problems with the spikes

[angelotronics]

Hi everyone and sorry for my English (I use a translator). I hope someone can help me ... I'm driving a coil (you can exactly define it as a transformer because I wound another winding as secondary) with a half-bridge mosfet circuit with frequencies from 0.5Mhz to about 1Mhz but on the primary of the coil I get unwanted oscillations that even reach 14Mhz or more...my circuit already includes snubbers but they are probably not sufficient or not of the right values...can you help me eliminate them or tell me how I can calculate the snubbers? Thank you

[MarkT]

What values are those "1m" capacitors?  1000uF for MHz sounds way way off. 

[angelotronics]

Excuse me, the values are from 1uF 630v

[Jwillis]

That's Ringing.The effects parasitics increase as frequency increases. Specifically the parasitic capacitance and inductance. Those series capacitors will compound those issues .
The series capacitors at C37 and C23 specifically .
I'm not an expert by any means but I would go back to the datasheet and follow the typical connection it suggests on page 1. https://www.infineon.com/dgdl/Infineon-IR2110-DataSheet-v01_00-EN.pdf?fileId=5546d462533600a4015355c80333167e
The connections you have don't look right.

[tggzzz]

Show a photo of your construction, and how you are attaching the probes.

Is there a solderless breadboard involved?

[jwet]

C37 is not wanted or needed.  The half bridge should drive the L directly (C23 and C17 will maintain the other side and will charged and discharged by the direct connection through the L's resistance.  As is, once the rise or fall of the circuit is over, the C becomes an open and the snubbers are invisible pretty much.  You've created an ideal way to get energy into an LC load to create a damped oscillation.

[angelotronics]

I tried to exclude the C37 capacitor and the signal dropped to a few mV and without eliminating the noise....I'm desperate...I don't understand why the snubbers don't work, maybe they need to be sized with different values? Would there be no way to calculate the parasitic inductance or capacitance that causes this?

[Jwillis]

I tried to exclude the C37 capacitor and the signal dropped to a few mV and without eliminating the noise....I'm desperate...I don't understand why the snubbers don't work, maybe they need to be sized with different values? Would there be no way to calculate the parasitic inductance or capacitance that causes this?

Not an expert here but I was experimenting with a similar set up of a Mosfet driving a transformer. I don't know what transformer your using so I just went with what I had. Oscilloscope probes were attached to secondary of transformer. 1Mhz square wave at 10Vpp at the Mosfet gate.
At 1Mhz signal at the gate of MOSFET I got a really close match to the distorted signal you did with no current or voltage across Drain to Source.  Changing the input frequency down to around 200Khz and the ringing started to disappear . I changed the value of the gate resistor to 270 ohm and the ringing was dampened down quite a lot.
I doubt my experiment is accurate, but I sure looks like the gate of the Mosfet might be ringing. Try increasing the gate resistor values and see if anything changes. I also removed the diodes parallel to gate resistors. Also at those frequencies you might try much lower values of decoupling capacitors if applicable. 10- 20 pF decoupling capacitor at the gate  was enough to filter out some of the high frequency ringing. With a low value inductor of 10uF parallel with a 1K resistor in series with the transformer primary I managed to achieve a trapezoidal wave  at the secondary with 15 volts at the Mosfet Drain.
This was only an experiment but I hope it may help in diagnosing the problem.   

[T3sl4co1l]

Would there be no way to calculate the parasitic inductance or capacitance that causes this?

There is. Listen to previous posts: please show your layout.

Snubbers don't work magically by putting them in the circuit just anywhere; in real circuits, stray inductance depends on distance, on wire length. Most likely you have loose wiring between elements, negating the snubber's effect.

Tim

[T3sl4co1l]

I tried to exclude the C37 capacitor and the signal dropped to a few mV and without eliminating the noise....I'm desperate...I don't understand why the snubbers don't work, maybe they need to be sized with different values? Would there be no way to calculate the parasitic inductance or capacitance that causes this?

Not an expert here but I was experimenting with a similar set up of a Mosfet driving a transformer. I don't know what transformer your using so I just went with what I had. Oscilloscope probes were attached to secondary of transformer. 1Mhz square wave at 10Vpp at the Mosfet gate.
At 1Mhz signal at the gate of MOSFET I got a really close match to the distorted signal you did with no current or voltage across Drain to Source.  Changing the input frequency down to around 200Khz and the ringing started to disappear . I changed the value of the gate resistor to 270 ohm and the ringing was dampened down quite a lot.
I doubt my experiment is accurate, but I sure looks like the gate of the Mosfet might be ringing. Try increasing the gate resistor values and see if anything changes. I also removed the diodes parallel to gate resistors. Also at those frequencies you might try much lower values of decoupling capacitors if applicable. 10- 20 pF decoupling capacitor at the gate  was enough to filter out some of the high frequency ringing. With a low value inductor of 10uF parallel with a 1K resistor in series with the transformer primary I managed to achieve a trapezoidal wave  at the secondary with 15 volts at the Mosfet Drain.
This was only an experiment but I hope it may help in diagnosing the problem.

