I have been very fortunate to have worked with some very brilliant, pioneers of the broadcast industry.  These include Bob Orban, the late Ron Jones, Gregg Oganowski, all audio processing experts that, along with Mike Dorrough, whom I also had the pleasure of meeting in 1984 and at several NAB conventions since, lead the way to modern audio processing, the late Ed Butterbaugh of CKLW - The BIG 8 - fame, the late Will Koeller of WTMJ who mentored me in broadcasting while I was in college, teaching me most of what I know about directional array design and operation, Arno Meyer of Belar, Hilmar Swanson of Harris Broadcast; designer of multiple high efficency AM broadcasting techniques, Norman Parker of Motorola and NTSC pioneer / creator of the "Parker Angles" that allowed accurate color reproduction in NTSC as well as co-inventor, along with Frank Hilbert, of the C-QUAM AM Stereo system, James "Jimmy" Weldon, founder of Continental Electronics and the Weldon-modified grounded-grid Doherty linear amplifier, and many more.  I will dedicate a separate page to many of these great men of engineering, but wish to single out herein one individual with whom I had the true pleasure of working with:  Joe Sainton of Continental Electronics.  Text continues below.........
        Joe was a very interesting individual.  When he was in his late teens, he wanted to fight in WW2 but the US was not yet officially involved in the war.  He found his way to Canada where he volunteered to fly for the Royal Canadian Air Force.  Being an electrical engineer interested in broadcasting, he landed a job with Continental after the war.  Continental Electronics was founded by Jimmy Weldon after he acquired the Western Electric Division of Bell Telephone after a court-ordered divestiture of that firm.  Weldon had worked with Doherty at WE and was an expert on the Doherty high efficiency linear amplifier.  Back in the 1930's he designed BIG transmitters for the "border blasters" in Mexico.  Weldon had improved the original Doherty / Terman design by operating one of the tubes grounded-grid.  This eliminated the need for neutralization as well as improving efficiency since the drive power, which was considerable, is passed on to the output.   
As broadcast technology matured, efficient designs such as the RCA Ampliphase implementation of the Chirex outphasing method (two independent phase-modulated transmitters are combined to produce envelope modulation) and the design of lighter, more efficient plate modulation transformers and tetrodes capable of generating the 25+kW needed to modulate a 50kW transmitter encroached upon the aging Weldon linear amplifier design.  Joe realized that the key to Continental Electronics survival and future was to improve upon the previous design.  He realized that the grid-modulated Doherty had reached it's sell-by shelf date.  Eimac had developed efficient power tetrodes, such as the 4CX35000, that were already pressed into AM broadcasting service.  While control grid modulation, as utilized in the Doherty / Terman / Weldon designs, results in a non-linear modulation characteristic unless the stage is operated Class-A and with substantial amounts of negative feedback, the screen (g2) of a tetrode can be utilized to produce highly linear envelope modulation of a Class-C RF amplifier stage.  Essentially, the modulated screen of a tetrode amplifier is analogous to a cascode stage; dual gate MOSFET low-noise amplifier stages of receiver front-ends operate identical to a screen-modulated amplifier.  
Before I describe the Sainton screen modulation method, I would like to talk about the mentoring I received from Joe.  I met Joe in 1983 when I began to work fulltime for Motorola Corporate Research in the Modulation Systems Laboratory mainly on AM stereo (C-QUAM).  I first met Joe in Hamilton, Ontario, Canada at  CKOC radio (where I also met Gus Sondermeyer; broadcast engineering consultant, John McCloy when he was chief engineer of CKOC, Bob Berger; chief engineer for CHML, and Dave Van Allen, audio processing consultant who was with Processing Plus of Allentown, PA at the time - they made a full matrix multi-band audio processor for AM / AM Stereo called the IMP).  We all bivouacked at a small motel "Gulliver's Travels" where we ate more pizza and drank more beer than any human should ever attempt.  After a few hours of drinking, the pizza crust was sacrificed for its topping; the crust was flushed (unsuccessfully) down the toilet which did not make the motel staff all to happy.  Fortunately, it was not my room that was sacrificed to the porcelain clogging gods! 
