I am researching high current power MOSFETs for my DC-DC converter and motor control applications. There are some with ON resistance as low as
0.0024 ohms (IRF2903Z) which is rated 260 amps (silicon limited) but package limited in TO-220AB to 75 A. An IRFBA90N20D has 0.023 ohms and is limited to 95 amps by its "Super-220" package, which does not have a mounting tab, and has three leads about 1.0 x 1.2 mm. The TO-220AB has leads about 0.6 x 0.9 mm. Wire of that size would probably be rated no more than about 10 amps, but I suppose the ratings assume the leads are very closely attached to a heavy PCB trace or other thermally conductive connector.
I would need something like #10 or #8 AWG to come even close to 75-95 amps. What is the best way to make such a connection? I would prefer not to resort to extra heavy PCB material. I would rather obtain or make some sort of copper connector. Probably something like 1/32" x 1/2" copper strip folded or rolled over the leads and then soldered would provide enough capacity, and then punch several holes in the copper for connecting several #10 wires with crimp lugs.
If anyone has had experience in high current applications with this package, I would appreciate any insight into the best way to make such a connection. Thanks.
Aren't you the fellow that we've been suggesting use higher battery voltages? Does your selection of a 200V part indicate agreement? Then can we assume more sensible lower currents?
Wire is allowed to self-heat; there's no sense requiring low copper wire temps if you're allowing high junction temps. Conventional wire ratings meant for conduit use, etc, are not aways appropriate.
I have multiple relevant experiences. One example, an 1800A switch, was made from similar-cased MOSFETs. I wired all the FET drains and sources together with a thick bus wire to promote current equality. Each FET was restricted to 38A nominal. Each group of six FETs (225A) was connected with nine #14 wires (25A each), soldered along the bus wire. The #14 wires were of equal length, and each set of nine was terminated in a compression lug and bolted to a wide thick copper bar carrying 900A. This setup runs reasonably cool with simple fans for the MOSFET heatsink assembly. The instrument (p/n RIS-254) has been in steady use for 9.5 years, in the famous light-stopping experiments.
Yes. I figure that the maximum power for a 12 VDC source would be about 2.5 kVA (200 amps), and then I would use 24 VDC, 48 VDC, and a maximum of 72 VDC for 15 kVA. That is about 20 HP, and should be fine for vehicular applications. Certainly it is easier to series connect batteries, and lower currents involve less copper loss. I would like to keep the voltage below
100 VDC, although even 48 VDC can be dangerous. There are many protective devices and components for 48 VDC (for telecom), so I might use that to make a 6 kVA DC/DC module for 360 VDC for a 240 VAC VF drive, and then use two in series for 720 VDC for 480 VAC VF drives.
Did you solder the #14 wire directly to the source and drain pins? I think I might use a crimp type butt connector without insulation, and use it to solder more easily to the pins. I will probably use #12 or #10 wire and a maximum of 50-60 amps per device. Since all the drains will be tied together to the positive supply, I was thinking about bolting them all to a copper bus bar attached to the heat sinks. Of course, this would not work for the Super-220 package, but that is for the low voltage device. I assume the drain tabs can carry the rated current.
I am not familiar with the light-stopping experiments. Is there a website with more information?
IR does this, and it's insane. The large print says X amps, and the fine print says X/3 or something. Who cares what the silicon is rated for?
Personally, I wouldn't run a TO-220 at more than 25 amps or thereabouts. TO-247's are better, but more cheap fets is prudent, as compared to trying to get huge currents through bleeding-edge parts.
Again, more small fets allows you to keep all the connections on a pc board and save a lot of hassle.
Why not use MOSFETs in the ISOTOP package. OK it costs more, but you don't have the expense of a pcb layout and the fiddling around to make decent-sized connections. ISOTOP is also easy to heatsink, with a lovely low overall thermal resistance.
Those values are textbook-theory for 25C junctions. Reality requires using corrections from the datasheet figure 10, plus wiring-resistance losses, e.g., at least a factor of two worse, 5 milli-ohms or more, and much less than 75A under continuous operation.
Again, apply the corrections, datasheet figure 4, use 3x worse, etc.
Indeed, so why not double or triple those voltages?
Just be careful. Most of the world uses 230 vac at home.
No, "I wired all the FET drains and sources together with thick bus wires" and "Each group of six FETs (225A) was connected with nine #14 wires (25A each), soldered along the bus wire."
. wiring scheme for 225A high-current paralleled MOSFET switch . . bus wires . ______ || || . | |===|| ########### The nine drain wires are shown, . | O |======|| same for the sources, not shown. . |______|=o || || The gates have individual small . ______ || ########### resistors to the gate-drive bus. . | |===|| || . | O |======|| . |______|=o || ########### . ______ || || . | |===|| ########### nine #14 AWG wired together for 225A . | O |======|| (each wire 9" long = 0.21-milliohms, . |______|=o || || Pd = 1 watt in each wire, at 20°C) . ______ || ########### . | |===|| || . | O |======|| . |______|=o || ########### . ______ || || . | |===|| ########### . | O |======|| . |______|=o || || . ______ || ########### . | |===|| || . | O |======|| . |______|=o || ########### . || ||
Perhaps I can post a photo to a.b.s.e., if you like.
