What is the purpose of the series diode?

[kt315]

do you mean the capacitance between GATE and SOURCE-DRAIN channel or SOURCE and DRAIN?

[Teleno]

do you mean the capacitance between GATE and SOURCE-DRAIN channel or SOURCE and DRAIN?

Te capacitance between drain and source (Coss).

It's in the order of nF at low voltages and is in parallel with the coil. See Coss in Fig. 5 of this datasheet https://www.vishay.com/docs/91070/sihf840.pdf

I've explained it in this post https://www.geotech1.com/forums/show...576#post226576

[kt315]

do you use special low capacitance probe with o-scope?

[moodz]

... may also consider the cascode switch configuration ( used in high speed high voltage switches since the valve era ) is just as fast or faster than the diode configuration. Also the series voltage drop with due selection of mosfet devices can be made lower than a series diode ... so more efficient & higher coil current. The cascode configuration mitigates the miller capacitance. NTX is a negative supply voltage.

DIODE TX



CASCODE TX

[moodz]

comparison of different damping configurations .... same voltage / coil ... different TX configs.

[Teleno]

... may also consider the cascode switch configuration ( used in high speed high voltage switches since the valve era ) is just as fast or faster than the diode configuration. Also the series voltage drop with due selection of mosfet devices can be made lower than a series diode ... so more efficient & higher coil current. The cascode configuration mitigates the miller capacitance. NTX is a negative supply voltage.

DIODE TX
[ATTACH]55639[/ATTACH]


CASCODE TX
[ATTACH]55640[/ATTACH]

The problem with the cascode is the lower MOSFET works at a low Vds where it's capacitance is at a maximum. In these conditions the capacitance seen by the coil is practically the full capacitance of the upper MOSFET, as if the lower MOSFET didn't exist.

Active damping tequires an exact balance of the polarization of the damping MOSFET to the energy of the coil. Even a small mismatch, such as grpund effect or a target throws the scheme off balance, requiring a complex feedback control loop.

[moodz]

The cascode is not something I invented ... it’s been used to reduce capacitance for years. The upper mosfet turns off as a charge is trapped by the lower mosfets drain when it turns off .... this charge is only a few volts compared to hundreds volts trapped by the diode tx configuration. So when the lower mosfet turns off the upper one also turns off but faster due to reduced charge thus reduced capacitance.

the name cascode originates from cascaded triode and allowed early valve amplifiers to reach 100s of mhz which was not normally possible due to excess capacitance ... this was back in the 30s or 40s as far as I know.

[Teleno]

The cascode is not something I invented ... it?s been used to reduce capacitance for years. The upper mosfet turns off as a charge is trapped by the lower mosfets drain when it turns off .... this charge is only a few volts compared to hundreds volts trapped by the diode tx configuration. So when the lower mosfet turns off the upper one also turns off but faster due to reduced charge thus reduced capacitance.

the name cascode originates from cascaded triode and allowed early valve amplifiers to reach 100s of mhz which was not normally possible due to excess capacitance ... this was back in the 30s or 40s as far as I know.

Actually the Miller effect takes place between the drain and the gate. The cascode maintains the drain at a constant voltage so the Cdg does not get amplified. It was invented to teduce the capacitance as seen from the input (base or gate).

From the drain of the upper MOSFET the capacitance looks like two capacitors in series. Since the Cds increases by orders of magnitude at low Vds, the lower capacitor has a a value in the 1000s of nF range, while the upper one is in the 100s range, so no significant reduction takes place when in series. The simulations do not produce a significant improvement, confirming this.

What I have observed in my experiments is a physical limit as to how early it can be sampled even with a fast coil (90 uH, 35pF) because of the initial sudden drop of the drain voltage, when using a diode, that induces a crosstalk in the signal still important after the coil's transient has decayed. It looks like a target signal in the 10s of millivolts that's enough to drive the amplifier into saturation despite the coil's transient being already gone. I'm talking about a 2us transient followed by a much longer, smaller decay from the drain's voltage.

