RF power amplifier, Class AB

Hi all,

I am not an RF engineer though I am starting to look at Envelope Trackers (ET)for RF Power Amplifiers as analog designer. I am wondering why class AB is not used in handsets RF PA, especially in multi-band LTE and 5G. Class E apparently still dominate. Class AB has higher efficiency than class A, can get even closer to efficiency of class B, and has the advantage to be fully differential, which means that supply ripple due to ET (DC DC converter with switching activity) is common mode and then rejected. It also exhibits superior linearity. Could anyone explain why class AB still can not make it to modern handsets?
Thanks

This is ASIC design methodology and Tools Thread. you posted this in wrong thread. go back and post it in analog or rf threads.

go back and post it in analog or rf threads.

Click to expand...

No. Preferably think first and then post in the right forum. In case of accident ask a moderator to move the thread.

Because Power Efficiency has the first priority in handset PA design, thefore Class-E is preferred due to superior efficiency over other linear classes.Especially where the Constant Envelope Modulation is used, Class-E is a good choice because when its linearity is optimized once, it will continue as long as desired.( no amplitude changing fluctuation)

Before 4G/LTE almost always Class-AB PAs were used in mobile phones (for 2G and 3G).

Class-AB has an issue named Gain Expansion, when the gain of the PA starts to increase few dB's before the P1dB, and decrease after the compression point.
For Envelope Tracking constant PA gain is mandatory from low input power levels till compression of the PA.

Also there is no reason class E amplifiers cannot be differential, though what you will end up with is usually closer to a class D amplifier. I don't see why making the amplifier differential would prevent supply ripple from modulating the output, though.

Thank you very much for your replies, and sorry as a newby I did not comply with thread rules. I understand in constant envelopes, class E is a better choice for efficiency though linearity is somehow sacrificed (compared to Class A and closely to B). Why class E is always used as single ended as mtwieg stated. Differential mode would lead to 20-30% gain in efficiency with only doubled complexity for active devices and same passive network. The only problem I see is that a transformer needs to convert the differential signal into single ended to feed the antenna that is referenced to ground. And probably rf transformers are still archaic.
It is known to analog designers that differential signaling is immune to any common mode disturbances (anything affecting equally sigplus and sigminus, is not seen when dealing with sig=sigplus-sigminus). This should apply also to differential RF and supply disturbance due to ET is common mode and rejected. We all know that ET easily doubles talk time (data transfer) in mobile handsets for the significant power it saves compared to constant supply. I was referring to class AB because I think that linearity is very important in 5G-LTE (both phase and amplitude modulation) and then Class E might not be suitable. Anyway as an ET designer my task with be much simpler and the power savings much bigger if the RF PA uses differential topology. Let alone the extra cost, the efficiency improvement in pre-ET design is also significant

- - - Updated - - -

sorry mtwieg, I now got your point. Conduction angle in class E is less than 180degrees. you are right, differential would not prevent supply disturbance from affecting the signal. But then, does not the LC network reshapes the choped signal to a 360 degree sinewave for which differential signaling would be effective? In my first post, I was thinking to class AB which is somehow a differential class-A with theoretical efficiency of 66%, single ended class A being 50% efficient

I think you're confused about how ripple on the drain bias impacts the output spectrum.

Consider a single-ended broadband RFPA amplifying a 100MHz carrier. If you put a 1MHz ripple on its drain bias, two things will happen to the output. One is that the the 1MHz ripple will be coupled directly to the output, so you'll get a 1MHz peak in your output spectrum alongside the original 100MHz carrier. The second thing is the 1MHz ripple will modulate the gain of the amplifier, causing intermodulation with the 100MHz carrier, and you will see spurs at 99MHz and 101MHz as well. The intermodulation products are much more problematic than the 1MHz peak, since they are difficult to remove with filtering.

Now consider a similar RFPA, but differential. The direct 1MHz peak will be eliminated from the output due to the balun. However, the intermodulation products will not go away (some simple math will reveal why). In practice, very linear amplifiers like A, AB, and B are fairly robust against intermodulation, but nonlinear classes like C, D, E, etc, are very susceptible to it, since their gain is directly proportional to the drain bias. But intermodulation is not corrected by differential amplifiers.

mtwieg said:

In practice, very linear amplifiers like A, AB, and B are fairly robust against intermodulation, but nonlinear classes like C, D, E, etc, are very susceptible to it, since their gain is directly proportional to the drain bias. But intermodulation is not corrected by differential amplifiers.

Click to expand...

Fair enough. Thank you. Then the question why not go back to class AB, which is very linear, lose probably 10% efficiency in constant envelope, but then gain back more than 90% (double talk time) with high speed high performance ET, the price to pay for such high performance ET being the ripple

Mister_hass said:

Fair enough. Thank you. Then the question why not go back to class AB, which is very linear, lose probably 10% efficiency in constant envelope

Click to expand...

Best case efficiency with class A/B is maybe 60%, class E should be at least 80%, which is a factor of two in dissipation.

The efficiency of a power amplifier can be maximized if the active device is
operated as a switch. When the transistor is turned on, the voltage is nearly zero
and high current is flowing through the device; that is, the transistor acts as a low
resistance (closed switch) during this part of a period. When the transistor is
turned off, the current is zero and there is high voltage across the device, i.e. the
transistor acts as an open switch during the other part of a period.