Yes, I remember that test well. I was skeptical of some of the FET's IR was bringing around with current ratings over 100 amps. The test I did was to clamp a huge copper clamp to the tab on the TO-220 case, and another clamp out at the very tip end of the middle (drain) lead. I then turned up the current until the lead turned red hot and melted, and Bob remembers correctly, it was at 150 amps.
The leads of a TO-220 neck down, and if you make a connection closer to the body where the leads are wider, you will be able to carry 100 continuous amps safely, if your connection at that point is heavy enough to be a good heat sink.
Of course it will be less than fun to make said conenction.
Lies, Damned Lies, and things IR say in their datasheets. I've seen them give power figures so high that run thru the devices own Rtheta they give dT > 200C. They like to spec things with a junction at 25C...
The first thing I do when comparing Rdson is to scale it to Tj = 125C
No, the nine wires are each 1.9 milliohms (according to my wire table), and together are equivalent to 0.21 milliohms. The dissipation is about 1.2 watts per wire; it runs cool.
I would like to find the least expensive overall solution to making an efficient high power converter. The TO-220 and its variants seem to be most economical and widely available from multiple sources. I can minimize the heat sinking requirements (and boost efficiency) by using a very low resistance MOSFET. Of course, at higher voltages you have higher resistance or higher cost. There are lots of 65 V MOSFETs that would be OK for up to
24 VDC supply, then 100 V for a 36 VDC supply. For a 48 VDC supply, I would need at least 150 VDC, and they are relatively rare. There are more again at 200 VDC, which would be OK for up to 72 VDC. Above that, the on resistance and cost go up. I'll supply a breakdown of what I have found so far:
I included the one ISOTOP device to show how much more expensive they are. However, I might be able to use just one device rather than four in parallel, and simpler assembly may make it worthwhile for any production.
I have some TO-3 versions of the 60N06 that I used for my prototype. Unfortunately I destroyed them because the overcurrent shutdown was not connected. I have two more, but I really need to use the TO220 or other inexpensive package for higher power testing.
I have 50 pieces of the 75645P coming in (won on eBay for $32+$6), and they should be good for supply voltages up to 36 VDC. I should be able to drive them to about 40 amps at 50% duty cycle for power dissipation of about 11 watts each (probably closer to 20W at actual operating temperature). This is a power input of 1440W, or about 2 HP. I should be able to use four in parallel to get 5.6 kW with 160 amps input. Approximate efficiency would be
1-40W/1440W =97.2%.
For 48 VDC 40A input, the IRF52N15D would provide 1920 Watts, but power dissipation would more than double. It would probably be OK on 60 VDC, for
2400 Watts, but the 150V would be marginal for 72 VDC. The efficiency would be 1-80W/2400W = 96.7%.
The IRFBA90N20D would work up to 72 VDC and up to 50 amps, for 3600 Watts. Power loss would be about 120 Watts, for efficiency of 96.7%. Three in parallel would give me 10 kW, which is about what I was looking for.
Of course, I would not expect efficiency that high, because of copper losses and transformer losses. However, I think 92-95% is realistic.
I may use a compression type lug on the leads, with solder, and then bolt the lugs to bus bars. This will make it easier to replace any devices that fail (and I'm sure they will, until I determine optimal snubbers and overcurrent protection).
Thanks for your comments. Now to get back to work on this beast.
Maybe that's why Id/Idm is specified on the data sheets rather than Is/Ism. ;-)
Best regards, Spehro Pefhany
--
"it\'s the network..." "The Journey is the reward"
speff@interlog.com Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog Info for designers: http://www.speff.com
I don't know about these specific parts, but many high-power devices use a spring-clip connection instead of wirebonds.
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus |
| Phoenix, Arizona Voice:(480)460-2350 | |
| E-mail Address at Website Fax:(480)460-2142 | Brass Rat |
| http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
The weak link is not the wire, but the wire-bond connection to silicon metallization.
If you check higher current mosfets, you may find unconventional internal connections that don't involve single wire bonds. It can be multiple wire-bonds or a pre-formed source lead that makes contact directly at multiple points.
Getting DC or low-frequency AC current through the connection is not the usual headache with these parts, but maintaining control for higher-frequency repetitive transitions. Your 400Hz application is close enough to DC that this may not be an issue.
Your concern should be with interconnection in the unit and thermal control. This is made easier if your topology allows direct bonding of the drain to heatsinks that can be used as conductors. Source contact should be with a suitable conductor that also makes early connection to a suitable bussbar.
Use of copper foil as gasketing or bussbars is effective, but is difficult to employ in a manufacturing process, without careful tooling.
My purpose in making the test described was to find out what current the external leads were capable of handling, since they are (nearly) the same on all devices that conform to the TO-220 spec; I didn't want the current to pass through any internal bond wires for my test. The internal conductors can be whatever the manufacturer chooses to make them, and lately some low-voltage FET's internals have gotten better than the external leads. The external leads are the limiting factor on those FET's, and, of course, the connection to the leads has to be very good to get maximum capability. On a higher voltage FET the external leads probably won't be the limiting factor.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.