[Teleno]

What I have observed in my experiments is a physical limit as to how early it can be sampled even with a fast coil (90 uH, 35pF) because of the initial sudden drop of the drain voltage, when using a diode, that induces a crosstalk in the signal still important after the coil's transient has decayed. It looks like a target signal in the 10s of millivolts that's enough to drive the amplifier into saturation despite the coil's transient being already gone. I'm talking about a 2us transient followed by a much longer, smaller decay from the drain's voltage.

Some scope screenshots relating to this.

Coil is 90uH, 35pF. MOSFET is IRF840. Series diode is a SiC Schottky with 1000V rating.

The decay of the blocked drain voltage looks like this during one PI period:




A close-up of the transient:





The orange line is there to highlight the non-linearity that shows at right at the beginning and extends into the 8us region when the coil already has decayed (2-3us). As a result it's not possible to sample as early as the coil's characteristics would theoretically allow us to.

I'm going to try an IGBT without the diode and see how it behaves. IGBTs have a tail current that might be a problem, and unlike MOSFETs they're destroyed by overvoltage.

[moodz]

hmm ... well I did not want to get into a P>>ing competition ;-) but my results differ somewhat.

I am using a "real" coil ... actually a commercial PI coil ... and I get the following results. The preamp gain is approx 10000/16

the coil is damped at just over 1 microsecond and the first sample is taken at 1.5 microseconds.

I am using an IRF9640 in the front end ... its possible to get even faster using more exotic mosfets.

[Teleno]

hmm ... well I did not want to get into a P>>ing competition ;-) but my results differ somewhat.

I am using a "real" coil ... actually a commercial PI coil ... and I get the following results. The preamp gain is approx 10000/6

the coil is damped at just over 1 microsecond and the first sample is taken at 1.5 microseconds.

I am using an IRF9640 in the front end ... its possible to get even faster using more exotic mosfets.

[ATTACH]55649[/ATTACH]

I'm not in competition mode, I'm trying to understand what happens when using a blocking diode (see the title of the thread) and why the results are worse than predicted in the simulations.

Not trying to invent anything, just pushing the limits of the configuration in the title and see how far it can be optimized.

The simulated circuit does not produce the secondary, slower decay. It does though when I connect a 100k-200k resistor in parallel with the MOSFET, so I assume the effect is caused by the faster rate of discharge of Cds right after the peak.

Sure, you can sample earler at 2V with low gain, but what I'm trying to achieve is a pure coil decay.with no crosstalk from the blocked capacitance.

Have you ever tried an IGBT?

[moodz]

I'm not in competition mode, I'm trying to understand what happens when using a blocking diode (see the title of the thread) and why the results are worse than predicted in the simulations.

Not trying to invent anything, just pushing the limits of the configuration in the title and see how far it can be optimized.

The simulated circuit does not produce the secondary, slower decay. It does though when I connect a 100k-200k resistor in parallel with the MOSFET, so I assume the effect is caused by the faster rate of discharge of Cds right after the peak.

Sure, you can sample earler at 2V with low gain, but what I'm trying to achieve is a pure coil decay.with no crosstalk from the blocked capacitance.

Have you ever tried an IGBT?

Understood ... I have not tried an IGBT ... had thought about them but I think at the time the spec I was reading indicated they had relatively slow switching speed and high conduction voltage drop ... but that was a few years ago. Maybe there are better ones ?

I would not use a Shottky diode though .... the reverse leakage current is a problem.

[Op04]

I'm trying to get lower flyback current using 800V mosfet, if I put a diode with lower 600V ultrafast is this will work to make higher volt flyback with lower current? Sorry if this not related

[Carl-NC]

Do you mean you want to reduce the flyback voltage? There is no flyback current.

[Op04]

Do you mean you want to reduce the flyback voltage? There is no flyback current.

I mean lowering coil discharge current but make flyback higher, previous mosfet IPI65R150CFD (with no diode) get stun (I believe coil discharge is too much high)