IMD problem with a RF amplifier and very linear supposed trt

EDAboard.com Forum Index -> RF, Microwave, Antennas and Optics -> IMD problem with a RF amplifier and very linear supposed trt

Author

Message

byteptr

Joined: 29 Apr 2008
Posts: 112
Helped: 11
Location: Europe

02 Oct 2008 14:49 mrf6v2300

Hello to all,

Im trying to desing a VHF amplifier for ose OFDM modulation, before apliyying predistortion techniques I have to improve the linearity the best possible. I was tried these 2 transistor from freescale (with well knowledge of efficiency vs linearity compromise): MRF6V2300, and Push-Pull 6VP2600H using 50V and differents IdQ.

I know how to well mach input/output and tunning the polarization gives me differents spectral regrowths, but the fact is that I can't achieve the specified performances in the datasheet for a 6Vp2600H nor MRF6V2300.

All this it gives me very poor linearity/efficiency performance compromise. I know for example that there are some amplifiers with pair of 6VP2600H that achieves 200Watts rms with 25% eff and shoulders < -30dbc, with this specs I can only achieve ~ 20-25 dBc and eff not better than 20%...

My desing process starts with an arbitrary point of polarization (one of the specified in the datasheets, frequently that gives the better linearity performance), then I try to optimize Input network, later to optimize the output matching network... in this process some re-tunning of input matching network may be needed... (Im not using design software, I have VNA, spectrum analyzer )

When I achieve the better efficiency performance, I try to optimize the linearity...

So my questions are:

What are the steps did you make for achieve the desired performances?
Did I something wrong? what?

Thanks In advance

Google
AdSense
Google Adsense

02 Oct 2008 14:49 Ads

biff44

Joined: 24 Dec 2004
Posts: 1835
Helped: 244
Location: New England, USA

02 Oct 2008 18:45 reverse intermodulation

Impedance matching, especially the load impedance, is a big effect.

Harmonic matching also helps.

You need a very stiff power supply--meaning lots and lots of very low esr capaciators near the devices it you want low frequency two-tones (like those 1 mhz apart) to be good.

Also, you really have to be carefull that the spectrum analyzer itself is not overloaded and causing the intermods! You need a lot of attenuation between the amp and the analyzer.

vfone

Joined: 10 Oct 2001
Posts: 2327
Helped: 326

02 Oct 2008 21:18 reverse imd

Poor efficiency and linearity in the same time could be that you have a problem finding the right output match. A load-pull system would help, or at least using a manual tuner.

Also check the gain of the PA over the RF bandwidth. Fast roll-off of gain at the edges of the band causes deterioration in the ACPR performance.

Another tip to keep good linearity, the 3dB bandwidth of the bias networks filtering needs to be at least two times greater than the modulation bandwidth.

kspalla

Joined: 19 Aug 2008
Posts: 848
Helped: 140

03 Oct 2008 8:12 imd amplifier

Linear amplifiers are bilateral in nature. so if you tune the input side it has the effect on output side too.
Have a design software to control over it.

byteptr

Joined: 29 Apr 2008
Posts: 112
Helped: 11
Location: Europe

03 Oct 2008 10:02 imd rf

Thanks to all for the responses.

Yes in effect, my hypotheses (I don't make theoretical calculations nor simulations yet) were in these directions: first about harmonic matching, and second bilaterality of the amplifier because I designed a very linear driver (< -50 dBc of intermod), then I watch power efficiency for estimate the matching and IMD between the driver and power stage.

Having a good matching (good matching means good efficiency/gain/gain flatness compared with theoretical) between stages and good matching in the power stage for the band of interest to 50 Ohm resistive load, I observed a rise of shoulders phenomena in the driver, this raise phenomena is mainly in the push-pull stage rather than single transistor stage, that may be because push-pull is "more bilateral" and this produces the called "reverse intermodulation products" effect.

I'm thinking to try the measurement with circulators but, I haven't expensive VHF circulators, and self-made circulator implies looking for anisotropic ferrites and waste more time..., so I will try the following thing:

In the final design two power stages in parallel wil be connected via 3dB quadrature Hybrid, is well known that one property of these hybrids is "two loads reflection coefficient masking effect" if the loads are the same. I think that the bilaterality (if the amplifier is stable), can be seen as load charge modulation by the driver, so, reflection coeffient modulation and therefore reverse IMD produtcs appears. If the two stages are equal, and they are connected by quadrature hybris, may be global linearity improvement...

What do you think about this?

I didn't speak yet about matching harmonics, but I belive this is also a tricky task, and they must be dissipated in an alternative load.

Added after 12 minutes:

P.S: Sorry for forget the other suggestions, about the matching is broad band. The coupling of power was made with lots of low ESR capacitors and good RF capacitors, (the test was also performed in a freescale evaluation board). The power source is a good laboratory source capable of handling 60V-50Amp. I adds extra ferrite beads to DC power, polarization and coaxial-RF cables and made EMC tests for induced and radiated.

kspalla

Joined: 19 Aug 2008
Posts: 848
Helped: 140

03 Oct 2008 12:33 6vp2600h

When I achieve the better efficiency performance, I try to optimize the linearity...
This is wrong approach. it should be reverse.
How is calculation for efficiency? is it pulsed power supply?Are you following the same substrate, art work of manufacturer etc?
The amplifier need to operate below 300W for -40dBc IMD. this giving maximum efficiency of 25% only:cry:
In my view Harmonic matching is not required as their levels are too low.

byteptr

Joined: 29 Apr 2008
Posts: 112
Helped: 11
Location: Europe

03 Oct 2008 13:04 rf amplifier imd

Hello
"This is wrong approach. it should be reverse."

Why?

(Although I will try this)

How is calculation for efficiency? is it pulsed power supply?

Are you following the same substrate, art work of manufacturer etc?
Yes.

"The amplifier need to operate below 300W for -40dBc IMD. this giving maximum efficiency of 25% only:cry:"

IMD with two tone test 100khz spacing, but I use OFDM 1.5 MHz Bw.

"In my view Harmonic matching is not required as their levels are too low"

May be, but... what about "inverse" IMD?

Added after 56 seconds:

PS: I don't take into account the power efficiency of supply, only the amplifier.

kspalla

Joined: 19 Aug 2008
Posts: 848
Helped: 140

04 Oct 2008 9:01 emc amplifier imd3

Quote:

"In my view Harmonic matching is not required as their levels are too low"

May be, but... what about "inverse" IMD?

As IMD levels are too low (-35 to 40dBc) the inverse IMD will be negligible. Hence I suggested for linearity firth then efficiency.

fjstuden

Joined: 06 Dec 2008
Posts: 1
Helped: 1

06 Dec 2008 4:20 imd load pull

could be many problems with your design. For a given drain load and drain voltage, the efficiency will be max only at max power as set by the drain load impedance and drop as RF power output drops. Therefore with complex waveforms, especially those with high peak to average ratio, efficiency must measure low

Nothing you can do except make drain voltage track the envelope of the modulation, but with widebandwith signals, its not easy task.

If push pull, is the ac bypass cap on the drain feed transformer extremely short run back to the sources? The RF current is circulating from drain feed point, through the drain feed transformers and the FETs and out the sources right back to the rf bypass on the feed choke. Don't believe the stories about"virtual grounds" etc and bypasses not being important in push pull. Sketch out where rf is flowing and keep all that extremely low impedance (inducive and capacitive).

I alway use 2 tone signals to measure and characterize IMD, its much faster, and repeatable than using real waveform.

Another easy idea- get one of the inexpensive laser pointing Infared temp sensors and point at all ckt components to see whats hot. Hot ones are where your missing power is going.