Relaxation oscillator issue, comparator self-oscillating at 25 MHz (40ns propagation delay)

Question

I have a relaxation oscillator circuit which is supposed to oscillate around 6 MHz. It mostly oscillates in a very stable manner around that frequency.

Sometimes, however, it starts to oscillate at 25 MHz which is the fastest oscillation available for this comparator (TLV3202) given its propagation delay is 40 ns. Some cases where this happens are:

1. On frequent on-off, I can put the circuit in this mode.
2. Bringing fingertip near inverting input, not touching just in the air.
3. Touching an open ended wire to non-inverting input.
4. During normal handling of running circuit we have seen this.

As this circuit will be used for further capacitive sensing, we are in trouble.

A few observations:

• Touching the non-inverting input with an open ended probe brings it back to 6 MHz
• A power reboot also fixes it.

Circuit diagram:

PCB layout:

A few things that I tried to fix this, with no success:

1. Putting a 10 pf capacitor between comparator-output <----> comparator-non-inverting input.
2. Putting a 200 ohm resistor on comparator-output.

Answer 1

A few things.

1 - You appear to have 3 decoupling caps on the comparator, on the left side of the IC. While the V+ lead is nice and short and compact, the ground trace for the two lower (presumably smaller) caps heads off to who-knows-where, and this will make them less effective. Especially the 0.1, which needs both traces as short and thick as possible to the IC pins. If you're not willing to redo the board, try running a small (as short as possible) jumper from the two smaller caps to the ground pins of the IC.

2 - Grounding both pins of the unused comparator is a bad idea. If you get a unit with extremely small input offset voltage, the unused portion may be primed to change state in response to very small inputs, and this transition can, in turn, affect the other section. Try lifting pin 6 and jumpering it to pin 8. The TLV3202 is rated for rail-to-rail inputs, so this should work.

Answer 2

From the symptoms ---- vulnerable to various pieces of metal near the pins 1,2,3 --- this seems like Electric Field feedback, thru the air.

Thus 2 steps to a cure

(1) capture some of the electric field coupling from output to non-inverting input; these are pins 1 and pins 2. Perform this capturing, by drilling 2 or 3 or 4 holes down thru the GND plane, and install some vertical metal pins on both sides of the feedback resistor. These grounded bits of metal serve as RETURN_PATHS for the Efield.

Your PCB, which looks good, by the way, has direct coupling between pins 1 and 2, INSIDE THE PCB, under pins 1 and 2, inside the FE-4 dielectric. NOthing to be done about that, unless you could bring the Grounded foil to be closer to those pins, which would capture more of the Efield underneath the solder-pads.

(2) add a small capacitor (5--10pF) across the +feedback resistor R13

(2a) reduce the 4.7K ohm resistors (3 of them) to 1Kohm

And where is the timing capacitor?