Datasheet question - ceramic board mount

[j]

While sifting through datasheets for transistors on mouser, I came across one that showed 2 different lines for absolute maximum power dissipation. One line for “Each terminal mounted on a recommended land” and one for  “Mounted on a ceramic board”. The line for ceramic board mounting showed 4 times the max dissipation.

 

My question then is, what exactly is ceramic mounting and is it comparable to installing mica wafers between the transistor and the heat sink?

 

Datasheet:  http://www.mouser.com/ds/1/348/2sar533p-44342.pdf

[Alex Hunt]

They are talking about ceramic circuit board material.  Think of replacing the fiberglass in a typical PCB with a ceramic substrate.  Ceramic boards are typically used in high frequency or high temperature applications.  In this situation a ceramic board has much higher thermal conductivity, so it is able to remove heat from the transistor's heatsink pad more efficiently.

It is not the same as a mica wafer.  The wafer just provides an insulating barrier - it does not improve the PCBs ability to conduct heat.

For instance:
http://www.bestpcbs.com/products/ceramic-pcb.htm

And yes - ceramic PCBs are more expensive than typical FR4.  :)

Alex

[j]

Ok, so in theory I could mount a heat sink directly to the transistor body w/ thermal grease and achieve the same results?

[Alex Hunt]

Do you realize that this is a surface mount part with a heatsink only 2x3mm?

Theoretically, yes.  Thermal grease, no.  This device is designed to be soldered to the heatsink (i.e. a big chunk of copper on a PCB).  Thermal grease can't match the thermal conductivity of copper.

How many watts do you need to dissipate?

Alex

[j]

Oops… I didn’t notice the surface mount. It came up on a search with the through-hole filter set. Weird.

 

Now that you ask, I realize my original calculations were incorrect, and a different part would work better. Max constant dissipation will be 4 – 5 watts. I’ll probably go with something that can handle 15W.

[j]

Ok I must be working too hard. Original calculations were correct. Just for completeness, attachment is a rough draft of what I’m thinking.

 

RGB LEDs draw 20ma according to the datasheet, so I estimate .5A & 1.6W @ 3.2V for the red & green channels, .5A & 1W @ 2V for the red channel. I can’t find a link for the datasheet now, but most of it’s in Chinese anyway.

 

This should do it? http://www.mouser.com/ProductDetail/Fairchild-Semiconductor/KSD1616AGTA/?qs=sGAEpiMZZMsTKkj12KWLXuW5oDYpP%252bzx4zCv16Y6MtU%3d

[Alex Hunt]

As long as you're saturating the transistor, that one will work fine.

Don't parallel the LEDs as shown in this schematic.  The diodes would have to be factory matched in order for that to work properly.

Assuming you have only one LED per transistor, I think you would be better off grounding the emitter and putting the diode in series with the collector.  That way it's a little easier to bias the transistor.

Usually the blue channel is 3V and the red & green are 2V.

Alex

[j]

Datasheet for the LEDs says typical voltages are 2.0 for red, 3.2 for green, and 3.1 for blue. http://www.dipmicro.com/?datasheet=HH-1000CRGBC810C-B.pdf

 

I’ll have 3 transistors (one for each color) for 25 10mm LEDs using 3 separate outputs from the arduino (actually it’ll be either a TLC5940 or a TLC5923 depending on the final numbers)

 

Good point on parallel vs. series. My original thinking was that I could use an adjustable regulator off the same 5V supply powering the arduino to get the 2 – 3 volts I need for the LEDs. In series, the power supplies would be 50V – 75V, and beyond tolerance for the 1616. I’ll continue the search tomorrow.

 

(updated schematic attached)

[Alex Hunt]

Interesting datasheet - I can't even tell if the LEDs are common cathode or common anode.

Things are easy with the TLC59xx chips - you have 16 channels to work with, so break your color strings into 5 serial sets for each color.  That way the supply voltage is manageable.  The chip has it's own driver FET and current limit, so you don't have to worry about an external transistor.

Alex

[j]

They’re common anode.

 

I burned out a 5940 chip already with just 1 LED attached, using the +5V source from the arduino (admittedly, I wasn’t watching the ERR pin).

 

I didn’t want to divide the group of 25 though because ultimately there will be more than one par can with 25 LEDs each. Will probably have 4 cans (100 LEDs total) for every 3 outputs on the chip. When I get to that point though, each can will have its own ac power supply – most likely just a transformer and full wave rectifier. For now, I just need to get a proof of concept without blowing up more chips.

[j]

They’re common anode.

 

I managed to fry a 5940 with just 1 led attached, driving it from the arduino +5V supply – which is why I want to use the 5940 to drive a transistor before moving forward. This is based on recommendations from TI per this document: http://www.ti.com/lit/an/slva280/slva280.pdf Also, I don’t want to break up the LEDs into smaller groups, but rather drive multiple transistors from each output. Ultimately, each par can with 25 lights will have its own power supply anyway. I estimate the finished product will be 16 par cans (400 LEDs!).

 

This one seems like it would do the job, with enough overkill that I wouldn’t have to worry about heat problems. http://www.mouser.com/ProductDetail/Infineon-Technologies/IPP114N12N3-G/?qs=sGAEpiMZZMvsEea7gdidXkQgIJODFu3sutGFAeM%252b%252btA%3d Is there something else that might be a better fit for 75v .5A continuous?

 

 

 

 

From: dhmn-di...@googlegroups.com [mailto:dhmn-di...@googlegroups.com] On Behalf Of Alex Hunt

Sent: Saturday, November 03, 2012 10:56 AM