Feedback current transformer doesn't work

[Egg]

How does this affect the feedback operation?


Did you identify this from the image with the blue wave?

Like david said it causes the duty cycle to go higher than 50%. You can see this both on the blue and yellow reading that the duty cycle is like %70/%30 percent. It seems easier to me to just get the right ic instead trying the resistor opiton. Can you scope the voltage on the 12v and 5v rails? See if there is any noise?

[davekni]

I believe I need to resolve this noise issue before proceeding. It's disturbing all my readings.

Yes, I'd agree.  Most of the scope traces can be interpreted even with noise present.  However, whatever is causing noise may be interfering with circuit operation at times.
Noise may be high frequency oscillation of some part of your circuit, which is aliasing with scope sample frequency.  Would be good to expand horizontal way out (low time/div) when looking at noise.
Try scoping ground on your board.  Both probe ground clip and probe tip connected to ground.  You may pick up enough of the noise that way.  Presuming so, noise is much easier to see when by itself and not on top of some other signal.

Wow, once again you got it right!
My chip is actually the HCT14. But I have some HC14 too. I'll make the change and test the result.

About the difference between the two, I found it briefly in a search that:
The basic difference is that HCT14 are designed to work at lower inputs (5v) and HC are rated for higher inputs (15v).

Neither can handle 15V.  Only the original 74C14, not HC nor HCT.  Both HC and HCT work well at 5V.  The difference is that HCT is designed for logic inputs coming from conventional TTL with input thresholds specified as being between 0.8V and 2.0V.  HC is expecting to be driven by other CMOS logic at the same 5V supply.

How does this affect the feedback operation?

Because HCT input thresholds are low, it takes a short time for the input capacitor to charge up the rising threshold, but a long time for it to discharge back below the falling threshold.

It seems easier to me to just get the right ic instead trying the resistor opiton.

Agree.  Even 74HC14 threshold voltages are typically somewhat lower than centered around 2.5V, though not nearly as low as 74HCT14 thresholds.  If you want to be perfectionist, a resistor still helps.  A higher value will work with the correct 74HC14 chip.

[ZakW]

Maybe I need to change some components on the board, maybe the 2N2222?

Before you change more parts can you scope the low voltage AC transformer output and the output of your interrupter? By scoping an AC signal you can compare it against the output of the interrupter. That helps confirm if the ZCD is triggering at or around zero crossing as well as making sure the interrupter is outputting a corresponding signal. From your included scope capture it looks like it is outputting a signal its just hard to tell if it is sync with the mains.

To reiterate, I was not able to get the coil to run consistently or at all without the self oscillation mod to the 74HC feedback section. If you're able to confirm that the above is working correctly, your issue may be more related to the 74HC chip as others have mentioned.

[TiagoBS]

Well, today I did some more tests.

Starting with the full bridge.

This first image is the reading of the rectifier output (using the 12v transformer).
-



Here after I put my finger between the bridge output. (It seems more correct).
-


The following two images are readings of the negative output of the rectifier bridge and the two phases of the 12v transformer.
-

-



I wanted to confirm whether this behavior is expected because when I carried out the same measurements in mains voltage (without the finger part) with the same bridge I got the expected result, but using the 12v transformer it did not.

After confirming this I will continue testing the other components to identify the source of the noise.

[davekni]

This first image is the reading of the rectifier output (using the 12v transformer).

To get a meaningful signal from the rectifier output, you need some load current, such as a resistor of 1k or lower.  Otherwise slight leakage current or capacitance of rectifier diodes overcomes tiny current of 10M scope input.

I can't quite read the low-res schematic image from LabCoatz.  The resistor on AC input after 1N4148 diode looks like perhaps 100k.  That may be too high if diode bridge has much leakage current.  I'd replace with 1k or lower and then scope again.  Diode leakage current and minimal load may be causing your line timing issues.

[ZakW]

TiagoBS,

Here are a couple old scope captures of the interrupter running relative to a 12v transformer. Ignore the blue trace. I would check this to verify if my interrupter was outputting the correct signal relative to the ac half cycle. Can you verify this as well? Were you able to implement the components that I shared in my previous post? I added an updated screenshot as well. I ended up using a 220pf cap and a 11k resistor to get the correct oscillation frequency. 1nf should work it is just a matter of choosing the corresponding resistor value to achieve the desired freq.

