Hello,

Currently trying to develop basically an analog switch with very low RDSon to switch a pulse up to 10V with a frequency of 20MHz. This 20MHz pulse will directly feed a load that can pull up to 1A. As you can tell RDSon needs to be very low <10mOhms for it to have virtually no impact on the signal being delivered. From my research there is no analog switches that are able to satisfy so I'm trying to do it with N Mosfet on the high side of the load. To test the circuit I use a PWM (Blue signal) to switch the low RDSon mosfet with a higher voltage.

At MHz frequencies I start seeing spikes and weird behaviors whenever the mosfet is not conducting. Selection of the mosfet seems crucial here due to their body properties. I've been able to minimize those weird behaviors from less than 2V replicas of the signal to these spikes seen below. What else can I do to stop having this unwanted behavior? Is there a better topology for a circuit with this function?

 

Switching any device at 20MHz is problematic.
What is the purpose of this signal?

 

Switching any device at 20MHz is problematic.
What is the purpose of this signal?

The 20MHz pulses are meant for characterization of a device. By delivering up to 1A to it.
I will test multiple devices, by having multiples of U1 which I will switch one at a time every other second

 

datasheet: https://www.diodes.com/assets/Datasheets/DMN22M5UFG.pdf

Watch out for the max gate-source voltage of +/- 12 volts.

The input capacitance is about 4nF, that with the 1K pull up in the drain of the 2n7002
limits the turn on time around 4uS much too slow. I'd also guess that the on resistance
of the 2n7002 is too high to get enough current out of the DMN22M5UFG gate to turn
it off fast too. You need V*C/t -> 5*4-n9/20e-9 -> about an amp of gate current each way.

 

The 20MHz pulses are meant for characterization of a device.

Can it be done with a sinewave or a rectified sinewave?

 

Does this help?

 

Can it be done with a sinewave or a rectified sinewave?

The source of the pulse will already be supplied as a square wave

datasheet: https://www.diodes.com/assets/Datasheets/DMN22M5UFG.pdf

Watch out for the max gate-source voltage of +/- 12 volts.

The input capacitance is about 4nF, that with the 1K pull up in the drain of the 2n7002
limits the turn on time around 4uS much too slow. I'd also guess that the on resistance
of the 2n7002 is too high to get enough current out of the DMN22M5UFG gate to turn
it off fast too. You need V*C/t -> 5*4-n9/20e-9 -> about an amp of gate current each way.

I believe I've been misunderstood. I don't need to switch the mosfet at those frequencies. Here is a block diagram for better interpretation

 

Does this help?

View attachment 320852
View attachment 320853View attachment 320854

Hi Irving,

Here V(out) looks much better. It definitely helps, but I wonder if I could achieve the same results without making use of the UCC21220A. Since I have to test several DUTs on board, I would like to keep the design as discrete as possible.
But if UCC21220A is really needed I can then maybe do a matrix topology, controlling both the high and low side of our DUTs.

 

"Device Characterization" is not always simple. This test looks more like a component source verification step in source approval.
At 20 MHz, driving and removing the gate signal will require a drive circuit with a much lower effective impedance.
So which device characteristic is it that will be evaluated with this check? With the arrangement described it could be running in a thermal chamber, which is always a bit more complicated.

 

I use GanFETs to switch fast. The gate drive current is amps when switching fast.
GanFet