8x8x8 multiplexed LED cube with an Arduino Mega 2560

Thanks for that explanation, its a bit difficult to understand, since I have no idea of how much I can accept of several factors. It would be nice to have a table of applications, and specific parts with specific details that were easily compared to other parts. Watt/package size is a nice relationship, that seems easy enough to factor.

Lets use 20 mA x 64 LED as an example. Assuming its running on 5v, using constant current driver to sink, (that sounds like the direction he is heading), it should use up to 1.28A. With this design, he is only doing 12.5% duty cycle. How do we apply the math to figure out what specific specifications to look for, and ranges maybe? It seems like a simple table would be easy enough to make.

Hippynerd:
Lets use 20 mA x 64 LED as an example. Assuming its running on 5v, using constant current driver to sink, (that sounds like the direction he is heading), it should use up to 1.28A. With this design, he is only doing 12.5% duty cycle. How do we apply the math to figure out what specific specifications to look for, and ranges maybe? It seems like a simple table would be easy enough to make.

Let's look at power dissipation first. If we assume that the software works so that the 1/8 duty cycle won't be exceeded, and we go for 0.5W maximum power dissipation in the mosfet, then the dissipation while the mosfet is conducting can be up to 8 * 0.5W = 4W. So we need Rds(on) <= 4/(1.28 * 1.28) = 2.4 ohms.

However, the voltage drop would then be 2.4 * 1.28 = 3.1V, and we certainly can't tolerate that in a LED driver running from a 5V supply as there wouldn't be enough voltage left to drive the LEDs and series resistors. So in this case, maximum Rds(on) is determined by maximum allowable voltage drop. If we were not concerned about the LEDs getting dimmer as more are turned on, we might allow 0.5V, allowing Rds(on) of up to about 0.4 ohms.

I search for MOSFETs like this at digikey.com:

x-channel mosfet (P or N)
in stock
Fets-single
logic level
thru hole (for prototyping anyway)
sort by price (click on unit price)

scroll down past the 1,000 qty min buy parts to the 1-lot, looking for a low Rds part,
which usually puts me at the start of the TO220 type parts.
Then I check to make sure the gate capacitance isn't ridiculous.

Right now, that search puts me at these two parts
p-channel - small list of parts, just 12 choices

n-channel, 18 pages of parts
after the price sort, right on the first page

Very nice explanation dc42! Now I understand why I need the MOSFETS that I need.
I bought the NDP620Ps from Farnell in the end: http://ro.farnell.com/jsp/search/productdetail.jsp?sku=1017724

In the datahsheet we can see this:
"Features
-24 A, -20 V. RDS(ON) = 0.05 W @ VGS= -4.5 V.
RDS(ON) = 0.07W @ VGS= -2.7 V.
RDS(ON) = 0.075 W @ VGS= -2.5 V."

I hope that's good news. What confuses me is that you used Ohms for the unit of Rds(on) and the datahseet uses Watts. But I guess V=IR and W=IV, so W = V/R*V = R. W=R ?!

Data sheet says 0.05 ohm, right on page 1.
http://www.fairchildsemi.com/ds/ND/NDP6020P.pdf

When I look at the datasheet, the characters that you describe as W are Greek omega-characters.

Rds(on) = 0.05 Ohms then, so it's good :slight_smile:

Thanks!

For anyone else who might be interested in the future to build a LED cube based on this design, I've put together a component list with prices based on what I have bought for the project. I tried to buy the components as cheap as I could, most of them off eBay, but for many of them I have bought more than actually necessary, because they sell in bulk or because I wanted some spare parts. Here's what I got:

NDP6020P: 14 pcs 33.3 EUR (43.8 USD)
3mm diffused blue LEDs: 1000 pcs 19.4 EUR (25.5 USD)
TPIC6B595N shift registers: 20 pcs 9.3 EUR (12.3 USD)
1K variable resistors: 100 pcs 7.4 EUR (9.8 USD)
5V 3A power supply: 1 pcs 7.1 EUR (9.4 USD)
Craft wire 0.8mm 6m: 3 pcs 7 EUR (9.2 USD)
Jumper cables 40 PCS: 2 pcs 6 EUR (7.9 USD)
Capacitors (0.1 uF) + resistors (82R, 220R): 3 pcs 4.6 EUR (6.1 USD)
PCB 18x12 cm: 2 pcs 3.9 EUR (5.2 USD)
Female pin headers: 200 pcs 3.8 EUR (4.9 USD)
TO-220 heatsink: 10 pcs 3.4 EUR (4.5 USD)
IC sockets DIP-20 for TPIC6B595N: 10 pcs 2 EUR (2.7 USD)
Crocodile clips: 2 pcs 0.8 EUR (1 USD)
Total: 108 EUR (142.3 USD)

Sorry, I just can't get the data into a well-formatted table.

