PopularFX
Home Help Search Login Register
Welcome,Guest. Please login or register.
2018-12-10, 02:17:58
News: If you have a suggestion or need for a new board title, please PM the Admins.

Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 [17] 18
Author Topic: Controller No5 With Protection - Looking for Explosions  (Read 90745 times)

Hero Member
*****

Posts: 1490

Quote
Is the ringing after the pulse, changing its frequency or is it my imagination ?
Does waving ceramic magnets in the vicinity alter it ?


The ringing frequency stays at 7.8Mhz, only the amplitude changes (minimum at 0ns phase (= really 15ns) and
max. at 260ns phase (= really 275ns).

Waving a stack of ceramic magnets along or perpendiculary to the bifilar coils does not show any effect.
Touching the bifilar coil with my hand decreases slightly the amplitude of the ringing signal.



Below screenshot shows both drain and gate signals of the both MOSFETs while sweeping from 0 - 190ns (just before
flipping to 0ns again) thus avoiding the glitches to occure.

Yellow / purple   MOSFET 1
Blue   / green    MOSFET 2

Itsu
   

Hero Member
*****

Posts: 1249
The below picture shows how the MOSFETs are connected to the coil, no choke is used.
What is this coil wound on ?
   

Hero Member
*****

Posts: 1490

The coil is wound on some length of 16mm diameter electra piping, see picture:


Itsu
   

Hero Member
*****

Posts: 1490

As i was working with my last pair of IRFP460 MOSFETs, it tried another type which i have more of, the SPW47N60C3.
This is a similar one as the IRFP460 (650V / 47A / 0.07 Ohm), but i was not able to get the effect with them.


So to protect my last IRFP460's, i added a common mode choke (CMC) in the 41V supply line and a 50 Ohm resistor.
The CMC more to protect the PS, and the 50 Ohm to create a voltage drop when amps start to flow.

Now i can see the glitches happening, but my PS does not get into current limiting mode and the glitches are much
more controlled (160V only area).

This allowed me to also scope the gate signal of a MOSFET when the glitch happens.

Screenshot shows in yellow the controlled glitch on the drain signal and in purple the gate signal that goes with it.

This looks to me that a glitch is causing (via what? induction / capacitance) a pulse on the gate which triggers the
MOSFET again etc. etc.

The drain spikes (160V) and the gate spikes (16V) are within the specs of the MOSFET, so no harm will be done.
But imagine what happens when the spikes of +700V we have seen earlier can do to the gate signal.

Not sure if any zeners are fast enough to protect the gate.

Itsu
   

Hero Member
*****

Posts: 1490

I lowered the driver voltage to 15V and put in 18V bidirectional tvs's (1.5KE18CA) across the gate / source, but when omitting the 50 Ohm resistor in the
41V supply lead, i again have the HV glitches which damaged 1 MOSFET.

So the tvs's are not fast enough, or the damage is caused at the drains.

Itsu
   

Hero Member
*****

Posts: 1249
This looks to me that a glitch is causing (via what? induction / capacitance) a pulse on the gate which triggers the
MOSFET again etc. etc.
Via the Drain-Gate capacitance (a.k.a. Miller's capacitance).
The purpose of that recent Miller testing setup was to test how much that high dv/dt waveform appearing on the drain, influences the gate voltage when that test gate is terminated by THE SAME impedance as the driving impedance of the other MOSFET gate.

Varying the gate driving impedance by several Ohms will exacerbate or minimize the effect of the Drain-Gate capacitance during high dv/dt Drain waveforms.  So this proposition can be empirically verified in that manner.

ANOTHER ISSUE:
From your scopeshot it is evident that there is a double, negative going, pulse when there is a large delay between the two channels.
According to the Lenz law the half of the bifilar winding, that is being driven, should induce a pulse of an OPPOSITE polarity in the other half of bifilar winding, that is NOT being driven.
...but this is not what is being observed.
Fortunately the mutual capacitance of these two halves of winding can be used to explain two pulses of the same polarity, ...meaning that in this coil the capacitve coupling is stronger than the inductive coupling.

