also seems more obvious that the current is flowing from a higher potential to a lower potential.

 

[Gab3Ru55]

Sounds good to me. The two sets of shift registers can be connected as a single string of shift registers as you'll need to update all registers for each column drive subcycle.

Oh, that makes sense, then the row mosfet will turn on as all of the columns do, so the sign syncs up automatically