This does not include, of course, the price of the Arduino driving the cube or other miscellaneous stuff like soldering materials or the prices of some tools that you use for other projects too.

Somehow I got lost on the math there DC. 1.28 * 1.28 = 1.6384, divide by 4 and you get .4096.
How do you come up with .5 watt?
He will need 8 mosfets, each one will run 1/8th of the time (12.5%), each one will need to be able to source between 0 and 1.28A. I see the 1.28 in there, and the 4 was from half watt times 8. So I see some of the numbers, and understand where they came from, Im still confused with the math, and where half a watt came from. it seems arbitrary.

I tried crossroads technique, and came up with 7 P-channels, I started by searching "p-channel mosfet", then clicked a link that said Fets-Single, That displayed a page with many options, I selected logic level gate, through-hole, and then I scrolled the package column, and selected the TO-220-3

From there, I notice that all of the parts have many mOhm RDS (way way over .4 ohm), not even close to mOhms.
I dont know what a good/bad gate capacitance is, but I can only find gate charge, and input capacitance.

I can also see that the NDP6020 is rated at 60 watts, and if using 1/2 watt, Im guessing you wouldnt need a heatsink.
How many 20mA LEDs could you run on that mosfet?

Also $150 for the parts for a 8x8x8 cube sounds like a lot, but I havnt built a cube that big.

6 Meters of wire seems like a low estimate to me too, I bet you use over 8. what kind of wire is it? soft wire is very difficult, but harder wire is a bit easier, and ends ups being sturdier. I find very hard steel works the best. For cube building, and general keeping wires straight, thinner and softer are more difficult, thicker and harder are easier, but for soldering, thinner is easier.

Hi!

I'm planning to run 64 20 mA LEDs from one of those MOSFETS.
That price of 142$ includes components for the big 8x8x8 cube, but also for the small 4x4x4 learning cube and some spares. I guess that if I tried to but onyl as many components as necessary for the 8x8x8 cube only, I could have pulled it off for about 100$.

6m of craft wire is definitely not enough. But I have bought 3x6=18m, which is probably twice as much as I need. I do not know how soft or hard it is, I bought it blindly from eBay based on the suggestion of somebody who has already used the same or similar craft wire to build a LED cube successfully. It is 0.8mm thick and is sold rolled up, I'll have to straighten it before soldering.

Hippynerd:
Somehow I got lost on the math there DC. 1.28 * 1.28 = 1.6384, divide by 4 and you get .4096.
How do you come up with .5 watt?

See reply #83. I suggested a maximum of 1W for a TO220 mosfet with no heatsink, and 0.5W for a mosfet in a smaller package.

CrossRoads, is there a special reason why the LED cube that we've discussed here uses the cathodes as columns (64 cathode columns) and the anodes as planes (8 anode planes) and not the other way around? Many other cubes that I've seen use anode columns and cathode planes and it seems to be more practical that way because the anodes are 2-3 mm longer, which compensates for the fact that the LED heads are taller than wide, so the horizontal and vertical distances between the LEDs could be longer and still equal if we used anode columns and cathode planes. Is it possible to do this reversing of anode/cathode roles? If yes, how?

Thanks,
Andras

Cathodes as columns because the TPIC6B595 can only sink current, not source it.
I have not seen a shift register that is designed to source 20mA (and not running at the Absolute Maximum as many people use 74HC595) outside of the 74AC299P which I think is not procurable in DIP form anymore. I bought a tube of them from Newark for not much when Newark was getting rid of the last of theirs.

I had not considered LED pin lengths - just figured cube dimension would adjusted to actual parts purchased, which no doubt vary by manufacturer.

74AC299.pdf (130 KB)

Well, I'll sacrifice those 2 millimeters then :slight_smile:
Thanks!

BTW, happy new year!

Thanks! Happy New Year to you as well!

Finally all my parts for the LED cube have arrived (except some minor things that I can start without) :slight_smile: I plan to start working on it soon, I will show you pictures when I'll have some results.
The only problem is that the LEDs that I bought seem to have very short legs. The Anodes are 18 mm and the cathodes are 15 mm. It will be a crowded cube...

I thought you were using craft wire? If you are using wire, you will be soldering the leads to wires, then clipping the leads.

My intention was to use craft wire only to strengthen the cube where necessary (diagonally relative to the LED anodes, inside the planes), but given the short legs, I might need to make the whole cube frame from craft wire. That would require twice as much soldering, though...

How do you keep from shorting things out with extra diagonal wires like that?

Well, all you have to make sure is that the craft wire inside the anode planes never touches any cathodes. The 64 (8x8) cathodes go down and all the anodes are soldered together in the same horizontal plane, including with craft wire. In other words, you solder the LED anodes together along the X axis. You strengthen the anode plane with craft wire along the Y axis (X and Y are in the horizontal plane) and the cathodes go down along the Z axis.