Mosfet Selection / Damping Resistor

[jon-ecm]

There has been some discussion on this topic but I had some questions

I ordered the following mosfets with mixed success
COSS VDS A RDS(ON)
1. FQPF6N60C 65 600V, 5.5 A 2 Ohm
2. FQPF7N65C 100 650V 7A 1.4 Ohm
3. FCPF9N60NT 140 600V 9A 0.3 Ohm

Number 1 never oscillates regardless of damping resistor, Decay is very slow and hardly any deflection in waveform when metal put near coil.
Not building enough current?
resonable em field created


Number 2 works good, decay's in about 20 us and provides good detection (requires 350 ohm damping resistor to stop oscillations [To low?] )
about 20% stronger em field produced than #1

Number 3 seems to provide good (if not best) waveform deflection when target put near coil
decays in about 25 us, also requires at least 350 ohms to damp reverberations
Has such high voltage spikes that it was freezing my tx circuit, had to play with it before getting it to work
about 40% stronger em field produced than #2

em field is a relative measure of the magnetic field being generated by the coil, obtained by
holding a speaker near the coil and comparing loudness

the coil is a 50' of 24 awg single strand teflon insulated with a diameter of about 8"

why do you think I need such a mean damping resistor, or dose 350 Ohms seem normal?

Why is mosfet #1 failing so miserably?

Thanks!
Jon

[Tepco]

Both #1 and #2 devices have too high Rds on resistance, #3 seems to be excellent choice. I don't know why #1 failed so miserably, but considering 2R Rds resistance at say, 2A coil current, 4V is lost, and another about 2V on coil and cable resistance, not much left to generate field. Simplified, of course, you never ramp up full coil current due to large resistive losses, device is just not suitable for this application.

Damping resistor value is about right.

You have to consider different coil design, 20-25uS is way too slow, when you add shield and coax this going to be even worst. Or at least, reduce number of turns on existing coil.

BTW Fairchild MOSFETs seems to be very useful for this, i did not even notice them before.

[porkluvr]

em field is a relative measure of the magnetic field being generated by the coil, obtained by
holding a speaker near the coil and comparing loudness

Jon

If you were measuring waveform decay time with the speaker nearby, that would have affected test results.

[mikebg]

Jon, the MOSFET operates also as damping element because at breakdown it dissipates energy.
See also
http://www.geotech1.com/forums/showp...61&postcount=4

[taliesin]

There has been some discussion on this topic but I had some questions

I ordered the following mosfets with mixed success
COSS VDS A RDS(ON)
1. FQPF6N60C 65 600V, 5.5 A 2 Ohm
2. FQPF7N65C 100 650V 7A 1.4 Ohm
3. FCPF9N60NT 140 600V 9A 0.3 Ohm

Number 1 never oscillates regardless of damping resistor, Decay is very slow and hardly any deflection in waveform when metal put near coil.
Not building enough current?
resonable em field created


Number 2 works good, decay's in about 20 us and provides good detection (requires 350 ohm damping resistor to stop oscillations [To low?] )
about 20% stronger em field produced than #1

Number 3 seems to provide good (if not best) waveform deflection when target put near coil
decays in about 25 us, also requires at least 350 ohms to damp reverberations
Has such high voltage spikes that it was freezing my tx circuit, had to play with it before getting it to work
about 40% stronger em field produced than #2

em field is a relative measure of the magnetic field being generated by the coil, obtained by
holding a speaker near the coil and comparing loudness

the coil is a 50' of 24 awg single strand teflon insulated with a diameter of about 8"

why do you think I need such a mean damping resistor, or dose 350 Ohms seem normal?

Why is mosfet #1 failing so miserably?

Thanks!
Jon

are we able to see the schematic ?

[jon-ecm]

If you were measuring waveform decay time with the speaker nearby, that would have affected test results.

I didn't test decay time at the same time.


25us is the final decay time, It's all setup on the metal detector with all necessary coax etc, dose that still seem slow?

About the mosfet limiting the coil surge voltage, I was aware of that when I chose a mosfet's with a breakdown voltage of over 600, I understand that the higher that surge voltage can get the faster it will end, is there any real disadvantage to letting the voltage peak so high besides dealing with it on the rx? Dose it reverse the field in targets making them harder to see or something weird like that?