In fact the entire bifilar coil can be characterized by the ratio of the capacitive coupling to the inductive coupling ...or treated as a distributed capacitance and inductance, which can be analyzed as a transmission line.
« Last Edit: 2018-11-17, 23:27:00 by verpies »
   

Hero Member
*****

Posts: 1490

I used that recent Miller testing setup again with my IRFP460 MOSFETs i use now, and changed the load resistor R4 to a 100 Ohm pot.
But varying the pot across its complete range (0 - 100 Ohm) did not show any change in the output gate signal, so it is not so obvious by looking at the output gate signal
what the correct input impedance should be.

So if the drain-gate capacitance is causing this feedback from the drain to the gate i should be able to combat that by a zener across the gate/source or as i did a TVS.
So probably my TVS is not fast enough as i did destroy a MOSFET again yesterday.


 


I calculated the length of my bifilar coils to be 570cm (114 turns on 16mm diameter).
The tdr measurement i did on this bifilar coil (transmission line) shows a round trip time (open ended) of 60ns, so a one way trip time of 30ns.

Speed of light in free space is 30cm/ns, so 900cm (9m) for my 30ns one way trip.
The velocity factor then of this bifilar coil is 570/900 = 0.63.

But the double negative pulse seen on the drain signal is much farther away, like 172ns farther, see screenshot.

There i have put up the white traces as being the signal from one MOSFET drain while the other MOSFET is not triggered (input signal to its driver disconnected).

White trace R1 is MOSFET 1, white trace R2 is MOSFET 2 (identical).
The yellow trace is the MOSFET 1 signal with both MOSFETs activated at phase 0 (really 15ns).

So this shows me that the double pulse (172ns apart) is not caused by the reflection on the bifilar coil.

Itsu

   

Hero Member
*****

Posts: 1249
So if the drain-gate capacitance is causing this feedback from the drain to the gate i should be able to combat that by a zener across the gate/source or as i did a TVS.
Even if the Zener or TVS had no capacitance and were faster than the MOSFET's gate (which they are not), they are several centimeters (and microhenries) further from the Drain than the drain-gate junction.
Even a REMOTE zero Ohm resistor might not be able to keep the gate down if several microhenries separate it from the gate.

I used that recent Miller testing setup again with my IRFP460 MOSFETs i use now, and changed the load resistor R4 to a 100 Ohm pot.
A 10 Ohm closely placed non-inductive pot would be better.

So this shows me that the double pulse (172ns apart) is not caused by the reflection on the bifilar coil.
OK, but the CDG can positively charge the gate of the undriven MOSFET ( thus turning it on *) unless it is grounded by a very low impedance... or better yet, when the gate is brought -15V below source by a very low impedance (by e.g. an AC gate driver)

*This can be seen by putting a 1 Ohm CSR in the Drain and scoping across it
« Last Edit: 2018-11-18, 08:14:09 by verpies »
   

Hero Member
*****

Posts: 1490
Even if the Zener or TVS had no capacitance and were faster than the MOSFET's gate (which they are not), they are several centimeters (and microhenries) further from the Drain than the drain-gate junction.
Even a REMOTE zero Ohm resistor might not be able to keep the gate down if several microhenries separate it from the gate.

Ok, so what you are saying is that in this case the zener/TVS is no cure for preventing the gate to develope spikes that will destroy the MOSFET.


Quote
OK, but the CDG can positively charge the gate of the undriven MOSFET ( thus turning it on *) unless it is grounded by a very low impedance... or better yet, when the gate is brought -15V below source by a very low impedance (by e.g. an AC gate driver)

So for the double pulse this AC coupled gate driver setup could prevent it.
Is it like shown on page 36 -> of this PFD:  www.logosfoundation.org/instrum_gwr/balsi/Texas_slua618.pdf

Itsu
   

Hero Member
*****

Posts: 1249
So for the double pulse this AC coupled gate driver setup could prevent it.
Is it like shown on page 36 -> of this PFD:  www.logosfoundation.org/instrum_gwr/balsi/Texas_slua618.pdf
No, I had in mind a driver that is supplied by -15V and -15V rails, so the gate is pulled up to +15V when the driver commands the MOSFET to be ON and pulled down to -15V when the driver commands the MOSFET to be OFF.

