IRF740 Coss parameter

[Ray-NM]

I was reading the tech specs on the IRF740 mosfet last night and noticed something about Coss which is the output capcitance. It looks like the output capacitance increases as the fet is turned off and starts to increase once its voltage across it goes below about 40 volts. I was wondering if this affects the delay time significantly as the reverse emf approaches zero volts.
Any thoughts on this and if it does have a negative effect then how can you overcome it. I have one thought but I will wait and see if it is worth pursuing.
RayNM

[bbsailor]

Capacitance Contributors

Ray-NM

COSS is typically rated at 25V on datasheets and represents the MOSFET drain to source voltage. But since most PI machines use 12V, the COSS curve is higher in the 10V-12V range. Check the D to S voltage and COSS graph on the MOSFET datasheet. On some MOSFETS the curves are pretty steep and the capacitance changes greatly with D to S voltage. Newer, low COSS types have curves that are less steep and may have 35 to 50pf at 25V but only raise to 60pf - 75pf at 10 to 12V. Some newer MOSFETS or new lower voltage MOSFETS may even have a lower COSS in the 20 to 30 pf range.

If we take an inventory of all the things that add circuit capacitance, the 740 MOSFET COSS adds the highest capacitance,
followed by the coil coax capacitance,
followed by the coil winding capacitance,
followed by the shield-to-coil capacitance,
followed by the clamping diode capacitance.

The total capacitance is reflected in the value of the damping resistor, higher means less capacitance.

Higher TX circuit capacitance causes the fall time of the flyback voltage to take longer and thus extend the delay until the signal stabalizes when the earliest sampling can occur.

The flyback voltage fall-time is based on the value of the coil inductance divided by the damping resistor (Rd) plus the value of the input resistor (Rin) to the first amp which is in parallel with Rd while the diodes are clamping. An Rd of 1K and and a Rin of 1K is a combined 500 ohms. 300 uh divided by 500 ohms is 0.6us for the first TC and additional TCs until the voltage falls below the diode clamping voltage of 0.6V to 0.7V.

As you can see, coils that have higher inherant capacitance in their design and that see higher capacitance in their TX circuit will take longer to settle down and take longer to allow the sampling to occur. Higher flyback voltages keep the diodes in clamping longer and keep the first amp in saturation longer.

Once you decide to put a series resistor to limit the flyback voltage and coil current, you can then use a lower voltage MOSFET which can have a lower COSS. This all translates into the ability to sampe earlier for more sensitivity to smaller, lower conductivity targets.

Use: lower COSS MOSFETS, shorter coax cables with the least pf/ft, coils with less capacitance (Teflon wire), less coil-to-shield capacitance (use Polyethylene (PE) spiralwrap spacer) and a coil shield that is not detected at the lowest delay.

These tips should move you in the right direction.

bbsailor

[gef12]

IRF740 MOSFET

hI aLL
I keep hearing about mosfet that are capable of a COSS less than 100 pf
Can anyone direct me to some fets... as it is a bit of a struggle searching the net and opening untold numbers of PDF files to search for an appropriate one.
What I think we a looking for is a MOSFET with Rds about .5ohms can handle switching of at least 5 Amps with Max V 400 and COSS less than 100pf would be nice.
Anything better than IRFI740 .. thats what I am currently using.
cheers
gef

[gef12]

mosfet

Hi again
That would be in TO-220 package preferably
thanks again
gef

[bbsailor]

Lower COSS MOSFETS

gef12

FQPF18N50V2 and others in that series.
IXTP3N120 and others in that series
FQP6N15 and others in that series
FQP6N25
2N7000 for a low power, high frequency PI design
IRF830 and others in that series

This should get you started in your research. Look at the chart to see the actual COSS at the D to S voltage in your circuit design, usually in the 10 to 12V range.

If you want to experiment with different MOSFETs, cut a flat-pin type IC socket to make a 3-pin MOSFET socket. Then you can plug in a variety of MOSFETS, adjust the damping resistor, and see which ones give the best performance for your application. Only use the flat pin type like those available in Radio Shack. The MOSFET leads will not fit into the round pin type sockets.

Run your PI through the full range of pulse widths to see if the MOSFET gets hot. If it does, you can add a 2 to 3 ohm resistor in series between the MOSFET and the coil, or add a heat sink.

You can lower your delay settings when there is less capacitance seen by the coil in the TX circuit. As the TX circuit capacitance goes down, the damping resistor value will be higher. Changing the MOSFET will give the best improvement when other factors in the circuit the contribute capacitance are also minimized, such as coil interwinding capacitance, coil-to-shield capacitance, coax length and total coax capacitance.

Keep good notes and share your findings.

bbsailor

[gef12]

mosfets

Yep thanks muchly for that. Now I can get down to business
cheers gef

[Ray-NM]

Series Fets

Here is an Idea I have about decreasing Mosfet capcitance in the circuit.
Not sure if it is valid with Mosfets but using the series cap formula it should work but here goes anyway. If you tie 2 fets in series with the gates tied together and drain from one to the source of the other would the effective Coss value be 1/2 or 103pf ? The Rds(on) would be doubled or 1.1 ohms. Plan on trying it tonight to see what the effect is on delay time. Any glaring misundertanding on my part feel free to let me know.
RayNM

[baum7154]

How did your series FET test work out? Isn't putting a series diode like an HER208 in the tx circuit doing the same thing with a device of far less capacitance? C1 X C2/C1 +C2.

[green]

How did your series FET test work out? Isn't putting a series diode like an HER208 in the tx circuit doing the same thing with a device of far less capacitance? C1 X C2/C1 +C2.

I'm using a MUR460 since it was available in LT spice and seemed to work good. Tried a 1N4937 in circuit, not as good. Has anyone tried the MUR460 and found a better diode?