Yellow = Interrupter output
Purple = reference AC signal


Purple = Interrupter output
Yellow = reference AC signal

[TiagoBS]

Well, I redid the entire board just in case.
It's still working like it does on CW.

During the process I checked the noise as I added the components.
The measurements were made only with the two 555 and the UCC37322 (as I only got this chip in SMD and had to solder it permanently)
When all components are connected, all the measurements are very noisy (all tests were only with the logic part connected)


I took some readings from the oscilloscope.

One phase of the mains voltage and pin 3 of the 555.
-

-

Pin 4 of the 555 at the highest and lowest duty cycle respectively.
-

-


Pin 6 of the 555 (here with little noise)
-

-


This reading here was done on the ground wire (disconnected from de board). Only with the positive test tip.
-


My apartment is very close to one of the energy distribution stations in my city. Many high voltage cables pass through my street, both on lampposts and underground. I'm using a large metal fence as the ground. Maybe this is picking up some interference.
When I took measurements using the connected ground the noise seemed to get much worse.

TiagoBS,
Were you able to implement the components that I shared in my previous post? I added an updated screenshot as well. I ended up using a 220pf cap and a 11k resistor to get the correct oscillation frequency. 1nf should work it is just a matter of choosing the corresponding resistor value to achieve the desired freq.

Yes, I added the components and switched from the 74HCT14 to the 74HC14. I left a 100k potentiometer to change the frequency.
In this capture the card was showing a little more noise.
-

-



At times the noise seems much more intense, changing the waveform. Here I took a measurement with the test tip loose with just the ground connected to the negative on the board.
-

-

I can't quite read the low-res schematic image from LabCoatz.  The resistor on AC input after 1N4148 diode looks like perhaps 100k.  That may be too high if diode bridge has much leakage current.  I'd replace with 1k or lower and then scope again.  Diode leakage current and minimal load may be causing your line timing issues.

Here is a link with better resolution:
https://content.instructables.com/ORIG/FFF/UMFD/L1AQ3MJU/FFFUMFDL1AQ3MJU.png
-


Is this resistor marked in yellow?

[davekni]

One phase of the mains voltage and pin 3 of the 555.

There are two 555 chips in the schematic.  Don't now how to interpret scope plots without knowing which 555 chip is being probed.  Perhaps call them "left" and "right" 555 chips per their position in schematic.  (Unfortunate that this schematic does not include reference designators, such as R1 or U1.)

Is this resistor marked in yellow?

Yes.  I recommend a lower value such as 1k instead of 100k.  As low as you have around without exceeding power dissipation limit of resistor.  That will reduce sensitivity to both diode bridge leakage current or noise.

[TiagoBS]

One phase of the mains voltage and pin 3 of the 555.

There are two 555 chips in the schematic.  Don't now how to interpret scope plots without knowing which 555 chip is being probed.  Perhaps call them "left" and "right" 555 chips per their position in schematic.  (Unfortunate that this schematic does not include reference designators, such as R1 or U1.)

I'm sorry about that. The measurements were taken at left 555.
Only the ramp waves were capture at 555 on the right.

[davekni]

This reading here was done on the ground wire (disconnected from de board). Only with the positive test tip.

If you have no other connection between your board and line ground, then that 60Hz line frequency signal is expected and would show up anywhere.  Capacitance through 12V transformer or other power supply to line is enough to make that +-5V line frequency signal.  The noise in many of your other scope traces is much higher frequency, above your ~350kHz SSTC operating frequency.  Measuring with scope probe ground clip disconnected will not be useful in finding that high frequency noise source.  Measuring with both scope ground clip and probe tip connected to ECB "ground" may be useful.

Pin 4 of the 555 at the highest and lowest duty cycle respectively.

Looks like duty cycle adjustment has no effect.  Perhaps 100k POT to the right of right 555 is the wrong value or shorted.
Also, did you change which side of 12V transformer is being scoped for line voltage reference?  Does not match previous plot of pin 3.

Yes, I added the components and switched from the 74HCT14 to the 74HC14. I left a 100k potentiometer to change the frequency.  In this capture the card was showing a little more noise.

Looks good, other than noise.  What does this look like with horizontal expanded as much as possible, say 10ns/div or how ever fast your scope can go?  The noise is likely so high frequency that it is aliasing with scope sample frequency.  I'd guess either you are near some RF communication equipment or a part on your board is oscillating at high frequency (driver chips would be most likely for that).

