Voltage controlled duty cycle with a CMOS inverter

The first mass-production switch mode converter I designed in my life also had to be dirt-cheap. So I took one out of the six Schmitt inverters of a CD40106, cap from input to ground, resistor from output to input, transistor from input to ground. Four parts. Sending current into the base skewed the duty cycle (all the way to off if needed) and the 600mV Vbe provided a zero-Dollar reference. if you don't need a reference but more voltage swing you could use a FET instead but those cost a cent more. Or use a "digital" transistor (pre-biased).

Can also be done with 5V logic Schmitt inverters but must be Schmitt. Hang a "For Lease" sign on the other five inverters, maybe a colleague can use some :-)

If you need the duty cycle to also go higher than 50% add another feedback resistor and a diode. Adding a resistor from Schmitt input to some noisy or at least erratic source in the circuit can help with EMC by smooshing the spectrum footprint.

Sometimes I design with ECL, at $3 to $30 per gate, and then have to get back to groveling for minimum gate count. In an FPGA, there's an infinite supply of gates.

I like single SOT23 opamps. Put them wherever they fit best.

About two minutes ago I designed a gated schmitt oscillator into my new box. There is a time when we need to run a small state machine and there's no convenient clock available. But it looks like I'll have to erase it (that's OK, I have an electric eraser) because it's too slow. Schmitts are always slow compared to other gates. The fastest inverting schmitt we have in stock is 3.7 ns typ at 3.3 volts, which would be marginal even with no capacitor.

I'll have to do something else.

I tossed the Schmitt and went for a 65LVDS2DBV in my recent designs (didn't you suggest that some time ago?), very fast propagation delay and you get to precisely set the threshold voltages and hysteresis, same sot-23 package.

I tried the one from Lasse in LTspice, works fine

It will be sensible to hysteresis drift of the inverter, but for example for the 74HC14 a min and max hysteresis band of 0.6V and 1.6V is defined.

Since I have a regulation loop wrapped around it, I do not care about hysteresis shift, it will be regulated out. It will however change the frequency, but for my application that is ok (I am driving a hyper cheap buck converter with it)

It's nice and straight forward, with little drift. But my application cannot have the cost of that added 555

I am scrambling for every cent, weighing off low cost versus reliability considerations.

Cheers

Klaus

Lasse's simulation works

Cheers

Klaus

I have that too, but it literally hurts my eyes to see a design with cost for no significant reason

The application is ok with the hysteresis band drift

I have a comparator at hand, so I could make the PWM in that one, but the 1.2us response time delay of the LM2901 in that implementation is troublesome, which causes the minimum pulse with to be non-zero.

Cheers

Klaus

I have never used a 555 in an actual product design

I am considering using the corner of the microcontroller as the PWM generator, but the review is going to be a heated discussion if I go down that road

Cheers

Klaus

I like the challenge of getting the last function out of a spare gate, or using a super cheap component in place of an expensive component

Life is to easy if I can just place whatever part I need and the company bottom line suffers.

Cheers

Klaus

Probably. Those are amazing parts for about 60 cents each. It's an rri comparator with sub-ns output risetime.

very fast

The 65LVDS2 is 2.6 ns typ, which might be OK. I could maybe make a delay-line oscillator, not an RC schmitt, to keep the frequency more accurate. Or even a triggered LC oscillator. This thing I'm designing is really, really hard, which is fun to do now and then.

The 65LVDS2 is a nice part, I like its sot-23 package, and its better then the comparators I was using before. But it's over 10 years old now, there are much faster LVSD receivers, better suited to your needs. So when you find the perfect super-fast part, let us know.

Oh, I agree there. I recently did a very sensitive (picoamps) front end usi ng an MCP6002 and an LM324. Works great.

It'll suffer a lot faster if penny pinching leads to field failures, which is where I was going. JL has shipped a lot more units than I have, but I'd be nervous about the super-loose logic threshold specs for CMOS--roughly 1/

Cheers

Phil Hobbs

Do you know of a faster receiver? The lvds-lvds parts are getting faster, lvds-cmos not so much.

Perhaps true when Jim Thompson was last susceptible to new ideas.

The TI web-site wasn't all that helpful

does seem to pick similar low-voltage digital signalling line receivers.

Motorola ECL line receivers are a more traditional choice.