I should be switching the mosfets pretty hard, I have a 6A mosfet driver on them.

Thanks!
Jon

[hobbes_lives]

There are several other parameters that come into play, in addition to what mikebg already mentioned about the breakdown voltage. The RDS(on) affects the TC of the circuit and the current flowing through the coil at switch-off. The total current flow possible will be calculated from the series resistance of the circuit, which includes the RDS(on) of the device. The current flowing at switch-off will be calculated by the total current flow possible, the TC of the circuit, and the pulse width (where are we at on the TC curve?). The Coss affects the total capacitance of the circuit, and will change the value of damping resistor required to achieve critical damping. Adding a diode in series between the MOSFET and the coil will greatly reduce the capacitance contributed to the circuit by the MOSFET. There are other things to consider as well, such as the speed at which the MOSFET can be switched off td(off) and tf, which is also affected by your MOSFET gate driver circuit and the Ciss of the device.

I'm by no means an expert on MOSFET devices, so please take this info for what it's worth, my 2¢

[Tepco]

Adding diode between MOSFET and coil is not a good idea at all. Trust me, i tried this many times, and in many different configurations. During the flyback period, diode will be forward biased most of the time when MOSFET breakdown voltage is reached, and will not "isolate "it, neither increase breakdown voltage. Finally, when it cuts off, at the end of cycle, what is remaining in the circuit then will be inadequately damped whit existing value of R dump. And as a result, this will ALWAYS be slower compared to MOSFET alone. Measured dozen times.

[Jose]

Adding diode between MOSFET and coil is not a good idea at all. Trust me, i tried this many times, and in many different configurations. During the flyback period, diode will be forward biased most of the time when MOSFET breakdown voltage is reached, and will not "isolate "it, neither increase breakdown voltage. Finally, when it cuts off, at the end of cycle, what is remaining in the circuit then will be inadequately damped whit existing value of R dump. And as a result, this will ALWAYS be slower compared to MOSFET alone. Measured dozen times.

-Is that what happens exactly when you install the diode between the coil and mosfet.
However, this modification allows to increase the value of damping resistance.
The end result is less time delay.
Jose

[hobbes_lives]

Adding diode between MOSFET and coil is not a good idea at all. Trust me, i tried this many times, and in many different configurations. During the flyback period, diode will be forward biased most of the time when MOSFET breakdown voltage is reached, and will not "isolate "it, neither increase breakdown voltage.

That's correct. The diode has no impact while the MOSFET breakdown voltage is exceeded. The diode's impact happens when the coil voltage drops below the MOSFET breakdown voltage and becomes reverse-biased.

Finally, when it cuts off, at the end of cycle, what is remaining in the circuit then will be inadequately damped whit existing value of R dump. And as a result, this will ALWAYS be slower compared to MOSFET alone. Measured dozen times.

As Jose mentioned, you must change your damping resistor value to compensate for the overall lower capacitance in the circuit, otherwise, you're over damping the circuit, causing the voltage across the coil to dissipate much more slowly than it can, due to the excess current flow. You need to adjust the resistor value higher until you achieve critical damping. Additionally, when adding a diode to an existing circuit, you may want to change the pulse width to a longer value to compensate for the loss of current through the coil, due to the diode's forward voltage drop when the MOSFET is on. It's not a bad idea to also add a high-value resistor from the MOSFET-diode junction to ground to bleed off the excess voltage prior to the next pulse being generated.

[Tepco]

Damping resistor is of course adjusted to optimum value every time, and this is still ALWAYS slower. Adding some high resistance "bleeder" across D-S terminal, but low enough to discharge Cds between pulses is mandatory. Diode forward drop is too small, no significant pulse width change is needed to compensate. And if pulse width is changed significantly, dumping resistor value must be too. At least in "normal" PI conditions. That is 200-500uH coil, 50-200uS pulse. Maybe this can work whit very short pulses, or very high voltage switches when breakdown voltage is not reached in process, but under normal conditions this is a bad idea. Another bad one is "active dumping", attempt to slow down MOSFET turn-off time in order to dissipate more energy on it. And this is why:

Ideal ringdown condition is pure LRC circuit, adding any nonlinearity to it (breakdown condition) will make it just slower. Adding even more nonlinearity, in form of diode or active dumping will just make things even worst. (I know US patent is given for this, John Hopkins university and active dumping, but will we ever see it in any MD, without being more complex than MD itself, and will it work?)

