Is IRF9640 the best choice for a high-side switching topology?

[kt315]

i just transfer part from N-channel in P-channel view. nothing new here. but if you do not see... i draw

[CrizzyD]

What's the purpose of the 1 Ohm? Just for current limiting?

[kt315]

sorry, my name is not Eric Foster. can not answer. i asked Eric to write a book... two times... he just answered me 'heh! i am undying'.
that's all. gone off. RIP.

[kt315]

---------avalanche-------
Re: Coil switch


Posted by: Eric Foster (---.ipt.aol.com)
Date: December 26, 2001 03:34PM


Hi Carl,
Even the method of switching the power transistor is interesting, as there are so many inter-related factors. Assuming for most applications nowadays that Mosfets are used, there are, as you say, various methods of driving them. Strangely, although Mosfets are devices with a very high input impedance, the drive to them is recommended to be from a low impedance source. This is because the parasitic input capacitance is quite high, at around 1500pf, for the IRF740 types, and you need a low impedance to charge and discharge this capacitance quickly for efficient switching operation. As you have outlined, there are various ways to do this, and the question, at the end of the day, does it matter?
In a PI detector, we are concerned with driving a current though a coil for a short period of time, and then switching it off cleanly and rapidly. Any switch-on delays do not really matter, as the growth of the current is limited by the coil’s inductance. This delay, before the current reaches its maximum, will likely be several orders of magnitude slower than any delay caused by the capacitance and drive impedance of the Mosfet. At switch-off, we need to examine things in a bit more detail. To get the best eddy current response from an object, the switch-off of the field from the coil, needs to be at least 5 times faster than the object time constant. Obviously, if the switch-off is the same length of time as the time constant of the object, then there will be no signal to observe. At the point where the drive pulse to the gate is terminated, nothing much happens initially, as the gate retains the voltage on its capacitance. This capacitance will then start to discharge through the drive circuit impedance. On many of my circuits, including the Aquastar, this is simply a 100R resistor to the source. The turn-on is achieved by a PNP transistor that connects the gate to the drain supply via a 47R resistor. The resistive divider pots down the drive voltage so that no more volts than is necessary for proper switching is applied to the gate, otherwise it will take longer to discharge the gate capacitance. The discharge time constant is 0.15uS, giving a total discharge time 0.75uS, so this is more than enough speed for a treasure hunting PI. However, the gate voltage does not have to fall all the way to zero, but only to the gate threshold voltage, where the Mosfet just ceases to conduct; about 4V max for the IRF740. Using a 15V supply to the Mosfet and using the above circuit arrangement to maximum gate turn-on voltage is 9.5V, hence the time to reach the turn-off volts is only about one time constant, which takes us back to 0.15uS. Now, if we look at what happens in the drain/coil circuit at the point of cut-off of the Mosfet. Inductors don’t like a change of current, so the energy stored in the magnetic field will collapse and try to induce a voltage to keep the current in the coil flowing at the value it had just before switch-off. This is why we get a high voltage appearing across the coil at this point in time. The Mosfet is now a very high resistance and, if unrestrained, the coil voltage could reach over 1000V. However, the Mosfet has a breakdown, or avalanche voltage, which in the case of the IRF740, is about 400V. The coil voltage quickly rises to this level and then the Mosfet drain/source junction breaks down, and further voltage rise is prohibited. The Mosfet is then a low impedance across the coil during this high voltage period, which gives another, rather longer time constant. If you insert a small (0.1R) current monitoring resistor in the coil circuit and look at the waveform on a scope, you can see the current ramping down linearly for the duration of the high voltage pulse. As soon as the coil voltage falls below the 400V, then the Mosfet becomes a high impedance again, and it is now the duty of the damping resistor to absorb the rest of the energy: yet another time constant. Only after everything has settled down, including the recovery time of the first receiver amplifier, can we then look at the microvolt signal levels from objects at some distance from the coil. The delays from whichever type of Mosfet drive circuit used, only make a very small contribution, in my experience.
The things that have a major effect on the field switch off time are 1) the coil inductance, 2) the coil current. 3) the breakdown voltage of the Mosfet, and 4) the capacitance of the coil/cable circuit. However, these will keep for another time.
Eric.