Joe was incredible - he could tweak a 50kW transmitter to produce a remarkable <1% envelope distortion and <3% 4:1 SMPTE IM performance while achieving 75% plate efficiency and very low incidental phase modulation.  It was no wonder that his third variant of the 50kW model 317C2 transmitter was known as Joe's modulating monster!  He took the time to explain every detail of the transmitter to me and actually let me do the adjustment and analysis of the transmitter - real on-the-job training and education.  We got along well and Joe asked me to join him in the Continental Electronics factory (with professional broadcast consulting engineer and equally incredible teacher Jack Sellmeyer of McKinney TX) to participate in the adjustment of a new 317C-2 50kW transmitter and stereo exciter as well as customer acceptance of that transmitter that was to be delivered to Canada.  Next, Joe asked me to join him to do conversions of CKCK and CKRC in Winnipeg and Regina, Canada.  These were 10kW classic Doherty transmitters that Joe had re-designed and coaxed perhaps the best distortion performance out of a tube transmitter that has ever been witnessed; it was common to get <0.3% envelope distortion from these transmitters at 95% modulation.  Stereo performance was equally stellar:  I saw one transmitter in Moose Jaw produce in excess of 50dB separation at 0.2% distortion into the antenna.  It was on the Regina trip that Joe and I searched for beer after finishing our work on a Sunday night.  Unable to find anyone that would sell package goods on a Sunday night, we finally succeeded in talking a local bartender into filling a couple of 1 liter Styrofoam soup containers with LaBatt's Blue for us which we took back to the hotel to consume.  It was that night that Joe told me of his long history with Canadian broadcasters, his fighting for Canada in the defense of England during WW2, and the plane crash he survived in Cairo, Egypt while there on business in the late 1950's.  He jumped from the burning plane, breaking his leg in the process.  The Egyptian police found him crawling from the burning wreckage and promptly through him in jail, without medical attention, for a couple of days thinking he was responsible for the crash and that he was trying to escape.  His leg had ever since bothered him.  Over the next several years, until he retired in 1985, he also invited me to join him to work at the CBC stations in Ottawa and Montreal, and the Rogers broadcasting stations in Vancouver, Saskatoon, Moose Jaw, Regina, Winnipeg and Toronto.  I also worked with him on one occasion at CFRB in Toronto.  My final work with Joe was during the preparation of a 317C-3 50kW transmitter for WGN in Chicago.  Joe had just retired at the time, but he helped diagnose a strange problem experienced with their transmitter.  I have had a long relationship with WGN and it was fitting that my last work with Joe was on that project.
Joe had worked with the CE 317B transmitter series that had utilized a screen-modulated, 5kW driver stage followed by a 50kW Weldon linear amplifier.  He knew that classic screen modulated transmitters were inefficient.  He also knew that impedance modulation, the concept in which the Doherty transmitter was wrapped around, would be desirable as plate voltage would remain constant:  PA stage components would not need to standoff the DC plate voltage plus the AC modulation component, thus reducing component cost.  In addition, the PA tubes could be operated at telegraphy plate voltages rather than the reduced telephony values.  This, too, allowed for reduced parts cost since higher voltage, lower current designs are less expensive to manufacture than higher current designs.  There is no big iron (modulation transformer and choke) would be needed, providing a big advantage over some of the competition. Finally, operating costs are reduced as no high level modulator tubes are required and substantial power savings exist since no high level Class AB or B modulator stage is required; such as stage would operate with a maximum efficiency of 75%; more likely in the 65% region.
Joe realized that a screen modulated Doherty would result in an efficient high power transmitter that could be produced cost-competitive.  The basic concept involves the use of two tetrodes connected in a Doherty configuration.  The plates of the "Carrier" and "Peak" tubes are connected via an impedance-inverting 1/4
l interplate network.  In this configuration, the peak tube is connected directly to the antenna (via a lowpass filter) while the carrier tube is connected to the peak tube through the interplate network.  To understand the operation of the PA, one must consider the action of the interplate network.  Let' to produce 67.114kW input power (50kW output) make a few assumptions:  The PA will operate Class C with a plate efficiency of 74.5%, output power will be 50kW and input power is, therefore, 67.114kW.  Now let's consider a tube (the carrier tube) operating at 18,500 VDC and 3.63 Amps.  The effective plate load of this stage, Rp = Ep/(Ip*k) where Ep is the plate voltage, Ip is the unmodulated plate current, and k is the correction factor for conduction angle of the tube:  For Class C the value of k is 2.  Therefore, Rp=18500/(3.63*2) = 3000 Ohms.  Next, assume a 4:1 impedance transformation in the interplate network.  To obtain a 750 Ohms "antenna load impedance" (this will be transformed to 50 Ohms later), the reactance values of the interplate network must be the geometrical mean of the transformed impedance:  (3000*750)^.5 = 1500 Ohms.  