What's the peak-average-ratio of RF switch controlled PA?

I have a PA works well if input is continuous wave, and the output power of the PA is 5dB back off from the P1dB.
But if I put a RF switch before the PA, that switch with a duty-cycle of 0.5% and on/off frequency of 1KHz, the PA is very easy to die.
Could anyone give any suggestion about that? Is this the peak-average-ratio problem?

Do you keep the input power the same with the switch as you did before? If so, then I do not see why the RF PA can be overdriven.
If you increase the power to obtain the average power with a varying duty cycle, semiconductor amplifiers sometimes tend to overheat.
In VHF and UHF transistors, internal leads are made of thin gold wires; for short duty cycles and high average power, some transistors oscillate at unexpected frequencies and certain gold links melt. Electrical fast response is short but thermal response is much slower. Overheating occurs locally, killing the device.

When the RF switch is off, does it change the impedance seen by the PA input? If so it may be causing instability during switch off time.

Will the thing also fail if you operate the switch, but with no source signal (just a 50ohm or whatever terminator)?

The reason of the PA failure is not the switch itself, is the bursty operation of the Power Amplifier, which results from switching its input power.
In this mode there is possible to appear overshoots in the output power (and collector/drain current) which might destroy some PA's if they were not biased or impedance matched for working in this mode.

Jiripolivkao you keep the input power the same with the switch as you did before?

Click to expand...

Yes, the power is the same in both case.

mtwieg: When the RF switch is off, does it change the impedance seen by the PA input?

Click to expand...

No, the impedance is also 50R. There is a buffer amp between the switch and PA, and the buffer amp works well, only the PA is die.

vfone: ......

Click to expand...

The PA is biased, and maybe the output impedance is not good. I will test it.
Thank all of you.

First what you are doing would generate a very wide spectrum.

You did not say what kind of PA it is. Is it intended for GSM or is it a linear PA. Most RF PA's have at least some biasing modification based on RF drive level. Linear PA have high DC bias idle (class AB). The bias control and its filtering will react differently to RF pulse input. You are likely driving the bias control crazy with on-off input keying. What you are doing creates a very high peak to average ratio. Most WCDMA/CDMA PA's can stand up to 10-15 db of peak to average without severe distortion. PA's intended for LTE can stand more.

A GaAs switch is open circuit when turned off so it provides a near open circuit impedance to disengaged port, only port capacitance. When closed it provides a low resistance path to conducted port so it takes on output port impedance, with some modification due to internal parasites of the switch, at the input port. So you are not terminating the input to the PA when switch is off.

The PA is Hittite HMC608LC4E, it is a linear PA with P1dB +27dBm. If the input is continuous wave, the output power is +22dBm. SO I think the peak power is +22dBm when switch is ON. The switch ON/OFF power ratio is more than 80dBc.
The configuration is as follwoing: Switch(IF) + Mixer + Filter + Driver Amp + Power Amp(HMC608LC4).

Who can give any suggestions?

I am not familiar with your half-watt amplifier you want to modulate with various duty-cycle RF input. Many such amplifiers have a gain-control or on/off pin. I would try to test this way to key the RF amplifier.
Various external components could introduce either a non-healthy mismatch when switching, or, the blocking capacitors around the power amplifier must be connected so that the amplifier response to keying pulses is correct.
I would try another switch at input, maybe a "matched" version that stays matched to 50 Ohms in both on and off states.
Reduce the RF input and carefully monitor all currents by an oscilloscope as well as the RF output response. By gradually increasing the RF output while following the possible glitch you should find out if there is a dangerous region killing your amplifier.
Also, test your RF output load; it may heat up and get mismatched, and then the RF amplifier has to dissipate more power than it can survive. Use a reflectometer to see if the load match changes.

---------- Post added at 22:40 ---------- Previous post was at 22:37 ----------

As you mentioned you put the switch between the buffer and final stage, this may force one of the amplifiers into oscillation... that kills the final stage.
Try to put the switch BEFORE the buffer amplifier, or, key its DC bias.

tony_lth said:

Who can give any suggestions?

Click to expand...

At this point you should probably give a schematic. If the output side of the switch is still 50ohm in both the on and off state, then there shouldn't be stability or reflection issues... another guess is that during the switching events, the transient from the switch may be disturbing the bias elsewhere in the circuit, causing damage. I asked before, does the amp still fail if you operate the switch, but have no signal going into the switch (just a 50ohm terminator)?

I would suspect some sort of oscillation is the problem. Try a ferrite isolator on the power amp input and output and see if that solves the problem. If that works, you will need to play around with the impedances so that it becomes stable when no rf power is applied.

What is Vgg doing during switching? Have you probed to make sure it is stable?

I had obeserved the Vgg change during the ON/OFF state. And I think the DC-decoupling is not so good.
The PA had worked well for 3 months with Agilent DC power supply, but after used our power supply, it failed more than 2 times. So I think it's the power de-coupling problem. I will test according to your suggestions, thank you very much.