At times the noise seems much more intense, changing the waveform. Here I took a measurement with the test tip loose with just the ground connected to the negative on the board.

Again, line frequency noise is normal if either probe tip or ground are not connected.

Here is a link with better resolution:

Thank you for the better link.  Makes analysis and discussion easier.

[TiagoBS]

Well, I did more tests this morning. I can't find what I changed, but the coil is now working with the CT!

Looks like duty cycle adjustment has no effect.  Perhaps 100k POT to the right of right 555 is the wrong value or shorted.
Also, did you change which side of 12V transformer is being scoped for line voltage reference?  Does not match previous plot of pin 3.

Here are some captures of the duty cycle changing and also the noise in the signal.
I changed the pin that was measured from 3 to 4. I believe this caused the difference in readings.
-

-

-

Looks good, other than noise.  What does this look like with horizontal expanded as much as possible, say 10ns/div or how ever fast your scope can go?  The noise is likely so high frequency that it is aliasing with scope sample frequency.  I'd guess either you are near some RF communication equipment or a part on your board is oscillating at high frequency (driver chips would be most likely for that).

Here is a capture at 10ns/div
-


In this video the 74HC14 is self oscillating at 100khz. I read this in ZakW's thread:

Ideally self-oscillation frequency will be close to operating frequency, about 3x higher than now.  330pF would be roughly there.  However, better to go down to 220pF and increase R13 value to get back down to 450kHz.  Better to have some margin away from self-oscillation dropping out.  (Likely going to very high frequency rather than dropping out, too high to get through driver chip.  Could cause problems, so better to stay away from failure threshold.)

Since the coil oscillation takes over the self-oscillation of CH14, what is the difference between leaving it oscillating at 100 kHz and increasing it to close to the coil's operating frequency?

I also read the ramp on pin 6 of the 555 on the right
-
 [ You are not allowed to view attachments ]

And here is the reading of the mains phase that is going to the interrupter and the Gate of one of the transistors.
-
 [ You are not allowed to view attachments ]

Yes.  I recommend a lower value such as 1k instead of 100k.  As low as you have around without exceeding power dissipation limit of resistor.  That will reduce sensitivity to both diode bridge leakage current or noise.

I haven't changed the resistor yet. But I will exchange it and measure the difference

[davekni]

Here are some captures of the duty cycle changing and also the noise in the signal.

OK, I now see that you were changing duty cycle by changing on time, not by changing off time.  Previous scope captures of left 555 pin 4 would be affected only by off time adjustment (100k POT to right of schematic).  Thus I'd presumed you were adjusting that POT.  Makes sense now, as left 555 pin 4 is not expected to change with left 20k POT.

I changed the pin that was measured from 3 to 4. I believe this caused the difference in readings.

I saw that.  However, still doesn't fit.  Pin 4 is a low-true reset to 555 chip.  When pin 4 is low, pin 3 is also low.  However, pin 3 goes high during low line half-cycles.  Pin 4 is high only during high line half-cycles.  So something else changed between those two captures.

Here is a capture at 10ns/div

That looks like ring more than noise.  Perhaps a capture at 100ns/div would be better for showing noise.

Since the coil oscillation takes over the self-oscillation of CH14, what is the difference between leaving it oscillating at 100 kHz and increasing it to close to the coil's operating frequency?

As long as a single edge is sufficient to start oscillation, 100kHz is likely fine, or even 0Hz (no self-oscillation mod).  Two advantages of self-oscillation closer to operating frequency.  First, in case feedback signal is weak, allows feedback to grow for a cycle or two until it is strong enough.  Second, higher frequency minimizes voltage across 1uF GDT coupling capacitor and minimizes chance of GDT core saturating during the initial half-cycle before oscillation starts.

I also read the ramp on pin 6 of the 555 on the right

Noise looks worse during time when coil is running (driver chips switching state).  Suggests at least some of the noise is coming from driver chips.  Does your ECB have a ground plane?

And here is the reading of the mains phase that is going to the interrupter and the Gate of one of the transistors.

Looks appropriate for a long (half cycle) on-time.  BTW, extending past 1/4 cycle usually does not add to arc length, but can make arc thicker by keeping it energized longer.

Might be a good idea to increase right 100k POT setting to reduce repetition rate.  Would allow playing with on-time while still keeping duty cycle low, so minimize stress on IGBTs and gate drivers in case of issues.