If the peak tube is cutoff, the carrier tube (assuming a 750 Ohms load termination) will be operating into a 3000 Ohms load.  As the peak tube begins to conduct, it will contribute power and modify the load placed on the carrier tube.  At 100% positive modulation, the peak tube will contribute equal power to the load; therefore, it must be operating with an equivalent impedance of 1500 Ohms.  Due to the impedance inversion characteristics of the 1/4l interplate network, a 1500 Ohm at the peak tube load must also present a 1500 Ohm load to the carrier tube rather than the 3000 Ohm load at unmodulated carrier conditions.  Interestingly enough, the plate impedance has dropped to ½ of it's original value; therefore the carrier tube must be producing exactly twice the carrier power.  Since 100% envelope modulation is accomplished when the RF envelope is exactly doubled in voltage and power is equal to the square of the voltage divided by the resistance (P=E2/R), the required peak envelope power is 4X the unmodulated carrier level.  The carrier tube provides ½ of that required power and the peak tube, since it is also working into a 1500 Ohm impedance, provides the other half of the required power.  As far as the load is concerned, it see the 1500 Ohm peak tube in parallel with the 1500 Ohm carrier tube load:  750 Ohms.  It can, therefore, be seen that the plate voltage remains constant at 18,500 Volts and modulation is caused by the variable impedance of the peak tube acting upon the carrier tube through the interplate network.  On the negative modulation cycle, the screen voltage of the carrier tube is reduced towards the cutoff voltage which is slightly below 0 volts.  Also note that at carrier level, the carrier tube is in saturation and additional screen voltage yields little additional power.  The required carrier tube power contribution on the positive modulation cycle comes solely from the variable impedance presented to the tube by virtue of the interplate network and the peak tube (consider these elements as analogous to a variable resistor).  To match the 750 Ohm output impedance to a 50 Ohm antenna load, an output network is utilized.  The network acts as a harmonic filter as well as performing the impedance matching.  In general, the output network is usually a combination of a pi and a Tee network.  The advantage is that an intermediate impedance is chosen so that the transformation is not too large per stage.  Each section also operates with a fairly low Q but the overall Q of the network is the product of each section. If the each section operates with a Q of 4, the overall Q is 16; thus providing necessary suppression of the harmonic content associated with Class C operation.  In the 317C series transmitters, the intermediate impedance is 100 Ohms; the final transformation allows matching from 50 Ohms (or, potentially, lower) up to about 230 Ohms.* 
I have always liked the Sainton design and have built ham transmitter based on the approach.  My first was a lash-up using a pair of 4-65A Eimac tetrodes to generate about 100W of carrier.  Recently, I have rekindled my interest in AM.  I have "collected" a number of 250, 500 and 1000 Watt broadcast transmitters but the smallest Sainton transmitter produced was the 50kW 317C series.  I did see a 1MW version in the CE factory being readied for a customer in the Mideast - a very impressive beast - but lower power versions were never built.  I decided to design three versions that are now in the construction phase.  The first is a 160 meter transmitter that utilizes a modified-Sainton approach.  I am using a pair of 803 pentodes in a Sainton configuration except that the suppressors, rather than the screens, are modulated.  I have also designed a 75M version, also suppressor modulated, using a quad of 4E27 / 5-125B Eimac pentodes.  That rig will produce 375 watts of output power.  Finally, I am using a Motorola Micor 375 watt high power base station as the basis for a Sainton screen grid-modulated 40M transmitter.  This transmitter will produce 250 watts output power.  I chose to use the Micor station as the basis for the rig since it uses a pair of 8560AS tetrodes (the conduction cooled version of a 4CX250), has no fans, has a complete power supply and safety circuits, is contained in a reasonable cabinet, and, most importantly, I already had one.
The schematic shown above is the basic circuitry for the modified Micor Sainton transmitter.  The carrier and peak tubes can be seen as well as the interplate network.  Connected to the peak tube, through a DC blocking capacitor, is the Pi and Tee matching sections.  They will be contained in small box above the PA deck.  All VHF components are removed from the deck making room for the interplate coil and the carrier tube plate coil and capacitor.  The peak tube tuning capacitor will be located in the output network box along with the harmonic filter / output impedance matching networks (Pi and Tee networks).  The plate circuitry is identical to the Sainton design. The screen circuitry will be similar; however, rather than using a small 1:1 audio modulation transformer to couple the peak and carrier screen tubes to the audio and DC supplies, a pair of 30H chokes are used.  These units, from Hammond, are reasonably priced and can handle the screen current as well as standoff the screen DC voltage.  Also, since the screen impedance is kept rather high (in excess of 10k Ohms), the 829B dual tetrode audio stage is operated as a Class A voltage amplifier stage; the plate load is a resistor.  The peak and carrier tubes have separate audio stages thus allowing clipping of the positive audio cycle that drives the carrier tube.
