Faster PMOS turn on time

[jonnyd42]

I'm taking a 3.3V microcontroller output (PWM14L), and using a common source single stage amp to drive the 6n137 optocoupler using 5V. Ignore the fact that things get inverted since I can account for that in code.

I want to be able to drive a load with a few possible configs:

• 5V, < 1A
• 12V, 1.5A
• 24V, 0.75A

For now, this is just handled by having a solder jumper. For this example, just ignore the implementation of the voltage source connected to the PMOS. On the bench, this circuit performs poorly. The PWM that comes out of the PMOS drain is different in duty cycle from the input because the rise time of the PMOS is so slow. I tested using a SI2371EDS-T1-BE3 since I didn't other PMOS parts to test on the bench. I observed that the gate gets pulled up very quickly through the 1k, so turn off is very fast and is well aligned with the input signal. PMOS turn on is super slow, and my guess is that the optocoupler pulling the gate to GND is not able to discharge the PMOS's equivalent gate capacitor quickly enough.

What ideas would you have for solving this? My preference would be to do this with the minimal # of parts, but I'd love to hear both gate driver (IC) and discrete ideas!

 

[KlausST]

Hi,

an optocoupler is made for galvanic isolation.
But in your circuit there is no galvanic isolation.

Thus my recommendation:
* omit the optocoupler
* use a MOSFET driver (it´s design to drive MOSFETs)

Klaus

 

[jonnyd42]

Hi,

an optocoupler is made for galvanic isolation.
But in your circuit there is no galvanic isolation.

Thus my recommendation:
* omit the optocoupler
* use a MOSFET driver (it´s design to drive MOSFETs)

Klaus

Yeah, initially those were supposed to be separate grounds but its TBD at the moment. If they end up the same as in the schematic, then there's no point.

Instead of going full driver IC, thoughts on the following simpler circuit? NMOS doesn't have to be the AOD4184a since there isn't a need for the NMOS to support a large I_d (just the PMOS since the current will go to the load), but I just copy pasted for speed's sake.

 

[KlausST]

Hi,

if possible, a low side driver is more simple and less expensive.

The headline talks about faster turn ON time
But in post#1 you are concerned about duty cycle.

What are you after?

If you just want a faaster turn On time, then the new circuit should be good.
But if you are after accurate duty cycle then it´s not optimal, becaue now I expect a slow turn OFF.

And if you ant fast edges you need to use capacitors to compensate for stray inductance.
--> Stabilize your power supplies with capacitors.

May opinion: a circuit without a capacitor is no (complete) circuit. (fur sure this is not true in either case)

Klaus

 

[jonnyd42]

Hi,

if possible, a low side driver is more simple and less expensive.

The headline talks about faster turn ON time
But in post#1 you are concerned about duty cycle.

What are you after?

If you just want a faaster turn On time, then the new circuit should be good.
But if you are after accurate duty cycle then it´s not optimal, becaue now I expect a slow turn OFF.

And if you ant fast edges you need to use capacitors to compensate for stray inductance.
--> Stabilize your power supplies with capacitors.

May opinion: a circuit without a capacitor is no (complete) circuit. (fur sure this is not true in either case)

Klaus

Well they are related. In the circuit I posted, if the PMOS is not able to turn on quickly enough the the duty cycle of the outputted waveform will change (since there will be a time delay that's significant when comparing the input PWM from the MCU to what comes out of this circuit).

Yeah - you're right that I would like edges to be as fast as possible. Can you suggest how to go about sizing the capacitors and where you'd suggest adding them?

 

[KlausST]

Hi,

I'm still unsure. Turn on? Or duty cycle?

I mean you could target for turn ON to be fast in the 10s of ns.
But if the turn OFF is 10us slow, it manipulates the duty cycle

If you need accurate duty cyvle, there is no need to be as fast as 10ns ... as long as both edges are symmetrically slow.
So if turn ON = turn OFF = 10 us .... the duty cycle is accurate

Capacitors: you did not give a single timing value. No frequency, no target for duty cycle accuracy, no duty cycle range, no delay timing...
.. so how can we calculate any capacitance value?

Klaus

 

[jonnyd42]

Hi,

I'm still unsure. Turn on? Or duty cycle?

I mean you could target for turn ON to be fast in the 10s of ns.
But if the turn OFF is 10us slow, it manipulates the duty cycle

If you need accurate duty cyvle, there is no need to be as fast as 10ns ... as long as both edges are symmetrically slow.
So if turn ON = turn OFF = 10 us .... the duty cycle is accurate

Capacitors: you did not give a single timing value. No frequency, no target for duty cycle accuracy, no duty cycle range, no delay timing...
.. so how can we calculate any capacitance value?

Klaus

Thanks for the patience, trying to learn how to do this better!

Duty cycle accuracy is what's important, so yeah if the wave is just shifted in time but the cycle is preserved that's fine! To be honest, I don't have an accuracy target...just "as good as possible". This isn't being designed to a spec so it's not like that exists at the moment.

Max frequency would be 500Hz.
Duty cycle range is 2% to 98%
Delay timing - would be good to have the delay between the MCU rising edge and output rising edge be < 10us, same with the MCU falling edge and output falling edge.

 

[KlausST]

Hi,

You current use two transustors as open drain with pull up / pull down.
This is unsymmetric. The transistor pulls to the one side (when ON with high current) and a resistor pulls to the other side (with relatively weak current).
To avoid this I recommend to use a totem pole drive method, where you have a transistor in either direction ... with rather symmetric current and rather symmetric timing.

You may build this with discrete parts .. or use a dedicated IC, like the TC4420. (Just as example, there are many others).

Pedantically ;-) "as good as possible" is the most extreme and thus the worst goal. Let's say you are down at 1ns .... and there is a peron that says: this is not as good as possible, I have an idea for 0.999ns ..... it will never end....

Indeed, even for hobbyists I strongly recommend to decide and write down the goal as value .. and do this before you start to write / look for a schematic.
This helps to be focussed. And this helps to not "over engineer" one part of the application.
Let's say you use standard power supplies and standard resistors. These may have a voltage output tolerance of let's say +/- 1%.

And you want the duty cycle accuracy in the same region (maybe you do PWM filtering to create something like a DAC).
Then given with your 500Hz (2ms) the duty cycle "symmetry" needs to be within +/- 20us.

Especially when you talk with other people (forum) a value is very helpful. Phrases as "fast" are random. Professionals when hear "fast turn on" often think in the low nanoseconds (not microseconds .. which means a factor 1000).

Back to your case:
So if you want more symmetry .. you could lower the (pull) resistor values, or make the transistor more "weak" by using gate series resistors.
Maybe even in the 10k range. Play around to see what happens. And you automatically will understand and learn....

Goid luck

Klaus

 

[d123]

Hi,

Some speed-up and slow-down options you might be able to adapt to your design.