[TiagoBS]

Yes.  I recommend a lower value such as 1k instead of 100k.  As low as you have around without exceeding power dissipation limit of resistor.  That will reduce sensitivity to both diode bridge leakage current or noise.

I replaced the 100k resistor with a 680ohm one and took this reading with the oscilloscope after the change. I don't know if this change affected it, but now the reading is much less noisy.

That looks like ring more than noise.  Perhaps a capture at 100ns/div would be better for showing noise.

-

-

Noise looks worse during time when coil is running (driver chips switching state).  Suggests at least some of the noise is coming from driver chips.  Does your ECB have a ground plane?

It does not have.

Looks appropriate for a long (half cycle) on-time.  BTW, extending past 1/4 cycle usually does not add to arc length, but can make arc thicker by keeping it energized longer.
Might be a good idea to increase right 100k POT setting to reduce repetition rate.  Would allow playing with on-time while still keeping duty cycle low, so minimize stress on IGBTs and gate drivers in case of issues.

Do you mean changing the 100k potentiometer that controls the frequency for example with a 500k one?
And how would this affect the duty cycle and allow playing with on-time?

[davekni]

I replaced the 100k resistor with a 680ohm one and took this reading with the oscilloscope after the change. I don't know if this change affected it, but now the reading is much less noisy.

Nice that the noise amplitude dropped a bit.  These two scope captures at 500ns/div and 2us/div are perfect for seeing the noise detail.  There are two frequencies present, one at ~6.5MHz and another around 600kHz.  The lower frequency ~600kHz component may be ringing from your coil upper pole or primary wiring or etc.  Upper frequency is more concerning.  Most likely some component on your board is oscillating at 6.5MHz.  (Unless you live near a radio transmitter or some other equipment generating 6.5MHz.)  You can use your scope probe as a magnetic pickup probe by connecting ground clip to probe tip and moving that loop around your circuit in search of 6.5MHz noise.  Or, for better location precision, wind a few small turns of wire between scope tip and ground clip.

Quote

    Noise looks worse during time when coil is running (driver chips switching state).  Suggests at least some of the noise is coming from driver chips.  Does your ECB have a ground plane?

It does not have.

Quote from: davekni on Se

That may be root of noise issues.  High current chips such as gate drivers pull large current spikes from power pins and large current spikes charging output load capacitance.  Good short routing to bypass caps helps.  Some boards without ground planes (and even breadboard hand wiring) end up working.  Ground planes aren't an absolute necessity.  However, much more chance of problems without one.

Do you mean changing the 100k potentiometer that controls the frequency for example with a 500k one?
And how would this affect the duty cycle and allow playing with on-time?

No.  I mean adjusting the existing 100k potentiometer to a higher value.  As is you are getting one half cycle every two line cycles.  Setting 100k potentiometer to a higher value will increase that time.  Max 100k value should give you between one and two half-cycles per second.

[yourboi]

I have been troubleshooting my zach armstrong MJOLNIR coil and I have developed the same issue. the feedback somehow ends up in the interrupter and exactly as you said it was working fine for a whole year of daily use.

Is the feedback signal overpowering the interrupter signal? I am not really sure how this could happen if the UCC is also connected to the interrupter. It would be helpful to see what the output of the interrupter looks like when it is exhibiting this behavior.

The only other thing I can think of at the moment is related to the ZCD. On several versions of my PCBs, my coils would start out working fine then I would have a failure and it would start exhibiting strange BPS/feedback issues similar to what is happening to you.

I don't have the time to read this thread too thoroughly right now. as I was troubleshooting my Mjolnir coil the past few weeks I have been debating posting here. I am pretty new to this all things considered. I've had this problem twice believe it or not on two separate boards. the antenna is finnicky to the touch placing my finger near the feedback antenna causes the interrupter to go super fast and ignore the enable pin. the interrupter seems to function fine with a perfect square wave. the Schmitt trigger looks good with its input being square. brand new ic and decoupling caps. sadly any readings I took show this on the output (Note: the noise goes away when I keep my hand away from the antenna, there is also a function generator connected to the antenna): https://imgur.com/a/XOgdTFT I've pretty much considered the coil as good as dead and am moving on to a new one because its taken abuse from troubleshooting and desoldering components. the problem started after pushing 2.2KVA through the thing and overheating and blowing the bridge up because the heatsinks I was using were even smaller then the typical ones. I went to rebuild and here we are.