[CrizzyD]

This is absolutely awesome stuff thanks a lot! I see Eric's name come up a lot as I'm researching PI, is he not around anymore?

[Carl-NC]

i just transfer part from N-channel in P-channel view. nothing new here. but if you do not see... i draw

The diode will not help at all with the avalanche issue. Remember that on a P-drive TX the flyback is negative. This application of the diode is normally used to reduce the effect of MOSFET capacitance, which is not the issue here.

What's the purpose of the 1 Ohm? Just for current limiting?

You normally use a super-low ESR cap with ~1Ω to the supply. This helps guarantee that the cap, not the supply, provides the turn-on current. During the off time the cap recharges.

[Carl-NC]

This is absolutely awesome stuff thanks a lot! I see Eric's name come up a lot as I'm researching PI, is he not around anymore?

Eric died last November.

[CrizzyD]

You normally use a super-low ESR cap with ~1Ω to the supply. This helps guarantee that the cap, not the supply, provides the turn-on current. During the off time the cap recharges.

Hmm, the original design had this but I removed it because the resistor was a constant point of failure. 1Ohm and 3300uF also forms a 50Hz low pass filter so I assumed it was for filtering purposes. I think that my only option is to just source a P-Channel FET with higher breakdown, although the flyback voltage seems to be much higher than what's possible with a P-Channel FET i.e. following the angle of the flyback, it looks like it would settle at around 800V. The best option seems to be the IXTX32P60P with -600V but it's crazy expensive compared to most other FETs!

[kt315]

This is absolutely awesome stuff thanks a lot! I see Eric's name come up a lot as I'm researching PI, is he not around anymore?

Eric mainly was in TECHOLOGY FORUM... you can still find his posts there. use SEARCH knob.
just example
(1) 2010 Lesson 1 - Flyback | Find's Treasure Forums (findmall.com)

[Carl-NC]

It's hard to extrapolate the peak because the flyback rounds over (it's an RLC response). The mathematical peak of the flyback is



Where I₀ is the peak current at turn-off and R_D is the damping resistor value for exact critical damping. This comes to about 600V/amp for R_D = 820Ω. If you use, say, a 600V PMOS and the flyback still clips then that is still better (less heat) than clipping at 400V.

[Carl-NC]

Here is a thread that includes illustrations:

Optimizing Target Responses

[CrizzyD]

Awesome thanks Carl! I wish I'd found this forum when I started work on this. The design I'm improving is purely analogue but I've added an ARM Cortex M4 microcontroller to process the signal and so many things that I've implemented in isolation I've now found on the forum. A median filter for example...

[CrizzyD]

Adding a series 5Ω resistor reduces the coil current, which is the opposite of what you want. Then, you have to increase the pulse width just to get back to where you were, at which point the FET will be just as hot.

The only practical way to deal with this is to use a higher breakdown FET to avoid avalanche. However, the flyback energy still has to go somewhere, and what you will find is that R58 (damping) and the input clamping resistor will get hotter.

The IRF9630 will not help you here, it has the same breakdown. If that was not the issue, then it would help reduce the FET capacitance and make recovery a little faster. I don't know where KT is suggesting to add a diode.

I agree that doesn't it makes sense to introduce a series resistance that reduces the back EMF to a value closer to what the FET can handle i.e. 600V -> 200V, as theoretically with a fixed coil design, one could reduce the pulse width and therefore reduce the energy in the coil (or reduce the FET turn-off i.e. decrease the slew rate). So I can't understand why the original designer of this PI transmit circuit would included a series (R27, R2 resistance between the switching element and the coil. Why not just reduce the pulse width? The coil connects to points (5) and (6)

[GeoMax]

Hi,

Have you tried a MTP2P50EG? This PNP MOS-Fet has 500V and a RdsOn of 6 Ohms.

[CrizzyD]

Hi,

Have you tried a MTP2P50EG? This PNP MOS-Fet has 500V and a RdsOn of 6 Ohms.

I believe that's an obsolete part now. I'd need something that has some kind of support from the manufacturer. Also, with an RdsOn of 6 Ohms, that would greatly affect the energy that could be put through the coil and I'd likely need to increase the pulse width substantially (like 5 or 6x longer).