Can I use an RF MOSFET in switching application?

[buenos]

Can I use an RF MOSFET in switching application?

I need to turn on and off about 200mA current at a rate of 100MHz.
Can I use an RF MOSFET for this? Those FETs that are marketed as switching FETs have turn on and turn off times in the range of 20ns or much more. To be able to nicely switch at 100MHz I need a FET that turns on or off in 1-2ns max.
Also RF FET datasheets don't mention turn on or turn off time. How to correlate the available data to turn on/off times?

 

[jiripolivka]

Can I use an RF MOSFET in switching application?

I need to turn on and off about 200mA current at a rate of 100MHz.
Can I use an RF MOSFET for this? Those FETs that are marketed as switching FETs have turn on and turn off times in the range of 20ns or much more. To be able to nicely switch at 100MHz I need a FET that turns on or off in 1-2ns max.
Also RF FET datasheets don't mention turn on or turn off time. How to correlate the available data to turn on/off times?

A RF MOSFET is a fast switch and possibly can do the job. 100 MHz corresponds to 10 ns. If you need a switch that switches within 1-2 ns, you need a device capable of 500 -1000 MHz operation which is not a MOSFET.

200 mA current at what impedance? If you use 50 Ohms, the power is 2 Watts, so you can find a PIN diode switch with a speed of 10-20 ns. No device I know can do this as fast as 1-2 ns.

MOSFET specifications include input and output capacitance, so you must design the source and load for a good match acordingly. RF MOSFETs and BJTs are often low-impedance devices that need wideband transformers from 50 to 1-5 Ohms.

 

[FvM]

The basic point is to understand that you are designing an analog circuit that must be analyzed as such. Various FETs and BJT with multiple GHz bandwidth are available and might be used for the design.

 

[buenos]

OK, any of these could do the job?
https://www.digikey.com/product-detail/en/PD55008L-E/497-6473-1-ND/1865507
https://www.digikey.com/product-detail/en/ATF-531P8-TR1/516-2239-1-ND/2404155

 

[mtwieg]

Can I use an RF MOSFET in switching application?

I need to turn on and off about 200mA current at a rate of 100MHz.
Can I use an RF MOSFET for this? Those FETs that are marketed as switching FETs have turn on and turn off times in the range of 20ns or much more. To be able to nicely switch at 100MHz I need a FET that turns on or off in 1-2ns max.
Also RF FET datasheets don't mention turn on or turn off time. How to correlate the available data to turn on/off times?

I think I can help you, since I've spent the past couple years working with switchmode RF amps in the 100MHz range. I typically use GaN FETs from EPC. LDMOS can work as well, but as you noted, RF FETs are almost never specified for switchmode operation, but those GaN FETs are.

edit: also I've used a BSD316N_L6327 MOSFET for lower power class E amps, but efficiency is much less.

 

[FvM]

What's intended source characteristic and load impedance?

 

[buenos]

I think I can help you, since I've spent the past couple years working with switchmode RF amps in the 100MHz range. I typically use GaN FETs from EPC. LDMOS can work as well, but as you noted, RF FETs are almost never specified for switchmode operation, but those GaN FETs are.

edit: also I've used a BSD316N_L6327 MOSFET for lower power class E amps, but efficiency is much less.

Sounds good. Can you give me a few part numbers as a reference?
[added]: I see, you provided a link. The datasheet does not specify switching timing, only capacitance and gate charge. How to correlate turn on and turn off time? Or are there other devices where they specify this directly?

What's intended source characteristic and load impedance?

OK, so I want to drive a laser diode that has no PD pin. 140mA-to-0mA switching, or 140mA-to-100mA switching. Not sure. I would use a series resistor to set the current. Previously I implemented this at 6MHz with a MOSFET I bought in a shop many years ago. Now I need to go up to 80-100MHz.

 

[mtwieg]

Sounds good. Can you give me a few part numbers as a reference?
[added]: I see, you provided a link. The datasheet does not specify switching timing, only capacitance and gate charge. How to correlate turn on and turn off time? Or are there other devices where they specify this directly?

If you only need <1A of output current then just go with the EPC2012. I can vouch that it switches in <2ns, though it helps to operate it in a soft switching circuit (class D/E/F). You'll likely need resonant gate drive as well, without exceeding the max gate voltage.

 

[buenos]

I don't think that I can use the EPC2012 as it only comes in a crazy BGA package with 0.4mm pitch.

 

[mtwieg]

It's not as bad as it looks. I often solder them on hand-etched boards with a cheap hot air gun with some tacky flux. And the small size is sort of necessary to keep parasitic inductance low for good switching.

 

[buenos]

so, what was the yield when you first tried to solder it?

Also, tell me more about the resonant gate drive.
Actually I have to run this at different frequencies, the freq will be selectable, 5MHz the lowest, and 100MHz the highest. Any AC spectrum optimization that is in the drive circuit has to work at all of those. Do I need a drive IC (like LM5114) for this, or can I drive it directly from a 50R coax cable's signal?

 

[mtwieg]

so, what was the yield when you first tried to solder it?

About 80% maybe, though after replacing burnt FETs on the same board it gets more difficult.

Also, tell me more about the resonant gate drive.
Actually I have to run this at different frequencies, the freq will be selectable, 5MHz the lowest, and 100MHz the highest. Any AC spectrum optimization that is in the drive circuit has to work at all of those. Do I need a drive IC (like LM5114) for this, or can I drive it directly from a 50R coax cable's signal?

Variable frequency makes things a lot more difficult. No commercial gate driver will be able to work past 10MHz or so (including the specialized GaN drivers). It might be possible to parallel a bunch of high speed, high current logic gates, but with a square wave drive overshoot on the gate due to parasitic inductance becomes a problem (with those GaNFETs anyways). If you really want wide bandwidth then a very small geometry MOSFET might be best, though its on state resistance and switching speed will be worse. At least MOSFETs are more tolerant of overdrive on the gate.

 

[FvM]

Related to usual laser driver technology, 100 MHz/200 mA isn't asking too much. I can just guess that you are probably thinking too "digital". It's definitely high speed analog design.

 

[buenos]

In the past I had some problems with laser driver ICs, the transistor based control worked nicely at 6MHz, so now I am building 2 different prototypes in the same time. One with a laser driver IC from Maxim, and one with transistors.