EEVblog Electronics Community Forum
Electronics => Beginners => Topic started by: cadr on April 01, 2021, 04:12:05 pm
-
I am interested in high-speed switching with MOSFET transistors (~7Mhz range, for use in class E or D RF amplifiers). I have been looking for transistors that have low rise/fall time and can dissipate a medium amount of power.
I found some Nexperia parts like the BUK9Y59-60E (pdf datasheet: https://datasheet.octopart.com/BUK9Y59-60E%2C115-Nexperia-datasheet-87153647.pdf) It has turn on/off and delay time under 10ns which should work for me, and I've found drivers that should work. My question is that it claims to be able to dissipate 37W of power, which seems very high to me for the size of package. Am I missing something there? Also, I'm not familiar with this "LFPAK56" package - would I just use the copper board for the heat sink? I found this application note (https://assets.nexperia.com/documents/application-note/AN90003.pdf), but I'm used to using big heat sinks on my PA finals, so this is a new world for me. (I've read some things on thermal design, but it didn't all go into my head.)
I'd like to know how much power I could (very conservatively) have these dissipate (and how), to see what I could reasonably design as the output power of the overall amp.
Thanks for any and all help!
-Ben
-
This might help you with some of the acronyms. AN11158 (https://assets.nexperia.com/documents/application-note/AN11158.pdf)
Quoted from page 5/29
Ptot (total power dissipation) is the maximum allowed continuous power dissipation for a device with a mounting base at 25 °C. The power dissipation is calculated as that which would take the device to the maximum allowed junction temperature while keeping the mounting base at 25 °C. In reality, it is difficult to keep the mounting base at this temperature while dissipating the 105 W that is the calculated power dissipation for the BUK7Y3R5-40H. In other words, Ptot indicates how good the thermal conductivity of the device is, and its maximum allowed junction temperature.Note that some other semiconductor vendors quote performance when mounted on a copper PCB usually 1 inch square. In practice, this information is rather meaningless as the semiconductor vendor has no control over how the device is cooled. See AN10874 - L FPAK MOSFET thermal design guide. AN10874 describes different techniques that can be used during the design phase to ensure that the PCB layout provides optimum thermal performance
-
Very tricky data sheet.
The 37 W dissipation is at Tmb, not Tamb. Tmb" being "mounting base temperature". Good luck with implementing that.
-
Very tricky data sheet.
The 37 W dissipation is at Tmb, not Tamb. Tmb" being "mounting base temperature". Good luck with implementing that.
I think they assume ambient (Tamb )to be 20oC. Operating temperature is between 25 and 175oC So I'm guessing the the 25oC is based on about 1W or 4.03 K/W with ambient of 20oC .
-
Thanks for the posts!
So, rereading that "AN90003" document, it seems to say on page 10/11 that the limiting factor is that FR4 PCB can't get over 130C, and that FR4 has a thermal resistance of ~50 K/W (not sure why they are saying Kelvin instead of Celsius). They work an example and say that you can't dissipate more than 1W if your ambient is 80C due to that. (I wouldn't be operating at that ambient temperature, but still seems to be the limiting factor.) Does that sound like I'm reading that correctly?
Any thoughts to what kind of heat sink I could use here to increase this? I'd really rather use something in more of a TO-220 package or something, but haven't found any with very low rise/fall times.
Cheers!
-
Right, they rate those, and D(2)PAKs and others, the same way: immersed in boiling freon or something like that -- holds Tc very nicely at 25°C. Implementation, is an exercise for the end user...
You can sink extra watts from SMTs using thermal pads on top and bottom, and thermal vias in the middle. Multilayer board helps, providing more total copper thickness with shorter distance through laminate (which has terrible thermal conductivity). 5, maybe 10W is reasonable this way. At least you don't need to worry about stray capacitance here: a switching transistor has many times more capacitance than the layout will. So don't be afraid to overlap GND planes and power pours.
Point of advice: don't pay attention to rise/fall times. They're entirely dependent on the test condition. Like, here's a completely random selection from Fairchild:
https://rocelec.widen.net/view/pdf/c0ogsfdyui/ONSM-S-A0003584205-1.pdf?t.download=true&u=5oefqw
notice (external) R_G in the test, and notice the insultingly large times (100s ns).