But before you start looking for a driver with > 30V supply rating and building a split power supply for it, put a 1 Ohm CSR in the undriven MOSFET's Drain and command its driver to turn it permanently OFF to see whether it becomes turned on by the high dv/dt waveform appearing on its Drain (caused by the switching action of the other MOSFET).

If it turns on ( current starts flowing in its Drain ) then disconnect the driver and sequentially connect the Gate to the Source with a 10Ω to 0Ω non-inductive resistors ...and if it turns on even after you get down to 0Ω then connect the gate to -10V to -18V using some kind of battery (e.g. two 9V batteries in series), etc...
   

Hero Member
*****

Posts: 1490

Ok, 1 Ohm csr in the undriven MOSFET's Drain and its driver turned off.

Ref3 is the drain signal from the driven MOSFET for comparison.
Ref4 is the csr signal of the undriven MOSFET.

Disconnected the driver and used 10 Ohm, 1 Ohm, 0 ohm (short) and -9V (battery) across the undriven MOSFET gate/source

Purple is the csr signal of the undriven MOSFET at -9V, but its the same as with the 10, 1 and 0 Ohm resistors.

So there seems to be no or very little effect pulling down or negatively bias the gate.

Itsu
   

Hero Member
*****

Posts: 1249
So there seems to be no or very little effect pulling down or negatively bias the gate.
I agree.
You just saved yourself a lot of useless work making an AC gate driver and split power supplies.

Ok, 1 Ohm csr in the undriven MOSFET's Drain and its driver turned off.
So, if the undriven MOSFET is not turning on due to the Miller effect, the only current that could be flowing through that Drain terminal can be caused by CDG and CDS.

Now, due diligence would require the calculation whether CDG + CDS can really cause these ~1A peak currents with these drain waveforms...but I feel too lazy to do that now.
   

Hero Member
*****

Posts: 1490
"lazy sunday af...  err  evening"  :)



First screenshot below shows both drain signals with the blue (MOSFET 2) one having no drive and its gate/source shorted.
So the blue (MOSFET 2) one is fully opened/activated without any obvious drive.


Second screenshot shows the drain signals with, additional to above, the blue (MOSFET 2) one disconnected from its coil (so no 41V drain voltage).
Obvious there is no pulse now, but we also see the yellow (MOSFET 1) one having only 1 pulse now.

Extending the sweep range to Peters box max. (1200ns) shows that the glitches keep on happening all the way up to this upper limit, so roughly from 170ns to 1200ns phase.


The Spectrum Analyzer shows massive noise up to 370Mhz when the glitches appear, but also on my 70cm Ham transceiver (432Mhz) i hear the noise still very well.


itsu
   

Hero Member
*****

Posts: 1249
First screenshot below shows both drain signals with the blue (MOSFET 2) one having no drive and its gate/source shorted.
So the blue (MOSFET 2) one is fully opened/activated without any obvious drive.
So the undriven MOSFET is conducting now !?

This is very strange...
   

Hero Member
*****

Posts: 1490

yes, i find it strange too.

Here a video of this situation where i have removed the MOSFET 2 (blue) driver and shorted its gate/source.
Scoping both MOSFETs drain voltage and the MOSFET 2 drain current.
https://www.youtube.com/watch?v=DhARKF4-bf4


See also the screenshot.

Yellow is MOSFET 1 (driven MOSFET) drain voltage
Blue   is MOSFET 2 (undriven / shorted MOSFET) drain voltage
green  is MOSFET 2 drain current.


Itsu
   

Hero Member
*****

Posts: 1249
Notice, that during the ringdown, current & voltage are in phase, so this means that the current is not flowing through a capacitance, as it would be 90º out of phase with voltage then.

What do you get on the undriven half of the winding when you ground it through a 1kΩ resistor instead through the undriven MOSFET 2 ?.
« Last Edit: 2018-11-20, 00:57:34 by verpies »
   

Hero Member
*****

Posts: 1490

Hmmm,   yes in phase, but opposing, so 180° off.

So i should not have inverted the green channel then, but if so, then the current is flowing out of the drain when the MOSFET is active and into the coil as my current probe "current arrow" points now towards the drain.

Will do the 1kΩ resistor test later this evening.