[TiagoBS]

Well, I was already moving towards finishing this project. But once again there was a problem.
The coil was already working well, the self-oscillatory circuit working as expected.
However, when I assembled the coil in its final configuration I had problems with secondary interference. So I made a grounded shield out of cardboard and aluminum foil to separate the secondary from the PCB. This solved the interference problem. To give a better finish, I designed and 3D printed a new shield. When I went to test this new shield, I removed the resonant capacitors, the coil remained working. So I added the capacitors again and when I connected the coil it blew both transistors and the fuse.

They were my last two 60N65SMD.
Now I'm doing tests with IRFP460.

I placed the new mosfets and removed the resonant capacitors and I also had to change the ucc37321. But this time the output was very weak, with a neon indicator I can see the coil oscillating according to the interrupter, but with no visible output, even at the highest duty cycle it is possible to see just a small spark. I tried adding the resonant capacitors but with them the coil simply doesn't work, when removing the capacitors it starts working again with a weak output.

By turning on just the low voltage circuit it is possible to hear the ucc clicking according to the interrupter, but the click seems weaker and less audible than before.

Does anyone have any idea what could be happening?

Some photos of how the coil looked:


-

-

-

-

-

-

-

-
 [ Invalid Attachment ]

[davekni]

So I added the capacitors again and when I connected the coil it blew both transistors and the fuse.

Such events are always frustrating.
Did you check the TVS diodes?  Sometimes a TVS diode will fail first, causing IGBTs to then fail.

They were my last two 60N65SMD.
Now I'm doing tests with IRFP460.

FETs and IGBTs are significantly different is several ways.  Not too surprising that behavior changed.

Some photos of how the coil looked:

Nice photos.  Helps to see details.

Shielding between coil and electronics is a good idea.  That shield is close enough to primary and bottom of secondary to change inductance and coupling and loss a bit.  Would require modeling and/or testing to see how much.  My rule-of-thumb is to have primary radius (half of primary diameter) of space between coil and shield to avoid such affects.  Closer may be OK if tuning matches reduced inductances and coupling factor.

With no ground planes and with control circuitry not far from power circuitry, there are many possibilities for unwanted feedback and noise.  One patch might help with the power section of your existing ECB:  There is a large loop in the power supply from low-side emitter to high-side collector.  Adding capacitor(s) directly from low-side emitter to high-side collector would provide a more direct path for the high frequency switching current in that path.

[TiagoBS]

Well, I did some more tests and discovered a few things.
The reason the output was very weak was due to the duty cycle potentiometer wire being broken and hidden under the insulation. However, I only discovered this because I checked the duty cycle time on the oscilloscope and realized that even at 20k it was no more than 10ms.
I switched to this potentiometer for a 50k one and now the output without the resonant capacitors is back to normal.

Despite having "solved" this problem, I wanted to understand what happened.

Now what I'm trying to solve is the fact that when I use resonant capacitors the coil has no output and just lights up a neon lamp.
I removed the TVS just in case, but the behavior didn't change.

[davekni]

Now what I'm trying to solve is the fact that when I use resonant capacitors the coil has no output and just lights up a neon lamp.

Can't think of anything useful to say without seeing scope captures.

[TiagoBS]

I did some more tests and it seems like I'm narrowing down the possibilities.

1 - When I touch the probe to pin 3 of the 555 on the left, it seems that the duty cycle is reduced and the coil stops having output and just lights up a small neon lamp. (without resonant capacitor)

2 - You can see the duty cycle being adjusted when I test pin 5 of the 555 on the left. Another point is that it is only possible to notice this change in the duty cycle after the potentiometer is turned around 60%.
However, you can see and hear the difference in output throughout all the potentiometer adjustment.(without resonant capacitor)
-

-



3 - Touching the probe to any of the Mosfet Gates causes the duty cycle to increase again. (without resonant capacitor)
[Video]

4 - Adding the resonant capacitor to the primary and measuring pin 3 of the 555 on the left, it is possible to notice that the duty cycle varies between 500us and 750us. This change is possible to notice by turning the 20k potentiometer (now 50k) but only during the first 15% of possible regulation, after that the cuty cycle remains close to 750us for the rest of the potentiometer adjustment.

Apparently I have problems with my osiloscope and that's why I'm using this portable one.
Using this other oscilloscope I didn't notice that much noise and it's even possible to see the ramps.
-

Perhaps some component of the switch is having problems?