Help me build a PA load-pull

[rfsurfer]

Hi,
It is always challenging to do load-pull on a power FET with big gate periphery like 18mm. Since the input/output impediance is as low as several Ohms, very often quarter of wavelength is used to bring the impedance higher lovel. However, this kind of structure has narrow bandwidth. Does anybody know better ways to do this?
Another issue is oscillation, since little matching exist with the loadpull board, oscillation is every where over smith chart. Any recommendation?

 

[vfone]

loadpull

**broken link removed**

 

[rfsurfer]

load-pull

Vfone, the paper is not quite relevant to load-pull.

Anyone can shed some light on the load-pull board issue I mentioned before?

Thanks

 

[khouly]

pa load pull

did u check about load pull in steve cripps books


khouly

 

[madengr]

steve cripps rule +load pull

There no reason you can't use lumped elements for a pre-matching transformer. You can then tune was required to shift the bandwidth. However each setup requires fixture charachterization. If you assume 1 or 2 load pull measuremnts are good enough for a lot of FETs then you must also assume that several FETs are identical enough that you can build multiple setups of different line lengths.

There are are some tuneable substrate materials but they are probably non-linear are large power levels and may not give the Er range to be worthwhile.

Vendors will sell you "pre-matching" tuners but that's usuually wastefull since the problem is the loss between tuner and DUT.

So.... Other than multiple fixtures there is really no way to get around your problem.

As far as oscillations are concerned, I would not use bias tees that feed through the tuners. The tees need to be part of the fixture, adjacent to the DUT. The gate network needs to provide a low resitance termination at low frequencies. You can get high gate currents. Sometiem the devices are so unstable you must use in-band stability such as series R||C. This network will now be part of the DUT.

 

[rfsurfer]

pa+load+pull

Thanks for the input which is really helpful. We did design multiple fixtures with different length for quarter of wavelength transmission. λ/4 transmission line is also used for dc feeding on our loadpull board, rather than using external bias Tee. It seems to me the λ/4 at dc path might be the trouble maker for oscillation we saw. We see much better stability when doing loadpull on a regular sample board where a lumped inductor followed by a damping resistor is used for gate bias.