TLC3702 PWM circuit (triangle generator + comparator output)

[mc172]

The apprentice at work needs a fixed PWM source which can be set with a potentiometer and then left there. I got him to build the circuit attached (taken from Linear's website), but using TLC3702s because I've got some stock laying around. The modulation input is replaced by a potentiometer. I have colour coded the input and output signals on the schematic to what they are on the oscilloscope photos later.

It works fine up to around 6 V supply voltage, but when you start turning the supply up, the trigger point moves up in voltage or right in time, presumably following the gradient of the triangle, depending on your perspective. I have attached images of what is happening.

The green trace is the set point, aka "input", which is the potentiometer, and I expect the PWM output (yellow) to have an edge whenever the blue trace (triangle generator) crosses the green trace. As you go past 12 V supply, the vertical edge of the PWM output sticks to the highest point of the triangle, and never goes past it. You can see this in the "12V supply" photo.

In all three scope pictures, I have not touched the scope settings at all, just turned the supply voltage up. you can see the yellow trace peaks at whatever the supply voltage is, or close enough. It's at 2V/div.

What I don't understand is why the bottom comparator is not triggering when in my mind, it should. I have simulated it in LTSpice (admittedly not using a TLC3702 because there isn't one in there), but it all works fine.

Any ideas please, anyone?

[grumpydoc]

Does it work if driven by a split, rather than single-ended supply?

[T3sl4co1l]

The pot, resistor dividers and amp are all supplied by the same (variable) supply?

The RC time constant is much lower than the response time of the amp, and the R value is much larger than the output resistance of the comparator?

Tim

[mc172]

Thanks for the replies.

I can try a split supply, but I don't see why it shouldn't just work on a single-ended one? The Linear schematic calls for a single-ended supply...

Everything is supplied by a single variable (TTI PL320) supply, with a single pair of wires going to the board (one for +V, one for return).

It's stated on the datasheet as around 50 ns rise, 125 ns fall time,  so yes, around 3 orders of magnitude lower.
The output of the TLC3702 is push/pull CMOS so presumably in the tens to hundreds of Ohms range.

[grumpydoc]

By the way - is the output driving anything (is the RC network connected, for instance) or is it floating?

[mc172]

It's floating for now. Out of interest I tried terminating the output into various resistances, between 4k7 and 47k and no change.

[mvs]

It's stated on the datasheet as around 50 ns rise, 125 ns fall time,  so yes, around 3 orders of magnitude lower.

Its just output stage specs, propagation delay is much worser.

PS Add some decoupling caps, this comparator is known to have propagation delay issues:
http://e2e.ti.com/support/amplifiers/f/14/t/662778

[mc172]

Thanks. I will get him to try it with two separate packages and report back, using only one comparator of each package and forcing the unused ones to a fixed state with a large differential voltage on the inputs. 

[spec]

Thanks. I will get him to try it with two separate packages and report back, using only one comparator of each package and forcing the unused ones to a fixed state with a large differential voltage on the inputs.

Hi mc172,

I would advise you to use a much faster comparator. The TLC3702, just can't keep up and, as mvs pointed out, it is not the most well-behaved chip.