Switching a 20 ns pulse with 2N3904 BJT transistor

Thread Starter

Wamor

Joined Feb 12, 2016
7
Hello,

First let me introduce myself. I am learning electronics combined with FPGA.
For some test I need to create a pulse generator with variable duty cycle and an output voltage of 24 V.
Minimum pulse width is 20 ns and I need to convert a 3.3 V FPGA signal to 24 V using a 2N3904 BJT.
I am not sure that the transistor can switch this short pulse but I know that I can improve transistor switching by using a speed up capacitor and Schottky diode. My question is: is it possible to archive this using a 2N3904?

Thank you all for your responses.

Regards,

Wamor
 

Thread Starter

Wamor

Joined Feb 12, 2016
7
A 2N3904 is not a fast switch.
Why not use a MOSFET?
They are generally much faster switches.
A 2N3904 is not a fast switch.
Why not use a MOSFET?
They are generally much faster switches.
Thanks for taking the time to reply.
Can a MOSFET switch a pulse of 20 ns? For a standard MOSFET the switching delays are already 10's of ns.
So making a pulse with a width of 20 ns would be difficult? Do I need a special MOSFET for that then?
 

AnalogKid

Joined Aug 1, 2013
8,259
In very round numbers, bandwidth equals 0.35 divided by risetime. So if the 20 ns pulse has 2 ns risetime, that works out to about 175 MHz. From 3.3V to 24 V is a gain of 7.3. Times 175 is a GBW product of about 1.3 GHz. That's a lot, and way beyond what the 2N2222/3904/4401 generic small signal transistors can do.

ak
 

RichardO

Joined May 4, 2013
2,271
I was going to suggest a 2N2369 instead of the 2N3904 and then noticed the 24 volt requirement. When you get above about 10 volts, finding high speed BJT' switches gets difficult if not impossible. :(

For a MOS-FET look at SD210, SD211 and the like. Fast, low capacitance and (barely) high enough drain-to-source voltage.
 

Thread Starter

Wamor

Joined Feb 12, 2016
7
I was going to suggest a 2N2369 instead of the 2N3904 and then noticed the 24 volt requirement. When you get above about 10 volts, finding high speed BJT' switches gets difficult if not impossible. :(

For a MOS-FET look at SD210, SD211 and the like. Fast, low capacitance and (barely) high enough drain-to-source voltage.
Thank you, I will have a look to these devices.
 

crutschow

Joined Mar 14, 2008
24,130
What is the drive capability of the FPGA output?
Even if you find a fast enough transistor, it can be a problem just driving it.
 

MrAl

Joined Jun 17, 2014
6,979
Hi,

With the mosfet you would have to use a high current gate driver.

With the bipolar you would have to find a faster one, and also make sure it does not go into saturation. If it is allowed to enter saturation the pulse time will stretch out quite a bit because it will not be able to turn off fast enough due to the storage time of the transistor.
 

DickCappels

Joined Aug 21, 2008
6,100
Another approach is to avalanche the transistor -you can get very large very fast pulses plus high peak powers with tiny transistors.


With the circuit above the output pulse width showed up as being a 28 volt peak Gaussian with a half-amplitude width of 4.2 nanoseconds.
Replacing the 20 pf capacitor with a length of coax an LC network that looks like a transmission line will stretch out the pulses (tried that myself). Notice that it uses a 2N2222, but a 2N3904 also works well. No semiconductors were harmed during the making of this experiment.)
 

Thread Starter

Wamor

Joined Feb 12, 2016
7
I looked at the suggested transistors by RichardO but they all look through hole.
Are there also smd-types available?
 

crutschow

Joined Mar 14, 2008
24,130
Below is the best I could come up with for a relatively simple circuit.
It uses a Schottky diode clamp to prevent Q3 from saturating.
The rise-time can be reduced by using a lower R1 value, but that increases the power dissipated. As is, the 400Ω resistor dissipates about 1.4W when the transistor is ON so it needs to be at least a 2W non-inductive (non-wirewound) resistor (or two 800Ω, 1W resistors in parallel).
On the real circuit you may have to experiment with the value of C1 to get the desired response.
Note that the circuit inverts the signal.

Pulse Gen.PNG
 

Attachments

Last edited:

crutschow

Joined Mar 14, 2008
24,130
Note that I didn't show any decoupling capacitors.
At a minimum, there should be a 0.1μF ceramic directly from the top of R1 to the emitter of Q3 and another from the collector of Q1 to the emitter of Q2.
 
Last edited:

crutschow

Joined Mar 14, 2008
24,130
See
It is necessary to take into account the load capacity. When the load capacitance of 10 pF collector resistor is 400 ohm too.View attachment 100749
Good point.
Just be aware that an R2 value of 150Ω is dissipating 3.8W when Q1 is conducting.

Below is the circuit with the addition of an emitter-follower output to improve the capacitive load drive capability.

Pulse Gen.PNG
 

Attachments

Last edited:

RichardO

Joined May 4, 2013
2,271
I looked at the suggested transistors by RichardO but they all look through hole.
Are there also smd-types available?
Oops. Those are really old parts...

A quick Digi-Key shows a lot of wierdly packaged power devices. More research needed. :(
 
Top