Outphasing Modulation

[kb3ouk]

I had some time to kill today so did some reading on outphasing modulation, or as RCA called it, Ampliphase. As I understand it, the way this system works is the RF from an oscillator is split into two phases that are 135 degrees apart (or anything under 180 degrees, as one source said), then one of the phases is phase modulated postive while the other is phase modulated negative (I'm guessing this is done by splitting the audio into two phases that are 180 degrees from each other?). Then the seperate phases are fed into seperate finals (which are of some nonlinear class, like class C), and the outputs are combined together in the output filter/network, forming the AM signal. So as the two signals get closer in phase, the amplitude increases, and as they get further apart, the amplitude decreases. Now, what I began to think of is what if we had a Class D/E transmitter, where the RF deck consisted of a pair of amplifier modules in push-pull. We split our signal from the VFO into two parts, 135 degrees out of phase, then phase modulate the two phases. The positive modulated phase is fed to the one module, and the negative modulated phase is fed to the other. Here's where I got stuck in figuring out how well this would work. First, according to the class E website, the voltage on the MOSFET should be 4 times less then the maximum rating of the device and the current should be 3 times less. Let's use an IRFP460, which is a 500 volt 20 amp device. the absolute maximums that one of these devices could handle would be about 120 volts 6 amps, but to allow for modulation, the actual voltage at carrier would be something more like 60 volts at 3 amps. Assuming we are using 2 devices with 90% efficiency, we are getting only about 325 watts of carrier. Whether that is what these devices will put out for real is a different subject. Now, since the Ampliphase finals are phase modulated and the only thing they are amplifying is a steady carrier that is constantly changing in phase, with the AM signal developing in the output network from the sum of the phases, could the devices used in the amplifier modules be run at higher voltages than in a normal high level modulated RF final? So in the case of the IRFP460s, instead of running at 60 volts 3 amps, run them at the maximum of 120 volts 6 amps? which would give a carrier output of about 1300 watts for the 2 devices together. Second, the only part I could not figure out on my own was how are the sidebands being developed in an Ampliphase system? Back to our IRFP460 amplifier, the 1300 watts is the maximum that the transmitter will put out. But is that what the resting carrier would be when measured at the output network, after the two 135 degree out of phase signals are summed up? The RCA 50H used a pair of 6697s in the final, one per channel, each which put out over 25 kw, which added up in the ouput network to be over 50 kw. so the IRFP460s, which would be putting out about 650 watts each, should actually be capable of 1300 watts of carrier when combined together in the output. Now, can someone tell me if I am on track with this, or am I way off?
Shelby

[steve_qix]

Hi Shelby,

Outphasing modulation is very interesting.  I built and operated an outphasing transmitter back in 1973, and in fact won 2nd place in the state science fair and got a scholarship to college (without this, I would not have been able to go to college) all because of that transmitter, so it is near and dear to my heart 

RCA ampliphase is not strictly outphasing modulation, but is a variation of it and uses some of the principals of the system.

The system I used employed 2 PI sections (one for each RF amplifier) and a common loading capacitor.  The carrier phase difference was 135 degrees.  So, at 180 degrees, the outputs would oppose and cancel.  At 90 degrees, the output voltage would be double what it was at carrier.

This all required careful adjustment because everything affects the phase.

There are other systems that use a balanced output (similar to push pull), where the RF amplifiers are 45 degrees out of phase with each  other at carrier.  When they are in phase, there is no output.  At 90 degrees, the output is double.  If you use "rectangular pulses" (AKA square waves), you essentially get "RF PWM".  In fact, one of the broadcast equipment makers had a patent on this way back.

The sine function is fairly linear from 0 to 45 degrees which is why this is the general range of phase swing.  One could, in theory, use the entire swing of the sine function from 0 to 90 degrees if the audio were "pre-distorted" to compensate for using the nonlinear portion of the sine function.

Anyway, probably more than you wanted to know 

I did want to make one correction.  The class E web site actually says the MOSFET rating should be more like 20 times the carrier DC voltage when you figure it out.  Here's the quote from the class E web site:

The peak voltage across the MOSFETs is going to be a little less than 4 x the DC applied voltage for a proper class E transmitter. This will vary somewhat with tuning and your exact circuit. If you have very low, or NO shunt capacitor, the ratio of peak RF voltage to applied DC can be 6x or 8x the DC or MORE. For class e transmitters with proper shunt capacitors, figure 4x the DC, plus a safety factor. If you're running 40V @ 5 amps of carrier, and expect to modulate 150% positive, your power supply voltage is going to be 100 volts. So, your MOSFETs are going to see 400V at a minimum. Use AT LEAST a 600 volt MOSFET, and I would personally use an 800V or 1000V MOSFET in this application.

Thought I'd mention that, because using under-rated MOSFETs in RF amplifiers will result in a very unreliable transmitter 

Regards,

Steve

[kb3ouk]

ok, well actually one other thing I had in mind would be if it would work with square wave drive or not. Actually, class D might be a little better for an outphasing rig since there would be nothing variable about the output network, which should keep everything in phase. How does the RF PWM work? I'm guessing the PWM frequency is at the RF frequency, then the PWM is fed to the RF deck and then filtered? So the RF deck would, in a way, be the modulator, too? Something I might as well bring up about MOSFET ratings, I recall seeing how one modern Class D BC transmitter (it was a 50 kw Nautel) used 20 modules with 4 devices per module, in an H configuration, where two of the devices would be in series with 400 volts on them, but each device received 200 volts. The modules put out about 2.6 kw a piece, which translates to around 700 watts per device, or 3.5 to 4 amps. they were only rated for 500 volts 41 amps. So are the ratings for Class D different from Class E?

