How can I slow down the rise and fall time of a power MOSFET without adding a delay before the switching begins?

Question

I'm building a simple 5 W HF CW transmitter, where the power to the final stages is controlled by a P-Channel Power MOSFET, whose gate is connected to a key.

I added a simple capacitor between the gate and ground, for two reasons:

1. To debounce the key (LTspice shows this will debounce up to 1ms bounces, which should be adequate)
2. To avoid key clicks on the transmitter - this requires that rise and fall time be on the order of 2ms. (See here for the standard approach.)

My circuit below achieves #1, and raises the rise and fall time to about 1ms, somewhat achieving #2. However, it injects a propagation delay of 2ms, presumably the time it takes the capacitor to cross the gate threshold. Is there any way I can decrease that propagation time, while keeping (or ideally increasing) the rise and fall time?

I've tried different values R2, R4, and C1, but nothing improved on the above. I also added an 100uF capacitor across the load R1. This may have increased the fall time, but didn't affect the rise time (presumably because the MOSFET can charge the capacitor quickly).

---

Clarification

To clarify my question, the circuit currently behaves as follows:

1. Assume key down at t = 0 ms
2. Power starts to fall at t = 2 ms (2 ms propogation delay)
3. Power falls for 1 ms, completing its fall at t = 3ms

What I'd like is to shorten phase #2 and lengthen phase #3. Something like:

1. Assume key down at t = 0 ms
2. Power starts to fall at t = 0.500 ms (500 ns propogation delay)
3. Power falls for 2 ms, completing its fall at t = 2.5ms

Re debouncing: The circuit as-is can debounce a key, no matter how long it bounces for, as long as the length it stays up or down on each bounce is less than a 1ms. Is there a term for that? My understanding is that while a switch may bounce for 5ms or 10ms, the length of each bounce is several orders of magnitude less than that.

Answer 1

No, you cannot. That's simply logically impossible: you want an N ms rise time.

Because the rise can't start before you press the button (otherwise, your circuit would need clairvoyance) the rise can only be finished N ms after you started to press the button. There's simply no effect before the cause – causality :)

You can of course make the rise steeper - simply reduce your R4 or C1. But that also makes it shorter. If you need a different kind of shape than the \$1-e^{-t/(R_4C_1)}\$ you get from an RC filter, you will need to build a more complex filter. That might be a ladder filter, it might be a more exponential ramp, or anything, including letting a microcontroller with a DAC control that FET, but: Whatever you do to make the shape "sharper" in time will make the shape "wider" in frequency – and you get the clicks back you wanted to get rid of. In the end, the same inherent bandwidth-time product limit applies to you as to Heiseberg's uncertainty principle: the steeper your filter in one domain, the wider it becomes in the other. Math is not negotiable there! So, you'll have to live with some delay if you want limit your bandwidth.

A few remarks:

1. I know Morse paddles can be very high-quality switches. Still, 1 ms is a very short debounce time.
2. there's absolutely no problem with a 2 ms delay between you closing the circuit and the tone happening – its stopping gets delayed by nearly exatly the same time, so the periods stay exactly as long. And: you might have excellent Morse reading, and latency might confuse you, but it's unlikely to be a problem at this level. From stage and recording technology we know that even professional musicians don't notice delays below ca. 4 ms (you'll find that 3 ms is kind of the magical delay that professional wireless microphones strive to attain). The nerves from brain to finger aren't even working that fast!