This is neither here nor there, without even part types, let alone a circuit.  270R is way too large for most power MOSFETs, unless you're intentionally slow-walking it (e.g. for a load switch with low repeat rate).

The time constant allows us to quickly exclude many combinations: for example, 10R + 10nF (a typical equivalent for a large-ish transistor of Qg(tot) = 100nC at Vgs(on) = 10V) in the gate loop is a 100ns time constant, and it would take L > R^2 C = 1uH to make that loop oscillate (and clearly, at a frequency of 1.6MHz, not 15).  That would be a LOT of stray wiring!

Gate voltage itself needn't look very clean, due to crosstalk and probing error.  One key is whether the trash appears "everywhere", i.e. it's common mode coming up the scope probe, not a differential measured between probe and ground clip.  Keeping the probe ground path short (use a contact spring instead of the ground clip, or even better still, a coaxial socket) helps with this.

Frequencies in the 10s of MHz up, are prime candidates for stray wiring, and probing error.  An inch of wire is ballpark 20nH, and 20nH at 20MHz is 2.5 ohms, hardly negligible in a power switching circuit.

Tim

[Jwillis]

This is neither here nor there, without even part types, let alone a circuit.  270R is way too large for most power MOSFETs, unless you're intentionally slow-walking it (e.g. for a load switch with low repeat rate).

The time constant allows us to quickly exclude many combinations: for example, 10R + 10nF (a typical equivalent for a large-ish transistor of Qg(tot) = 100nC at Vgs(on) = 10V) in the gate loop is a 100ns time constant, and it would take L > R^2 C = 1uH to make that loop oscillate (and clearly, at a frequency of 1.6MHz, not 15).  That would be a LOT of stray wiring!

Gate voltage itself needn't look very clean, due to crosstalk and probing error.  One key is whether the trash appears "everywhere", i.e. it's common mode coming up the scope probe, not a differential measured between probe and ground clip.  Keeping the probe ground path short (use a contact spring instead of the ground clip, or even better still, a coaxial socket) helps with this.

Frequencies in the 10s of MHz up, are prime candidates for stray wiring, and probing error.  An inch of wire is ballpark 20nH, and 20nH at 20MHz is 2.5 ohms, hardly negligible in a power switching circuit.

Tim

Fair enough. But the effect is still there when probe is connected directly to a 40N60 mosfets source with no other connection applied other than the 1Mhz square wave signal to gate through a 4 ohm resistor . Granted it is not the 16N60 used in the design but should be close enough. Even a IRE3205, 75343P3 exhibit similar effects.
Parasitics  become more pronounced as frequency goes up. Reduce the frequency the effect goes away. Increase the gate resistance and the effect diminishes  These are just the observations I've made and have to conclude that the parasitics in the mosfet may be a part of the problem. It's not an attempt exclude other possibilities or dismiss other peoples experience. Just saying that checking couldn't hurt.

[angelotronics]

thank you all for your help, because it is difficult to find competent people. However I also tried to delete the C37...I attach some photos of my pcb, and since I had already understood how to optimize the connections and make shorter and thicker tracks, a month ago I created another pcb and also a ground plane on the top layer, the pcb should arrive in a few weeks I hope... I'll put the oscilloscope probe on the primary. I tried decreasing the snubber capacitor and almost nothing changes, what should I do in these situations? What tests can I do? I add that the frequency range I am interested in using is from 450khz to 1.5 MHz

[T3sl4co1l]

Yeah that's pretty hopeless... haven't even traced what the gate driver sees, it might be toast anyway, but if you're lucky you can salvage what you have now by shortening these paths, massively, like so:

(1) scratch off soldermask and connect to trace here.
(2) no connection, make node in air.

If the primary peak current is some 10s of A (as the transistor type seems to suggest), 3A axial diodes (UF5404 etc.) will suffice.

You can remove the antiparallel diodes; I don't know what those are doing there, the MOSFETs have internal body diodes rated for forward drain current at least.

Tim