[steve_qix]

You can do RF PWM with a very high speed PWM controller, or you can do it with 2 phase modulated RF amplifiers producing a square wave output, and putting these across a balanced circuit (transformer primary). 

A pulse will be produced in the output only when the RF amplifiers across the transformer primary are out of phase.  The more out of phase they are, the longer will be the duration of the output pulse until you reach 180 degrees out of phase and maximum output.

Class D works well here in this application, and in fact that was what the broadcast manufacturer (don't remember the company - probably Continental) used in their patent.

[kb3ouk]

Sounds like the phase modulated way would work, so if I used two MOSFETs, and had one one one side of the primary and the other on the other side, and then that was feeding the output filter, that would produce the RF PWM? Now to get the squarewave, would the devices have to be driven squarewave, and therefore the driving squarewave be pahse modulated (is this even possible, phase modulating a squarewave?), or would you have to use sinewave drive and phase modulate that, then drive the devices to the point where the output is a squarewave? And for MOSFETs in Class D service, do they follow the same voltage and current rating rules as they would for class E? Since the amplifier would be passing a carrier only, you wouldn't have to allow for the headroom like you would with a transmitter that the final is being modulated by some kind of modulator that would produce the AM in the finals, would you? I'm beginning to like the idea of the RF PWM, might try it someday, just have to figure out how to do it.

[John K5PRO]

Continental did build an RF PWM transmitter, the XL-301 1 Kw model on 540-1610 KHz, in 1988. It was touted as excellent for AM stereo due to low incidental  phase noise. The transmitter didn't have exceptional efficiency, but ONLY weighed 560 lbs. It was taken off the market, for various reasons, and I am not sure
if any were sold to stations. They replaced it with a couple of different transmitters that they bought from others, including the Omnitronix PWM transmitters. I am quite familiar with those, having designed the modulator stage as a contract job. Continental eventually gave up on lower power AM rigs, and now even high power AM 50 kW rigs. They continue to make higher power shortwave transmitters though. Omnitronix disappeared, sold their line to LPB, also in Pennsylvania, and they too closed their doors.

In the past 5 years Broadcast Electronics introduced their 4MX series of high power transmitters (they were also first b'dcast co to market class E AM transmitters over a decade earlier). It doesn't say explicitly RF PWM anymore, but a look at their brochure shows essentially a synthesis of similar signals in the RF amplifier blocks.

A New Zealand company was planning to build an Outphasing solid state AM about a decade ago, and abandoned it for a more conventional class S PWM modulated class D or E amplifier. I cannot even remember their name. However, the engineer who was going to sell them in Australia is coming by to visit me in 2 weeks, on his way to NAB, so I will try to ask him more about what happened to that.

The 5 part series in High Frequency Electronics magazine is a good general summary of the various types and classes of RF amplifiers, including the last installment devoted to RF PWM, probably archived on their website. It was in the Jan. 2004 issue. The authors are the heavy's on high efficiency modern amplifiers, Cripps, Raab, Sokol and more. Check that out.

[steve_qix]

Continental did build an RF  . . . [snip]

That is a very interesting and informative post!  Thanks!!

[W7TFO]

I serviced several RCA Ampliphase boxes over the years, from 10 to 50 kW.  Most of the station engineers had one term for them:  Amplimush.

They were really tricky to keep sounding good, the end of that type tube era was a good thing.

Unlike the WECo / Continental Doherty jobs that always sounded good and were not a lot of hassle.

Obviously there is a gap 'tween design and technology of the day.  A phasing rig ought to be easier to build today, just like the freq counter used to be unobtanium and is now a $20 hamfest item.

73DG

[kb3ouk]

I believe that probably the whole reason the Ampliphase system didn't work out so well was partly because it was a tube design. Building the same thing using modern solid state devices would probably be a lot more stable.

[The Slab Bacon]

Amplifuzz................  

[flintstone mop]

Thanks Shelby and Steve.
A lot of grey matter moving in my brain cells. I like the plain and simple approach to modulate a TX. PWM is super, but there are some things that require more care in the building and set up.
And Steve, congrats on your achievements.

[kb3ouk]

Ok, so I got to looking at a tube outphasing transmitter http://www.rossrevenge.co.uk/tx/koh5.jpg, and noticed that it used a tripler after the phase modulator to get the carrier to the desired frequency. Is any kind of multiplier neccessary, or could you simply have a vfo running on the frequency you want to transmit on, and phase modulate that directly , or do you have to phase modulate at a lower frequency, then multiply it up to where you want to transmit? I think as long as you can get a -45 to +45 degree swing on either side of 135 degrees, then it should work.

[John K5PRO]

I had a simple phase modulator at work a decade ago, using a pair of mixers and quadrature RF LO. It used an Analog Devices trig generator IC, to create a sin and cos of the control voltage in real time. It was essentially a SSB modulator approach. I saw the design in an old RF Design magazine. The thing had a 360 deg linear phase control, with reasonable control voltage bandwidth, at 2800 KHz RF. I was thinking at the time of an outphasing transmitter experiment, but never built one. The trig IC was discontinued from AD, something like an AD639 rings a bell. It was a really neat analog chip, but better all done digitally these days.