Itsu
   

Hero Member
*****

Posts: 1249
Hmmm,   yes in phase, but opposing, so 180° off.
Getting 180º with passive elements is tantamount to finding the Holy Grail - the negative resistance.
   

Hero Member
*****

Posts: 1490

Undriven half of the winding grounded via a 1KΩ resistor (getting hot quick), current probe arrow pointing to ground, see picture.
Green channel NOT inverted (perhaps the yellow channel should).

White trace is the driven MOSFET drain voltage (notice NO double pulse now)
yellow trace is across the undriven half of the winding.
green trace is the current through the 1kΩ resistor.

Itsu
   

Hero Member
*****

Posts: 1490
Not sure what went wrong with the above screenshot, but the current trace value seems odd (253.6mApp) as the
vertical setting shows 50mA/div.

So i did a fresh calibration of the scope and redid the test, but now with the yellow probe across the coil reversed.
No more ringing there (ringing appears when disconnecting the ground tip, so..).

See below screenshot.

edit:  added the diagram


Itsu
« Last Edit: 2018-11-20, 20:03:40 by Itsu »
   

Hero Member
*****

Posts: 1249
I was thinking a long time and I could not think of a good reason for this MOSFET to become turned on with the gate shorted to source.
Normally, this happens when there is a excessively high dv/dt on the train (rising edge for N-ch MOSFET) or when the D-S junction breaks down due to overvoltage, ..but none of these things seem to be happening here, unless your scope is missing some huge subnanosecond kV pulse.

The voltage waveform across the 1k resistor is almost identical to the voltage waveform on the driven winding, suggesting simple capacitive coupling, the small differences that are visible could be due to delayed transmission-line like reflections.

Since this is strange, it should be investigated further. Does Tinsel look at his thread ?
   

Hero Member
*****

Posts: 1490

Quote
I was thinking a long time and I could not think of a good reason for this MOSFET to become turned on with the gate shorted to source.
Normally, this happens when there is a excessively high dv/dt on the train (rising edge for N-ch MOSFET) or when the D-S junction breaks down due to overvoltage, ..but none of these things seem to be happening here, unless your scope is missing some huge subnanosecond kV pulse.

Not sure to which post / situation you point with this remark, as the MOSFET with its gate/source shorted is not in the circuit in my last post situation.
It was in post #414.   
I doubt i would not detect any huge subnanosecond kV pulse

Quote
The voltage waveform across the 1k resistor is almost identical to the voltage waveform on the driven winding, suggesting simple capacitive coupling, the small differences that are visible could be due to delayed transmission-line like reflections.

On this i agree if pointing to my last post situation.  The strong coupling of this bifilar coil could do that.


Quote
Since this is strange, it should be investigated further. Does Tinsel look at his thread ?

Not sure if besides you and me anyone else is looking at this thread.


Itsu
   

Hero Member
*****

Posts: 1249
Not sure to which post / situation you point with this remark, as the MOSFET with its gate/source shorted is not in the circuit in my last post situation.
It was in post #414.
Yes, I was referring to post #414.
I still don't understand what could have turned that MOSFET on...or was it simply a matter of the capacitance between Drain and Source.
   

Hero Member
*****

Posts: 1490
ok, thats clear.

The IRFP460 output capacitance is 870pF

So for a MOSFET (IRFP460 here) to activate with a shorted gate/source and no driver chip, it could be due to the its output capacitance (870pF here).

There should be a way to simply test this.

Perhaps some other EE's could confirm this or come up with another cause.
For that, start at post #412 for explanation and screenshots.

Thanks for now,   Itsu


added the circuit diagram as it was from post #412
« Last Edit: 2018-12-06, 20:48:51 by Itsu »
   

Group: Administrator
Hero Member
*****

Posts: 2645
It's not as complicated as it may seem...
In fact the entire bifilar coil can be characterized by the ratio of the capacitive coupling to the inductive coupling ...or treated as a distributed capacitance and inductance, which can be analyzed as a transmission line.

 O0

...or treated as a distributed capacitance and inductance, which can should be analyzed as a transmission line.
   
Pages: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 [17] 18
« previous next »


 

Home Help Search Login Register
Theme © PopularFX | Based on PFX Ideas! | Scripts from iScript4u 2018-12-10, 02:17:58