The grid drive is significantly different than that used in the Sainton design.  Joe used a leading 90 degree network to generate the required RF phase lead to the carrier tube to compensate for the lagging plate network) and a single RF driver stage (a 4-400 developing about 400 Watts in the 317C series).  I chose to develop the 90 degree-separated RF signals by using a Johnson counter; starting with RF at 4 times the carrier frequency.  The 0 degree (I) reference signal is amplified by a small FET to generate about 2 W of drive for the carrier tube while the 90 degree lagging signal is amplified to the 2W level by an identical FET stage.  The power supplies for each FET are separately adjustable using LM317 regulators thus allowing exact drive levels to be generated and drive power to each tube to be matched.  Finally, each FET RF amplifier has a Pi network that transforms the output of the FET (about 12.5 Ohms) to 8k Ohms at each grid.
The audio is simple; op-amps are used to drive the Class-A 829B stages.  Each section of the 829B has separate bias and audio drive controls so that distortion due to PA tube mismatch can be minimized.  There is nothing special about the 829B; they are a convenient dual tetrode and I happen to have a good supply of them.  Plus, they were the RF PA in Buck's (K9RYW) 6M AM transmitter and I always thought they were a cool tube! The audio demands are low enough that a 6360 (9-pin miniature dual beam tetrode) or an 832A could be used.  Even old audio tubes such as the 6AQ5 or 6V6 could be used as well; with the miniature tubes care would be needed to make sure that the DC plate voltage would not exceed the prescribed limits of the tube.  The required RMS power for the carrier tube stage is about 5W while the peak tube is somewhat less.  Each tube is biased to dissipate about 20watts while having the capability to produce 10W of audio power to the screen grid load.  This far in excess of what is needed, but it is still a far cry from the 125 watts of audio needed for plate modulation of the same RF power if conventional plate modulation were utilized.
The schematic shown is for the original 75M design but the only difference in the 40M version is the use of a 15uH variable inductor and a fixed plate capacitor for the carrier tube.  The peak tube is still tuned by means of a variable capacitor. This saves the need of using a variable capacitor that can standoff the DC plus RF voltage present at the carrier tube.  The RF voltage at the peak tube is ½ that found at the carrier tube at carrier.
When the 40M version is complete, I plan to build a second one for 75M.  Finally, I am collecting the parts to build one "from scratch" using a pair of 4-1000A tetrodes for 75 meters.   It will be capable of "in excess" of 375W output carrier power.
Finally, I also present an analysis of the output network and the design centers for impedance.  If you have any questions, please feel free to email me using the link on my home page.  I also attach a photo of Joe at CKOC (holding his obligatory cup of coffee - the approved daytime beverage of choice) and some pictures of a very young me at WCCO along with Jim Stanley, Jerry "The Greek" Miller, Mark Persons; broadcast engineering consultant from Brainerd, MN, and some of the WCCO staff.

*As a side note, I did work on a pair of 317C-1 transmitters at WCCO in Minneapolis, MN with the great Jerry Miller, now SK, that matched to 230 Ohm open wire feeder line.  I also worked on the 317C-2 at KFAB in Omaha (with John Bruna and Jerry McKenna; both former chief engineers for the blowtorch of Omaha), but that transmitter fed the phasor at 50 Ohm, which, in turn, feed the 3-tower array with 230 Ohm open wire transmission line.  Open wire feeder lines are uncommon these days, having been replaced with either 50 / 51.5 / 52 Ohm concentric hardline or Heliax.  But, I do remember Ed Butterbaugh telling me of having his 50kW transmitter at CKLW knock off the air one night while he was doing night work on a very early Harris MW-50 transmitter.  That transmitter was always a source of concern for him; he always treated it with respect and caution.  He thought he had done some serious damage, but the transmitter came right back up upon resetting it.  Nothing seemed to be wrong, so he relaxed.  The next morning, he walked to the towers to see if he had done some damage in one of the tuning units.  On his way to one of tuning huts he noticed the upper half of an owl directly below the open wire feeder line.  Legend has it that the owl had crapped out the remains of his last meal while perched on the upper cage of the transmission line.  In doing so, he completed a path between upper ground cage and the center feeder pair, thus completing a path for the 7200 V RF signal, blowing his bunghole clean off!
Click on any image for a fullsize pop-up.
317C Simplified Schematic
Click image for a readable view

Doherty IEEE Paper

Doherty Sales Brochure

Doherty Transmitter Ad
WGN 317C-3 PA Cabinet
(no longer on site/in service)
WGN Radio - Chicago
317C-3 Doherty Transmitter

(No longer in Service)