I need to know how to calculate the maximum frequency from the information given on the data sheet for N-channel MOSFET's. I'm looking for a switching transistor to turn on and off LED's at a maximum frequency of 5 MHz and need to find an N-channel MOSFET that can handle that frequency. The data sheets only give rise times and fall times and on and off times. I'm including sample data from a data sheet.

Thanks,
Mike

 

Well, add up all those times and you get 51nSec so, if at 20Mhz it would fail to turn on and off fully.

But the real question is: what do you hope to accomplish by switching LEDs at 5MHz?

Bob

 

All by itself the datasheet may not give you what you want without some effort on your part. First question: "Do you understand that switching the gate of a MOSFET is not a straightforward proposition? Once you understand how to do it you'll be on your way to understanding how fast you can do it. Simulation is a great help in understanding this behavior. I recommend LTSpice.

 

Normally, for a switch, you want the rise and fall time to be a small percentage of the total switching frequency period, since, during the switching time, the MOSFET is dissipating more power than when it's fully on. This allowable percentage depends upon the application and what switching efficiency you want. This determines the maximum switching frequency you would use.

 

Well, I'm experimenting with a LED array of 10 controlled from a decade counter with a maximum frequency of 5 MHz at each output. The purpose being to light up a room while using the least amount of power. The LED's are Cree CXA1304 high power.

Crutschow, would you mind giving me an example?

Papabravo, that's a good idea, but I'm not familiar with LTspice. I am familiar with Pspice though. I've used MOSFET's for this purpose before, but you are right that I don't have a complete understanding. I'm not sure how to simulate this though. Would you mind giving me a good place to start?

Thanks a lot,
Mike

 

If you want to "light LEDs using the least amount of power" then you need to reduce the frequency right down.

Once the frequency rises, the switching losses (because of slow turn on/off of FETs and LEDs) get very high because switching losses are proportional to frequency.

And why the need for a decade counter? Is that to have only one LED on at a time, to reduce peak current of the entire circuit?

 

Yes the idea is for only one LED to be on at a time. Each CXA1304 LED will run at 18 V and 400 mA, which gives 7.2 watts each plus whatever switching losses there are. Or am I looking this wrong?

Thanks,
Mike

 

....................
Crutschow, would you mind giving me an example?
....................

Example of what? I gave you some general rules for determining the maximum switching frequency from rise and fall times, and desired efficiency. As a general rule you probably want the total of the rise and fall time to be no more that 10% of the period. So, for example, if the rise and fall time were 100ns (200ns total) then, based upon this 10% criteria, the maximum switching frequency would be 1/2μs = 500kHz.

 

Yes the idea is for only one LED to be on at a time. Each CXA1304 LED will run at 18 V and 400 mA, which gives 7.2 watts each plus whatever switching losses there are. Or am I looking this wrong?
....

I think the eye tends to integrate switched light to give an average intensity so I don't think switching the LEDs will save power over just running them at a steady, but lower current (if you use an efficient switching constant-current regulator to power the LEDs).

But if you do still want to switch them you don't need to do that at more than a 100Hz or so since the eye can't perceive flicker above that frequency.

 

That's the kind of example I was looking for. Thanks!

So, am I only looking at the rise and fall times on the data sheet? For the transistor data I attached in my first post the total is 25ns times 10 gives 250ns. This would give me a maximum frequency of 4 MHz. Am I to understand that if I switch them too fast, they will get too hot? There's going to be one transistor connected to each output of the decade counter, so each transistor will be on 10% of the time. This would seem to suggest that I can go to a higher frequency than 4 MHz.

Thanks a lot,
Mike

 

I think the eye tends to integrate switched light to give an average intensity so I don't think switching the LEDs will save power over just running them at a steady, but lower current (if you use an efficient switching constant-current regulator to power the LEDs).

But if you do still want to switch them you don't need to do that at more than a 100Hz or so since the eye can't perceive flicker above that frequency.

Ten LED's on at the same time is 72 watts versus 7.2 watts when just one LED is on at a time. That's how I'm seeing it. Also, the faster they flash, the brighter they will be. I know that we cannot see beyond about 10 frames per second.

Thanks,
Mike

 

..............
So, am I only looking at the rise and fall times on the data sheet? For the transistor data I attached in my first post the total is 25ns times 10 gives 250ns. This would give me a maximum frequency of 4 MHz. Am I to understand that if I switch them too fast, they will get too hot? There's going to be one transistor connected to each output of the decade counter, so each transistor will be on 10% of the time. This would seem to suggest that I can go to a higher frequency than 4 MHz.

It's the sum of the rise and fall times so 10 times 50ns gives 500ns or 2MHz maximum.

What is the counter frequency?

 

.......................
Also, the faster they flash, the brighter they will be. ..............

There's no basis for that statement. Once you are beyond the frequency where the flicker is not discernible by the eye, further increase in frequency will have no visible effect.