So far i can offer a sixpack of beer to anyone who can prove that series diode or active dumping can work faster than normal circuit under normal conditions (put aside uS pulses and sub uS sampling etc, we are not going to build UWB stuff, just normal detectors)

OK i will have to buy brewery if i'm not right on this, but this is tried so many times...

[hobbes_lives]

Well, I'm already using a series diode in a Surf PI Pro as one of my mods. I can sample down around 10uS, as opposed to 15uS stock. I'm not where I can verify the components right now, but I believe the diode I used was a 1N4937, and I believe my damping resistor is somewhere around 470 ohms. I'll have to double check these values later, as this is all from memory at the moment.

[porkluvr]

I won't enter the fray about a series diode. It seems apparent that there are pros and cons either way. I have a design in the works that generates an interleaved waveform (such as 1:3, 1:4, 1:5, etc.). Two FETs are used, with one FET transmitting the single pulse and the other FET transmitting the multipulses. With two series diodes and two different values of series resistances I can have two different levels of coil current from a single coil. The diodes serve the purpose of isolating the two series resistance from the coil when their associated FET is not active, but that is not exactly related to your previous discussion about series diodes. My point is, limiting coil current with series resistance can enhance flyback speed. (See where I mention the CS6-PI below.)

Something that has been bothering me (how do I put this in words...) is that Teflon wire seems to be a trap.

What is needed for a faster coil is to use stranded wire. Stranded wire (as compared to a single conductor of equivalent gauge) reduces eddy currents as a pulse is transmitted.
Tin plating seems to work better than silver because silver gives higher conductivity between individual strands, causing them to act more like a single conductor.
Teflon insulation is highly desirable because of its low dc.

Ok... we all know these things.

But a huge problem arises because it is next to impossible to find tin plated, stranded, Teflon insulated wire. At the temperatures necessary to apply and cure Teflon insulation, individually tinned strands will fuse together. If you want stranded tin plated wire, you had better look for some other insulation besides Teflon.

I read that Eric Foster used (or uses) PVC insulated wire in some of his PI coils, but maybe he has some filler material in between individual turns to help reduce the higher dielectric constant. I don't know, just a guess. Eric does not reveal all his cards.

One other thing he did (in his CS6-PI, notably) is use a series resistance (but no series diode) between the tx FET and the coil. That series resistance limits the value of peak current that the coil will charge to. That probably reduces overall depth of detection, but it also enhances flyback speed. He also used appropriately short TX pulses so as to not waste battery power. He did NOT use large gauge wire. That is another trap... and I know about traps. (Save the large gauge wire for monster size coils??)

"Fast" and "deep" are subjective terms, but doesn't it seem apparent that "fast" coils and "deep" coils are somewhat mutually exclusive? We can have one or the other, but how can we have the best of both worlds?

I wonder if bifilar wound coils like Moodz has been playing with might be something in the right direction. Or, maybe that would explain why Minelab uses (as I have read) Litz wire in some of their PI coils. Either method seems like drastic action to me but maybe that's what is required.

[jon-ecm]

So far i can offer a sixpack of beer to anyone who can prove that series diode or active dumping can work faster than normal circuit under normal conditions (put aside uS pulses and sub uS sampling etc, we are not going to build UWB stuff, just normal detectors)

I don't drink but I'll do some tests with my oscilloscope and post pictures when I get home, I can't come to logical conclusion why it wouldn't work.

If results don't add up I'll put a piece of steel in the coil when I test it without a diode. :P j/k

[hobbes_lives]

I believe the diode I used was a 1N4937, and I believe my damping resistor is somewhere around 470 ohms. I'll have to double check these values later, as this is all from memory at the moment.

The numbers on the diode are unreadable, but I'm 99% sure it's a 1N4937. The damping resistor value is actually 750 ohms, so I was way off on that.