R_G (internal gate spreading resistance), when given, gets closer to the true limit of device performance. It's not necessarily the ultimate limit, because it depends on how it's distributed: some transistors, there's a fraction of resistance between the gate bond pad and the first MOS cells, then some more resistance to the next, and so on; gate drive kind of diffuses out over the die.
Diffusion has a sqrt(F) dependency, so that in this case, gain drops off rather gently -- effectively you might have only some tiny fraction of the transistor still actively amplifying, at a high enough frequency; it might still have net gain, but the maximum undistorted output is pitiful, meanwhile you're burning a ton of supply current.
Other (newer?) transistors may have more of a fractal gate wiring pattern, i.e. a low resistance connects the gate bond pad to a bunch of feeders, which have higher resistance to even more, even thinner feeders, which have even higher resistance to the actual gate elements. Because the resistances and capacitances are in a geometric series, they all have a similar RC time constant, and more of a dominant-pole response is had. In other words, the equivalent gate circuit looks like just a single lumped RC. (These types can oscillate at 400MHz or more, if you put external capacitance between gate and source -- forming a series resonant tank between stray lead inductance and that capacitance. This includes the capacitance of a mere zener diode, by the way. And again, the amplitude isn't huge, it's not going to be a full like 500V from an offline switching transistor, but it's more than enough to ruin your radiated emissions. For these cases, you need a damping resistor.)
Anyway, shortwave frequencies aren't anything special, hams have been doing that with IRF510-540 and such for decades -- again, mind the lead inductance, don't blow up the gate with excessive drive power, and put it on a heatsink.
If you want to deep dive into class D amps at these frequencies, consider GaN MOSFETs -- you can get dev kits that already solve layout and drive for you, or inverter ICs that do much the same. They are very fast: frequencies in this range are comparable to Si MOSFETs at fractional MHz. I recommend this over discrete parts, as layout is even more critical, so it's nice to not have to figure it out yourself as a beginner. It's still your responsibility to provide reasonable signals (PWM, dead time, whatever) and protection, and mind that, because the dies are even smaller, they take very little overload indeed before popping. Likewise, little or no linear mode operation is allowed, or avalanche breakdown.
Tim
-
Thanks Tim! I'll try to absorb all of that :)
-
Thanks for the posts!
So, rereading that "AN90003" document, it seems to say on page 10/11 that the limiting factor is that FR4 PCB can't get over 130C, and that FR4 has a thermal resistance of ~50 K/W
FR4 has horrible thermal conductivity, several hundred times worse than copper. You can basically assume it is a total thermal insulator. If you are actually trying to dissipate anything near 37 W, you need a device that interfaces directly to an aluminum heatsink, possibly with fan cooling.
You can dissipate a few Watts with large copper pours on PC boards, but are limited by the thin copper foil on the board.
Jon
-
FR4 has horrible thermal conductivity, several hundred times worse than copper. You can basically assume it is a total thermal insulator. If you are actually trying to dissipate anything near 37 W, you need a device that interfaces directly to an aluminum heatsink, possibly with fan cooling.
You can dissipate a few Watts with large copper pours on PC boards, but are limited by the thin copper foil on the board.
I'm learning so much :)
Not planning on trying to get near the limit, more trying to understand what that is/what it would take to get there. But would actually want to operate it really far under the limit. Regarding heatsinks, still unsure how you heatsink a device in this form factor *other* than the pcb. I've generally built things so far in a manhattan/island pad style, and used packages where I could attach a heat sink to the device, so this is all new territory.
Thanks for the advice!
-
Low profile surface mount heat sinks Dpak/LFPak (https://ca-en.alliedelec.com/m/d/a9418620c5818658d7c8c2d17b0d1380.pdf)
Another method is to attach a heat sink to the pad of the PCB next to the component by soldering or thermal adhesive. I've seen some that bolt right through the PCB .
-
If you add up all the thermal resistances from Tja to the heat sink you can get a good estimate of how hot the heatsink will get when it's outside the box.
The total K/W or C/W multiplied by the power dissipation you have calculated for your device. Well that's the starting point see https://sound-au.com/heatsinks.htm
-
Careful usage of "C", “oC”, “Co”, and "K": The measured temperature (air temperature, case temperature, etc.) can be expressed in either "oC" or "K". A temperature rise should be expressed in "Co" or "K", although few are careful enough to maintain the Celsius distinction. That is an advantage of "K" for thermal calculations. Without a degree sign, "C" means Coulomb, the unit of electric charge.