Issue with TL3116 comparator for Tesla Coil

[bmboucher]

Hello all, I'm having an issue with my Tesla Coil's logic board - I'm the rankest of amateurs so I'm hoping someone can point out to me what I'm doing wrong.

Specifically, my board's design is based on this one which is very popular: https://www.loneoceans.com/labs/ud27/
My board is at https://github.com/bmboucher/drsstc
I basically borrowed the schematic wholesale, dropped the extra MOSFETs that boost the driver output, and replaced the external fiber optic interrupter with an Arduino on the board so the whole thing is self-contained. I also added some ICM7216 frequency counter chips because.... I'm not gonna lie, I thought they were cool It's a traveling science exhibit that a local museum uses for outreach.

Anyhow, the part of the board that's giving me the problem is the TL3116 comparator; the role of this circuit is to convert the incoming feedback signal (sine wave) and convert it to a 5V logic square wave via zero-crossing detection. The UD2.7 website says to test it by hooking up a frequency generator to the inputs and probing the drive output, but no matter what I do the output doesn't look right. At some point I realized I had gotten the pinout wrong in the PCB, and corrected that with a new chip and some minor surgery, but still no dice. Eventually I built a breakout circuit on a breadboard (yes, I know that should have been step one but PCBs are so cheap these days I got cocky).

The test circuit schematic is attached below - I dropped everything except the input voltage clamp to 1.6V, the 1.6V divider for the reference and the hysteresis feedback. The INPUT is connected to a frequency generator producing a roughly +/-3V sine wave (with AC coupling). When I scope the two inputs to the TL3116 I get exactly what I expect (see test A below), which is a clamped sine wave on the inverting input and a flat line on the non-inverting.

I would expect the output then to be a 5V square wave at the same 250kHz frequency I'm driving the frequency generator, but instead I get.... a 1V square wave at ~60 hertz? The test B images below are taken with different time scales to show that but note that the voltage scales are aligned and the same as the test A images.

I'm sure its possible I just damaged the chip somehow, I originally tested this circuit without the 1.6V clamp. But this chip was otherwise just used for this circuit and the one on the board has also been replaced once without showing any changes. Is there something I just don't understand here? The circuit I'm attempting to copy is very widely used but I don't see any discussion of problems like what I'm having.

BTW, here's the chip I bought: https://www.digikey.com/product-detail/en/texas-instruments/TL3116CD/296-6626-5-ND/372679

[Hiemal]

Sounds kinda like you're getting interference from mains if it's 60 hz. Also make sure you have the scope properly grounded and connected well, since sometimes funky ground connections can cause issues like this from what I've experienced.

[bmboucher]

That was my first thought as well - I've got the ground leads clipped to a pin header on the ground rail of the breadboard, right next to where I attached the ground from the bench power supply. I believe both are mains-earth referenced but I'll check the continuity the next time I'm in the lab.

I'd feel a lot better if I was seeing the pattern I wanted with 60 Hz noise, this is a pretty clean square wave. I'll try moving the probes around physically and see if it changes.

[bmboucher]

UPDATE: I fixed it, but I'm still not sure why it wasn't working. Although every schematic I've seen has the hysteresis circuit hooked up "normally", i.e. connecting the regular output to the non-inverting input (i.e. pin 7 to pin 2). When I switched it so that the complementary output was connected to the inverting input (i.e. pin 8 to pin 3) it worked as expected! I tested that on the breadboard, then ordered a new PCB and confirmed it worked in circuit as well. I just.... don't know why it didn't work the other way. Hope this helps someone who hits the same problem, and if anyone can teach me a little EE and explain what the problem was I'd much appreciate it.

[profdc9]

If It's any help, you can look at some of the designs I drew up

https://github.com/profdc9/DRSSTC-PCB-Pack

especially

https://github.com/profdc9/DRSSTC-PCB-Pack/tree/master/ud29
https://github.com/profdc9/DRSSTC-PCB-Pack/tree/master/ud27c%20thru%20hole

I definitely had to put my controller into a metal enclosure when running the tesla coil and I used common-mode chokes on the power and external transformers also to prevent RF from getting into the case.  But it works:



In general, it's a good idea to use the hysteresis resistor because it might prevent the drivers from a rough switch between states.  If you look at my version of the UD2.7C in the attached schematic you will see R16/R10/C11.  R10/C11 provides a brief deadtime after switching preventing the comparator from switching back-and-forth during transitions.   After C11 charges after a transition, R16 provides for a little residual hysteresis.  But the deadtime is important to prevent spurious switch and the deadtime depends on the IGBT you use, with bigger IGBT bricks generally requiring more deadtime and therefore more capacitance for C11.   Also the pins 2 and 3 of the comparator are biased to about 1/3 of the supply voltage as the TL3116 works has a narrow common-mode voltage range.  I think pins 2 and 3 are around 2 volts or so without a signal.  But there should definitely be positive, not negative, feedback in the comparator to achieve some hysteresis.

That's a nice board you have there and I hope to see it in action on a coil!  If you need information from a helpful crowd, check out highvoltageforum.net.  By the way the UD2.9 I designed adds skip pulse functionality as well, and it's easy to modify UD2.7C to add this feature.

Dan