Why in the power stage,many designs use two NMOSs'instead of one PMOS +one NMOS ? what is the advantage of using 2 NMOSs rather than 1 PMOS +1NMOS?
N channel mosfets are an inherently/intrinsically more efficient device. They use electron flow instead of hole flow (faster, less resistance) lesser parasitics etc.
Aside from those bonuses and largely because of them, it's far easier to create nice symetrical switching with two N channels. It's a little more complex to drive though, but well worth it.
Aside from those bonuses and largely because of them, it's far easier to create nice symetrical switching with two N channels. It's a little more complex to drive though, but well worth it.
NMOS vs PMOS
NMOS has better R on
NMOS has better switching characteristics
NMOS is cheaper
That's why NMOS very often is preferred in switching power designs
NMOS has better R on
NMOS has better switching characteristics
NMOS is cheaper
That's why NMOS very often is preferred in switching power designs
I believe the other and more main issue is that making PMOS in Si takes more space and is harder than making NMOS. Many companies do not have a good complementary process so the NMOS might be really good but the PMOS transistors are crappy.
NMOS vs PMOS
You are right and that's why the practical consequence is that Ron is lower of NMOS, they're faster and cheaper
You are right and that's why the practical consequence is that Ron is lower of NMOS, they're faster and cheaper
Hi,
Let's for a moment assume what you've said about the manufacturing process not being as good for P channel as N channel is true.
Let's now assume for a moment we're living in an ideal world and their processes are fully optimized for manufacturing either device, so optimized in fact that the devices could not be made any better by any other means.
In that world, P channel devices will still be less efficient because they use hole flow as the majority charge carriers which have less mobility than electron flow as majority charge carriers like N channel devices use, and therefore you end up with a higher on resistance (for a given area) and so less efficiency.
You can easily get around that by using a bigger P channel, with larger area, that way you can attempt to match R on with the N channel in use, but then you still have a higher threshold, higher parasitic capacitances... and this results in distortion causing asymetrical switching, which, is the real concern.
However, as I'm sure JohnW would tell you, they can still be used for lower power applications with good results. I believe he uses them for every circuit up to 120 watts and gets seriously respectable THD results in open loop! EDIT: (Well he uses them for up to 120W in his "cheaper" circuits anyway, I shouldnt' have said "every circuit").
Typically I'd imagine what you'd save by using P channels (bootstrap circuit/auxiliary supply, start up circuitry ==more complexity in the drivers and house keeping supplies==extra board space) will easily make up the difference in cost for a P channel that's a mm or two larger.
Regards,
Chris
Let's for a moment assume what you've said about the manufacturing process not being as good for P channel as N channel is true.
Let's now assume for a moment we're living in an ideal world and their processes are fully optimized for manufacturing either device, so optimized in fact that the devices could not be made any better by any other means.
In that world, P channel devices will still be less efficient because they use hole flow as the majority charge carriers which have less mobility than electron flow as majority charge carriers like N channel devices use, and therefore you end up with a higher on resistance (for a given area) and so less efficiency.
You can easily get around that by using a bigger P channel, with larger area, that way you can attempt to match R on with the N channel in use, but then you still have a higher threshold, higher parasitic capacitances... and this results in distortion causing asymetrical switching, which, is the real concern.
However, as I'm sure JohnW would tell you, they can still be used for lower power applications with good results. I believe he uses them for every circuit up to 120 watts and gets seriously respectable THD results in open loop! EDIT: (Well he uses them for up to 120W in his "cheaper" circuits anyway, I shouldnt' have said "every circuit").
Typically I'd imagine what you'd save by using P channels (bootstrap circuit/auxiliary supply, start up circuitry ==more complexity in the drivers and house keeping supplies==extra board space) will easily make up the difference in cost for a P channel that's a mm or two larger.
Regards,
Chris
You can also parallel up different numbers of FETs. For example, you could have 1 NMOS and 2 parallel PMOS, to lower the Rdson of the PMOS. Of course, this increases the gate capacitance, so it's a trade-off.
As classd4sure points out, P ch MOSFET is inherently 'worse' than N ch. If I remember my microelectronics right, all other things being equal, you get a bit less than 2x the Rdson of N ch, and also a bit less than half the transconductance, as well as different threshold voltages.
In class D where efficiency is king, this makes a whole lot of a difference. However, sometime raw Rdson figures can be sacrificed if you get better switching characteristics as shoot-through and dead time also affect efficiency adversely, along with creating other undesirable effects.
The key to avoiding these problems, much like in class AB amplifiers, is to have symetrical Gm, threshold and capacitances.
While I have seem may attempts to match Rdson with 2 parallel P-ch MOSFETs, at the expence of greatly increased capacitance (and no help with regards to threshold voltages), it still puzzles me that no-one seems to use the fact that higher voltage N-ch parts can be VERY good matches for lower voltage P-ch parts.
Rdson is sacrificed, because you use the Pch part as a reference, but it is actually possible to find complements that differ from each other on the order of differences between two batches of the same type part.
The trick to it is knowing that MOSFETs are actually arrays. Manufacturers create larger current parts by simply enlarging the die proportionally, i.e. enlarging the array. For a 2x higher voltage N-ch part the effective channel length has to be longer due to the ability to withstand a higher voltage. The oxide layer also tends to be thicker for the same reason. The net result is a MOSFET with similar capacitances but twice the voltage rating and half the current rating. P-ch parts also obay the same rule, just starting at lower voltages. This is actually simple to see from the datasheets, especially if they include die sizes (only from manufacturers that offer a KGD a.k.a. Known Good Die product).
In class D where efficiency is king, this makes a whole lot of a difference. However, sometime raw Rdson figures can be sacrificed if you get better switching characteristics as shoot-through and dead time also affect efficiency adversely, along with creating other undesirable effects.
The key to avoiding these problems, much like in class AB amplifiers, is to have symetrical Gm, threshold and capacitances.
While I have seem may attempts to match Rdson with 2 parallel P-ch MOSFETs, at the expence of greatly increased capacitance (and no help with regards to threshold voltages), it still puzzles me that no-one seems to use the fact that higher voltage N-ch parts can be VERY good matches for lower voltage P-ch parts.
Rdson is sacrificed, because you use the Pch part as a reference, but it is actually possible to find complements that differ from each other on the order of differences between two batches of the same type part.
The trick to it is knowing that MOSFETs are actually arrays. Manufacturers create larger current parts by simply enlarging the die proportionally, i.e. enlarging the array. For a 2x higher voltage N-ch part the effective channel length has to be longer due to the ability to withstand a higher voltage. The oxide layer also tends to be thicker for the same reason. The net result is a MOSFET with similar capacitances but twice the voltage rating and half the current rating. P-ch parts also obay the same rule, just starting at lower voltages. This is actually simple to see from the datasheets, especially if they include die sizes (only from manufacturers that offer a KGD a.k.a. Known Good Die product).
Thanks - I've learned that lesson the hard way with class AB amps. If I had a $ for every time I heared that IRFP240 + IRFP9240 sounded bad compared to the ubiquitous 2SJ/2SK pairs, I would be rich. I mean, even in a simulator, comparing Gm will reveal a 1:2 difference between the two IRFP parts, yet people still use them that way... which does not mean that you cannot find pretty good complements out of the IR catalog 😉 as well as other manufacturers.
- Status
- Not open for further replies.