@Ufo, oh boy it is surely is addictive
I surely hope all turns out well for you eventually.

You have a kind an generous soul.

Regards, Bert

Hi torpex,
the switch off time behaves similar.
LM339 output is supposed to sink 6mA minimum / typical 16mA. In fact it will drain more than data sheet points out because the manufacurers give secure data they never will exceed.
In this case the discharge cycle is difficult to calculate because we have no constant current but a transistor with dynamic behaviour. Don't forget that the LM393 needs to sink the current from the pullup resistor additionally to the discharge current out of the gate capacitance.
This is the reason why I suggested to source about 10mA only by the pullup resistor. So you have some freedom to sink both currents (pullup + gate cap) at same time.

BTW: Professional FET drivers source and sink up to several amps for short time!
Anyway it is no gimmik at all if there is the advice to use multiple comparators in parallel.

Alternatively you might decide to follow the advice using a NE555 for every FET. Previous post They sorurce and sink 8 times the current of a LM393. The solution is somewhat more straight forward and separates the oscillator from the drivers.

BTW: The 10 Ohm resistor at output will dampen parasitic oscillations between gate capacitance and the inductance of the wires. see
This is a problem at IGBTs as well!

The current suggestions for control circuitry are difficult to understand by newbies or can not be adjusted in a linear manner.

I confess I was lazy and bought a PWM kit. I like this circuit and I share it for your convenience. The circuit features nicely separated functional blocks - so it can be built and debugged easily - step for step!
10Hz up to 100Khz in 4 overlapping ranges.

AND MOST IMPORTANT: IT WORKS!

1. It runs from 5V so use any 5V regualtor available (LM317, 7805 .....) Modifications for 12V possible but not advisable. With 5V we can start with a single 12V battery for test in order to prevent blowing FETs in case of failure.
2. Most left is a variable current source - charging capacitors - linear behaviour! The LM385 reference diode (2.5V) can be replaced by 4 diodes in series for now. This part can be built first and and measured with a simple DVM. You might decide to use a 100µf cap first in order to be able to verify teh basic function in very low frequency mode with your simple DVM.
3. The 393 is dedicated to discharge control. It is a very smart circuit and I will elaborate more on it if requested.
4. Output 7 of LM393 outputs a sawtooth signal.
5. The most right IC is a super fast comparator with FET in ns range. You do not need this spec just now. It can be replaced by another double LM393 being nicely separated from the oscillator!!!! This comparator converts the saw tooth signal into a pulse width signal.

I do not want to seduce anybody for using this circuit. If you are content with your circuit - use it! If you are not content you might decide to try this one. On request I will elaborate more reagarding the function and will draw an appropiate schematic.

JohnStone
If you can supply a full circuit and a parts list if possible, I will build it. What size board should it be set on? I see output voltage to 20V on the diagram and guess that this is the max?
Dana

• Please give me some days in order to modify the original schematic.
• You can build it on 2" x2" but I reccomend to use about 4" x 4" in order to have good chances for measurement and additions.
• 20V is because of special voltage regualtor and special logic and fast FET but I will replace them with standard parts as I posted obove.

BTW: LM317 calculator for stock resistors: see