Linear regulator

Well, sort of...

In the early 1980's when the programming voltage for a 2716 EPROM was

25 volts and the read voltage on Vpp was 5 volts. I used an LM7805 linear regulator to do the job. The ground leg was in series with a variable resistor which was set to provide 25 volts. The junction of the variable resistor and the 7805 was connected to an open collector TTL buffer.

I was able to switch cleanly between 25V and 5V at a rate of over 10KHz.

I hope this information helps.

An LM317 wants/needs 4 mA to bias normally; if you have that much drive, why not just use a resistor and zener diode? For better precision, maybe a TL431 shunt regulator?

Thanks, well I need to have a stable 12V PWM output since the input can vary +/- 8V. I have a load of 100mA on the output.

A quick and dirty method if your load is relatively stiff is to use a 60/120 voltage divider. This will give you 100mA into your 120Ohm equivalent load if it is purely resistive. The resistors will dispate about 2.4W and 1.2W each.

This assume your inpute is 24V and doesn't fluxuate. You said Input "signal" of 24V then say "the input can vary +/-8V".

If your input does very and/or your load is not stiff(which it can be made stiff but essentially follows whit3rd suggestion) then you can try to use a linear regulator with the PWM after the regulator.

The idea is to put a mosfet between the regulators output and the load with your 24V pwm driving the mosfet.

This way you get regulation and PWM.

Pwr->LReg->Mosfet->Load->Gnd or Pwr->LReg->Load->Mosfet->Gnd

This will pretty much suit your needs but is fairly more complex if you don't know about mosfets and such.

Two main things to note. The mosfet cannot be driven directly from the 24V PWM since it is too high so some method must be used to reduce the voltage. Since you have such a low PWM rate a simple voltage divider would probably work just fine.

High side NCh requires a floating gate which can be achieved through several means. Because of the low PWM rate you can simply this using a few passives and zeners.

You can use a low side Nch and probably can achieve the same result without any issues(depends on what you are driving though as it might not work well).

The LReg might require a small load for biasing which is easy enough by adding an appropriate resistor in parallel with the the load + mosfet.

Also, the LReg could be replaced with a switching regulator or other method since it's only purpose is to supply power.

The main point is that your approaching it wrong and hopefully I've cleared up how you should go about it. It is possible to use whit3rd method if you don't wanna go about it this way. It all really depends on what your trying to do.

It is a bad idea to PWM the regulators power though as it is not designed for such things. Although I have no real idea if it would work or not. If you really want to try it though then go ahead and see. Just don't expect it to work nor expect it to work in general.

If the power dissipation is bothersome, add a transistor; collector of NPN to the 24V pulse input, base to 12 V zener, emitter drives the load with 12V maximum...

A linear regulator is slow to initialize, you'll get a few microseconds of unspecified behavior at each rising edge. If there's any capacitance on the output (as most regulator datasheets recommend), the reverse bias at shutdown is also a problem (clamp diode from output to input is the usual recommendation). One could use a linear regulator from steady 24V with a logic-disable input, but that wastes the high-current driver you already have.