Re: [CAD_CAM_EDM_DRO] Re: Was Chopper circuit - Now ver 3.2

mariss92705 wrote:

> Jon,
>
> Couple of good questions, let me address them in order.
>
> MOSFET circuits require RF layout techniques because of their
> extremely fast switching times. This means a pair of board layers
> dedicated to ground plane and power distribution. That leaves two
> circuit interconnect layers for a total of 4 layers. Multi-layer
> boards are more expensive than double-sided boards by about 50%. If
> the boards are very small, this is not much of a factor.

Well, I built a 10 A, 80 V full-bridge PWM servo amp that runs at 100
KHz. Definitely RF! And, it is all on a 2-layer board, with high gain
differential amps about 2" away. I would have liked extra layers to make
the traces even wider, but I made it work with the 2 layers. It might
be more compact if 4 layers were available. But, the prototyping
costs make that an expensive proposition. As it was, with a number
of other aspects of the design being developed at the same time,
I got it fully implemented and working on the 3rd revision (and one
of those was based on the awful Harris HIP4080 chip).

>
> The IR2104s are rated at 600VDC making them suitable for 440 volts
> RMS circuits, so 80VDC is pretty much under-utilizing them. There are
> other factors that get in the way for higher voltages though.
> Primarily it is the MOSFET drain to source resistance increasing as
> the square of the Vds avalanche rating.

If you can live with 25 KHz swithcing frequency, fast IGBTs are amazing!
Where $18 TO-247 HEXFETs ran so hot they had to be paralleled, a single
$6 T)-220 IGBT ran almost cold! You have to drive them hard, and may need to
have the control chip powered on before power is fed to the bridge,
but they are vast improvements over 400+ V FETs. (The trick is you
must NEVER let them enter the linear mode, as they have the reverse
transconductance/temperature effect similar to bipolar transistors, and
the current will hog to the hottest part of the transistor.) IR's driver
chips are fine for driving them, but the Collector-Gate capacitance
can turn them partially on during the power up ramp, and if the driver chip
can't hold the gate at Emitter potential, then they blow on power-up.

> The IR540 contains a silicon die that is about as big as you can get
> in a TO-220 package. It is rated at 100 volts. A 200 volt rated
> device would have to have silicon with 4 times the area to keep the
> same on-resistance. It would not fit in a TO-220 package and would
> have to go in the much larger and more expensive TO-247 package.

The advantage of the IGBT is it doesn't have simple resistance, it has
a characteristic Vce (sat) which usually runs 1.8 V for small ones and
2.5 V or so for the big ones. In some inverter and power factor correction
circuits I was working on a few years ago, I was getting 15 V Vds at
about 75% rated current of the HEXFET. That will run up a big power
loss very quickly!

> If a 200 volt TO-220 part (IR640 perhaps) was used, current would
> have to be scaled back 50% to keep dissipation normalized. Twice the
> voltage at half the current still yeilds the same output power.
> Nothing would be gained by it.
>
> If someone was so inclined, they could use IR640s instead of IR540s
> and run it at 160VDC. I just don't see the point unless someone wants
> to kill themselves because they don't want to use a transformer.

It isn't the transformer issue, but the current needed to run 1 Hp+ spindle
motors. The amps are sky high if you can even get such motors below
90 V. (The aircraft motors run abot 45 A / Hp at 24 V.)

You ought to check into IGBTs, anywhere over 100 V rated devices, they
look much better than MOSFETs.

Jon